*Review* **Label-Free Electrochemical Biosensor Platforms for Cancer Diagnosis: Recent Achievements and Challenges**

**Vildan Sanko <sup>1</sup> and Filiz Kuralay 2,\***


**Abstract:** With its fatal effects, cancer is still one of the most important diseases of today's world. The underlying fact behind this scenario is most probably due to its late diagnosis. That is why the necessity for the detection of different cancer types is obvious. Cancer studies including cancer diagnosis and therapy have been one of the most laborious tasks. Since its early detection significantly affects the following therapy steps, cancer diagnosis is very important. Despite researchers' best efforts, the accurate and rapid diagnosis of cancer is still challenging and difficult to investigate. It is known that electrochemical techniques have been successfully adapted into the cancer diagnosis field. Electrochemical sensor platforms that are brought together with the excellent selectivity of biosensing elements, such as nucleic acids, aptamers or antibodies, have put forth very successful outputs. One of the remarkable achievements of these biomolecule-attached sensors is their lack of need for additional labeling steps, which bring extra burdens such as interference effects or demanding modification protocols. In this review, we aim to outline label-free cancer diagnosis platforms that use electrochemical methods to acquire signals. The classification of the sensing platforms is generally presented according to their recognition element, and the most recent achievements by using these attractive sensing substrates are described in detail. In addition, the current challenges are discussed.

**Keywords:** label-free electrochemical detection; electrochemical sensor; cancer diagnosis

#### **1. Introduction**

Cancer, which causes premature death in almost all countries of the world, maintains its position at first place even if it is sometimes replaced by cardiac disease. In particular, due to demographic effects and the trends of these effects in cancer incidence in different locations, it is expected that instances of cancer will approximately double in the next 50 years globally. However, cancer does not affect the population of all countries at the same rate, and it is predicted that there will be a higher increase in countries that can be classified as low–middle income [1,2]. The Global Cancer Statistics 2020 report shows that the most common cancer in men is prostate cancer, followed by lung cancer, colorectal cancer and liver cancer, whereas breast cancer and cervical cancer are the most commonly diagnosed cancers in women. In addition, according to the same report, what is striking is that an estimated 19.3 million new cancer cases were detected worldwide and approximately 10.0 million deaths were calculated due to cancer only in 2020 [3].

Regardless of the type, the diagnosis and treatment of cancer at an early stage is very important to reduce both cancer incidence and mortality rates. As the traditional cancer detection method, enzyme-linked immunosorbent assay (ELISA), which detects cancer-specific protein biomarkers and is called the gold standard, is widely known [4]. Also, genomic- and proteomic-based molecular methods such as polymerase chain reaction (PCR), immunohistochemistry (IHC) and radioimmunoassay (RIA) are used for cancer diagnosis [5]. In addition, various clinical tools such as magnetic resonance imaging (MRI), positron emission tomography (PET), endoscopy, sonography, X-ray, computed tomography (CT) and biopsy are extensively utilized [5–7]. However, although the mentioned

**Citation:** Sanko, V.; Kuralay, F. Label-Free Electrochemical Biosensor Platforms for Cancer Diagnosis: Recent Achievements and Challenges. *Biosensors* **2023**, *13*, 333. https:// doi.org/10.3390/bios13030333

Received: 1 December 2022 Revised: 17 February 2023 Accepted: 23 February 2023 Published: 1 March 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

methods and technologies are efficient, most of them are expensive, time-consuming, invasive and limited to the laboratories of some hospitals. Especially with imaging methods, detecting cancer tumors below the millimeter size may be inconclusive. Similarly, invasive methods such as biopsy have the same problems and difficulties in diagnosing early-stage cancer tumors [6].

Early-stage cancer diagnosis increases the survival rate of the patient [8]. In addition, early diagnosis offers several advantages that lead to more appropriate treatment for the patient and even reduce the severity of the cancer [9]. One of the biggest problems limiting early diagnosis in cancer detection is the nonappearance of obvious symptoms in the early stages of cancer; the other is not detecting sufficiently sensitive biomarkers [10]. From an economic point of view, it is known that the costs used for cancer treatment are increasing rapidly, and this cost is expected to increase up to USD 246 billion by 2030. Therefore, detecting cancer at an early stage can reduce the potential economic burden for the patient and society [11]. There is a crucial need to develop low-cost, sensitive, non-invasive (bio)sensors for early-stage cancer diagnosis. In general, biological biomarkers show the genetic characteristics of cancer cells and diagnosis/monitoring of cancer with a biomarkerbased biosensor is seen as one of the most promising approaches [12]. These biomarkers can be deoxyribonucleic acid (DNA), ribonucleic acid (RNA), hormones, protein, enzymes and specific cells that can be found in human bodily fluids such as urine, serum, plasma and blood [13,14]. For this aim, electrochemical sensors have been widely used in the field of cancer diagnosis. There are valuable studies in the literature that include various approaches to detect different cancer types such as breast cancer [15], ovarian cancer [16], prostate cancer [17], pancreatic cancer [18] and lung cancer [19]. Electrochemical sensors are prominent tools because they are sensitive, selective, fast, cost-effective, instrumentable and can be performed as on-site analysis [20,21]. Different electrochemical sensing methods such as potentiometric, amperometric, conductometric, impedimetric and voltammetric are used to convert the obtained signal into useful analytical data. Detection methods in biosensors can be grouped as labeled or unlabeled depending on the use of labels as electroactive molecules or nanomaterials. However, labeled systems are complex and expensive as they require an extra labeling process. Conversely, label-free biosensors have shorter analysis time and simplicity and they offer good advantages [22–24].

In this review, current label-free cancer diagnosis platforms in the literature, including the last three years, in which the electrochemical method is used as a signal converter, are detailed. Biorecognition elements and mechanisms used in biosensor design for cancer diagnosis are emphasized. In addition, the immobilization method and immobilization matrices, which are important parameters for the activity and stability of a biorecognition element, are also the subject of this study. Finally, current challenges and future perspectives are discussed.

#### **2. Electrochemical Techniques as a Sensing Mechanism**

Electroanalytical studies are included as a sub-discipline of analytical chemistry, which includes charge transfers in addition to oxidation–reduction reactions [25]. Biorecognition elements, which are one of the parameters that make up the biosensor, are important components for the analyte to be detected. This component needs to be used with a converter so that a meaningful signal can be generated according to the analyte concentration [26]. In an electrochemical transducer system, detectable signals such as current, potential, impedance and conductivity are obtained as a result of the interaction of samples with a bioreceptor. In connection with these signals, electrochemical biosensors are included in various classifications such as amperometric, potentiometric, impedimetric and conductometric. In addition, voltammetric techniques are important and sensitive techniques to help analyte determination [27,28]. Electrochemical detection systems, which provide analytical advantages such as low cost, simple design and portable features, are platforms that can make sensitive and selective detections even in body fluids with complex matrices such as serum [29,30]. Therefore, these detection systems have attracted great attention in biosensor technology owing to their unique properties.

Voltammetric techniques have been commonly utilized. For example, differential pulse voltammetry (DPV), where a pulse is applied to the electrode and provides current measurement. Before the pulse is applied and at the end of each pulse, the current is measured and the difference between the currents is calculated. This procedure effectively reduces the background current due to linear increase, thus resulting in a faradaic current with no capacitive current. The biggest advantage of DPV is a low capacitive current, which leads to high sensitivity. Small steps in DPV also lead to narrower voltammetric peaks, and therefore, DPV is often used to distinguish analytes with similar oxidation potentials. Thus, this technique is preferred in electrochemical cancer biosensors as it exhibits very sensitive properties against the reduction and oxidation of bio-electrochemical species [31,32]. Cyclic voltammetry (CV) is one of the most common methods to obtain information about redox potentials and to investigate the mechanisms and kinetic parameters involved in the reactions of electroactive analytes. In this method, the current between the working and counter electrodes is monitored, but changes in the potential of the working electrode due to the reference electrode are also controlled [33]. In the electrochemical impedance spectroscopy (EIS) technique, the impedance change in both faradaic and non-faradaic modes is measured. As an example, in the measurement system in the faradaic mode, the change in the electron transfer rate caused by the aptamer–analyte interaction is examined. In measurement systems taken in non-faradaic mode, the surface capacitance change due to the aptamer–analyte connection is detected [34]. In the amperometric technique, the working electrode is kept at a constant potential that is sufficient to reduce or oxidize the analyte of interest and the resulting current is monitored over time. Potential selection is critical as only one potential is applied in this technique. Due to the monitoring of current over time at a constant potential, all dynamic changes in the current can be observed [31]. On the other hand, in a potentiometric system based on potential measurements, the principle of changing the potential with the concentration of the analyte is used in the measuring system with the help of a reference electrode with a fixed electrode potential. Besides cancer diagnosis, electrochemical techniques are also highly preferred in routine laboratory analysis and clinical and environmental monitoring analysis [35].

When electrochemical techniques are compared with each other, it is observed that each of them can have limitations in different aspects. For example, the sensitivity of the potentiometric method depending on the environment and temperature is an important limitation. For the limitations of other methods, it can be said that redox elements are needed in the amperometric technique, whereas EIS is sensitive to the environment and requires theoretical stimulation for data analysis [36]. Voltammetric techniques show high selectivity and sensitivity due to the voltammetric peak potential applied to the analyte. However, one of the major problems encountered with these techniques is obtaining overlapping voltammetric responses due to very similar oxidation peak potentials. Various recently developed materials and protocols are used to overcome this problem [37]. Besides this, choosing an appropriate sensing technique for analyte detection can minimize the limitations. Additionally, parameters such as pretreatments applied to the working electrode and the biofunctionality of the electrodes can have a great impact on the precise and effective determination [34].

The electrochemical transformations occurring at the interface of the label-free sensing platform are determined by the affinity between the analyte and the biorecognition elements, regardless of the use of labels [29]. Thanks to the detectable signals obtained by electrochemistry, these techniques are widely preferred not only for cancer detection and follow-up but also for the accurate and sensitive detection of analytes in areas such as the detection of different diseases and environmental and food control [38–43]. In an electrochemical biosensor, two different reactions can be observed as a result of the interaction of the electrode surface and the analyte: the first is the positive read signal called "signal-on", and the other is the negative read signal called "signal-off" [44,45].

Label-free electrochemical biosensors are particularly interesting and important for studies in the biomedical field. In this type of electrochemical biosensor, the information in the reaction is converted into an electrical signal by the direct transfer of electrons between the electrode surface and the biorecognition elements as a result of the interaction between the biomolecule and the analyte [46]. Additionally, the surface characteristics of the electrodes significantly support improving the sensitivity of the biosensor. Therefore, surface modification is also important for good analytical performance. At this point, nanomaterials have been in the scope of scientists. The use of nanomaterials of different sizes, shapes and morphologies together with electrochemical transducers makes it possible to improve properties. Nanowires, metal/metal oxide nanoparticles, carbon nanotubes, graphene or graphene-like structures and conductive nanostructures such as polymers have provided more sensitive biosensors with high surface/volume ratios [47,48]. The scope of this study mainly covers the discussion of the technological developments and also problems/limitations in the development of label-free biosensors containing different biorecognition elements to serve cancer diagnosis.

Despite a lot of effort and good progress in the field of biosensors, it is seen as an inconsistency that only a few of them find a place in the commercial market. The first example of commercial biosensor is the enzymatic glucose biosensor, which is expected to have a market of USD 38 billion by 2027 [49]. This biosensor currently holds approximately 75% of the global biosensor market. There are still outstanding challenges, both to overcome the current constraints and to making the products available commercially. Firstly, understanding the mechanisms of biocatalytic work and charge transfers and also improvements in the properties of biorecognition elements that provide selectivity should be considered. In addition, the use of various nanoparticles and hydrogels has been reported to improve existing deficiencies, although not completely [50,51]. For this purpose, researchers are conducting detailed studies about the effects of parameters on biomolecule (such as enzymes) immobilization and the effect of these parameters on the performance of the biosensor platforms [52]. However, since the biomolecule redox reaction processes are still not fully known, in situ inspection techniques are used for evaluation [53]. Some of the obstacles in the transformation of biosensor studies from laboratory to commercial products are performance and nonspecific surface interaction problems in various body fluids, which have complex matrices [49].

Although electrochemical methods provide several advantages, each method may also have limitations. It is particularly important to focus on and discuss these limitations to put the developed technologies into clinical practice. Reducing or overcoming all the disadvantages could help to develop more accurate and sensitive electrochemical cancer biosensors. More effective platforms for early diagnosis can be created with a multidisciplinary study. In addition, the detection of new cancer biomarkers will greatly benefit the facilitation of early-stage diagnosis and thus the management and control of the cancer disease process. It is expected that the label-free electrochemical methods will increase in reliability after the difficulties we have mentioned have been overcome. As a result, they will find a regular use in the clinical field. To strengthen this reliability, novel and advanced electrochemical cancer biosensors with different perspectives need to be developed.

#### **3. Importance of a Label-Free Electrochemical Sensing Platform**

A typical electrochemical biosensor is expected to convert signals that are related to the presence of the analyte molecules into measurable quantities with the help of the biorecognition unit. In some cases, various markers/labels or tags are used for the detection of the analyte and the signal is obtained in conjunction with them. These biosensor systems are called label-based biosensors. The use of these labels, which are commonly classified as radioactive-, fluorescent- or chemiluminescence-based, is time consuming and laborious because it requires an extra process. More importantly, it is thought that in this case, the affinity between the biorecognition element and the analyte may be adversely affected. To eliminate these limiting factors, unlabeled detection systems have become highly preferred in recent years. If a direct measurement is made with the biorecognition system, this is called a label-free biosensor system [54].

In a typical label-free biosensor design, sensing can be performed by converting it to optical [55], mechanical [56] or electrical [57] signals and more accurate information can be provided as biorecognition systems are directly used. Within this classification, electrochemical label-free biosensors can be used actively in the field and can be also implanted in the body to detect biological analytes, increasing their future potential [58]. Various electrodes with different biorecognition elements and composite designs have been developed for analytes such as gliotoxin [59], microRNA (miRNA) [60], bacterial pathogens [61] and aflatoxin-B1 [62] in this biosensor group, which combines the advantages of both the electrochemical method and the label-free platform. For the continuation of the remarkable progress of the mentioned electrochemical label-free biosensors, a better understanding of the current working processes is required for the creation of sensitive and selective biosensing systems that find application in wider use. Based on this idea, we have detailed and discussed cancer studies classified on different biorecognition elements.

#### **4. Biorecognition Elements for Label-Free Electrochemical Cancer Diagnosis**

Basically, antibodies, aptamers, nucleic acids and cells are immobilized to surfaces/ interfaces to achieve affinity and selective biorecognition. In this part, the classification of the label-free electrochemical cancer detection systems is divided into categories according to the type of the biorecognition element. Besides this classification, electrode material and the detection technique are also highlighted. Figure 1 demonstrates the schematic presentation of the label-free electrochemical cancer biosensors with successful electrode modifications, such as nanotechnology-based materials, biorecognition immobilization protocols and some of the powerful electrochemical detection techniques.

**Figure 1.** Label-free electrochemical cancer biosensors: electrode modifications such as nanotechnology-based materials, biorecognition immobilization protocols and some of the powerful electrochemical detection techniques.

#### *4.1. Nucleic-Acid-Based Label-Free Cancer Biosensors*

Nucleic acids are natural biopolymers that store genetic information in humans and almost all organisms [63]. Nucleic acids include DNA and RNA, which are composed of nucleotides. The well-known specific hybridization feature between nucleic acid chains also constitutes the main detection principle of DNA biosensors [64]. The development of biosensors for the detection of DNA sequences is important because of its application in gene identification, molecular diagnosis and drug screening [65]. Nucleic acids can be affected by environmental conditions such as temperature and pH [66]. Nevertheless, in many studies electrochemical signal amplification by means of nucleic acids has been successfully developed for cancer applications [67,68].

Studies in recent years show that excessive secretion of microRNAs is associated with malignancies that cause cancer [15,69–71]. In one study, Zhao et al. proposed MXenemolybdenum disulfide (MoS2) constructs with thionine and gold nanoparticles for the label-free electrochemical detection of microRNA-21, which plays an important role in the emergence of cancer associated with proliferation/differentiation in cells. The modification of the prepared nanocomposite on glassy carbon electrode (GCE) was performed by drop casting. Then, the hairpin capture probe was dropped onto the modified electrode. The hybridization event was carried out in the presence of the target and a hairpin probe 2. The detection method was square wave voltammetry (SWV). Thanks to this structure, the capture probe immobilization was improved, the amplification of the electrochemical signal was achieved and microRNA-21 detection in the linear measurement range of 100 fM to 100 nM was obtained with a detection limit of 2 fM [72].

Pothipr et al. described a gold nanoparticle-dye/poly(3-aminobenzylamine)/twodimensional molybdenum selenide (MoSe2)-based electrochemical label-free biosensor for breast cancer diagnosis that could detect cancer antigen 15-3 and microRNA-21 simultaneously. Based on the complexity of the immune system in the human body and therefore the inadequacy of cancer assays using single biomarker systems, they introduced this bidirectional detection platform produced on a two-screen printed carbon electrode. DPV was used for the evaluation of the electrochemical performance of the biosensor and the detection limit was found to be 1.2 fM for microRNA-21 detection [73]. Jafari-Kashi et al. presented a DNA biosensor for the detection of cytokeratin 19 fragment 21-1, which is associated with lung cancer. They preferred DPV as an electrochemical technique to examine the interaction between the capture probe and target using GCE modified with reduced graphene oxide, polypyrrole, silver nanoparticles and single-stranded DNA (ssDNA). With this technique, no peak was detected before DNA hybridization, but a distinctive peak was obtained after hybridization according to the oxidation of guanine. They declared that the label-free DNA biosensor showed a good result for detection of cytokeratin 19 fragment 21-1, with a wide linear measurement range and a 2.14 fM limit of detection [74]. Avelino et al. presented a polypyrrole film containing DNA immobilized chitosan/zinc oxide nanoparticles for the diagnosis of myelocytic leukemia by BCR/ABL fusion gene detection. Oxidation and reduction steps were observed in line with the voltammetric measurements taken in 10 mM [Fe(CN)6] 3–/4–. It is also stated that the biosensor was designed as a result of bioactivity tests and could be used as a new biosensing platform that enabled the identification of early-stage cancer [75].

#### *4.2. Aptamer-Based Label-Free Cancer Biosensors*

Aptamers are single-stranded DNA or RNA molecules that can usually be synthesized using an in vitro method. In fact, RNA-based aptamers were first found in 1990, followed by DNA-based aptamers, with the development of in vitro selection/amplification for the isolation of RNA sequences that could specifically bind to molecules [76]. In aptamerbased electrochemical sensors, it is necessary to be able to detect the conformational changes caused by the presence of the aptamer on the electrode surface for obtaining a signal [77]. Aptamers are widely used in the development of biosensors due to their high specificity, easy synthesis, simple modification and high chemical stability [78]. They offer the advantages of more cost-effective production, easy modification and thermal stability, especially when compared with monoclonal antibodies. After the aptamers are immobilized on a conductive matrix, their redox-active moieties allow the formation of aptamer–target complexes and thus the design of various electrochemical biosensors with the realized electron transfer properties [76]. The most important problem in this electrochemical process can be the generation of a determinable signal between the target

analyte and the aptamer. In order to solve this problem, electrochemically active labeling units such as hemin [79], ferrocene [80] and methylene blue [81] have been introduced. However, labeling of aptamers introduces known disadvantages such as time consumption, poor affinity performance and cost [82].

In recent years, aptamers have attracted great interest in electrochemical label-free biosensor design, which has applications in the diagnosis and follow-up of various cancers. Label-free aptasensors also require an increased surface area to improve weak signal intensity. Nanomaterials contribute greatly to increasing the surface area because they act as electron-transfer tunnels, which increase the electrical communication between the redox regions of the aptamer and the electrode surface [83]. Zhang et al. developed a label-free aptasensor for the detection of cancer antigen 125 by immobilizing aptamer on the surface of nickel hexacyanoferrate nanocubes/polydopamine functionalized graphene. DPV was utilized for electroanalytical studies in this work, which was designed to provide a detectable electrochemical response with the help of increasing surface area and conductivity. Thanks to the insulating structure formed as a result of the combination of aptamer and cancer antigen 125 (CA125), or in other words aptamer–CA125 complex, the peak current value decreased as the analyte concentration increased. The linear measurement range and limit of detection were calculated as 0.10 pg mL−1–1.0 μg mL−<sup>1</sup> and 0.076 pg mL−1, respectively. The measurements were carried out in phosphate buffer solution (PBS) [82]. In another study, a paper-based electrochemical label-free aptasensor was fabricated for the detection of epidermal growth factor receptors. Interestingly, the concept of origami as a valve for a paper-based biosensor was used in this study. As a result of the biochemical reaction, the data became an electrochemical response with the presence of the nanocomposites containing amino functionalized graphene/thionine/gold. This system in the form of origami was designed to increase the penetration of the liquid and shorten the time taken for flow, resulting in a shorter test time. The linear concentration range obtained with the sensor was from 0.05 ng mL−<sup>1</sup> to 200 ng mL−<sup>1</sup> and it had a detection limit of 5 pg mL−<sup>1</sup> [84].

#### *4.3. Antibody-Based Label-Free Cancer Biosensors*

Antibodies are protective proteins produced by the immune system in response to the presence of antigens, including pathogens and toxic materials [78]. Biosensors that offer the advantages of high binding affinity and specificity and use antibodies for biorecognition take the advantage of the high affinity between antibodies and antigens for detection and are called immunosensors [85,86]. However, there are some parameters that limit their use. Apart from being adversely affected by environmental conditions and having difficulties for storage, it can be said that the production of polyclonal antibodies in animals is difficult and costly. Moreover, polyclonal antibodies may lack selectivity as they can have affinity for different epitopes [87]. With the help of the new and improved sensor interfaces developed in recent years, some disadvantages have been overcome and many antibodybased sensitive and selective label-free electrochemical biosensors have been designed. Also, these limitations pave the way for the development of new forms of biorecognition units that can replace antibodies, thus introducing new biosensor projections to the field.

Various electrochemical techniques have been used for antibody-based biosensors for gastric cancer [88], breast cancer [89–92], ovarian cancer [93–96], bladder cancer [97], colorectal cancer [98], lung cancer [99], prostate cancer [100–105], liver cancer [106] and more. In a study for a label-free electrochemical immunosensor developed for early-stage detection of prostate cancer, the surface of the indium tin oxide electrode was firstly coated with chitosan and reduced graphene oxide, and then the specific polyclonal anti-prostatespecific antigen (PSA) antibody as a recognition element was immobilized on the surface. It was determined that a linear decrease had been observed in the peak current values of the redox probe by using DPV with increasing concentrations of the antigen. It is reported that the linear measurement range determined for prostate-specific antigen detection was between 1 pg mL−<sup>1</sup> and 5 ng mL<sup>−</sup>1, and the limit of detection was 0.8 pg mL−<sup>1</sup> [107].

CA125 was detected by DPV using a layer-by-layer assembly of ordered mesoporous carbon, gold nanoparticles and MgAl-layered double hydroxides containing ferrocene carboxylic acid composite. It is explained that the conductivity increased significantly with the addition of the ferrocene component to the composite. The electrochemical performance of the biosensor was determined based on the change of the peak current observed in the voltammogram at +0.27 V according to the ferrocene in the presence of different CA125 antigen concentrations. It is stated that the peak current value obtained with the increase in the CA125 concentration changed inversely, since the complex formed between the antigen and the antibody. The linear measuring range and limit of detection of the biosensor were described as 0.01 U mL<sup>−</sup>1–1000 U mL−<sup>1</sup> and 0.004 U mL<sup>−</sup>1, respectively [108]. A label-free sandwich type biosensor was developed for the electrochemical detection of cytokeratin fragment antigen 21-1 (CYFRA 21-1), a lung cancer biomarker. An antibody–antigen– antibody sandwich structure was formed between the 4-(2-trimethylsilylethinyl)benzoic acid gold electrode used as a bridge and the poly(ε-caprolactone)-b-poly(ethylene oxide) copolymer. The linear concentration range and limit of detection for the sensor determined by electrochemical impedance spectroscopy were declared as 1.0 pg mL−<sup>1</sup> to 10 ng mL−<sup>1</sup> and 0.125 pg mL−1, respectively. According to the impedance results, the electrochemical responses showed a linear response with the concentration of CYFRA 21-1 [109].

Liu et al. developed a gold nanoparticle/polyethyleneimine/reduced graphene oxide nanocomposite for the electrochemical detection of matrix metalloproteinase-1, a cancer biomarker, based on the knowledge that gold nanoparticles were supportive in maintaining the reversibility of redox reactions in electroanalytical reactions. They determined that the biosensor performance obtained by DPV had an operating range of 1 ng mL−<sup>1</sup> to 50 ng mL<sup>−</sup>1. In this work, the peak current value obtained from voltammetry decreased due to the increased antigen concentration blocking on the electrode surface. In the electrochemical measurements taken in 5 mM Fe(CN)6 <sup>3</sup>−/4<sup>−</sup> medium, it is stated that an insulating layer was formed due to the antigen–antibody complex, and therefore, a repulsive electrostatic interaction occurred between the antigen and Fe(CN)6 <sup>3</sup>−/4<sup>−</sup> [110]. Zhu et al. also developed a carbon-based nanocomposite to take advantage of its high surface area and good conductivity properties. The surface was used for the construction of an immunosensor for the detection of alpha-fetoprotein, which is a liver cancer biomarker. They calculated a linear measurement range of 0.10 ng mL−<sup>1</sup> to 420 ng mL−<sup>1</sup> and a limit of detection of 0.03 ng mL−<sup>1</sup> using square wave voltammetry, a method that could suppress background current and provide sensitivity to the biosensor system [106].

#### *4.4. Cell-Based Label-Free Cancer Biosensors*

The use of cells as a biorecognition element dates back to the early 1970s and it is still preferred today. Cells offer an interesting alternative to other biorecognition units such as antibodies, enzymes and nucleic acids thanks to their relatively easy production and lower cost than antibodies and purified enzymes. As an example, since whole cells offer a multi-enzyme alternative, they can be preferred in the development of biosensors for the simultaneous determination of various analytes. In addition, cell-based biosensors enable in situ monitoring using suitable substrates [78,111,112]. However, some limitations such as maintenance and immobilization of cells can arise [113].

Human cervical carcinoma (HeLa) cells were used as a biorecognition unit in an electrochemical label-free cytosensor to evaluate the anticancer activity of pinoresinol, which had biological properties such as anticancer, anti-inflammatory and antifungal effects. HeLa cells were immobilized on a GCE surface modified with multi-walled carbon nanotubes and gold nanoparticles, and the performance of the biosensor was evaluated by electrochemical impedance spectroscopy with different pinoresinol concentrations. The limit of detection value for the biosensor, which showed a linear correlation with the pinoresinol concentration range of 102 to 10<sup>6</sup> cells mL<sup>−</sup>1, was reported as 10<sup>2</sup> cells mL−<sup>1</sup> [114]. Another cell-based label-free electrochemical biosensor was developed to investigate the interactions of cancer cells (HepG2 cells and A549 cells) with molecules and to screen anticancer

drugs. Cancer cells were immobilized on the GCE coated with N-doped graphene–Pt nanoparticles–chitosan and polyaniline. It is stated that this electrode surface might be suitable for examining different cell lines by changing the targeted cells as a result of the electrochemical properties examined by DPV with its large surface area and catalytic properties [115].

Liu et al. carried out the detection of cell surface glycan that played an important role in processes such as cancer cell metastasis by means of a nano channel ion channel of porous anodic alumina hybrid combined with an electrochemical detector. Thus, the enhanced ionic current caused by the array nano channels along with the ionic current rectification gave a precise current response. The alumina was functionalized with aminopropyltriethoxysilane and glutaraldehyde to immobilize the cell surface glycan. The linear working range was obtained from 10 fM to 10 nM, and the limit of detection was calculated to be approximately 10.0 aM. It is stated that this biosensor was a promising alternative that could be used in cancer diagnosis and an important platform for label-free detection of cell surface glycan [116].

Despite the advantages of cell-based electrochemical biosensors, there are also various disadvantages faced by designers such as reproducibility and inability to selectively place cells at detection sites [117]. In addition, some difficulties in terms of electrochemical techniques such as amperometric and impedimetric have been reported in the literature. For example, the difficulties often observed in electrochemical impedance spectroscopybased studies are that the measured electrochemical response is the total change produced by a set of cells and poor selectivity. Emerging technology, nanomaterial selection, new immobilization matrices, integration of different transducer mechanisms and advances in the control of the sensor interface are some of the promising approaches to overcome these challenges [105,106].

#### **5. Immobilization Strategies of Biorecognition Elements**

Biorecognition element immobilization or its integration is one of the important processes to be considered, since this step thoroughly affects the analytical performance of all types of biosensors. The efficient immobilization of the biorecognition element is a process applied to overcome the problems such as loss of activity and stability by integrating biomolecules into a suitable support material. The immobilization methods are classified as adsorption, covalent bonding, cross-linking, etc., according to the type of the biomolecule to be immobilized and the structure of the immobilization surface [118]. These methods are illustrated in Figure 2.

In Table 1, the immobilization methods used by some of the studies within the scope of this review are indicated. Some cancer detection studies in the literature for recent years, different biorecognition units, other biosensor components and the parameters used in these studies are listed. Metals, metal oxides, conductive polymers, biopolymers, carbon-based structures, quantum dots and their composites [93,100,107,109,119,120] have been used as the immobilization matrices for label-free electrochemical cancer biosensors. In general, electrostatic interactions can have negative effects on the stability of the biorecognition element or the repeatability of the biosensor [121,122]. However, these methods, which have very simple processes, are still actively used in the surface immobilization of many electrodes. The entrapment method also offers specific properties and contributes to the improvement of chemical and thermal stability. However, leakage and low biological activity limit this method. To overcome the leakage problem, crosslinkers are preferred in the immobilization step. However, at this stage, excessive chemical requirements are necessary [123].

**Figure 2.** Various immobilization methods for the biorecognition elements.

In the study of Yaiwong et al., an immunosensor for label-free electrochemical cancer detection was developed. Electrostatic interaction was carried out for the immobilization of the anti-metalloproteinase-7 (MMP-7) capture antibody, which was used as a biorecognition element, on the surface of the screen-printed carbon electrode (SPCE) coated with twodimensional (2D) MoS2/graphene oxide [124]. More commonly, immobilization methods by covalent or cross-linking over carboxyl or amine groups are robust and reproducible ways to obtain an effective biosensor interface. Glutaraldehyde or carbodiimide structures that act as bridges in these binding reactions are preferred [121]. As an example, Yan et al. coated the surface of an indium tin oxide electrode with chitosan-modified reduced graphene oxide nanocomposite for prostate cancer detection. In order to detect prostatespecific antigens with this biosensor, they immobilized the recognition antibodies onto the electrode surface by covalent bonding. Chitosan naturally provided a large number of amino groups to the electrode surface, and glutaraldehyde, a bifunctional bridge, was used for covalent immobilization of the anti-PSA antibody with amino groups. Thus, a label-free electrochemical immunosensing platform based on antibody–antigen affinity was developed [107].

Echeverri et al. immobilized the anti-β-1,4-galactosyltransferase-V (β-1,4-GalT-V) antibody biorecognition element on the self-assembled monolayer (SAM)-coated SPCE by covalent bonding for the detection of colorectal cancer. The SAM provided a carboxylic acid group that allowed for antibody binding [98]. Generally, N-(3-dimethylaminopropyl)- N- -ethylcarbodiimide hydrochloride (EDC) and N-hydroxysuccinimide (NHS) pairs are used for this type of covalent bonding. In this way, a bridge is formed between the amine and carboxyl groups and a high binding efficiency is achieved [121]. Although covalent bonding seems to offer good efficiency and is an advantageous method, it can also have various disadvantages in some cases. For example, denaturation may occur due to the undesirable site orientation of the biorecognition element, and in addition, the bridging compounds are needed to use in the covalent bonding reaction. Therefore, there can be a decrease or disappearance of the biocatalytic effect expected from the biorecognition unit [125]. Moreover, covalent bonding, which causes a tight binding, can also restrict the movement of the biorecognition elements, which may also cause a loss of activity [126].


**Table 1.**

Electrochemical-based

 label-free biosensors for cancer detection:

biorecognition

 elements, sensor platforms,

immobilization

 methods




*Biosensors* **2023**, *13*, 333

Although the immobilization of biorecognition elements on the surface of the biosensing platform is a very important step for the design of sensitive, selective and long operational lifetime biosensors, it is clear that each method has several advantages and disadvantages. Various factors such as the immobilization matrix and the charge or functional groups of the biorecognition units guide the selection of the appropriate method, and thus, effective interfaces are created.

#### **6. Label-Free Electrochemical Cancer Biosensors for Point-of-Care Applications**

Label-free electrochemical biosensors have a high capability of being adapted into point-of-care (POC) systems that can be used for outside the laboratory testing to minimize the need for healthcare services such as hospitals [14,143–145]. In POC testing particularly, microfluidic devices have attracted great attention lately for effective and accurate cancer diagnosis owing to their ability to separate analytes at a good resolution in a rapid reaction time and to minimize the handling errors and costs [143]. As a result, promising detection systems with high performances are acquired with the elimination of the need for trained personnel. Recently, in the study by Keyvani et al., a POC sensing device for the detection of cervical cancer was developed for whole blood. This system identified cancer circulating DNA with high purity by the help of a graphene oxide-dependent electrochemical sensor platform by using differential pulse voltammetry [146]. In another study, Ming et al. fabricated a cellulose-paper-based POC testing with the modification of amino redox graphene, thionine, streptavidin integrated gold nanoparticles and chitosan for the detection of biomarker 17β-estradiol, which may be associated with breast cancer. The detection strategy, realized with differential pulse voltammetry in phosphate buffer solution, was carried out via the interaction of the target biomarker and its biotin-modified aptamer on the surface of the paper. The linearity of the label-free sensor was between 10 pg mL−<sup>1</sup> and 100 ng mL<sup>−</sup>1, with a limit of detection value of 10 pg mL−<sup>1</sup> [147].

Besides microfluidic devices, multiplex systems that can detect multiple analytes associated with cancer have several advantages in terms of label-free point-of-care testing. As an example, Kuntamung and his colleagues achieved simultaneous detection of breast cancer biomarkers: mucin1, cancer antigen 15-3 and human epidermal growth factor receptor 2 depending on the formed antibody and antigen interactions. For this purpose, redox species and antibody-conjugated polyethylenimine-modified gold nanoparticles were utilized as the modification elements of a SPCE. In addition to multiplex detection performance, the label-free biosensor kept 90% of its initial responses obtained via voltammetry [92]. In another approach that contained the fabrication of a flexible screen-printed electrode system, carcinoembryonic antigen was detected on graphene–ZnO nanorods deposited on a polyethylene terephthalate substrate with a screen-printed electrode by Chakraborty et al. ZnO nanorods were functionalized with aptamers and the resulting surface improved the mass transport through the electric field application. This system was integrated into smartphone interface technology and a handheld potentiostat. The linearity of the label-free sensor was between 0.001 pg mL−<sup>1</sup> and 10 pg mL−1, with a limit of the detection value of 1 fg mL−<sup>1</sup> by using electrochemical impedance spectroscopy. The results were also validated using a commercial ELISA kit [148].

The use of label-free POC testing in cancer diagnosis is in increasing demand in recent years since POC systems yield rapid decisions, more frequent testing to monitor wellness, eliminate the need for trained staff and utilize small specimen volumes. In addition, they are cost-effective. Despite these advantages, they are still more open to false positives or negatives and incorrect interpretations. Also, these sensing platforms have a risk of external interference since the environment is not as well controlled as in laboratories. In some cases, the sampling procedure can be inconvenient, such as in cancer diagnosis protocols. Indeed, POC-based electrochemical cancer biosensors are not yet available on the market. One of the additional reasons for this issue could be the distance between physicians and electrochemical biosensor developers. It is believed that multidisciplinary studies between them will improve the quality of the developed platforms. Additionally, shelflife and production control are important parameters to improve their commercialization capacity [149–152]. However, electrochemically based POC systems are promising tools for the accurate and fast detection of cancer with their overall characteristics.

#### **7. Conclusions and Future Perspectives**

In the current review, we have summarized the recent achievements and progresses around label-free electrochemical biosensors that are utilized for cancer detection. Since the type of biorecognition element is an important key parameter to enhance the selectivity of the detection, the classification of the biosensors is made according to the types of recognition elements. Besides the achievements, the current challenges are also outlined in detail. Label-free detection systems are in urgent demand owing to their properties, including reducing labored modification steps and interference effects.

The growing demand on clinical research and the medical industry for cancer studies has pushed scientists to perform early detection with practical analytical tools instead of time-consuming and back-breaking methods. In addition to detection, isolation of the cancer cells is also important to increase the survival rates and quality of life. The design and development of early-cancer diagnosis platforms has been one of the hot topics of the last decades. The recent advances in the field of cancer diagnosis show that electrochemical sensing methodologies have an important impact on the accurate, rapid and sensitive detection of cancer types. Particularly, label-free electrochemical biosensors maintain predominant features to obtain reliable, cost-effective and selective cancer diagnosis that can serve for future implementations. With the addition of advanced materials such as nanomaterials, not only sensitivity of the biosensors but also the selectivity of them can be significantly improved. Surface modification makes bare electrode substrates available and suitable for biorecognition element immobilization. Recent studies on label-free and electrochemical biosensing of cancers indicate how promising and operational these biosensors are. It is certain that their advantages will certify more powerful medical applications in the near future with the support of growing materials science technology.

**Author Contributions:** V.S.: Conceptualization, Writing—Original draft preparation, Review and Editing, Supervision; F.K.: Conceptualization, Writing—Original draft preparation, Review and Editing, Supervision, Project Administration. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** F.K. acknowledges Turkish Academy of Sciences as an associate member.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

**Lauren S. Puumala 1,2,\*, Samantha M. Grist 1,2,3, Jennifer M. Morales 4, Justin R. Bickford 4, Lukas Chrostowski 3,5,6, Sudip Shekhar 3,5 and Karen C. Cheung 1,2,5,\***


**Abstract:** Silicon photonic (SiP) sensors offer a promising platform for robust and low-cost decentralized diagnostics due to their high scalability, low limit of detection, and ability to integrate multiple sensors for multiplexed analyte detection. Their CMOS-compatible fabrication enables chip-scale miniaturization, high scalability, and low-cost mass production. Sensitive, specific detection with silicon photonic sensors is afforded through biofunctionalization of the sensor surface; consequently, this functionalization chemistry is inextricably linked to sensor performance. In this review, we first highlight the biofunctionalization needs for SiP biosensors, including sensitivity, specificity, cost, shelf-stability, and replicability and establish a set of performance criteria. We then benchmark biofunctionalization strategies for SiP biosensors against these criteria, organizing the review around three key aspects: bioreceptor selection, immobilization strategies, and patterning techniques. First, we evaluate bioreceptors, including antibodies, aptamers, nucleic acid probes, molecularly imprinted polymers, peptides, glycans, and lectins. We then compare adsorption, bioaffinity, and covalent chemistries for immobilizing bioreceptors on SiP surfaces. Finally, we compare biopatterning techniques for spatially controlling and multiplexing the biofunctionalization of SiP sensors, including microcontact printing, pin- and pipette-based spotting, microfluidic patterning in channels, inkjet printing, and microfluidic probes.

**Keywords:** silicon photonics; evanescent field biosensor; SOI biosensor; biofunctionalization; functionalization; bioreceptor; immobilization chemistry; biopatterning; microfluidics

#### **1. Introduction**

Biosensors, which comprise a transducer and biorecognition element, aim to meet increasing demands for medical diagnostics by permitting rapid testing, guiding personalized care, and reducing healthcare costs in decentralized and low-resource settings [1–3]. Silicon photonic (SiP) sensors are one class of optical refractometric sensors with promise as sensitive, rapid, and inexpensive transducers for point-of-care (POC) biosensing [4]. Compared to other types of transducers employed for biosensing, such as electrochemical [5], piezoelectric [6], and mechanical (e.g., microcantilever) [7] sensors, some advantages of SiP sensors are their high sensitivity, wide dynamic range, compatibility with label-free operation, mechanical stability, and insensitivity to electromagnetic interferences [8]. SiP devices can be patterned with wafer-scale semiconductor fabrication techniques, allowing for reproducible, inexpensive, and highly scalable production [1,9,10]. These devices consist

**Citation:** Puumala, L.S.; Grist, S.M.; Morales, J.M.; Bickford, J.R.; Chrostowski, L.; Shekhar, S.; Cheung, K.C. Biofunctionalization of Multiplexed Silicon Photonic Biosensors. *Biosensors* **2023**, *13*, 53. https://doi.org/10.3390/ bios13010053

Received: 8 November 2022 Revised: 10 December 2022 Accepted: 23 December 2022 Published: 29 December 2022

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

of nanoscale patterned silicon or silicon nitride structures that can guide and manipulate light, owing to the high refractive index contrast between the structures themselves and the surrounding media [1,11]. In SiP sensors, near-infrared light is confined in silicon or silicon nitride waveguides [1,12]. A portion of the light's electric field, known as the evanescent field, extends outside the waveguide and interacts with the surrounding medium to create a refractive index-sensitive region (Figure 1a) [1]. A change in the refractive index within this region due to analyte capture on the waveguide surface, for example, perturbs the evanescent field and changes the effective refractive index, *neff*, of the guided optical mode [1,4]. This translates to a shift in the optical phase, and in the case of resonant circuit architectures, leads to a resonance wavelength shift that is proportional to the amount of bound analyte, yielding a quantifiable change in the device's optical spectrum [1,4,12]. This change is typically read out using benchtop-scale optical inputs (e.g., broadband optical source or tunable laser) and outputs (e.g., spectrum analyzer or photodetector) [12–16].

**Figure 1.** (**a**) Illustration of cross-section of silicon photonic (SiP) sensor, showing the SiO2 substrate, Si strip waveguide (height: 220 nm, width: 500 nm), and approximate evanescent field decay distance (~40–200 nm, depending on waveguide geometry and light polarization). (**b**) Illustration of four different SiP sensing architectures, including (i) microring resonator (MRR), (ii) Mach-Zehnder interferometer (MZI), and (iii) Bragg grating sensor. (**c**) Visual depiction of a multiplexed SiP MRR sensor chip, showing different rings functionalized with different antibodies (different antibodies are represented by different colors). Antibodies in (**a**,**c**) are not to scale.

Interferometers, microring resonators (MRR), and Bragg gratings (Figure 1b) are among the SiP biosensing architectures that have been demonstrated for disease biomarker detection at concentrations down to the pg/mL scale [17,18]. Readers are directed elsewhere [1] for a detailed description of the principles of operation of each of these sensing architectures. Porous silicon sensors, which are fabricated with electrochemically etched crystalline silicon, have also been widely used in Bragg reflector and PhC configurations for biosensing since the late 1990s and are compatible with many of the same functionalization approaches [19]. This review, however, will mainly focus on planar SiP sensors, which permit greater optical confinement and guidance. Dozens of these individually addressable planar SiP sensors can be fabricated on a single millimeter-scale chip [10]. This permits multiplexed sensing, which is the simultaneous detection of multiple analytes from a single sample. Some benefits afforded by multiplexed biosensing are (1) the opportunity to diagnose multiple conditions/diseases from the same sample, (2) more selective and

reliable diagnosis of a single condition by using multiple biomarkers to inform decisionmaking [20–22], and (3) the opportunity to include controls and reference sensors (e.g., to control for temperature fluctuations) to improve measurement accuracy [23–26]. In addition to these benefits afforded by multiplexed functionalization with different bioreceptors, multiple sensors on the same chip with identical functionalization offer the benefit of replicate measurements to improve accuracy and replicability (e.g., serving as technical replicates allowing for exclusion of failed measurements and averaging out the effects of sensor-to-sensor variability and some assay issues) [27].

The process of functionalizing the sensor surface with biorecognition elements (also called bioreceptors) that selectively bind target analytes is essential to accurate SiP biosensing. The performance characteristics of the biosensor, such as sensitivity, reproducibility, and stability, are inextricably linked to the biofunctionalization chemistry [28]. Here, we broadly characterize biofunctionalization in terms of bioreceptor selection, bioreceptor immobilization strategy (attachment to the sensor surface), and biopatterning technique. Designing antifouling surface modifications is also often included in biofunctionalization procedures to prevent non-specific binding. However, this topic has been reviewed in detail elsewhere [29] and will not be a major focus of the current review.

Many different biofunctionalization strategies are available and should be carefully chosen and optimized to suit the application and sensor architecture. In general, the selected bioreceptor should have good selectivity toward the target analyte to ensure low cross-reactivity with non-target molecules in the sample, high affinity toward the target to achieve fast, sensitive detection, good stability to retain consistent binding activity over time, and reproducible production to ensure predictable and replicable sensor performance across batches/lots of reagents [30]. The strategy used to immobilize bioreceptors on the sensor must not damage the sensor surface or the bioreceptors, and it should be compatible with any system-level integration required for the sensor chips (e.g., chipmounted lasers and detectors, photonic wire bonds, etc.). It should also allow for oriented bioreceptor immobilization to optimize target accessibility and binding activity, permit uniform bioreceptor coverage on the sensor surface to ensure predictable and consistent target binding across all active sensing areas, have good stability to prevent bioreceptor detachment, and be reproducible [29,31]. The patterning strategy refers to the method by which bioreceptors are deposited on specific locations of the sensor surface (Figure 1c). This is required for multiplexed sensing and to confine bioreceptors to active sensing areas, thus preventing target depletion from dilute samples during sensing [32–34]. The selected patterning technique should not damage the sensor surface or bioreceptors. It should also have sufficient resolution for the selected application, be multiplexable so multiple different bioreceptors can be handled and deposited on a single substrate, produce uniform patterns with good spot-to-spot reproducibility, be compatible with the immobilization protocol (e.g., patterning under conditions that preserve functional groups on the silicon surface), and have low reagent consumption to conserve costly and precious reagents.

In addition to the general biosensor functionalization needs outlined in the previous paragraph, SiP devices have unique needs that distinguish them from other biosensors. Many immobilization techniques (e.g., covalent crosslinking) and bioreceptor types (e.g., antibodies, aptamers, etc.) [35] are shared across an array of sensing technologies including lateral flow assays [36], electrochemical probes [37], piezoelectric sensors [38] and other optical sensors like SPR [39]. While these sensing technology applications can provide valuable insight to inform functionalization strategies for SiP devices, only some of the findings are relevant because they utilize a variety of surfaces including glass, paper, polymers, specialized membranes (nitrocellulose), quartz, nanomaterials, alloys, metals (gold), and ceramics. Here, we focus specifically on immobilization techniques for silicon, silicon nitride, and other like materials.

Among these other transducer types, SiP sensors likely share the most similarities with SPR sensors, which employ a similar evanescent field-based detection principle. Nevertheless, SiP and SPR sensors exhibit differences in their surface chemistries, evanescent field propagation distances, miniaturizability, and multiplexability, as summarized in Table 1 [4,11,12,40–53]. Due to these differences, SiP devices have unique biofunctionalization needs, providing the main motivation for this review. For example, as SiP surfaces typically consist of 90–220 nm-thick silicon or silicon nitride nanostructures patterned on a silicon dioxide substrate [11,40], while SPR sensors typically have gold surfaces, the efficient thiol self-assembled monolayer-based strategies often used to modify metallic SPR biosensor surfaces are not suitable for SiP devices; instead silane-based chemistries are typically used [31,41]. Another unique consideration is the evanescent field penetration depth. For SiP devices, this is ~40–200 nm, depending on waveguide geometry and polarization (Figure 1a) [4,12]. Consequently, SiP sensors require a very thin biofunctional layer that brings target analytes within ~40–200 nm of the sensor surface. The size of this refractive index-sensitive region must be considered when choosing both the biorecognition element and the immobilization chemistry.

**Table 1.** Comparison of SiP and surface plasmon resonance (SPR) sensors, including SPR imaging (SPRi) and localized SPR (LSPR) devices.


Broadly, the more well-established field of SPR sensing offers a few advantages over SiP sensing. For example, SPR permits the use of simple thiol-based self-assembled monolayer functionalization strategies [29,31]. SPR variants (e.g., SPRi and LSPR) are also compatible with excitation via direct illumination and simple colorimetric readout, which are attractive for portable sensing [42,53,54]. Multimodal SPR-SERS (surface-enhanced Raman scattering) sensing is also possible for highly sensitive and reliable analyte detection [55–57], while multi-modal sensing strategies based on SiP still require further research and development [58,59]. Nevertheless, large-scale and low-cost production remains a challenge for widespread use of SPR-based sensors outside of the laboratory environment [12,53].

SiP biosensor chips, themselves, are uniquely suited to reliable point-of-care (POC) use owing to their ease of miniaturization, low cost, and ease of multiplexing [1,12]. POC biosensing not only permits accessible diagnosis in decentralized and resource-limited settings, but also facilitates treatment decision-making in situations like stroke and sepsis where rapid confirmation of clinical findings is required and conventional lab-based assays may be too time consuming [20,60]. Further, wearable sensors that can be interfaced with flexible electronics may permit real-time and noninvasive monitoring of physiologically relevant analytes (e.g., in sweat) [61,62]. However, one major challenge associated with the translation of SiP biosensors to POC applications is that SiP devices are typically operated with expensive benchtop-scale fluidics and optical readout systems [13]. Miniaturized system-level integration is possible in principle, though, and work to integrate SiP sensors with microfluidics, CMOS electronics, and on-chip lasers and detectors via photonic wire bonds is underway to produce low-cost and portable complete-system PCB-mounted sensors [13,63–66]. Another major challenge with this translation is biofunctionalization.

Given the potential of SiP devices for POC biosensing, a major focus of this work is benchmarking SiP biofunctionalization strategies against needs pertaining to their multiplexed use at the POC. Some of these needs include good environmental and temporal stability to ensure predictable performance after transport and storage at ambient conditions, scalability and manufacturability to permit large-scale deployment, low cost to ensure accessibility, compatibility with easy-to-collect biological samples, such as whole blood, urine, and saliva, and biopatterning resolution on the order of 10 μm to complement the sensor miniaturization afforded by SiP technologies [2]. Reusability is another desirable feature for POC devices that could further reduce sensing costs and improve the accessibility of diagnostic tests in remote and low-resource settings [30]. Chip-level integration of SiP sensors introduces additional biofunctionalization needs. Not only must the biofunctionalization workflow be compatible with the SiP chip architecture, but it also must be compatible with attached optical inputs/outputs and electronics. For example, the immobilization chemistry and patterning technique must not damage electrical or photonic wire bonds, chip-mounted lasers, or PCB materials. Additionally, the immobilized bioreceptors need to be stable through any processing and packaging that needs to be done after immobilization.

To date, numerous existing reviews provide an overview of SiP biosensing technologies, focusing largely on the transduction techniques [1,12,14,19,42,67], with limited discussion about surface biofunctionalization. Others have focused on a single class of bioreceptors for biosensing applications (e.g., antibodies [68,69], nucleic acid probes [70,71], and molecularly imprinted polymers [72]), often including discussion about immobilization chemistries specific to that bioreceptor, and others have focused solely on the comparison of multiple bioreceptor classes for biosensing [30,73]. Several reviews have provided detailed discussion about bioreceptor immobilization chemistries for SiP sensors [31,69] and other biosensing technologies [43,74–76]. A number of works have explored different patterning techniques for the preparation of microarrays and the multiplexed functionalization of biosensors [32,77,78]. Finally, some reviews have discussed at least two of the three key aspects of biofunctionalization (bioreceptor selection, bioreceptor immobilization strategy, and biopatterning technique) for SiP [29] and other sensor technologies (e.g., SPR [41,79,80] and electrochemical sensors [80,81]). Distinct from these existing works, the current review (1) focuses on the unique functionalization needs and strategies of multiplexed SiP biosensors, (2) discusses all three key aspects of biofunctionalization (bioreceptor selection, immobilization chemistry, and patterning technique) and how they are interrelated, and (3) includes a review of biofunctionalization strategies that have been previously implemented on SiP biosensors. To our knowledge, our review is the first contribution to comprehensively summarize and categorize the biofunctionalization strategies previously demonstrated for SiP biosensors (from 2005 to present) as well as present a critical analysis of the various existing (demonstrated on SiP) and potential (demonstrated on similar sensor types) strategies towards the goal of meeting the performance criteria most relevant to SiP biosensors.

Here, we benchmark biofunctionalization strategies against the needs outlined in Table 2, with specific focus placed on biosensor design for multiplexed POC use [82,83]. First, we critically discuss several bioreceptor classes as biorecognition elements for SiP biosensors. Examples of SiP biosensors employing these bioreceptors are highlighted, including their demonstrated sensing performance and assay format. Strategies for bioreceptor immobilization on SiP platforms are discussed along with their advantages and

limitations, with particular focus on gold standard silane-mediated covalent chemistries. Finally, contact and contact-free techniques for patterning bioreceptors on SiP sensors are identified and their performance characteristics are discussed. This review aims to present a balanced discussion of the tradeoffs of a range of biofunctionalization strategies to help guide those designing SiP biosensors in selecting a biofunctionalization approach that meets the unique needs of their intended application.

**Table 2.** Biofunctionalization needs for SiP biosensors. Please note that the performance metrics included in this table are general guidelines and designers should tailor these metrics based on their application. Interdependencies between the different columns of this table should also be considered (e.g., more expensive bioreceptors may still be suitable when combined with patterning techniques that permit very low reagent consumption).


KD: dissociation constant, PCB: printed circuit board.

#### **2. Bioreceptors**

In this section, we introduce several classes of bioreceptors that have been used for SiP sensor functionalization and benchmark them against performance criteria outlined in Table 2. A high-level comparison of these bioreceptors is provided in Table 3. We have included subsections for each bioreceptor class to provide details about the opportunities and tradeoffs associated with each of these bioreceptors. For each bioreceptor class, tables summarizing their key advantages and limitations, and categorizing their use in SiP sensor functionalization approaches demonstrated in the previous literature are provided. Because strategies to improve sensitivity, specificity, stability, and other performance metrics are in many cases dependent on the bioreceptor class, within each subsection we have outlined strategies for these types of improvements as well as provided comparisons with other classes where relevant and available. Where appropriate, comparisons between bioreceptor subtypes are also tabulated according to these performance metrics.


**Table 3.** Comparison of different bioreceptor classes based on biofunctionalization needs for SiP biosensors.





and are not endorsed or suggested by the authors.

#### *2.1. Antibodies*

Antibodies (Figure 2) are the most commonly used bioreceptors for diagnostic assays [91,153]. Antibodies are Y-shaped proteins of ~150 kDa in size, which consist of two identical Fab regions (fragment, antigen-binding), and a single Fc region (fragment, crystallizable) [30,33,87]. The Fab regions specifically bind with high affinity to target molecules called antigens via binding sites called epitopes on the antigen surface. Antigens comprise a diverse range of biological molecules including simple sugars, hormones and lipids, complex macromolecules like proteins, nucleic acids, phospholipids and carbohydrates, and even viruses and bacteria [29,33,84]. On the other hand, the Fc region typically interacts with effector molecules and cells in biological systems and may be targeted for antibody immobilization on a solid substrate in biosensing applications [33,87,154]. Millions of antibodies have been validated for tens of thousands of antigen targets, making them a widely-available and flexible bioreceptor option for many different use cases [92–95]. Antibody production starts by immunizing animals against an antigen to stimulate the production of antigen-specific antibodies by the animals' B cells [88,155]. Then, the antibodies can be obtained directly from the animal immune-sera. Alternatively, antibody-producing B cells can be immortalized by fusion with hybridoma cells for long-term production.

**Figure 2.** (**a**) Illustration of an antibody and bound antigens. Illustrations of different antibody subtypes, including (**b**) polyclonal antibodies, (**c**) monoclonal antibodies, and (**d**) a Fab fragment. Note that polyclonal antibodies are produced as heterogeneous mixtures in which different antibodies may bind to different epitopes of the same antigen. Monoclonal antibodies are produced as homogeneous samples in which all antibodies bind to the same epitope.

There are two major classes of antibodies: polyclonal and monoclonal. Polyclonal antibodies are produced as heterogeneous mixtures from animal serum and individual antibodies in a serum sample may bind to various epitopes on a single antigen [87]. Polyclonal antibodies exhibit significant batch-to-batch variability, partly owing to their animal origin [156]. Antibody quality can vary from animal-to-animal and even throughout an individual animal's lifetime [156]. Conversely, monoclonal antibodies are produced from immortalized cell lines, are homogeneous in nature, and bind to a single epitope on the target antigen surface [88,156]. Monoclonal antibodies offer excellent specificity and reduced cross-reactivity and variability compared to their polyclonal counterparts; as a result, monoclonal antibodies have been widely used in diagnostic assay applications [86–88,90]. More recently, molecular engineering has also been used to generate shorter antibody variants including Fabs, single chain variable fragments, and single domain antibodies that can be produced more easily in vitro and used for applications that solely require epitope binding [29,75,157]. A comparison of polyclonal antibodies, monoclonal antibodies, and Fab fragments as bioreceptors for SiP biosensors is provided in Table 4.


**Table 4.**

Comparison

 of antibody subtypes as

bioreceptors

 for SiP biosensors.

Numerous SiP biosensing platforms using antibodies as bioreceptors have been reported in the literature. Conventional ELISAs are typically done in sandwich or competitive assay formats, requiring labeled secondary antibodies or labeled analyte molecules, respectively [160]. SiP platforms, however, permit label-free assays [1]. In the label-free format, binding of a target analyte to surface-bound antibodies is directly monitored, offering the advantages of real-time detection and simple sample preparation [14,161]. Nevertheless, sandwich formats using an unlabeled secondary antibody [18] or labeled antibody combined with subsequent enzymatic amplification [17,162] or protein-based multilayer signal enhancement [163] have been used to achieve more sensitive and specific detection for low-concentration and low-molecular weight analytes. To tether the capture antibodies to the sensor, these antibody-based SiP platforms typically rely on randomly oriented covalent immobilization strategies that target abundant amine or carboxyl groups on the antibody surface [75]. However, other covalent and non-covalent immobilization strategies have also been used [75].

SiP biosensors using antibodies as bioreceptors (Table 5) have been proposed for the biomarker-based diagnosis of cancer [17,18,22,161,163], cardiac disorders [164,165], inflammation [166], and viral infection [167], in addition to the detection of toxins [25,168], viral particles [169–171], and bacteria [172]. Such antibody-based SiP platforms have achieved LoDs as low as the pg/mL range using enzymatically or layer-by-layer-enhanced sandwich assay formats [17,163]. Other antibody-based SiP platforms have achieved label-free analyte detection with LoDs in the low-ng/mL range [161,169]. While most of the aforementioned examples employ whole polyclonal or monoclonal antibodies, Chalyan et al. [25] functionalized thiolated silicon oxynitride microring resonators with Fab fragments obtained from protease digestion of polyclonal antibodies for the detection of a carcinogenic mycotoxin, Aflatoxin M1, with a LoD of ~5 nM. The functionalization strategy used in this work targeted sulfhydryl (–SH) groups present on the Fab surface that were liberated from splitting the intact antibody; since these sulfhydryl groups are located opposite to the antigen-binding sites, this strategy ensures highly oriented bioreceptor immobilization, making it an attractive alternative to amine- and carboxyl-targeting strategies [75,173]. Shia and Bailey [168] functionalized silicon microring resonators with recombinantly derived single domain antibodies for the detection of ricin, a lethal protein toxin. The single domain antibodies exhibited improved specificity and lower cross-reactivity compared to a commercial polyclonal anti-ricin antibody.

Despite their excellent sensitivity and specificity, antibody-based biosensors present notable challenges regarding POC sensing. Namely, antibody discovery is achieved by months-long in vivo screening processes, which are expensive and laborious [89]. Antibody production largely relies on mammalian cell lines, which means that these bioreceptors are costly and require highly trained personnel to produce, precluding their use in highly scalable and low-cost sensors [2,97–99,157]. Moreover, among antibody vendors, there is a lack of consistency in the context-specific validation and reporting of antibody specificity and reproducibility for different applications [92,156,178]. The use of animals and cell colonies in antibody production makes these bioreceptors susceptible to sample contamination [89]. This means that choosing successful antibodies for biosensors is often an expensive and time-consuming task involving troubleshooting and returning failed antibodies to suppliers [156,178]. Antibodies are also susceptible to denaturation and require carefully controlled storage conditions, which may be difficult to maintain in POC settings [24,91]. Further, antibody immobilization on a solid substrate is known to reduce antibody binding activity, making the optimization of immobilization strategies using mild chemistries a particular challenge in the design of highly sensitive biosensors [75]. The key advantages and limitations of antibodies as bioreceptors are highlighted in Table 6. Given the limitations of antibodies discussed here, several classes of synthetic affinity reagents have been developed as alternatives to antibodies and have been demonstrated as bioreceptors on SiP platforms [2].




**Table 6.** Advantages and limitations of antibodies as bioreceptors.

#### *2.2. Aptamers*

Aptamers, which have been referred to as "synthetic antibodies", are short, singlestranded DNA or RNA molecules that are systematically selected to bind to a given target molecule (Figure 3) [87,89]. These single-stranded oligonucleotides fold into unique sequence-specific three-dimensional structures that bind to targets with high specificity and affinity via non-covalent effects, including electrostatic interactions, van der Waals, and hydrogen bonding [89,100]. Aptamers are generated using an in vitro process called SELEX (systematic evolution of ligands by exponential enrichment), which allows for the selection of unique target-binding DNA or RNA molecules from a large library (Figure 3c) [100]. The SELEX process begins with a library of around 1015 single-stranded oligonucleotides, each containing a different random sequence of 20–60 nucleotides, flanked by fixed sequences on the 3 and 5 ends [89,100]. This library is amplified by the polymerase chain reaction (PCR), then strand-separated to yield ssDNA or transcribed to yield RNA, depending on whether a DNA or RNA aptamer is desired [100,102,179]. These amplified products are then incubated with target molecules and target-bound DNA or RNA are separated from unbound sequences, followed by elution of the bound species. The amplification and target-binding stages of this process are repeated with the enriched pool of target-binding sequences. The process is repeated for a total of 8–20 cycles during which competitive binding causes high-affinity binding sequences to outcompete lower-affinity ones, eventually yielding a pool dominated by sequences with the strongest affinity to the target [100–102,179]. An additional negative selection step can also be included in the SELEX process to reduce cross-reactivity of aptamers to structurally similar targets, thus enhancing selectivity [102]. The selected oligonucleotides can subsequently be sequenced and synthesized for analysis and use [100]. The resulting aptamers can achieve comparable, or even better, affinity to their targets when compared to monoclonal antibodies, with typical dissociation constants (KD) in the low nanomolar to picomolar range [85,100,101].

Since their discovery three decades ago, aptamers have been generated against inorganic ions, metabolites, dyes, drugs, amino acids, peptides, proteins, cells, and even tissues [89,100,101,105]. Because the production of antibodies relies on the immune response, antibodies can only be generated for immunogenic and non-toxic targets [89,100]. Conversely, the in vitro SELEX process theoretically allows for the generation of aptamers against any target. Further, given the small size of aptamers (5–30 kDa) compared to antibodies (150–180 kDa), aptamers can be designed against small molecule targets that are inaccessible to antibodies [89]. In evanescent field-based sensing applications, the smaller size of aptamers can allow for greater surface immobilization density and can bring captured analytes closer to the sensor surface, potentially improving sensitivity [113,114]. The selection environment (e.g., buffer type, ionic strength, pH, temperature, etc.) during aptamer generation can also be tailored to the binding conditions required for the intended use case [89,100,180]. This is contrasted to antibodies which are limited to target recognition under physiological conditions.

**Figure 3.** (**a**) Illustration of aptamer and bound target. (**b**) Visual representation of aptamer subtypes: DNA and RNA aptamers. (**c**) Illustration of SELEX (systematic evolution of ligands by exponential enrichment) process to design aptamers against a target. In (**c**), different colors in the oligonucleotide pool represent different nucleic acid sequences, while different colors in the sequencing step represent different nucleic acid bases identified by Sanger sequencing or high-throughput sequencing methods. Part (**c**) is reprinted from Ref. [101] in accordance with the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Other advantages conferred to aptamers by the nature of the SELEX discovery process include fast discovery and low batch-to-batch variability [89]. While antibody discovery requires upward of 6 months, the SELEX process can be completed in a matter of days if high-throughput automated methods are used [89,102]. Additionally, since antibody synthesis relies on animals or cell cultures, batch-to-batch variability can be high; this variability is avoided in aptamer samples because they are generated via chemical synthesis procedures with a low risk of contamination [89]. Aptamers also exhibit better environmental stability, especially thermal stability, and long shelf lives compared to antibodies [89,105]. Namely, aptamers are resistant to high temperatures up to 95 ◦C and cycles of denaturation and renaturation, while they can also be lyophilized and stored at room temperature [89]. This makes aptamers attractive bioreceptors for point of care devices and opens opportunities for surface regeneration and reusable sensors [96,105]. Finally, aptamer discovery and manufacture are generally lower cost than for antibodies. For example, CamBio offers custom aptamer discovery down to USD 5000 per target [181]. After the aptamer has been selected and sequenced, it can be manufactured at low cost using common oligonucleotide synthesis techniques. For example, Aptagen offers aptamer manufacture at USD 1–4 per milligram for microgram-scale synthesis, US \$300 per gram for milligram-scale synthesis, and USD 50 per gram for gram-scale synthesis, while IDT offers DNA oligonucleotide synthesis at CAD 1.40–2.40 per base for 1 μmol quantities for sequences of 5–100 bases in length [89,182]. However, the manufacture of RNA sequences, especially those exceeding 60 bases can be more costly. For example, Bio-Synthesis, Inc. manufactures RNA sequences of 10–30 bases in length for USD 14.50–50 per milligram

for 50–1000 mg-scale synthesis, while IDT manufactures RNA sequences of 5–60 bases in length for CAD 24.00 per base at 1 μmol quantities and RNA sequences of 60–120 bases in length for CAD 23.00 per base at 80 nmol quantities [89,106,182]. Table 7 provides a high-level comparison of DNA and RNA aptamers for SiP biosensing.

Table 8 summarizes aptamer-functionalized SiP biosensors that have been demonstrated in the literature. In all of these aptamer-based SiP sensors, label-free sensing formats were used. Park et al. [24] demonstrated IgE and thrombin detection on an aptamerfunctionalized silicon microring resonator and demonstrated reproducible surface regenerations for up to 10 cycles after IgE and thrombin binding using a NaOH solution. Byeon and Bailey [174] compared thrombin binding on aptamer-functionalized silicon microring resonators to antibody-functionalized resonators and demonstrated aptamer-functionalized surface regeneration using proteinase K. The authors found that the aptamer had a lower affinity toward thrombin (KD = 8.2 nM) compared to the antibody (KD = 3.3 nM), suggesting a poorer limit of detection for sensing applications relying on steady-state binding. However, the aptamer-functionalized sensors demonstrated faster thrombin-binding kinetics, which could produce a theoretically lower LoD for the aptamer-based sensor in applications that leverage binding kinetics measurements to generate a calibration curve (e.g., by linearly fitting the initial slope of the binding kinetics curve to quantify analyte concentration [161,184]). Christenson et al. [164] presented a comparative study in which aptamer- and antibody-functionalized Photonic Crystal-Total Internal Reflection biosensors were investigated for the detection of cardiac troponin I. The aptamer- and antibody-functionalized sensors achieved detection limits of 0.1 ng/mL and 0.01 ng/mL, respectively. While the aptamer-functionalized sensor demonstrated poorer sensitivity, both sensors achieved clinically relevant limits of detection, and the aptamer sensor was lower cost and did not require refrigeration during storage. Chalyan et al. [25] compared the performance of aptamer- and Fab-functionalized silicon oxynitride microring resonator biosensors for the detection of Aflatoxin M1. A limit of detection of 5 nM was reported for both the aptamer- and Fab-functionalized sensors, though the Fab-functionalized sensor was deemed preferable due to its superior reproducibility. Both Chalyan et al. [25] and Guider et al. [185] reported effective sensor regeneration after Aflatoxin M1 binding using glycine solutions.

While aptamers offer notable advantages over antibodies in the context of POC diagnostics, they still face challenges such as degradation in biological fluids, low SELEX success rates, lower availability, and highly variable costs. Firstly, aptamers, especially RNA aptamers, are susceptible to nuclease degradation in biological fluids [100,102]. For example, in human serum, the half-life of an unmodified aptamer is about one minute [180]. This limits the use of unmodified aptamers as bioreceptors in diagnostic devices using blood or serum samples. RNA aptamers are also more susceptible to hydrolysis than DNA aptamers at pH > 6 [183]. However, chemical modifications, such as the incorporation of 2- -fluoro or 2- -amino-modified nucleotides, are often introduced to aptamers either at the beginning of SELEX or during chemical synthesis to improve their resistance to nuclease degradation [89,186]. These types of modifications can increase an aptamer's half-life in biological fluids to multiple days [180], but modifications introduced during and after SE-LEX can add complexity to the SELEX process or change the folding structure and binding properties of the aptamer, respectively [89]. As such, careful optimization is required to achieve effective nuclease resistance without compromising binding performance.


**Table 7.** Comparison of aptamer subtypes as bioreceptors for SiP biosensors.

 Molecular weight and size for aptamers consisting of 15–100 nucleotides. Prices are listed in CAD for products available as of July 2022, unless specified. ‡ Readers are directed elsewhere [89] for a more comprehensive list of key companies in the global aptamers market. Vendors listed are not endorsedor suggested by the authors. § Readers are directed elsewhere [103,104] for a more comprehensive list of key companies in the global oligonucleotide synthesismarket. Vendors listed are not endorsed or suggested by the authors.


**Table 8.** Demonstrations of SiP biosensors using DNA aptamers as the biorecognition element and their sensing performance. All demonstrations tabulated here used label-free assay formats.

Next, the success rate of SELEX aptamer generation is lower than in vivo antibody generation, likely due to the lower structural diversity of nucleotides compared to amino acids and the small size of aptamers [101,180]. This increases the time and resources required to optimize aptamers for new targets. However, this <30% SELEX success rate could be improved through the use of specialized SELEX technology variants, personalized protocols, optimized oligonucleotide libraries, and quality control measures [180,187]. The target-binding performance of an aptamer depends on its structural conformation, which can be influenced by pH, ionic strength, and temperature [180]. Therefore, to ensure predictable binding, aptamer selection must be carried out in buffer systems similar to those used in the final application. However, this may also mean that an aptamer that performs well in solutions of a purified target in buffer may not perform as well in complex biological samples. Lastly, aptamers lack the type of extensive commercial infrastructure and investment seen in the antibody market and usually must be custom-synthesized by a handful of companies [89]. A summary of the key advantages and limitations of aptamers as bioreceptors is provided in Table 9.

#### *2.3. Nucleic Acid Probes (Hybridization-Based Sensing)*

Short, single-stranded nucleic acid probes have been widely used for the detection of nucleic acid targets via hybridization-based SiP sensing (Figure 4) [80,107,188]. Both ssDNA and RNA sequences can be immobilized on a biosensor surface, where they bind complementary nucleic acid target sequences through hydrogen bond formation, yielding DNA-DNA, DNA-RNA, or RNA-RNA duplexes [33,70,189]. Such biosensors are often called genosensors [81]. Compared to aptamers, which can be designed to bind many different types of target molecules, nucleic acid probes can only bind other nucleic acids [30]. Additionally, the function of nucleic acid probes depends primarily on their nucleotide sequence, not on their three dimensional structure: once the target gene sequence is known, the complementary probe can be designed directly [30]. This means that nucleic acid probes can be designed against a new target very quickly compared to antibodies and aptamers. Short nucleic acid probes of 100 nucleotides or less can be synthesized using well-characterized phosphoramidite chemistry [103,104,111,112]. This synthetic method of nucleic acid synthesis is highly reproducible, allows for the incorporation of functional groups like thiols and amines to aid in probe immobilization on solid substrates, and is typically low-cost [81,111,112,190]. Another key advantage of nucleic acid probe-based biosensors is that they can be thermally or chemically regenerated with good reproducibility between sensing cycles [80].


**Table 9.** Advantages and limitations of aptamers as bioreceptors.


[180]

• Less widely available than antibodies and usually require custom synthesis [89]

**Figure 4.** (**a**) Illustration of nucleic acid bioreceptor and bound nucleic acid target. Comparisons of the chemical structures of different nucleic acid subtypes, including (**b**) DNA, (**c**) RNA, (**d**) PNA, (**e**) LNA, and (**f**) morpholino, shown as line structures informed by Refs. [73,191].

In addition to conventional ssDNA and RNA probes, synthetic nucleic acid analogues with functional chemical modifications to improve binding performance and biostability have recently been explored for biosensing applications. These include peptide nucleic acids (PNAs), locked nucleic acids (LNAs), and morpholinos [23,30,73,81,110,115–117,191,192]. PNAs (Figure 4d) are synthetic DNA mimics that can hybridize to complementary DNA and RNA, but have a backbone consisting of *N*-(2-aminoethyl)-glycine units linked by peptide bonds, rather than the sugar-phosphate backbone usually found in DNA [81]. Unlike natural nucleic acids, PNAs are uncharged, giving them improved hybridization stability [73]. Their hybridization stability is also impacted to a greater extent by single base mismatches than DNA-DNA hybridization, making PNAs more selective than DNA probes and a good choice for detecting single nucleotide polymorphisms [193]. PNAs also exhibit ionic insensitivity and improved pH, thermal, and enzymatic stability [73]. LNAs (Figure 4e) are another class of synthetic DNA mimics in which the ribose is locked in the 3- -endo conformation, resulting in reduced conformational flexibility, improved biostability, and enhanced binding affinity toward the target sequence [30,81,101]. Morpholinos (Figure 4f) are synthetic nucleic acid analogues in which the sugar-phosphate backbone is replaced by alternating morpholine rings, connected by phosphoramidite groups [110]. Morpholinos are uncharged and possess many of the same characteristics as PNAs, but morpholinos exhibit improved solubility, poorer stability at low pH, and improved flexibility of synthesis regarding sequence length, offering the opportunity to bind longer DNA and RNA target sequences, compared to PNAs [108]. Table 10 provides a comparison between these nucleic acid subtypes and benchmarks them against functionalization performance criteria for SiP biosensing.

Numerous SiP sensing platforms have been demonstrated in the literature using nucleic acids or nucleic acid analogues as biorecognition elements for the detection of ssDNA and RNA biomarkers with applications in the detection of cancer [23,192,194–197] and bacteria [198,199] (Table 11). Often, a label-free assay format is used on these sensing platforms. For example, Sepúlveda et al. [200] demonstrated label-free detection of short ssDNA targets down to 300 pM using a silicon nitride Mach-Zehnder interferometer sensor functionalized with ssDNA probes, while Shin et al. [197] demonstrated specific and label-free detection of longer ssDNA targets (>100 nucleotides) on ssDNA-functionalized silicon microring resonators down to 400 fmol, which corresponds to 16 μL of a 25 nM sample. A silicon nitride slot waveguide Mach-Zehnder interferometer functionalized with methylated ssDNA probes was demonstrated by Liu et al. [192] to quantify the methylation density of a DNA-based cancer biomarker at sample concentrations down to 1 fmol/μL or 1 nM. Nucleic acid-functionalized SiP sensors have also been used for microRNA detection, as demonstrated by Qavi and Bailey [194], who used a ssDNA-functionalized silicon MRR sensor for the rapid and label-free quantification of microRNAs. In this work, the authors reported a limit of detection of 150 fmol, which represented the minimum quantity of microRNA that could be reasonably detected in solution with the reported biosensor. Based on the supporting information provided for this work, this detection limit corresponded to a 75 μL analysis volume of 2 nM microRNA. Synthetic nucleic acid analogues have been demonstrated as receptors and targets for SiP sensors. Yousuf et al. [110] recently demonstrated the detection of short ssDNA targets on morpholino-functionalized suspended silicon microrings down to 250 pM, while Hu et al. [201] demonstrated PNA detection using ssDNA-functionalized planar SiP sensors.

In contrast to these label-free methods, Qavi et al. [109] amplified the detection of microRNA on a ssDNA-functionalized silicon microring resonator sensor using S9.6 anti-DNA:RNA antibodies. The S9.6 antibody selectively binds to DNA-RNA heteroduplexes and was shown here to effectively amplify the signal after microRNA hybridization, achieving a limit of detection of 350 amol, corresponding to 35 μL of a 10 pM microRNA sample. This was a 3-fold improvement compared to label-free microRNA detection on the same sensor. This work also demonstrated preliminary results demonstrating that LNA probes could be used to capture the microRNA targets, followed by successful, albeit slightly less effective, amplification with the S9.6 antibody. Kindt and Bailey [196] improved the limit of detection of a ssDNA-functionalized silicon microring resonator sensor for the detection of mRNA using streptavidin-coated beads. This bead-based amplification improved the sensor's limit of detection to 512 amol, compared to 32 fmol without bead-based amplification.

To date, most nucleic acid hybridization-based biosensors have been demonstrated for the detection of short target sequences due to the tendency of longer sequences to fold and obtain secondary structures [70,198]. These secondary structures significantly slow down binding kinetics, thus increasing sensing times. This challenge can be mitigated by pre-treating the targets via thermal denaturation, fragmentation, or the use of short nucleic acid chaperones which disrupt the nucleic acid target's secondary structure [196,198]. In one work [198], the folded structures of long transfer-messenger RNA (tmRNA) targets were modified using one of the three following strategies prior to detection: (1) chemical fragmentation, (2) thermal denaturation, or (3) thermal denaturation in the presence of chaperone probes. Subsequently, the treated tmRNA targets were detected in a label-free format on ssDNA-functionalized silicon microring resonators. Chemical fragmentation was found to be the most effective RNA pre-treatment strategy for increasing the binding kinetics and magnitude of the sensor response. In another work [196], short DNA chaperone molecules were used to disrupt the secondary structure of full length mRNA transcripts prior to detection on ssDNA-functionalized silicon microring resonators. This effectively improved the sensing assay's binding kinetics.


**Table 10.**

Comparison

 of nucleic acid subtypes as

bioreceptors

 for SiP biosensors.



\* All prices are listed in CAD for oligonucleotides with no chemical modifications. Prices are provided for μmol-scale production for comparative purposes. Note that lower prices are available for large scale synthesis. † Price provided for DNA oligonucleotides containing 1–20 LNA bases. ‡ Readers are directed elsewhere [103,104] for a more comprehensive list of key companies in the global DNA and RNA oligonucleotide synthesis market. Vendors listed are not endorsed or suggested by the authors. § These are not exhaustive lists of PNA, LNA, and morpholino vendors. Vendors listed are based on an exploratory search and are not endorsed or suggested by the authors.


**Table 11.** Demonstrations of SiP biosensors using nucleic acid probes as the biorecognition element and their sensing performance.

\* Does not include PCR amplification prior to introduction to sensor surface.

Indeed, the greatest limitation of nucleic acid-based bioreceptors is their limited applicability: they are only suitable for applications requiring nucleic acid targets [30]. Further, nucleic acid targets usually require significant sample preparation prior to detection [188]. For DNA targets, the sample usually must undergo fragmentation to ensure that the target sequence is accessible to the capture probes, followed by denaturation to yield single-stranded sequences. Depending on the abundance of the target, it may also require amplification through PCR or isothermal strategies prior to detection [188,195,199]. For RNA targets, sample preparation may be simpler, but still typically requires a fragmentation step [188]. Finally, DNA and RNA carry an inherent negative charge, making them susceptible to non-specific binding due to electrostatic interactions with non-target molecules [30]. This also poses challenges regarding nucleic acid probe immobilization. For example, nucleic acid probes are repelled by an unmodified SiP sensor's negatively charged native oxide surface, which means that the SiP surface must be modified with a cationic film should passive adsorption be used for probe immobilization [204]. When covalent immobilization strategies are used, this negative charge increases steric hindrance between adjacent nucleic acid probes, which affects the maximum density of probes that can be immobilized on the sensor surface and the number of available binding sites for targets, potentially limiting sensor sensitivity [201]. This effect, however, can be reduced by employing in situ synthesis of nucleic acid probes on the SiP surface. Hu et al. [201] demonstrated a greater than 5-fold increase in ssDNA probe surface coverage and a greater than 5-fold increase in detection sensitivity for SiP microring resonators and photonic crystal sensors when functionalized via in situ probe synthesis, compared to the covalent immobilization of full ssDNA sequences. Conversely, if the immobilization strategy is optimized and the density of immobilized nucleic probes on the surface becomes too high, hybridization of targets to the surface-bound probes is hindered by steric crowding and electrostatic repulsion, also limiting sensor sensitivity [71]. As such, careful tuning of the spacing between immobilized probes is required for optimal performance. Some of these limitations can be mitigated by the use of uncharged synthetic DNA analogues including PNAs or morpholinos [30]. For example, in a study investigating DNA- and PNA-functionalized electrochemical sensors for the capture of DNA targets, the PNAfunctionalized sensors exhibited stronger target capture and demonstrated optimal sensing performance at higher probe surface density than the DNA-functionalized sensors, likely due to reduced steric and electrostatic effects [205]. This contributed, in part, to a greater sensitivity for the PNA-functionalized sensor, which had a very wide dynamic range from pM to μM and a LoD that was 370 times lower than that achieved when using DNA probes. However, the lack of electrostatic repulsion between uncharged DNA analogues can lead to local clustering on the sensor surface, creating a heterogeneous layer of these uncharged probes, thus hindering the reproducibility of the functionalization strategy [193,206]. A summary of the key advantages and limitations of nucleic acid probes for SiP biosensing is provided in Table 12.

#### *2.4. Molecularly Imprinted Polymers (MIPs)*

Molecularly Imprinted Polymers (MIPs) are a type of label-free synthetic receptor for binding a broad spectrum of analytes from small molecules and viruses to larger proteins and cell membrane structures (Figure 5) [72]. The first imprinted polymers were developed in the early 1990s and demonstrated the ability to change impedance in response to target binding. Later, more developed MIP films exhibited changes in refractive index upon binding, making them ideal for optical sensors.


**Table 12.** Advantages and limitations of nucleic acid probes as bioreceptors.

**Figure 5.** MIPs can be templated with an array of targets including: RNA, DNA, amino acids, peptides, proteins, lipids, glycans, viruses, and bacterial or cell epitopes. Reproduced from Ref. [207] in accordance with the Creative Commons Attribution 4.0 International license (CC BY 4.0).

Several strategies for MIP preparation on SiP platforms and representative surfaces have been summarized in Table 13. MIPs are created via template assisted synthesis where an analyte is cured within a polymer making a 3D impression in the form of a binding pocket (Figure 6a) [119,208]. There are two main methods of MIP polymerization or "templating" for optical sensors: solution based (Figure 6b) or surface stamping (Figure 6c) [119,209]. In solution-based MIPs, a target, or template, is solvated in organic solvents with precursors, initiators, and monomers [72]. Smaller molecules are primarily used directly as a template, whereas larger targets (proteins, peptides, etc.) use a smaller binding epitope for imprinting. These formulations are specific to the template and form complexes of reversible covalent or noncovalent interactions with the template's chemical structure. Next the solution is deposited on a surface and cured by ultraviolet (UV) or thermal polymerization. Solution-based MIPs can be templated onto many shapes such as coatings, thin films, and nanoparticles [210]. This is advantageous since they can conform to many different fiber and waveguide topologies. MIP films can be grown on a variety of photonic sensor designs, dipped on optical fibers, or developed in solution on microspheres [211]. Following MIP synthesis, the template molecules must be extracted, which is often achieved by washing or soaking in solution [211–215] or by plasma-treatment [211], though physically assisted solvent extraction (e.g., microwave- or ultrasound-assisted extraction) and extraction using supercritical or subcritical fluids have also been used [216]. This produces a distribution of exposed binding site geometries due to the template's random orientation on the polymer surface (Figure 6b). Surface stamping using support molds was the first method of casting [209,217]. Template molecules are crosslinked to a surface mold and pressed onto the polymer surface over the sensor prior to curing. Removal of the mold leaves imprinted binding sites stamped on the surface of the polymer. This method produces more regular pockets in comparison to solution-based MIPs due to the added

control over the depth of imprinted binding sites and the opportunity to control template orientation on the surface mold (Figure 6c) [218].

**Table 13.** Strategies demonstrated for the preparation of MIPs on SiP sensors and representative surfaces.


These methods produce specific binding pockets on the polymer surface that match the three-dimensional molecular structure of the template. Targets primarily bind via hydrogen bonding, electrostatic interactions, and Van der Waals forces. Reversible covalent bonding is less common since it is dependent on the template's molecular structure, available specialized monomers, and more complex synthesis [220]. Direct adsorption of analytes into the binding pockets produces a change in refractive index or electrochemical (impedance) signal that can be read out by optical and amperometric sensors, respectively.

MIPs are considered an alternative to antibodies since they are highly sensitive, reversible and have both chemical and mechanical stability. They are synthetic making them robust, scalable, low-cost, and shelf-stable [118]. They have been shown to be stable over months in a large temperature range (up to 150 ◦C) with over 50 adsorption/desorption cycles in organic solvents, acids, and bases [221]. Divinylbenzene MIP bases are twice as robust (up to 100 cycles) in comparison to methacrylate- or acrylamide-based polymers over a larger pH range. Although MIPs are an excellent synthetic method of producing a non-refrigerated product with a long shelf life, there are several limitations to the technology. Currently, synthesis is developed for one target at a time and requires computational studies to downselect polymer precursors and benchtop chemistry to optimize formulation. Computational studies include quantum mechanics/molecular mechanics (QM/MM) calculations (ab initio, molecular dynamics, etc.) between possible precursors and the template molecule [222]. These calculations determine which reagents interact with the chemical structure of the template molecule. Then, MIPs are formulated based on the set

and ratios of precursors are empirically tested. The final MIP formulation is selected to maximize sensitivity and specificity, based on these empirical data.

**Figure 6.** Illustration of MIP templating approaches. (**a**) MIP templating begins with a template mixed with polymer precursors followed by curing and the template removal. (**b**,**c**) Illustration of molecularly imprinted polymer (MIP) showing the random and oriented nature of template orientation on the surface of solution based (**b**) and stamped (**c**) MIPs, respectively. In solution-based MIP preparation (**b**): templates are first solvated in organic solvents with precursors, initiators, and functional monomers (i), followed by deposition on the sensor surface (ii), curing (iii), and template extraction (iv), after which the MIP can be used for target capture (v). Note that the small pieces of color left behind in the binding sites after template extraction, as seen in (iv,v), represent sites where the functional monomers formed non-covalent or covalent bonds with the template. In surface stamping based MIP preparation (**c**): templates are immobilized on a surface mold (i) and pressed into a polymer film on the sensor surface (i,ii) prior to curing. After curing, the surface mold is removed (iii), leaving imprinted binding sites on the sensor surface, which can be subsequently used for target capture (iv). Part (**a**) is reproduced with permission from Ref. [210]. Copyright 2016, American Chemical Society.

MIPs have limited specificity in complex solutions due to the imprinted nature of the polymers, which include an array of heterogeneous binding pocket orientations [30]. Smaller or like molecules can fill the binding pockets, producing a background signal or affecting the MIP's affinity toward its target [223]. Formulations thus need to be thoroughly optimized for the template (as described above) and tested against non-imprinted polymers (NIPs) [224]. NIPs are the same composition as the MIPs, only formulated without the template. They are used as a control to determine the sensitivity of the MIPs against nonspecific adsorption. Further studies testing MIPs in real bioanalyte samples are essential to validate their specificity [30]. A summary of the advantages and limitations of MIPs as bioreceptors is provided in Table 14.

**Table 14.** Advantages and limitations of MIPs as bioreceptors.


The use of refractive index sensing with MIPs in silicon photonics (Table 15) is limited, although they have been well demonstrated with SPR-based sensors [209]. MIPs can be drop-cast, spray-coated, spin-coated and inkjet printed on the sensor surface. Chen et al. [212] demonstrated thermally polymerized, drop-cast ultrathin film MIPs on a passive SOI microring resonator sensor for testosterone. This method is highly sensitive for sensing ultralow concentrations with a sensitivity of 4.803 nm/(ng·mL). First the template solution is premixed to promote self-assembly between the template and monomers specific to its chemical structure. This produces a pre-polymerized layer surrounding the template in solution that is further complexed with the addition of carboxyl-terminated monomers. This matrix is then drop-cast on the sensor's surface and thermally treated for 12 h. The combination of the pre-polymerized matrix and dilute solution results in an ultrathin assembled monolayer of MIPs on the surface with a limit of detection of 48.7 pg/mL. Multiple cycles of MIP regeneration (using a 1:1 acetic acid-ethanol rinse) and sensing with a solution of 1 ng/mL testosterone were tested on this platform to assess reproducibility. There was a drift in the sensor response and corresponding decrease in sensitivity as the number of regenerations increased, which the authors attributed to damage to the MIP during testing. Selectivity was also assessed by introducing the small molecule toxin, microcystin-LR, to the sensor, which produced a negligible response.

**Table 15.** Demonstrations of SiP biosensors using MIPs as the biorecognition element and their sensing performance.


Photopolymerization can be achieved all at once by direct UV polymerization or in stages by pre-polymerizing in a dilute crosslinking solution followed by the addition of a UV initiator for a final cure. Xie et al. [219] used this process with cascaded microring resonators for sensing progesterone. They used an SU-8 cladding and a slightly larger ring diameter to match the free spectral range of the reference ring to the MIP-coated sensing ring. The MIP is prepared by pre-polymerizing acetic and methacrylic acid with progesterone for 3 h followed by adding UV crosslinkers in a specialized tank for UV curing. This produces a thin self-assembled film on the sensor surface. Their results showed a limit of detection of 83.5 fg/mL which is approximately 3 orders of magnitude lower than enzyme-linked immunosorbent assays (ELISA). The sensor shows good selectivity to progesterone with little to no response with testosterone and the NIP. Eisner et al. [213] used

MIP sol-gels to compare airbrush versus electrospray ionization deposition techniques. These sol-gels are formed by hydrolysis and polycondensation of a colloidal liquid into a gel at low temperatures. The colloid includes metal oxides, salts, or alkoxides suspended in solvents. This ceramic-based MIP was designed for the detection of trinitrotoluene (TNT) vapor and was coated on passive silicon racetrack resonators with thicknesses of 500–700 nm to minimize resonant wavelength shift artifacts due to changes in the bulk refractive index surrounding the MIP. The results showed a ~10× increase in response and sensitivity in the electrospray MIP in comparison to airbrushing. The MIP-coated sensors showed a nonspecific response to other nitro-based explosives (2,4-dinitrotoluene (DNT) and 1,3-dinitrobenzene (DNB)); however, the device's sensitivity was about an order of magnitude greater for TNT than for DNT and DNB.

Hydrogel-based MIP thin films are less successful since they expand and contract based on water content and salinity, producing unwanted effects. Reddy et al. [225] sensed hemoglobin on silicon oxynitride waveguides for dual polarization interferometry. The gels initially increased in thickness and mass upon injection of a control solution, but the response was transient suggesting adsorption and desorption of the control on the hydrogel surface. In contrast, the target, hemoglobin, produced a continuous signal and remained selective in solutions containing <1% pooled bovine serum.

#### *2.5. Peptides and Protein-Catalyzed Capture Agents*

Synthetic and native peptides are an attractive method of capture for chemical and biological targets in SiP due to their small size in comparison to antibodies, aptamers, and other larger components (Figure 7) [226–228]. Peptides are differentiated from proteins by their size (2–70 amino acids) and flexible structure. There are two main types of peptides for attachment: native and synthetic [120]. Native peptides are small binding epitopes or ligands found in nature that selectively bind to a specific site on the target of interest. They are primarily recombinant and produced by cloning the peptide in an organism. The peptide sequence is inserted into a plasmid, expressed in bacteria, insect or mammalian cells and purified for processing [229].

**Figure 7.** (**a**) Illustration of peptide bound target. (**b**) Comparison of peptide, aptamer, Cas9 enzyme and antibody relative sizes, informed by protein data bank crystal structures 2AU4, 4OO8 and 1IGY [226–228]. Peptides are smaller than aptamers, antibodies, and many other bioreceptor classes discussed here, offering potential improvement in SiP biosensor sensitivity by bringing the binding interaction into a region of the evanescent field with higher field intensity.

Synthetic peptides are chemically synthesized using solid phase peptide synthesis (SPPS) or solution-based synthesis (SPS) [230]. Synthetic peptides are made using D-amino acids instead of the more naturally occurring L-amino acids seen in native peptides. Dand L-amino acids are enantiomers, or the same amino acid sequence with a mirror image structure. This change in configuration makes D-amino acids less susceptible to enzyme degradation without changing their biological function. SPS was the first synthesis method, developed in 1901, where a chain of amino acids is grown one residue at a time in solution [231]. SPPS followed in 1963 and uses a solid support for anchoring the peptide chain that enables washing steps between the addition of successive amino acids. Both methods start from a primary amino acid using selective protecting groups (FMOC, BOC) where successive amino acids are added in a step-by-step fashion to form a chain [232]. Generally, SPPS is the most common method since it is a well-established commercially available process and contributed to the Nobel Prize in Chemistry in 1984 [231]. Its use of a support and wash cycles results in a higher production of correctly formed peptides, removes reaction byproducts, as well as decreases the tendency of aggregation and incomplete reactions. However, SPS is still used since the lack of a support enables more challenging structures (cyclic), nonstandard components, and a larger array of coupling conditions (acidic, oxidative) [233].

Protein-catalyzed capture (PCC) agents are specialized, short (20 amino acid), synthetic peptides optimized to capture a target of interest [234]. They are considered "synthetic antibodies" due to their comparable high specificity and affinity for a target without the temperature sensitivity or stability issues common in enzymes, aptamers, and antibodies [235]. PCCs are highly selective since they are computationally designed based on the binding sites of proteins and other targets. Screening of chemical peptide libraries, such as one-bead one-compound (OBOC), identifies peptide components with high specificity and selectivity to the target of interest [236,237]. Due to this design, their affinity can be tailored to the specific dynamic range needed for sensing. Agnew et al. [234] evaluated the epitope binding sites and affinity of PCCs to those of monoclonal antibodies of the same target using principal component analysis. Their analysis covered 14 different protein targets as well as considered their physicochemical properties and molecular binding interactions. The results showed that PCCs are able to match and surpass antibody affinities with the majority of the binding driven by electrostatic interactions and hydrogen bonding.

In the literature, peptides and PCCs have been demonstrated as bioreceptors against antibodies [238], cancer cells [239], viral proteins [91], and streptavidin [240] on SiP platforms (Table 16). Angelopoulou et al. compared recombinant SARS-CoV-2 spike protein peptide on silicon nitride MZI sensors to conventional ELISA assays [238]. Silicon nitride MZI sensors were crosslinked via glutaraldehyde to the spike peptide against SARS-CoV-2 in a manner that selectively attached the peptides to only the silicon nitride waveguides and not the surrounding silicon dioxide. The reference was blocked with bovine serum albumin as a control for non-specific binding. The label-free peptide MZI showed a 80 ng/mL limit of detection and correlated with the ELISA results of 37 diluted serum samples. The addition of an antibody as a label improved the limit of detection to 20 ng/mL.

**Table 16.** Demonstrations of SiP biosensors using peptides as the biorecognition element and their sensing performance. All tabulated demonstrations used a label-free assay format.


Martucci et al. [239] used idiotype peptides to determine the surface capture efficiency of tumor cells on silicon surfaces. Idiotype peptides are ligands from the binding site of receptors on the surface of immune cells that bind to antigens on the surface of lymphoma cells. They are specific to a subset of B-cells and can specifically identify lymphoma cells. The authors functionalized the surface of porous silicon microcavities by submerging in a 5% amino-terminated silane solution, crosslinking with a double N-succinimidyl terminated linker to crosslink to a primary amine on the peptide. The authors moved away from crosslinking antibodies to silicon surfaces since antibodies are known to have problems assembling monolayers in the same orientation due to their large size and multiple crosslinking sites [241,242]. Their results showed that the idiotype peptide covered 85% of the sensor's surface with a uniform, oriented layer and had a detection efficiency of 8.5 × <sup>10</sup>−<sup>3</sup> cells/μm2.

PCCs are starting to become a more well-known method for biological sensing using silicon photonics due to their good stability and long shelf life. They are temperature stable, showing little to no change in affinity after heating to 90 ◦C, and resistance against protease degradation [91,243]. Layouni et al. [91] showed a PCC specific to Chikungunya virus E2 protein on porous silicon microcavities and with positive detection in response to 1 μM E2 viral protein. In addition, their results showed no statistical significance in sensor response between previously heated (90 ◦C, 1 h) and unheated PCCs. This stability was further confirmed by PCCs for vascular endothelial growth factor maintaining 81% of their affinity after 1 h using standard ELISA assays. Another work with porous silicon microcavities for PCC sensing of streptavidin showed detection of 5 μM streptavidin using PCCs immobilized via click chemistry crosslinking [240]. A summary of the advantages and limitations of peptides and PCCs as bioreceptors is provided in Table 17.

**Table 17.** Advantages and limitations of peptides and PCCs as bioreceptors.


#### *2.6. Glycans and Lectins*

Both glycans and lectins have been employed as biosensor recognition elements on SiP devices (Figure 8). Glycans are carbohydrates which are covalently conjugated to proteins (glycoproteins) and lipids (glycolipids) [122,244]. In biological systems, glycoconjugates are typically found on cell surfaces, in the extracellular matrix, or in cellular secretions, and participate in intermolecular and cell–cell recognition events. Glycans consist of monosaccharides linked together in linear or branched structures by glycosidic bonds [244]. The diversity of their constituent monosaccharide residues and the position and configuration of their glycosidic bonds give glycans significant structural variability [128,244]. Lectins are non-immune proteins that recognize and bind glycoconjugates and non-conjugated glycans via carbohydrate recognition domains (CRD) [121,122,134]. Specific lectin-glycan binding is affinity-based and facilitated by hydrogen bonding, metal coordination, van der Waals and hydrophobic interactions [121]. The CRDs of lectins may target monosaccharide residues or they may show poor affinity toward monosaccharides and, instead, preferentially bind oligosaccharides based on their glycosidic linkages [121,122,244]. The affinity of individual CRD-glycan interactions are weak, with dissociation constants in the micromolar to millimolar range [121,122]. Multivalent binding between lectins and glycans, however, allows for higher-avidity interactions, with dissociation constants that are multiple orders of magnitude lower [122,123]. Namely, some lectins possess multiple CRDs that bind to multiple monosaccharide residues on a polysaccharide or to multiple proximal

carbohydrates immobilized on a densely-coated solid substrate [121–123]; moreover, lectins can recognize homogeneous carbohydrate-coated surfaces or mixed glycan patches. Conversely, in the case of lectins with only one CRD, higher-avidity binding may be achieved by the clustering of many lectin molecules [122]. While many lectins have been identified and their glycan-binding characteristics have been characterized, these only encompass a small fraction of the diverse set of glycans that are found in nature [123]. Compared to proteins and nucleic acids, the functional study of glycans lags far behind [129].

**Figure 8.** (**a**) Illustration of a glycan and bound lectin. (**b**) (i) SEM image of a microring resonator and (ii) cross-section of microring resonator waveguide using glycans as bioreceptors. The glycans are immobilized using an organophosphonate linking strategy and used for lectin (protein) capture. Part (**b**) is adapted with permission from Ref. [126]. Copyright 2012 American Chemical Society.

Glycans can be immobilized easily on biosensor surfaces in an oriented manner; for example, the terminal amine group of a glycan derivative can be targeted for site-directed covalent amine coupling to a surface [244]. In comparison, lectins possess more complex structures, making oriented immobilization more challenging.

Homogeneous glycan samples for biosensing applications cannot be synthesized easily in large quantities using biological systems, making chemical and chemoenzymatic synthesis the preferred routes of production for structurally defined glycans and glycoconjugates [127,245]. Multi-milligram quantities of polysaccharides up to 50 mers in length can be rapidly and reproducibly synthesized and optionally conjugated to nonglycan entities, like proteins, to yield glycoconjugates [127,128]. Nevertheless, chemical glycan synthesis is in its infancy and is inherently more challenging than oligonucleotide and oligopeptide synthesis because glycans are often highly branched and their biosynthesis is not template-driven [129]. Chemical glycan synthesis requires the modification of one monosaccharide hydroxyl group at a time in the presence of many others and the careful control of glycosidic linkage positions [127]. Currently, the synthesis of complex and highly branched glycan structures remains a major challenge [129].

Lectins may be purified from various organisms, though yields, especially for animalderived lectins, are often too low for practical use [130]. Consequently, recombinant techniques are usually required for the production of lectins in multi-gram quantities [130]. Notably, anti-carbohydrate antibodies can be generated for glycan capture, but, due to the poor immunogenicity of carbohydrates, these antibodies typically have poor affinities toward their targets and limited versatility, making lectins preferable for carbohydrate detection [121]. In comparison to antibodies, the cost of lectin production is also lower. However, similarly to antibodies, the commercial synthesis of lectins is cell-based, and samples may vary in purity, properties, availability, and activity within and between vendors [121]. An overall comparison of glycans and lectins as bioreceptors for SiP biosensors is detailed in Table 18.


**Table 18.**

Comparison

 of glycans and lectins as

bioreceptors

 for SiP biosensors. listed are based on an exploratory search and are not endorsed or suggested by the authors.

Glycan-coupled SiP biosensors can be used for lectin capture and have applications in toxin [132] and virus [126] detection. For example, Ghasemi et al. [132] covalently immobilized GM1 ganglioside glycans on the surface of a TM mode silicon nitride microring resonator sensor for label-free detection of Cholera Toxin subunit B. The authors reported an absolute limit of detection of 400 ag, which corresponds to a surface coverage of 8 pg/mm2. Shang et al. [126] used an organophosphonate strategy to tether glycans and glycoproteins to silicon microring resonators for label-free detection of various lectins and norovirus-like particles. The authors reported a limit of detection of 250 ng/mL for the norovirus-like particles. The functionalized sensors also demonstrated excellent stability, retaining strong binding performance after one month of storage at ambient conditions and after multiple cycles of surface regenerations with high-salt and high- and low-pH solutions. Indeed, the good chemical stability of glycans, even at ambient and dry conditions for prolonged periods of time, is an attractive characteristic of glycan-conjugated biosensors [124,125]. Other publications have demonstrated glycan- and glycoconjugate-functionalized SiP sensors for the label-free detection of common lectins, with limits of detection down to the ng/mL range [133,246].

Given that various diseases, such as cancer, autoimmune diseases, infections, and chronic inflammatory diseases are associated with glycan aberrations, glycans are valuable disease biomarkers [121,129]. Lectin-coupled biosensors have, therefore, been proposed for glycan biomarker-based disease diagnosis [121,129]. While lectin-coupled SiP sensors have seldom been reported in the literature, Yaghoubi et al. [131] reported a lectin-coupled porous silicon sensor using reflectometric interference Fourier transform spectroscopy for label-free detection of bacteria. The authors functionalized sensors with three different lectins, concanavalin A (Con A), wheat germ agglutinin (WGA), and ulex europaeus agglutinin (UEA), and found that the Con A- and WGA-coupled sensors demonstrated the greatest binding affinities for *E. coli* and *S. aureus*, respectively and demonstrated limits of detection of approximately 10<sup>3</sup> cells/mL. Table 19 provides a summary of SiP biosensors demonstrated in the literature that use glycans or lectins as bioreceptors.

**Table 19.** Demonstrations of SiP biosensors using glycans or lectins as the biorecognition element and their sensing performance. All tabulated demonstrations used label-free assay formats.


The greatest limitation of biosensors using glycan-lectin binding is their specificity. Unlike antibodies, lectins often bind to more than one glycostructure and demonstrate broader specificity, thus requiring extensive selectivity and cross-reactivity characterization prior to use [121,123]. The poorer selectivity of glycan-lectin interactions complicates their detection

in complex biological samples and makes it difficult to detect small aberrations in the target structure [121,244]. Moreover, the avidity of glycan-lectin interactions is highly variable and depends not only on the structure of the biomolecules, but also on their multivalency and packing density on the sensor surface [122]. While glycans offer simple oriented conjugation to sensor surfaces and improved stability compared to antibodies, the discovery and production of biologically relevant glycans, especially complex and highly branched ones, is limited by current structural characterization and synthesis techniques [129]. Commercially available glycans are very expensive, at roughly CAD 200–1200/10 μg [135], while custom glycan synthesis is also costly [247]. On the other hand, lectins can be characterized and produced using mature and cost-effective techniques, but these proteins suffer from the same batch and vendor variability and pH-, temperature-, and buffer-sensitivity issues as antibodies [121,136]. While lectin regeneration is possible, it is likely to result in activity loss [121]. Table 20 highlights the key advantages and limitations of glycans and lectins as bioreceptors for SIP biosensors.

**Table 20.** Advantages and limitations of glycans and lectins as bioreceptors.


#### *2.7. Other*

#### 2.7.1. High Contrast Cleavage Detection (i.e., CRISPR Cleavage Detection)

Recently, high contrast cleavage detection (HCCD) has been proposed as a detection mechanism for optical biosensing employing CRISPR-associated proteins as the biorecognition elements [137–139]. This is a clustered regularly interspaced short palindromic repeats (CRISPR)-based biosensing approach that can be used for sensitive detection of nucleic acid (DNA or RNA) targets. CRISPR systems contain CRISPR-associated (Cas) proteins, which possess endonuclease activity to cleave targets via guide RNA [140,248]. Most reported CRISPR based biosensors use Cas9, Cas12, or Cas13 effectors, which demonstrate different cleavage activities [249]. Namely, CRISPR-Cas9 cleaves target dsDNA based on guidance from single guide RNA [249]. CRISPR-Cas12 captures target DNA that is complementary to its guide RNA, activating non-specific collateral cleavage (or trans-cleavage) of nearby ssDNA [250]. Similarly, CRISPR-Cas13 captures target RNA that is complementary to its guide RNA, activating non-specific collateral cleavage of nearby ssRNA [250]. In the HCCD technique, Cas12 or Cas13 effectors can be used [138].

Most SiP biosensors rely on affinity-based detection, whereby low-index bioanalytes are captured on the sensor surface upon introduction of the target analyte. The HCCD method, however, adopts a different architecture and relies on the removal of high-index contrast reporters from the sensor surface upon introduction of the target analyte (Figure 9). In HCCD, the sensor surface is first decorated with high-index contrast reporters, such as silicon nanoparticles, gold nanoparticles or quantum dots, tethered to the surface by single-stranded oligonucleotides [137–139,141,142,251]. Then, the analyte is combined with CRISPR-Cas12 or CRISPR-Cas13, which have guide RNA complementary to the target [137]. Once activated, these CRISPR-Cas complexes cleave the reporters from the surface, leading to a change in the local refractive index that can be transduced by the SiP device.

**Figure 9.** Illustration of HCCD, showing (**a**) activation of the CRISPR-Cas12a/13 effector by the target nucleic acid sample, (**b**) high index contrast reporters (e.g., gold nanoparticles) tethered to the sensor surface by single-stranded DNA or RNA prior to cleavage by the activated CRISPR-Cas12a/13 complex, and (**c**) non-specific collateral cleavage of single-stranded DNA or RNA by the activated CRISPR-Cas12a/13 complex, leading to the removal of reporters from the sensor surface.

The first experimental implementation of this method was reported in 2021 by Layouni et al. [137] on a porous silicon interferometer platform. This was a proof-of-concept study in which the sensor surface was decorated with nucleic-acid-conjugated quantum dot reporters, then exposed to a DNase solution, which cleaved reporters from the surface. While this work did not report specific analyte detection, it demonstrated the ability to detect a large shift in the sensor's reflectance peak upon enzyme-mediated removal of reporters from the porous silicon surface. This work paved the way for another preliminary study in which Liu et al. [139] demonstrated the detection of SARS-CoV-2 target DNA on a silicon microring resonator chip using HCCD (Table 21). The authors reported a ~8 nm blue shift in the resonance wavelength upon cleavage of gold nanoparticle reporters from the sensor surface by CRISPR-Cas12a activated by a 1 nM sample of target DNA in buffer solution. To our knowledge, SiP sensors using HCCD have yet to be demonstrated for the detection of nucleic acid targets in complex biological samples like whole blood, serum, and plasma. Chung et al. [251] proposed an inverse-designed waveguide-based integrated silicon photonic biosensor for HCCD-mediated sensing. However, this biosensing architecture has not yet been demonstrated experimentally.

**Table 21.** Demonstration of SiP biosensor using HCCD its sensing performance.


The HCCD technique touts several advantages compared to traditional hybridizationbased nucleic acid sensing. On a SiP sensor using hybridization-based sensing, signal generation relies on the small difference in refractive index between the sample buffer and the target nucleic acids. Typically, to achieve a detectable signal, the nucleic acid sample needs to be PCR amplified prior to detection or a secondary amplification molecule must be used after hybridization [138]. In HCCD, the refractive index contrast between the highindex reporters and background fluid is greater, leading to a greater signal change upon reporter removal compared to the binding of unlabeled targets [137,138]. Each activated CRISPR-Cas complex can perform up to 10<sup>4</sup> non-specific probe cleavages after activation, leading to multiplicative signal amplification, thus enhancing sensitivity [138,251]. Further, since HCCD relies on the removal of reporters from the surface, the SiP sensor experiences a blue shift in resonant frequency for a positive result; this is in contrast to affinity-based sensing in which a positive result causes a red shift. This means that HCCD is less susceptible to false positives caused by non-specific adsorption of biomolecules to the sensor surface [137]. Another beneficial feature of the HCCD method is that it derives its specificity from the CRISPR-Cas12 or -Cas13 complexes, which are activated in a highly specific manner by their nucleic acid targets. Since specificity is conferred by the CRISPR-Cas complexes rather than biomolecules immobilized on the sensor surface, there is an opportunity to develop universal reporter-functionalized SiP sensors which can be used with application-specific CRISPR-Cas reagents, thus reducing the costs of sensor development and production [138].

While the sensitivity of this detection strategy is bolstered by the collateral cleavage of the activated CRISPR-Cas complexes, this non-specific cleavage also makes multiplexing challenging. To the best of our knowledge, multiplexed nucleic acid detection based on HCCD has not yet been demonstrated. Another limitation of HCCD is that the irreversible cleavage of reporters from the SiP surface prevents facile regeneration of the functionalized sensor for repeated use. Further, HCCD is only suitable for the detection of nucleic acid targets, limiting its versatility. Finally, while the nucleic-acid-based reporter-modified surface has improved storage stability compared to antibody-functionalized surfaces, Cas enzyme activity is sensitive to storage conditions, complicating POC use [33]. This could potentially be addressed by lyophilizing the assay reagents [145]. Overall, while HCCD remains in its infancy and is yet to be validated for detection in complex media, this method addresses some of the limitations of hybridization-based nucleic acid detection schemes and offers potential as a highly sensitive and specific strategy for SiP sensing. Table 22 highlights the key advantages and limitations of HCCD.

**Table 22.** Advantages and limitations of HCCD.


#### 2.7.2. CRISPR-dCas9-Mediated Sensing

CRISPR-associated proteins have also been used as a biorecognition element for signal amplification in silicon photonic sensors, in combination with nucleic acid probes. In 2018, Koo et al. [143] proposed a CRISPR-dCas9-mediated SiP biosensor for highly specific and sensitive detection of pathogenic DNA and RNA fragments for the diagnosis of tick-borne diseases. Broadly, this sensing method relies on twofold signal enhancement. Firstly, recombinase polymerase amplification (RPA) is used to amplify nucleic acid targets. RPA is a rapid enzyme-mediated DNA amplification technique that can be completed

isothermally at mild temperatures [147,252]. This isothermal strategy obviates the need for power-intensive thermal cycling, which is required for conventional DNA amplification via PCR [147]. As such, RPA has been identified as an attractive alternative for POC use [145]. Additionally, reverse transcriptase (RT) can be added to the RPA reagents to facilitate isothermal amplification of RNA targets and reverse transcription of cDNA from RNA [143,147]. Secondly, nuclease-deactivated Cas9 (dCas9) is used in this sensing method as a labeling molecule. Like its active form, dCas9 binds to target dsDNA based on guidance from a target-specific single guide RNA (sgRNA) sequence [143,249]. Unlike its active form, dCas9 cannot cleave target sequences.

Koo et al. demonstrated this sensing method on SiP microring resonator sensors for the detection of pathogenic DNA and RNA sequences for scrub typhus (ST) and severe fever with thrombocytopenia syndrome (SFTS), respectively (Table 23) [143]. The sensor surface was first functionalized with single-stranded nucleic acid probes, complementary to the target sequences (Figure 10a) [143]. RPA or RT-RPA reagents were prepared, then added to the pathogenic DNA or RNA samples, along with dCas9 effectors and sgRNA. This mixture was incubated on the sensor chip in acrylic wells at 38 ◦C (for DNA targets) or 43 ◦C (for RNA targets). During this on-chip incubation, three key events took place: (1) the target DNA or RNA was amplified via RPA or RT-RPA, respectively, (2) amplified targets bound to complementary probes immobilized on the sensor surface (Figure 10b) and (3) dCas9 effectors bound to the hybridized targets to increase the refractive index change associated with each bound target (Figure 10c).

**Table 23.** Demonstrations of SiP biosensors using CRISPR-dCas9-mediated sensing and their performance.


**Figure 10.** Illustration of CRISPR-dCas9-mediated sensing. (**a**) Single-stranded nucleic acid probes are immobilized on the sensor surface and the nucleic acid targets (amplified by recombinase polymerase amplification) are introduced to the sensor surface. (**b**) The nucleic acid targets hybridize to the surface-bound probes. (**c**) Deactivated Cas9 (dCas9), guided by single guide RNA (sgRNA) specifically binds to the nucleic acid duplex to amplify the signal, without cleaving the nucleic acid duplex.

The authors reported the detection of pathogenic DNA for ST with a detection limit of 0.54 aM and the detection of pathogenic RNA for SFTS with a detection limit of 0.63 aM [143]. The platform effectively discriminated between ST and SFTS in clinical blood serum samples in just 20 min. Indeed, this platform allows for exceptional sensitivity as a result of the aforementioned twofold signal enhancement. Further, specificity is

ensured in three ways. Firstly, the nucleic acid probes immobilized on the sensor surface facilitate selective hybridization of complementary targets. Secondly, RPA or RT-RPA nucleic acid amplification is guided by primers to selectively amplify target sequences in the sample [252]. Thirdly, dCas9 solely binds to double-stranded target sequences based on sgRNA guidance, so dCas9 signal enhancement can only occur after targets have hybridized to complementary probes on the sensor surface. As such, this is a promising method for applications requiring highly sensitive detection of nucleic acid targets in complex samples.

To our knowledge, this is the only example of CRISPR-dCas9-mediated biosensing on a SiP platform in the literature. Because this sensing method uses dCas9, which does not demonstrate collateral cleavage, it may offer more straight-forward multiplexing compared to the HCCD technique, but at the cost of increased assay complexity [248,253]. Multiplexing may be possible if multiple microrings on a single chip are functionalized with different target-specific nucleic acid probes in a spatially defined manner, and multiple target-specific RPA primers and dCas9/sgRNA complexes are used [254].

Regarding costs, the short synthetic nucleic acid probes and CRISPR-Cas reagents required for this detection method can be produced at moderate cost, but the RPA reagents are more expensive [144,145,248,249,255]. For example, a single CRISPR-based diagnostics reaction involving RPA pre-amplification costs an estimated USD 0.61–5.00 in a laboratory setting, with RPA reagents making up the majority of this price [145–147]. Nevertheless, given the microlitre-scale reagent and sample volume requirements of SiP-based assays, these costs are unlikely to be prohibitive for POC use.

In this detection method, the sensor surface is prepared similarly to conventional nucleic acid hybridization-based biosensors, as described in Section 2.3, which allows for superior sensor stability compared to antibody-functionalized devices. However, one key limitation of this method is the requirement for many different assay reagents, including RPA enzymes and primers, dCas9 enzymes, and sgRNA. This increases the complexity of the assay preparation and requires environmentally controlled storage of the assay reagents, especially the enzymes, making POC use less feasible. However, lyophilization of environmentally sensitive reagents for transport and storage before use is a potential solution to this challenge [145]. The use of such a platform in a POC setting is further complicated by the need to implement careful thermal control over the RPA reaction. Finally, as is the case with classic nucleic acid hybridization-based biosensors, this platform only allows for the detection of nucleic acid targets, limiting its breadth of applications. A summary of the advantages and limitations of CRISPR-dCas9-mediated sensing as a biodetection technique for SiP biosensors is provided in Table 24.



#### 2.7.3. Lipid Nanodiscs

Lipid nanodisc-functionalized SiP sensors have been proposed to study signaling and interactions at cell membranes [148–150]. Lipid nanodiscs are 8–16 nm scale discoidal lipid bilayers, held together and made soluble by two encircling amphipathic protein belts, called membrane scaffold proteins (Figure 11) [150,151]. These nanodisc structures recapitulate the native cell membrane environment and allow for the precise control of lipid composition. This permits the study of biochemical processes that occur at cell membranes, and which require specific lipid compositions for full functionality [256]. Lipid nanodiscs also solubilize and stabilize membrane proteins, which typically demonstrate loss of activity and function outside of the phospholipid membrane environment [151]. Given that membrane proteins are involved in vital regulatory cell functions and are

often the target of therapeutic drugs, lipid nanodiscs are a valuable tool for studying cell membrane interactions involving these proteins. Compared to other structures, such as liposomes and detergent-stabilized micelles, which are used to mimic the cell membrane environment, nanodiscs offer improved consistency, monodispersity, production yield, and control over lipid and protein composition [150,151].

**Figure 11.** Illustrations of lipid nanodiscs with bound targets. The nanodiscs consist of lipid bilayers, held together by two encircling membrane scaffold proteins. The nanodiscs may be prepared without (**left**) or with (**right**) embedded membrane proteins.

SiP sensors are an appealing platform on which to investigate interactions between lipid nanodiscs and other biomolecules. The multiplexability of SiP sensors permits highthroughput screening of cell membrane interactions. Further, membrane proteins are challenging to produce and typically have low yields, making the low reagent volume requirements of SiP sensors particularly attractive [148]. Finally, nanodiscs physisorb directly onto silicon dioxide, permitting their facile immobilization onto the native oxide surfaces of silicon and silicon nitride waveguides [150].

In the literature, lipid nanodisc-functionalized silicon microring resonator sensors have been used to probe interactions between soluble proteins and lipids, glycolipids, and membrane proteins embedded in nanodiscs (Table 25) [148–150]. In a study by Sloan et al. [150], lipid nanodiscs prepared with varying compositions of the phospholipids, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) and 1-palitoyl-2-oleoylsn-glycero-3-[phospho-L-serine] (POPS), were used to probe the binding of annexin V, a lipid-binding protein. Nanodiscs prepared with glycolipids (1,2-dimyristoyl-sn-glycero-3 phosphocholine/monosialotetrahexosyl ganglioside, GM1), biotinylated lipids (*N*-(biotinoyl)- 1,2-dipalmitoyl-sn-glycer-3-phoaphoethanolamine, biotin-DPPE), and enzymes (cytochrome P450 3A4, CYP3A4) were also used to probe binding interactions with cholera toxin subunit B, streptavidin, and anti-CYP3A4, respectively. A 4-plex assay was prepared by microfluidically patterning the sensor chip with POPS, GM1, biotin-DPPE, and CYP3A4 nanodiscs, then exposing the whole sensor surface to annexin V, CTB, streptavidin, and anti-CYP3A4 solutions in sequence. This multiplexed assay demonstrated effective binding with minimal cross-reactivity for each specific protein-nanodisc combination.


**Table 25.** Demonstrations of SiP biosensors using lipid nanodiscs as the biorecognition element and their sensing performance. All tabulated demonstrations used a label-free assay format.

PC: phosphatidylcholine, PS: phosphatidylserine, PA: phosphatidic acid, POPC: 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, POPS: 1-palmitoyl-2-oleoyl-sn-glycero-3-(phospho-L-serine), GM1: monosialotetrahexosyl ganglioside, biotin-DPPE: *N*-(biotinoyl)-1,2-dipalmitoyl-sn-glycer-3-phosphoethanolamine, CYP3A4: cytochrome P450 3A4, and PE: phosphatidylethanolamine.

In another work by Muehl et al. [149], a SiP microring resonator platform was used to investigate interactions between four different blood clotting proteins (pro-thrombin, factor X, activated factor VII, and activated protein C) and lipid nanodiscs prepared with seven different binary lipid combinations of phosphatidylcholine (PC), phosphatidylserine (PS), and phosphatidic acid (PA). A 7-plex sensor was demonstrated using these seven nanodisc preparations to obtain dissociation constants for binding between the coagulation proteins and lipid surfaces. All of the coagulation proteins studied in this work bind in a Ca2+ manner, so the nanodisc-functionalized surfaces were regenerated with good replicability using a Ca2+-free buffer after protein binding. In a subsequent work, Medfisch et al. [148] used a SiP microring resonator platform to study the binding interactions of seven different protein clotting factors (prothrombin, activated factor VII, factor IX, factor X, activated protein C, protein S, and protein Z) and lipid nanodiscs prepared with nine different phospholipid compositions involving PS, phosphatidylethanolamine (PE), and PC. The effect of PE-PS lipid synergy on the membrane binding of clotting factors was investigated. Again, surface regeneration after binding events was achieved using Ca2+-free buffer.

So far, SiP sensors functionalized with lipid nanodiscs have demonstrated value in the study of binding interactions at cell membranes. This is in contrast with other classes of bioreceptors discussed in this review, which have primarily been proposed for toxin and pathogen detection and/or diagnostic applications. The nanodisc-protein interactions demonstrated by Muehl et al. [149] and Medfisch et al. [148] have limited specificity, with all of the investigated clotting proteins binding, albeit to different extents, to the lipid nanodiscfunctionalized surfaces. Hence, these nanodisc-functionalized sensors are likely unsuitable for selective discrimination between multiple targets. The incorporation of embedded membrane proteins or glycolipids into the nanodiscs, however, may offer more selective detection of soluble proteins, as demonstrated by Sloan et al. [150]. Lipid nanodiscs are typically custom-synthesized in the laboratory setting, allowing for the precise control of lipid composition and membrane protein content; while this leverages the flexibility of lipid nanodiscs, it limits their accessibility for assay-development and widespread use [151]. Overall, nanodisc-functionalized SiP sensors offer an excellent opportunity for high-throughput laboratory-based cell membrane interaction studies, but their potential in

POC diagnostics may be limited. A summary of the advantages and limitations of lipid nanodiscs as bioreceptors for SiP biosensors is provided in Table 26.

**Table 26.** Advantages and limitations of lipid nanodiscs as bioreceptors.


#### *2.8. Summary and Future Directions*

Given the myriad of potential applications for SiP biosensors and the complex tradeoffs of each bioreceptor class, there is no simple formula for selecting an optimal bioreceptor as each class has its own set of advantages and limitations that must be balanced with the needs of the application. For studies specifically probing carbohydrate-protein or cell membrane interactions, the choice is simple, with glycans/lectins or lipid nanodiscs typically being the most appropriate options, respectively. For other applications, the choice of bioreceptor can initially be narrowed down based on compatibility with the target of interest (see Table 3). Beyond this, the specific functionalization needs for the application of interest must be identified and used to guide further bioreceptor selection. For example, for non-nucleic acid targets, one must choose between antibodies, aptamers, MIPs, PCCs, and peptides. For non-POC applications where stability, regenerability, and cost are less important, monoclonal antibodies may be a suitable option due to their widespread availability and good binding affinity and selectivity. For POC applications, antibodies may not be suitable, and the choice between aptamers, MIPs, PCCs, and peptides will likely depend on the availability of pre-designed and validated products for the target of interest, or access to the relevant expertise and resources to design a custom bioreceptor for the target of interest. Trade-offs between affinity, selectivity, and stability should also be considered as relevant to the desired application. For nucleic acid targets, nucleic acid probes may be the best option for applications where assay simplicity, cost, stability, and/or multiplexing are the most important considerations. The opportunity to choose between different nucleic acid analogues (e.g., DNA, RNA, PNA, LNA, morpholinos) and chemical modifications can be used to tailor the stability and affinity of the nucleic acid probes for the application of interest. Applications requiring exceptional sensitivity and selectivity may benefit from the use of the more complex and early-stage HCCD or CRISPR-dCas9-mediated sensing strategies.

Regarding future directions, further research and development are required to improve the availability of pre-designed synthetic antibody analogues (e.g., aptamers, MIPs, PCCs) against various biomarkers. The availability of successful MIP formulations may be enhanced by increased use of computational methods. Such computational methods can aid in the development and optimization of MIP formulations for targets of interest and reduce experimental effort by guiding researchers toward promising systems [119]. Future SiP biosensing studies should focus on biomarker detection in complex biological fluids to quantify bioreceptor selectivity and to ensure reliable detection performance when using clinically relevant samples. For instance, the validation of aptamers for target detection in complex biological samples is essential for their translation to real-world sensing applications due to the sensitivity of their three-dimensional conformation and binding affinity to the ionic strength and pH of the sample [180]. HCCD and CRISPR-dCas9-mediated sensing are in their infancy and future studies should focus on validating these strategies for sensing in complex biological samples. Moreover, future work should focus on multiplexing these CRISPR-based methods to enable simultaneous detection of multiple targets.

#### **3. Bioreceptor Immobilization Strategies**

The surface of unmodified SiP sensors consists of a native silicon dioxide layer, which grows on silicon and silicon nitride upon exposure to air and moisture [31,257]. This oxide surface is hydrophilic in character [258,259] and has a negative surface charge density above pH 3.9 [260]. Strategies for immobilizing bioreceptors on SiP devices generally rely on non-covalent interactions between bioreceptors and the native oxide surface or target surface silanol groups for covalent attachment. In this section we discuss bioreceptor immobilization strategies for SiP biofunctionalization, focusing on passive adsorption, bioaffinity binding, and covalent immobilization (Figure 12). We discuss methods relevant to antibody, aptamer, nucleic acid probe, peptide, PCC, glycan, lectin, and lipid nanodisc immobilization and present tables categorizing bioreceptor immobilization strategies that have been used in functionalization approaches in the previous literature. It should be noted that strategies discussed in the following subsections generally are not relevant to MIP-based bioreceptors, which are immobilized on SiP surfaces during synthesis via casting and/or in situ polymerization; as such, MIPs are not discussed in detail here. Table 27 provides a summary of bioreceptor immobilization strategies that have been employed on SiP devices in the literature and benchmarks these strategies against biofunctionalization needs for SiP biosensors.

**Figure 12.** Illustrations of different strategies for immobilizing bioreceptors (antibodies are shown as an example) on SiP sensor surfaces. The depicted immobilization strategies include (**a**) non-covalent passive adsorption, (**b**) covalent attachment, (**c**) bioaffinity-based oriented immobilization using antibody-binding proteins adsorbed to the surface, (**d**) bioaffinity-based oriented immobilization using antibody-binding proteins covalently linked to the surface, and (**e**) bioaffinity-based immobilization in which the surface and bioreceptor are covalently conjugated with biotin and streptavidin is used as a linking molecule.


**Table 27.**

Comparison

 of different bioreceptor

immobilization

 chemistries

 based on SiP

biofunctionalization

 needs.


**Table 27.** *Cont.* poly(thymine)/poly(thymine

 cytosine).

#### *3.1. Passive Adsorption*

Adsorption (Figure 12a) is the fastest and simplest method by which bioreceptors can be immobilized on a biosensor surface [29,68,69,80]. Adsorption-based bioreceptor immobilization has been widely used, especially in preliminary demonstrations of novel sensing architectures [31]. Bioreceptors may adsorb to a bare or modified SiP surface due to electrostatic, hydrophobic, polar-polar or Van der Waals interactions, or some combination of these non-covalent interactions [29,262]. Nevertheless, covalent and affinity-based strategies are typically preferred to adsorption-based immobilization.

One major disadvantage of adsorption is that it provides little control over the orientation of immobilized bioreceptors [31,69,204,263,274]. This may render binding sites unavailable for target capture, reducing the target binding capacity and, therefore, the sensitivity of the sensor. This random orientation, combined with intermolecular interactions, may also lead to poor bioreceptor loading density on the sensor surface [263]. Adsorption-based immobilization may lead to reduced bioreceptor activity due to folding or denaturation. This is especially relevant for protein-based bioreceptors, like antibodies, which are known to denature when adsorbed to surfaces, potentially changing the structure of their Fab fragments and diminishing their antigen-binding capacity [29,69,262,263]. Further, adsorbed bioreceptors are susceptible to desorption, leading to poor surface stability [31,69,261]. This is particularly relevant when the sensor is operated in flow conditions or when surface regeneration involving the release of targets from the sensor for multiple cycles of reproducible binding is desired. For example, Jönsson et al. [261] demonstrated that antibodies physisorbed onto chemically modified silicon dioxide surfaces were unstable toward changes in the surrounding medium, demonstrating significant desorption upon exposure to low pH, low surface tension, detergent, urea, and high ionic strength solutions. Finally, surfaces allowing for strong adsorption of bioreceptors may also be amenable to the adsorption of other biomolecules present in a complex biological sample, such as blood, leading to non-specific adsorption and high background signals [33]. Similarly, if other proteins possessing higher adsorption affinities to the sensor surface are present in the fluid, the bioreceptors may leach off the sensor [75]. This, in turn, compromises the selectivity of the sensor.

Despite the numerous limitations of adsorption-based functionalization, SiP sensors functionalized with lipid nanodiscs have demonstrated good stability and selectivity using adsorptive immobilization [148–150]. Indeed, lipid nanodiscs are especially amenable to adsorption-based immobilization because, like lipid bilayers, they are known to adsorb well to silicon dioxide surfaces [275,276]. This allows for simple nanodisc immobilization without the need to chemically modify the SiP surface or the nanodiscs. These nanodiscfunctionalized sensors were regenerated after target binding with good reproducibility and no appreciable nanodisc desorption using Ca2+-free buffer, indicating stable immobilization [148,149]. These sensors were also used for multiplexed detection of soluble proteins with minimal non-specific binding [150]. However, the native silicon dioxide surface of SiP sensors is negatively charged at physiological pHs, such as the buffered systems employed in these nanodisc studies, meaning that nanodiscs with lipid compositions containing a high percentage of anionic lipids show a lower affinity for the sensor surface, leading to poorer surface coverage [148,149]. Fortunately, this reduced affinity is predictable and could be counteracted, at least in part, by using higher spotting concentrations [149].

#### *3.2. Bioaffinity-Based Immobilization*

Bioaffinity-based receptor immobilization involves the creation of multiple noncovalent interactions between a bioreceptor and biomolecule(s) acting as linker to the substrate [41]. The sum of many weak interactions yields a strong link between the bioreceptor and the surface. Two of the most common bioaffinity-based receptor immobilization strategies used for SiP sensor functionalization involve antibody-binding proteins and the biotin-avidin system, both of which can achieve oriented bioreceptor immobilization [29,69,74,75].

Antibody-binding proteins, including Protein A, Protein G, Protein A/G and Protein L, have been widely used for the oriented capture of antibodies on biosensor surfaces [29,75,263,277]. Protein A is derived from *Staphylococcus aureus*, while Protein G is derived from *Streptococcus* species, and Protein L is derived from *Peptostreptococcus magnus* [75,277]. Both Proteins A and G reversibly bind the Fc region of antibodies, binding a maximum of two antibodies at a time, and have variable antibody-binding affinities that depend on the immunoglobulin (Ig) subclass and the species of origin [75]. Protein A can capture mammalian IgGs with dissociation constants as low as the 1–10 nM range, while Protein G can achieve slightly higher affinity capture of mouse and human IgGs with dissociation constants as low at the 0.1–1 nM range [278,279]. The oriented capture of antibodies by these proteins ensures that the antibody's Fab fragments are accessible for antigen capture, significantly enhancing the functionalized surface's antigen binding activity [277,278]. For example, Ikeda et al. [278] demonstrated a 4- to 5-fold increase in antigen binding capacity for antibodies immobilized on silicon wafers using Protein A, compared to antibodies immobilized via physisorption alone. This was attributed to the improved steric accessibility of the antigen binding sites of the well-oriented Protein A-immobilized antibodies. In addition to its Fc-binding regions, native Protein G has additional sites for albumin and cell surface binding; however, recombinant Protein G, containing only Fc-binding domains, has been produced using *E. coli* to prevent this nonspecific binding [74,75]. Protein A/G is a recombinant protein that contains the Fc-binding domains from both Protein A and G [74]. Similarly to Proteins A and G, Protein L also binds antibodies in an oriented manner, but instead of binding to the Fc region, Protein L binds to antibodies' κ-light chains outside of the antigen-binding site with dissociation constants as low as ~10 nM [277,279]. As a result, Protein L can bind any class of antibody, in addition to Fab fragments, which lack an Fc region, though its binding affinity is species-specific [277,280]. Indeed, a significant challenge associated with antibody-binding protein-directed bioreceptor immobilization is this Ig subclass and/or species-based variation in antibody-binding affinity; further this technique cannot be used to immobilize any bioreceptors aside from antibodies.

While Proteins A, G, and L allow for optimal orientation of immobilized capture antibodies, oriented immobilization of these antibody-binding proteins remains a challenge [75]. Fortunately, these antibody-binding proteins have several high affinity binding sites for antibodies, making their orientation on sensor surfaces less critical [41]. In the literature, antibody-functionalized SiP microring resonator sensors have been prepared using Protein A physisorbed on the sensor surface (Figure 12c) [1,264]. It has been reported that Protein A adsorbs onto silicon surfaces in a two-step process to yield a ~3–4 nm-thick adlayer [259,264]. First, a monolayer of Protein A is rapidly adsorbed on the surface; this first monolayer is denatured due to very strong non-covalent binding to the surface. Next, a second and third monolayer of Protein A are slowly adsorbed on the surface; these layers consist of non-denatured proteins which retain their ability to effectively bind the Fc region of antibodies. This strategy of passive Protein A adsorption followed by oriented antibody capture was used on sub-wavelength grating SiP microring resonators by Flueckiger et al. [264] and Luan et al. [1] to immobilize anti-streptavidin for model streptavidin-binding assays.

Others have tagged antibody-binding proteins with small molecules or other proteins to achieve higher-affinity binding to SiP sensor substrates. For example, Ikeda et al. [278] constructed a fusion of Protein A and bacterial ribosomal protein L2, which is termed "Si-tag" and binds strongly to silicon dioxide surfaces [281]. The authors demonstrated that the fusion protein was strongly immobilized on silicon dioxide surfaces in an oriented manner with a dissociation constant of 0.31 nM. The Si-tagged protein A also strongly bound mouse IgGs with a dissociation constant of 3.8 nM. The fusion protein immobilized 30–70% more IgG compared to physisorption of IgGs on bare silicon dioxide surface. Further, the fusion protein-immobilized IgGs demonstrated a 4- to 5-fold increase in antigen binding performance compared to the physisorbed IgGs. This functionalization strategy was subsequently demonstrated on a SiP microring resonator platform [282]. Christenson et al. [164]

leveraged the strong bioaffinity interaction between biotin and streptavidin to immobilize recombinant Protein G on silicon photonic crystal-total internal reflection sensors. In this work, the sensor surface was modified with silane-PEG-biotin molecules, followed by streptavidin, then biotinylated recombinant Protein G. Antibodies were immobilized on this surface and used to detect cardiac troponin I. Covalent immobilization of antibody-binding proteins to silicon-based substrates (Figure 12d) may also be facilitated via methods such as surface modification with silane and a crosslinker, followed by Protein A or G attachment, as demonstrated by Anderson et al. [283] or via click chemistry, as demonstrated by Seo et al. [277] on glass substrates.

Interactions between antibodies and antibody-binding proteins are reversible and can be disrupted by variations in pH [29,75]. This limits biosensor stability and complicates sensor regeneration because antigens cannot be easily eluted from the sensor surface without also removing the capture antibodies. In this way, sensor regeneration is possible, but requires that both the antigen and capture antibody be eluted from the antibodybinding protein-functionalized surface, followed by reapplication of the capture antibody for another round of detection [75,263,283]. For example, Seo et al. [277] covalently bound Protein A onto glass slides, followed by the immobilization of receptor antibodies (rabbit anti-goat IgGs). These antibody-functionalized slides were used to capture target antibodies (goat anti-human IgGs) and were then treated with a low pH glycine-HCl buffer to remove the receptor and target antibodies. After this wash step, only the covalently bound Protein A remained. The surfaces were then successfully regenerated for a second round of binding by reapplying the receptor antibodies. Similarly, Anderson et al. [283] covalently bound Protein A to silicon dioxide optical fibers, followed by the immobilization of capture antibodies (rabbit anti-goat IgGs). Then, the functionalized fibers were used to capture fluorescently labeled targets (Cy5.5-goat IgG). The surfaces were regenerated using a pH 2.5 glycine-HCl, 2% acetic acid solution, followed by re-application of the capture antibody. Four cycles of regeneration were performed successfully with no appreciable reduction in Protein A's Fc-binding capacity. However, the authors also reported unsuccessful regeneration of Protein A and G for an assay detecting plague F1 antigen, showing that regeneration of antibody-binding proteins may depend on the selected capture antibody and antigen. Here, the necessary reapplication of the receptor antibody also increases the cost and complexity of sensor reuse compared to strategies in which the functionalized surface can be regenerated solely by the removal of the target.

Another common bioaffinity interaction coupling method used in biosensor functionalization is based on the biotin-avidin/streptavidin complex, whereby the sensor surface is coated with avidin or streptavidin and used to immobilized biotinylated receptors (Figure 12e) [41]. Biotin is a small vitamin and avidin is a glycoprotein found in egg whites, which contains four biotin binding sites [265]. The biotin-avidin interaction is one of the highest affinity non-covalent interactions known in biology, with a dissociation constant on the order of 10−<sup>15</sup> M [75,80,265]. This nearly irreversible non-covalent interaction is extremely resistant to variations in temperature, buffer salt, pH, and the presence of denaturants and detergents [74,265]. Streptavidin is a biotin-binding protein, derived from *Streptomyces avidinii*, which shows similar biotin-binding activity to avidin [265]. Streptavidin, however, has a pI of 5, while avidin has a pI of 10.5; as such, streptavidin is less susceptible to nonspecific interactions at physiologic pH, often making it the preferred choice [80,265].

The high-affinity nature of the biotin-streptavidin interaction means that biosensor regeneration via target removal can be achieved without disrupting the link between the receptor and the surface [284]. This means that the sensor can be used for multiple cycles of target binding without reapplying receptors. For example, Choi et al. [285] functionalized silicon nitride chips for reflectometric interference spectroscopy by covalently linking biotin to the surface, followed by avidin, and biotinylated concanavalin A. This lectin-coupled chip was used to reproducibly capture ovalbumin, a glycoprotein, over multiple binding cycles by regenerating the surface with a 10 mM glycine-HCl (pH 1.5) solution, which removed captured glycoproteins, while leaving the lectin-functionalized surface intact. In another work [284], SPR surfaces were functionalized with a biotin analogue, desthiobiotin, followed by streptavidin, and biotinylated IgGs. The authors reported that the functionalized surface was stable throughout multiple cycles of regeneration with solutions commonly used for target removal from bioreceptors, including HCl, Na2CO3, glycine buffer, and SDS solutions. Lü et al. [286] functionalized optical fiber probes by covalently linking streptavidin to the exposed silicon dioxide core, followed by the oriented immobilization of 5- -biotinylated DNA probes. These surfaces were used to bind complementary DNA targets, followed by thermal regeneration via washing for 2 min in hybridization buffer at 70 ◦C, or chemical regeneration via washing in 4 M urea solution. The surfaces demonstrated no appreciable loss in hybridization ability over six cycles of thermal or chemical regeneration. Efforts have also been made to break biotin-streptavidin interactions for complete surface regeneration whereby receptors are completely removed from the surface [266,284]. This has been achieved using a pH 7 chemical buffer solution [266], sequential rinsing with free biotin, guanidinium thiocyanate, pepsin, and sodium dodecyl sulfate [284], and washing with water at 70 ◦C [287]. These strategies require the reapplication of biotinylated receptors and sometimes streptavidin/avidin between binding cycles, but also open the possibility for sensors to be reused with different bioreceptors for each cycle.

The biotin-avidin/streptavidin-based immobilization strategy is more flexible than antibody-binding proteins, in that many different classes of receptors can be tagged with biotin and immobilized on avidin/streptavidin-coated sensor surfaces. On SiP platforms, this biotin-avidin/streptavidin bioaffinity functionalization strategy has been used to immobilize antibodies [288,289] and nucleic acid probes for both hybridization sensing [203] and CRISPR-Cas-modulated high contrast cleavage detection [137,139]. Similarly, it has been used to immobilize lectins on a silicon nitride sensor using reflectometric interference spectroscopy as the transduction technique [285]. This strategy can achieve unoriented or oriented receptor immobilization. For antibodies, amine, carboxyl, sulfhydryl, and carbohydrate groups can all be targeted for biotinylation, depending on the choice of biotin derivative; this can lead to unoriented antibody capture in the case of amine and carboxyl targeting or oriented capture in the case of sulfhydryl or carbohydrate targeting [29,69,265]. Optimally oriented nucleic acid probe immobilization has been achieved through biotinylation at terminal groups [137,139,286].

#### *3.3. Covalent Immobilization*

Covalent strategies are the gold standard for bioreceptor immobilization on SiP biosensors. Covalent immobilization (Figure 12b) is versatile, robust, and can be used to tether many different types of bioreceptors to SiP surfaces, yielding irreversibly bound functional layers [29,31]. This irreversible immobilization is beneficial for stable sensor performance under flow conditions and across multiple cycles of regeneration. Covalent methods may yield a higher density of immobilized bioreceptors compared to physisorption and bioaffinity techniques, which may, in turn, increase sensitivity [263]. Designing and optimizing a suitable covalent immobilization chemistry, however, can complicate assay development and preparation. Surface pre-treatments, reagents, and reaction conditions must be carefully chosen to yield reproducible and homogeneously thin surface modifications, while avoiding damage to the biosensor surface and bioreceptors [31]. For the design of POC sensors, further considerations may include selecting a scalable chemistry and designing a workflow that is suitable for SiP chips integrated with electronics and optical inputs and outputs.

#### 3.3.1. Silane-Mediated Immobilization

Most covalent immobilization strategies for SiP sensors involve silanization with organosilanes. Organosilane-based methods have been used for antibody, aptamer, nucleic acid probe, glycan, and lectin immobilization on SiP devices. Silanes consist of a silicon atom bonded to four other constituents [290]. Organosilanes include silane reactive groups and at least one functional organic group. The silane reactive groups covalently couple to the sensor's native oxide surface by forming siloxane linkages with surface hydroxyls (Figure 13) [31,290]. A surface pre-treatment step is typically performed prior to silane deposition to remove organic contaminants and increase the number of surface hydroxyl groups available for silane grafting (Figure 13a) [31,74]. This pre-treatment step, which often involves oxidation via piranha, UV radiation and ozone, or plasma treatment, is essential to improving the silane grafting density and reproducibility of silanization. The silanization reaction can be performed using solution- or vapor-phase processes, with solution-phase processes being more widely used on SiP devices. However, no consensus on optimal reagents or reaction conditions exists, with significant variations in solvent choice, reagent concentrations, reaction time, and reaction temperature existing in the literature. After the silane is attached to the sensor surface, the silane's organic groups can react with other organic molecules to facilitate bioreceptor attachment. While it is possible to directly attach bioreceptors to the organosilane surface [25,172,185,291], it is more common to attach bioreceptors using a bifunctional crosslinker that is highly reactive toward both the silane and the bioreceptors, as the most commonly used organosilanes lack sufficient reactivity toward bioreceptors [29].

When attaching the bioreceptor, native reactive groups or non-native reactive groups introduced during synthesis are targeted for immobilization. These may include amine, carboxyl, thiol, or carbohydrate groups. The choice of targeted functional group affects the orientation of the immobilized bioreceptor. Antibodies, for example, possess native amines in their lysine residues and native carboxyls in their aspartate and glutamate residues [69]. These residues are abundant on the antibody surface, so targeting amines or carboxyls leads to unoriented antibody immobilization. Conversely, thiol groups present in cysteine residues of the hinge region can be targeted for site-directed antibody immobilization [69,75]. However, creating reactive thiol groups to target requires reduction of the hinge disulfide bonds, which may lead to undesired reduction of other disulfide bonds, potentially reducing the antibody's activity toward its target [75]. Native carbohydrate moieties present in the Fc region of antibodies can also be targeted for oriented capture [69,75]. Synthetic bioreceptors, including aptamers, nucleic acid probes, and glycans can be immobilized on silanized SiP surfaces by targeting terminal amine or thiol groups introduced during synthesis; this allows for oriented immobilization.

The most commonly used silanes for SiP functionalization are aminosilanes, particularly 3-aminopropyltriethoxysilane (APTES) (Figure 13). Aminosilanes contain organic groups that terminate in a primary amine, which can be targeted by amine-reactive crosslinkers for bioreceptor conjugation [290]. In order to initiate the reaction between the silane reactive groups of APTES and the hydroxyl groups present on the SiP surface, APTES must be hydrolyzed by moisture or water (Figure 13b) [74,290]. In the literature, APTES silanization of SiP sensors has been performed in anhydrous solvents such as toluene [201,239,292], acetone [17,195,293,294], and ethanol [132,202]. In these reactions, APTES hydrolysis is initiated by trace amounts of moisture present in the solvent [74]. APTES silanization has also been performed on SiP sensors using aqueous reaction solutions that contain a small quantity of water (e.g., ~5%) to catalyze APTES hydrolysis, combined with an organic solvent, typically ethanol [23,24,161,166,192,194,197,199,295]. These aqueous reactions are simpler than anhydrous ones, as they typically do not require drying the solvent or carrying out the reaction in a rigorously controlled inert atmosphere and/or under reflux [290,296]. However, in aqueous solutions, APTES is susceptible to copolymerization in the liquid phase prior to attachment to the solid substrate [258,268,297]. This can lead to the formation of thick and uneven films containing large silane aggregates (Figure 13d). Consequently, using an anhydrous solvent or maintaining low water content (~0.1%) in the reaction solution may yield thinner and more uniform silane layers [261,297,298]. Aside from solvent choice, an APTES concentration of 1–5% is typically used for solution-phase deposition [166,195,292], while reaction times vary significantly from several minutes [166,195] to overnight [202]. An alternative approach is vapor-phase

silanization, which has been used to create uniform monolayer aminosilane films on silicon substrates [261,267,268,298]. In vapor-phase techniques, APTES is hydrolyzed by atmospheric moisture [74]. Compared to solution-phase reactions, vapor-phase aminosilane deposition has been reported as more reproducible, less sensitive to reagent purity and atmospheric conditions, and less likely to deposit polymeric silane particles [261,267–269]. Further, vapor-phase silanization may be more suitable than solution-phase methods when functionalizing SiP chips integrated with chip-mounted electronics and optical inputs/outputs, as vapor-phase processes do not require solvents that may degrade PCB or photonic wire bond materials. The final step of APTES silanization is typically a curing step at elevated temperature, which aids in the removal of moisture and the formation of siloxane bonds between the silane and surface [192,197,202,290].

**Figure 13.** Silanization of SiP surface using 3-aminopropyltriethoxysilane (APTES). (**a**) The native oxide surface of the Si waveguide is pre-treated to remove organic contaminants and activate the surface hydroxyl groups, (**b**) APTES is hydrolyzed to form reactive silanols, and (**c**) adjacent APTES molecules are covalently linked together via silanol condensation and APTES is covalently bound to the surface. This yields a covalently bound APTES monolayer presenting functional amine groups for linker or bioreceptor immobilization. (**d**) Undesirable formation of large silane aggregates on the surface. (**e**) Attachment of a homobifunctional crosslinker to the aminosilane-coated surface, showing (i) ideal homobifunctional crosslinker attachment whereby one reactive group reacts with the silanized surface and the other remains available for conjugation with the bioreceptor, and (ii) undesirable crosslinker-mediated bridging whereby both ends of the homobifunctional crosslinker react with functional groups on the silanized surface, becoming unavailable for bioreceptor immobilization. BS<sup>3</sup> is used here as an example of a homobifunctional crosslinker.

Once the SiP surface has been modified with an aminosilane, bioreceptors are covalently linked to the surface via functional linkers such as glutaraldehyde (GA), bis(sulfosuccinimidyl)suberate (BS3), or 1-ethyl-3-[3-(dimethylamino)propyl]-carbodiimide/Nhydroxysuccinimide (EDC/NHS) [74]. GA and BS<sup>3</sup> are homobifunctional linkers, which crosslink amine groups on the silanized substrate to amine groups on the bioreceptor. GA

contains aldehyde groups which form imine bonds with amines via the formation of Schiff bases [31,74]. GA has been used to link antibodies [170], amine-terminated aptamers [24], amine-terminated DNA probes [23,197,199], and amine-terminated morpholinos [110] to aminosilane-modified SiP sensors.

GA linking has also been combined with SiP surface modification strategies whereby hydrofluoric acid (HF) is used to produce primary amines on silicon nitride waveguide surfaces [177]. These HF crosslinking approaches are particularly attractive for use with silicon nitride waveguides since they can be designed so that the amines are only produced on the nitride and not on the surrounding oxide [299]. This method uses basic cleaning methods followed by a HF dip to produce primary amines on the waveguide surface without the need of an additional aminosilane surface coating step. Next the sensor is immersed in a 2.5% GA crosslinker solution and washed. Bañuls et al. [299] developed this process to increase and localize biotarget capture to waveguide surfaces. The authors hypothesized that oxide comprised 98% of their sensor surface area with only 2% of the surface belonging to the silicon nitride sensing waveguides. This suggested that nonselective bioreceptor immobilization would lead to the majority of the target being captured by bioreceptors immobilized outside the sensing region. To show selective attachment to silicon nitride, slot waveguide ring resonator biosensors were modified with BSA and anti-BSA using the HF/GA procedure. Their results showed a detection limit of 28 pg/mm<sup>2</sup> for anti-BSA antibody immobilization on the surface and 16 pg/mm<sup>2</sup> for BSA. A similar procedure was used by Angelopoulou et al. [238] who modified MZI sensors with HF and GA, then spotted mouse IgG on individual sensors using an inkjet printer for multiplexing, followed by incubation with fluorescently labeled goat anti-mouse IgG antibodies and washing steps. The authors tested this direct attachment method in comparison to physical adsorption of the bioreceptors on amine-terminated silane (APTES) coated waveguides. The silane protocol yielded fluorescently tagged antibodies attached to both the waveguides and the surrounding oxide, whereas the HF procedure only functionalized the silicon nitride waveguides (Figure 14). Next, both sensors were spotted with a peptide, Receptor Binding Domain (RBD) of SARS-CoV-2 Spike 1 protein, and a BSA blocking protein on the sensing and reference waveguides, respectively. The HF method produced well-coated waveguides with the response of the reference sensor showing little change compared to the baseline signal upon exposure to anti-RBD antibodies. In comparison, the APTES modified reference sensor response could be clearly distinguished from the baseline signal. This suggests that BSA did not fully coat the APTES coated waveguides.

BS<sup>3</sup> consists of sulfo-NHS esters at either end of an 8-carbon spacer arm [300]. The NHS esters react with primary amines to form stable amide bonds. BS<sup>3</sup> has been used to conjugate antibodies [17,166], amine-terminated DNA probes [195], and peptides [239] to APTES-modified SiP sensors. When applied to antibody immobilization, GA and BS<sup>3</sup> target native amine functional groups that are abundant on the antibody surface, leading to random antibody orientation. Moreover, as these immobilization strategies target functional groups that are abundant on the antibody surface, they may result in the formation of multiple bonds between the antibody and the surface [74]. This may lead to conformational changes of the antibody and render binding sites inaccessible for target capture. As such, spacer molecules or hydrophilic polymers can be incorporated into the linking chemistry to reduce steric hindrance and the risk of bioreceptor denaturation. The hydrophilic polymer, oligo(ethylene glycol), which can be used for this purpose, has also shown antifouling properties with short chains (≤7 repeats), which create a less ordered surface and decrease non-specific adsorption [301]. When applied to bioreceptors modified with terminal amine groups, such as 5 amine-modified aptamers or nucleic acid probes, linking strategies using GA and BS<sup>3</sup> permit site-directed immobilization. Another notable limitation of homobifunctional crosslinkers like these is that they may form bridged structures where both reactive ends are linked to the substrate, limiting the number of binding sites available for bioreceptor attachment and thus reducing bioreceptor density (Figure 13e) [302]. This can be avoided with heterobifunctional crosslinkers. EDC/NHS is a heterobifunctional

crosslinker combination using carbodiimide chemistry, which links carboxyl groups on the bioreceptor to amine groups on the silanized substrate via the formation of stable amide bonds [74,176]. This linker chemistry has been used to covalently attach antibodies to APTES-modified MRRs [166] and silicon photonic crystals [165]. Since this strategy targets abundant carboxyl groups, which are also abundant on antibody surfaces, it results in unoriented antibody immobilization and may cause conformational changes, as described above. A similar carbodiimide chemistry was used by Peserico et al. [202] in which an APTES-modified MRR chip was carboxylated with succinic anhydride, then EDC was used to covalently link 5 amine-modified DNA probes to the carboxyl-presenting surface, this time in an oriented manner.

**Figure 14.** Silicon nitride waveguides from a MZI sensing window with fluorescently tagged (Alexa Fluor 488) antibodies (goat anti-mouse IgG) attached by (**a**) covalent HF/glutaraldehyde-based immobilization and (**b**) APTES functionalization followed by passive adsorption [238]. Parts (**a**,**b**) are adapted with permission from Ref. [238]. Copyright 2022 Elsevier. (**c**) Chemical reaction mechanism for selective silicon nitride functionalization by HF and glutaraldehyde crosslinking [299]. In (**c**), the subscript of "3" on the glutaraldehyde structure indicates that only one of the three carbons between the formyl end groups has been drawn for brevity. Part (**c**) is adapted with permission from Ref. [299]. Copyright 2010 Elsevier.

Despite their popularity, GA, BS3, and EDC/NHS linker chemistries pose reproducibility challenges. GA polymerizes in aqueous solutions and the extent and nature of this polymerization depends on the age of the solution and can be difficult to control and reproduce [300]. BS3 and EDC/NHS linker chemistries both involve NHS ester groups which rapidly hydrolyze in aqueous solutions [31,74]. This rapid hydrolysis competes with biomolecule conjugation and is highly sensitive to reaction conditions, hindering reproducibility and limiting the yield of the conjugation reaction.

Bioreceptor conjugation using SoluLink chemistry is another silane-based strategy which offers good reproducibility and has been extensively used on SiP devices, namely the commercial Genalyte MRR platform [31,295]. In the literature, this chemistry has been used to covalently immobilize antibodies [18,22,161,168,169,174,295], 5 amine-modified aptamers [174], and 5- amine-modified DNA probes [109,194,196,198,303] on MRRs. Using this strategy, the bioreceptor is reacted with succinimidyl-4-formylbenzamide (S-4FB), which targets primary amines via succinimide coupling. The substrate surface is either modified with an aminosilane, followed by reaction with 6-hydrazinonicotinamide (S-HyNic) [161,194], or the bare SiP surface is directly

reacted with HyNic-silane [18,22,109,168,169,174,196,198,295,303]. The 4FB-conjugated bioreceptors are introduced to the HyNic-modified surface, leading to bioreceptor immobilization through hydrazone bond formation. This reaction proceeds slowly, but aniline can be used as a catalyst to increase the rate of reaction, improve bioreceptor loading on the substrate, and allow for lower reagent consumption [295]. Despite its good reproducibility, chemically modifying bioreceptors with 4FB prior to immobilization adds time and complexity to this technique. More recent demonstrations on the Genalyte platform have instead used APTES silanization and BS<sup>3</sup> to immobilize unmodified amine-containing bioreceptors for simple and flexible assay design [17,27,195,304,305].

Others have used 3-mercaptopropyltrimethoxysilane (MPTMS) to install thiol groups on SiP sensor surfaces to mediate bioreceptor immobilization. Thiolated bioreceptors can be directly conjugated to MPTMS-modified surfaces without an intermediate crosslinker through the formation of disulfide bonds [31]. For example, Chalyan et al. [25] directly immobilized Fab fragments on a MPTMS-modified SiP sensor. In this work, the Fab fragments were generated from protease digestion of polyclonal antibodies, followed by the reduction of hinge disulfide bonds to generate reactive thiol groups [25,69]. A similar strategy omitting the protease digestion step can also be used for site-directed antibody capture on MPTMS-modified surfaces [306]. However, covalent immobilization via thiolbearing cysteine residues, which are usually internal to the antibody structure, and the unintentional reduction of non-target disulfide bonds may disrupt antibody conformation and binding affinity [69,75]. In addition to antibodies, this thiol-directed covalent strategy has been used for nucleic acid probe immobilization. Sepúlveda et al. [200] modified silicon nitride Mach-Zehnder interferometer sensors with MPTMS, followed by covalent and oriented immobilization of 5thiol-modified ssDNA probes.

Bioreceptors that lack reactive thiols can also be conjugated to MPTMS-modified surfaces using maleimide linkers. For example, Xu et al. [175] covalently immobilized antibodies on a MPTMS-modified planar silicon nitride optical waveguide interferometric biosensor using m-maleimidobenzoyl-N-hydroxysuccinimide ester as a thiol-to-amine crosslinker. Ghasemi et al. [133] covalently attached amine-derivatized glycans to MPTMS-modified silicon nitride MRRs using a SM(PEG)12 linker. SM(PRG)12 contains a polyethylene glycol (PEG) chain terminated by NHS ester and maleimide reactive groups. As such, it acted as a heterobifunctional linker between the thiolated surface and amine-derivatized glycans, while the PEG chain prevented nonspecific interactions between non-target molecules and the sensor surface.

3-Glycidoxypropyltrimethoxysilane (GPTMS) is another silane that can mediate direct covalent immobilization of bioreceptors on SiP sensors. GPTMS installs epoxy groups on silicon surfaces, which are reactive toward amine, thiol, or hydroxyl groups [31,290]. Ramachandran et al. [172] conjugated monoclonal antibodies and 5 amine-modified ss-DNA probes to GPTMS-modified glass (Hydex) MRRs. Using this strategy, the bioreceptors were covalently linked to the surface via amine reactive groups, resulting in unoriented and oriented antibody and ssDNA probe capture, respectively. Chalyan et al. [25] and Guider et al. [185] covalently immobilized amine-terminated aptamers on GPTMSmodified silicon oxynitride MRRs in an oriented manner.

#### 3.3.2. Organophosphonate-Mediated Immobilization

Organophosphonate chemistry presents a promising alternative to silane chemistry. Compared to silanes, phosphonate films can achieve greater monolayer density, surface coverage, and stability, and have a lower tendency to form multilayered structures [270,307]. Shang et al. [126] demonstrated covalent immobilization of amine-bearing glycan and glycoprotein bioreceptors on silicon MRRs using an organophosphonate surface coating and an amine-vinyl sulfone linker (Figure 15). After treating the surface with piranha solution to increase the number of available surface hydroxyl groups for organophosphonate grafting, the sensor surface was coated with a monolayer of 11-hydroxyundecylphosphonic acid (UDPA). This was achieved using the "T-BAG" method whereby UDPA is adsorbed onto the

substrate, then heated to 120–140 ◦C to activate the formation of covalent linkages [126,307]. After the sensors were modified with UDPA, divinyl sulfone (DVS) was used to link the hydroxyl-terminated organophosphonate film to the amine-bearing bioreceptors [126]. In this work, the MRRs demonstrated excellent stability and reproducibility across multiple cycles of chemical regeneration and long-term storage at ambient conditions. A similar strategy was used to functionalize the surface of silicon nanowires with cysteine-modified PNA oligonucleotides [270]. Here, 3-maleimidopropionic-acid-N-hydroxysuccinimidester was used instead of DVS as a heterobifunctional linker to attach the thiol-containing PNA oligonucleotides to the UDPA-modified nanowires.

**Figure 15.** Organophosphonate-based surface functionalization scheme whereby the surface is coated with a film of UDPA using the T-BAG method and a DVS linking strategy is used for the immobilization of aminated bioreceptors. Reprinted with permission from Ref. [126]. Copyright 2012 American Chemical Society.

#### 3.3.3. Click Chemistry

Click chemistry is a widely used crosslinking technique for simple, fast, and selective attachments with high efficiency. This method is attractive for biorecognition components since it uses physiological reaction conditions (neutral pH, buffered solution). Briefly, click chemistry involves linking molecules via heteroatom links (C–X–C) [271]. There are three main click procedures based on Cu(I) catalyzed azide-alkyne, strain promoted azidealkyne, and tetrazine-alkene ligation reactions. The reaction is simple, more efficient than EDC/NHS chemistry, selective to only click reagents, has many commercially available modular components, and is not sensitive to oxygen or water [271].

This method has been used to immobilize ssDNA probes [59] and PCCs [91,240] on the surfaces of silicon-based optical biosensors. Juan-Colás et al. [59] demonstrated a novel silicon electrophotonic biosensor consisting of silicon MRRs fabricated with a thin n-doped layer at their surface to combine high-Q-factor photonic ring resonance with electrochemical sensing (Figure 16). In this work, the MRRs were covalently functionalized with ssDNA probes using the popular copper-catalyzed azide-alkyne click reaction. Firstly, two electrophotonic MRRs fabricated on a single chip were modified by electrografting azidoaniline or ethynylaniline onto the rings to install azide or alkyne groups, respectively (Figure 16a). The two electrophotonic MRRs were individually addressable, allowing for site-directed electrografting of azide groups on one ring and alkyne groups on the other. Next, the copper-catalyzed azide-alkyne click reaction was performed to conjugate azide-modified ssDNA probes to the alkyne-modified ring and alkyne-modified ssDNA probes to the azide-modified ring (Figure 16b,c). This unique strategy permits high-density multiplexed functionalization with submicrometer- to micrometer-scale precision, though it is not suitable for traditional SiP sensors that lack electrochemical control. Click chemistry was also used by Cao et al. [240] and Layouni et al. [91] to covalently link PCCs to porous silicon surfaces. In these works, the surfaces were modified with alkyne moieties by thermal hydrolyzation with 1,8-nonadiyne, followed by copper-catalyzed azide alkyne cycloaddition to attach azide-modified PCCs. This method requires removal of the substrate's native oxide layer by exposure to HF prior to hydrolyzation. Consequently, this method may not be suitable for SiP devices patterned with extremely fragile silicon structures like sub-wavelength gratings, which may be partially etched or delaminated upon exposure to HF. Overall, some of the key advantages of click chemistry compared to silane-mediated strategies are its insensitivity to oxygen and water and its chemoselectivity, which prevents side reactions with other bioreceptor functional groups and preserves bioreceptor activity [29,91,308]. However, a limitation is the requirement for prior surface and bioreceptor modification with functional tags, like azide and alkyne groups, which adds complexity to the functionalization process [29,234].

**Figure 16.** Click chemistry-mediated immobilization of nucleic acid probes on an electrophotonic ring resonator. (**a**) Two different diazonium salts (azidoaniline and 4-ethynylbenzene) are electrografted on the electrophotonic rings, which are electrically isolated. The individual microrings are then functionalized with alkyne- and azide-modified DNA probes using copper-catalyzed azide-alkyne click reaction. (**b**) First, the azide-modified sensor is functionalized with the alkyne-modified singlestranded DNA probe (ssDNAalkyne). (**c**) Next, the alkyne-modified sensor is functionalized with the azide-modified single-stranded DNA probe (ssDNAazide). The target sequences complementary to ssDNAazide (**d**) ssDNAalkyne (**e**) (labeled cDNAazide and cDNAalkyne, respectively) are introduced and hybridize to the functionalized sensors. Adapted from Ref. [59] in accordance with the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

#### 3.3.4. UV-Crosslinking

Direct UV-crosslinking of nucleic-acid-based bioreceptors has been demonstrated on planar glass and silicon dioxide wafers. This is a simple and inexpensive method that could be extended to SiP biosensors. Gudnason et al. [273] linked poly(T)10-poly(C)10-tagged ssDNA probes to unmodified glass surfaces using UV light irradiation. The immobilized probes demonstrated similar hybridization efficiency when compared to ssDNA probes immobilized on an amino-silane surface via traditional chemical crosslinking. The UV-linked probes showed no appreciable decrease in hybridization performance after incubation in water at 100 ◦C for 20 min, demonstrating strong thermal stability. In this work, the hybridization assay was performed in PerfectHyb Plus buffer to obviate the need for a surface blocking step. A similar strategy was used by Chen et al. [272] to covalently link thrombin-binding DNA aptamers with poly(T)20 tails to unmodified glass and silicon dioxide wafer surfaces using UV irradiation, while maintaining strong target affinity. Note that in this work, thrombin binding was performed in the presence of BSA and Tween-20 surfactant to reduce non-specific binding of the target to unmodified regions of the substrates. This UV-linking strategy is both simple and rapid because it requires no prior chemical modification of the substrate. Additionally, the nucleic acid-based bioreceptors do not require chemical modifications with reactive functional groups, lowering synthesis costs. However, to our knowledge, this strategy has not yet been demonstrated on patterned SiP sensor surfaces.

Tables 28, 29, 31 and 33 and Section 3.4 outline strategies that have been demonstrated on SiP sensors and representative surfaces for the immobilization of antibodies (Table 28), aptamers (Table 29), nucleic acid probes (Table **??**), peptides and PCCs (Table 31), glycans and lectins (Table **??**), HCCD reporters (Table 33), CRISPR-dCas9-mediated sensing probes (Table **??**), and lipid nanodiscs (Table **??**) in the previous literature.

#### *3.4. Summary and Future Directions*

We have discussed adsorption, bioaffinity, and covalent strategies for immobilizing bioreceptors on SiP surfaces. While adsorption-based strategies offer excellent simplicity, their poor stability and lack of control over bioreceptor orientation limit their suitability for SiP biosensing applications. However, novel polymeric coating materials, such as PAcrAm™ and AziGrip4™ from SuSoS AG, are available and replicably self-assemble as stable monolayers on silicon substrates by adsorption from solution [309–311]. These polymeric coatings have customizable functional binding groups and allow for covalent and electrostatic capture of bioreceptors on the adsorbed coating [309–311]. This may create the opportunity for bioreceptor immobilization with similar simplicity to passive adsorption, but with improved stability and more controllable bioreceptor orientation, making this a potentially valuable future research direction. To the best of our knowledge, such functionalization techniques have not yet been demonstrated on SiP platforms.

Bioaffinity and covalent strategies typically offer improved stability and control over bioreceptor orientation compared to adsorption, but at the cost of increased complexity [41]. Bioaffinity strategies involving antibody-binding proteins permit controlled antibody orientation, but have limited stability compared to biotin-based and covalent methods [265,278,279]. Covalent strategies, especially those using silanization, have been widely used on SiP platforms, as they can permit very stable and tailorable bioreceptor immobilization [29,31]. When designing a covalent immobilization protocol, surface pre-treatment must be carefully considered to ensure that the sensor surface is free of organic contaminants prior to applying the immobilization chemistry, and to activate surface functional groups (e.g., hydroxyls) that will be targeted by the immobilization chemistry [31,74]. Such pre-treatments improve grafting density on the sensor surface, while also improving the reproducibility of the immobilization protocol [74]. Pre-treatment approaches that have been used in SiP bioreceptor immobilization protocols, such as piranha, UV radiation and ozone, plasma, and HF treatments, have been comprehensively summarized in Tables 28, 29, 31 and 33 and Section 3.4. Future work should focus on optimizing standardized silanization protocols that can be used for highly replicable, scalable, and robust surface modifications with limited silane aggregation. In parallel with future work focusing on the system-level integration of SiP sensors for POC use, immobilization chemistries that are compatible with these integrated sensor architectures should be designed and tested. For example, translating solution-phase surface modification protocols to vapor-phase ones may reduce the risk of damage to the sensing system during functionalization, while improving scalability, reproducibility, and film uniformity [261,267–269]. Immobilization strategies using UVcrosslinking of bioreceptors directly to unmodified surfaces should also be explored on SiP sensors as a potentially simple, low cost, and scalable immobilization technique [272,273].

In designing immobilization protocols, potential steric crowding effects should also be considered in the context of bioreceptor immobilization and target capture. For example, crowding of bioaffinity linkers on the sensor surface may hinder subsequent bioreceptor immobilization [289]. These steric effects can be counteracted by using a higher bioreceptor concentration in the immobilization protocol or by using long linking molecules to increase the distance between the sensor surface and bioreceptors, providing more flexibility for the receptors to optimize steric crowding. When using these longer linking molecules, however, the potential sensitivity trade-offs associated with moving the binding reaction farther away from the sensor surface should also be considered. Immobilization approaches using these longer linking molecules may be most suitable for SiP architectures with greater evanescent field penetration depths (e.g., those based on ultra-thin [40] or sub-wavelength grating [16,46,264] waveguides). Similarly, dense receptor packing on the sensor surface may not always enhance target binding. Steric hindrance effects due to target molecule binding can reduce the rate of the forward binding reaction for neighboring receptors and affect the dynamic range of the sensor [312]. Thus, these steric effects should be accounted for when optimizing bioreceptor immobilization protocols.


**Table 28.**

Strategies

demonstrated

 for the

immobilization

 of antibodies on SiP sensors.


**Table 28.** *Cont.*





#### **4. Patterning Techniques**

In this section, we introduce several patterning techniques that can be used for SiP sensor functionalization and benchmark them against the critical patterning performance criteria relevant to SiP biosensing, as outlined in Table 2. A high-level comparison of these patterning techniques is provided in Table 36. The subsequent subsections provide further details about each patterning technique, outline their opportunities and limitations for multiplexed SiP biofunctionalization, and highlight demonstrations from the previous literature in which these patterning techniques have been used to deposit bioreceptors on SiP biosensors. For each patterning technique, tables categorizing these demonstrations from the previous literature are provided.

#### *4.1. Microcontact Printing*

Microcontact printing (μCP), also called microstamping, is a soft lithography method whereby geometrically defined 2D patterns of biomolecules are transferred to a substrate using an elastomeric stamp (Figure 17a) [320,321]. This technique has been used to prepare patterns of bioreceptors like antibodies [32,322], DNA [323–325], MIPs [326], and carbohydrates [327] on solid substrates.

**Figure 17.** (**a**) Illustration of the process to pattern a surface using μCP. (i) First, the elastomeric stamp is inked with a bioreceptor solution whereby bioreceptors adsorb to the stamp surface. Inking may be achieved using soak, spray-on, or robotic feature-feature ink transfer methods. Subsequently, the stamp can be rinsed and dried or used wet for stamping. (ii) The stamp is contacted with the sensor surface and gentle pressure is applied to transfer the bioreceptors from the stamp to the surface at the regions of contact. (iii) The stamp is released to reveal the bioreceptor-patterned surface. (**b**) Graphical representation of the functionalization of a SiP waveguide using tip-mold microcontact printing, showing the (i) waveguide cross-section of a reference microring, (ii) waveguide cross-section of a control microring, (iii,iv) application of ssDNA probes on the waveguide surface using a PDMS tip-mold μCP tool, and (v) hybridization of ssDNA targets to the immobilized ssDNA probes on the waveguide surface. (**c**) Images of (i) the SiP MRR sensor chip functionalized by Peserico et al. [202] via tip-mold μCP, (ii) the optical fiber tip with an unpatterned PDMS cladding used as the μCP tool, and (iii) example of bioreceptor application on MRR using μCP. Parts (**b**,**c**) are adapted with permission from Ref. [202]. Copyright 2017 The Institution of Engineering and Technology.


*Biosensors* **2023**, *13*, 53

**Table 36.**

Comparison

 of bioreceptor

 patterning techniques based on SiP

biofunctionalization

 needs.

the authors.

The first step of μCP is fabricating the elastomeric stamp. Polydimethylsiloxane (PDMS) is the most popular stamp material for μCP because it is easy to mold, flexible, chemically inert, and impermeable to biomolecules like proteins [32,320,328]. In μCP, the geometry of the stamp is defined by casting it in a master mold, prepared by photolithography or micromachining [320,321]. Once the stamp has been cast, it is "inked" with the bioreceptor solution to be deposited on the substrate. The ink adheres to the stamp via passive adsorption, which can be tuned by modifying the stamp's surface wettability with plasma or ozone treatment [32,328]. The inked stamp can be dried prior to stamping or used wet [329]. Next, the stamp is contacted with the substrate under a load, which can be achieved robotically or using a micropositioner to ensure precise alignment. The stamp is removed, leaving behind a 2D pattern of bioreceptors. The transfer of ink from the stamp to the substrate depends on the differential wettability between the stamp and substrate; in particular, the substrate must have greater wettability and, therefore, greater affinity toward the ink compared to the stamp [32,316].

A notable advantage of μCP is its excellent resolution. Patterns with critical dimensions down to 0.1–0.5 μm can be achieved [32,77]. This resolution is more than sufficient for patterning biomolecules on SiP surfaces, where the patterned sensing structures, like MRRs, typically have dimensions on the order of 10 μm. Some other advantages of μCP include its procedural simplicity, low cost, and good reproducibility [77,321,328,329]. PDMS stamps are robust and can be reused many times without significant loss of performance, but they are also sufficiently inexpensive and easy to fabricate that they can be treated as disposable when sample contamination is a concern [77,321]. Compared to printing techniques that address one spot on a substrate surface at a time, μCP is high-throughput, as a complex 2D pattern can be printed with only a single inking and application step [329].

While μCP is suitable for efficiently creating complex 2D patterns of a single bioreceptor, it is poorly suited to creating multiplexed arrays with many different bioreceptors [32]. Multiple cycles of inking and printing and careful stamp alignment would be required to print multiple bioreceptors, making this a time-consuming and cumbersome process. Another challenge is that bioreceptor immobilization strategies often include surface modifications, like silanization, which increase surface hydrophobicity prior to bioreceptor attachment [258]. This can reduce the differential wettability between the stamp and substrate, which may, in turn, reduce the efficiency of bioreceptor transfer to the substrate. Materials like PDMS can also transfer unwanted materials like residual uncured oligomers to the regions of the chip that they contact during stamping, potentially contaminating the surface and complicating bioreceptor patterning and subsequent assay steps [330–332]. Other limitations of μCP include a potential reduction in bioreceptor binding activity due to drying [322,329], patterning accuracy issues due to PDMS deformation under loads and swelling in the presence of some solvents [328], the requirement for cleanroom facility access to fabricate stamp master molds [77,329], and potential damage to the fragile sensor surface resulting from direct contact with the stamp.

To date, μCP has not been widely used to pattern SiP biosensors, though Peserico et al. [202] used a "tip-mold microcontact printing" technique to functionalize silicon nitride MRRs with ssDNA probes in a spatially defined manner (Table 37, Figure 17b,c). Instead of using a traditional stamp, a PDMS μCP probe was prepared by casting a thin layer of PDMS over the tip of a 125 μm-diameter optical fiber. The probe tip was treated with hydrochloric acid and hydrogen peroxide to enhance its hydrophilicity, then inked in a solution of amine-modified ssDNA. Using a micrometric positioner, the inked probe tip was contacted with the MRR of interest for 45 min in a humidified environment. This allowed sufficient time for the probes to covalently link to the amine-reactive sensor surface, which had previously been modified with APTES and a succinic anhydride/EDC linker. The authors reported that the printed ssDNA probes retained good hybridization efficiency toward their targets. Overall, a resolution of 100 μm was reported for this μCP method, which was suitable for the 200 μm-diameter MRRs used. While this variant of μCP could be used for multiplexed functionalization if parallelized with multiple tip-mold probes or multiple cycles of inking and printing, such a process would be cumbersome, time-consuming, and generally unsuitable for high-throughput biosensor preparation.

**Table 37.** Demonstration of bioreceptor patterning using μCP for the functionalization of SiP sensors.


#### *4.2. Pin and Pipette Spotting*

Nano- and micropipettes filled with a bioreceptor solution can be used in contact mode to deposit small drops of reagent on a substrate by capillarity [33]. Manual spotting of bioreceptor solutions with a micropipette, potentially accompanied by a microscope or stereoscope for improved positional accuracy, is a simple and low-cost technique for spatially controlling the deposition of different bioreceptor solutions on specified regions of a SiP chip in the research setting. However, this low-resolution technique has limited reproducibility, accuracy, and throughput. This technique could be adapted to a high throughput multiplexed dispense format using a pipetting robot [333]. Commercially available pipetting robots, however, typically have minimum dispense volumes of 200–500 nL [334–336], which is approximately three orders of magnitude greater than the dispense volumes achievable with pin- and inkjet-based dispensing. Thus, this strategy would still be limited by poor resolution.

Pin-based spotting or pin printing is a similar technique whereby a robotically controlled pin is loaded with the printing solution, then tapped on the sensor surface to deposit picoliter- to nanoliter-scale droplets (Figure 18) [77,329]. Pin printing has been widely used for the preparation of DNA microarrays, and commercial arrayed pin printers are available for this purpose [33]. This technique offers low sample consumption and good resolution, with minimum spot sizes in the range of 1–100 μm, depending on the pin geometry [33,77,313].

**Figure 18.** (**a**,**b**) Illustration of pin printing, showing (**a**) pin loading with the bioreceptor solution (antibody solution illustrated here as an example) and (**b**) printing of bioreceptors on a sensor surface using the loaded pin. (**c**–**e**) Different pin geometries, including (**c**) a solid pin, (**d**) split pin, and (**e**) quill pin.

Variations of pin printing include contact printing with solid, split and quill pins (Figure 18c–e) [77,329,337]. Solid pins are usually fabricated from micromachined stainless steel, tungsten, or titanium, and have convex, flat, or concave tips. They are loaded by dipping the pin tip in a reservoir filled with the bioreceptor solution and must be reloaded every few spots [77]. Commercially available solid microarraying pins available from Arrayit Corporation can print spots down to ~90–100 μm in diameter [315]. Solid pins are suitable for printing viscous liquids. This is valuable for protein solutions which are often prepared with viscous additives like glycerol, concentrated sugars, or high molecular weight polymers [77,329]. However, the requirement for frequent pin reloading makes solid pin printing very time-consuming. This limitation is addressed by split and quill pin designs, which permit serial printing of many spots from a single load. Split pins are fabricated with a 10–100 μm-diameter microchannel that is filled by capillary action during sample loading [77]. During printing, the split pin must be impacted on the substrate to overcome surface tension and eject picoliter- to nanoliter-scale droplets [77,337]. Quill pins have a similar design to split pins, but with a larger fluid reservoir [338]. Consequently, they can print hundreds of spots from a single load. Unlike split pins, quill pins only require a small tapping force to eject sample droplets onto the substrate [329,337]. Commercially available split and quill microarraying pins can achieve spot volumes down to ~350 pL and spot sizes down to ~37.5 μm in diameter [339–341]. Split and quill pins are best suited to low-viscosity solutions because they are susceptible to clogging with viscous liquids, which hinders spot reproducibility [77,329].

Split and quill pins can be micromachined from metal, but they have also been fabricated from silicon using standard microfabrication techniques that offer lower cost and smaller pin dimensions for improved resolution [33,77]. The BioForce Nano eNabler (Bioforce Nanosciences, Virginia Beach, VA, USA) is a commercial automated pin-based printer which uses a microfabricated silicon cantilever, called a Surface Patterning Tool (SPT), to deposit 1–60 μm droplets with 20 nm positional accuracy in the x, y, and z directions [313,314]. The SPT cantilever includes an integrated microfluidic network consisting of a reservoir to hold 0.5 μL of sample and a microchannel through which the sample flows to the tip via capillary action [313]. The droplet size is controlled by the contact time and contact force of the cantilever tip with the surface [313].

Pin-based functionalization of biosensors can be multiplexed by replacing or washing the printing needle when switching solutions [77]. In general, solid pins are easier to clean than split or quill pins, which usually require ultrasonication (for metal pins) or heating with a propane torch (for silicon pins) to thoroughly remove contamination [337]. Regarding its suitability for patterning SiP sensors, pin printing is inherently a contact technique that may damage fragile SiP structures [329]. A major challenge associated with pin printing is that optimizing spot size and reproducibility is a highly multifactorial problem [338]. Namely, the printing performance is highly dependent on the fluid properties, surface wettability, pin geometry, surface contact force, robotic controls, and environmental conditions [338]. Temperature and humidity control are typically required to slow evaporation of the sample, lower the risk of pin clogging, facilitate covalent bioreceptor immobilization on the surface, and preserve bioreceptor activity [329]. Further, spot reproducibility may deteriorate over time as a pin deforms from repeated contact with the substrate or as a split or quill pin's reservoir is depleted [77]. All of these considerations must be accounted for when designing a protocol for reliable SiP biosensor functionalization.

In the literature pipette and pin spotting have been widely used to pattern bioreceptors on SiP biosensors (Table 38). Several works have used spotting with a micropipette to pattern SiP sensors with 0.1–10 μL-scale droplets of antibodies [17,166,168,170], ssDNA probes [109,194–196,198,303], and lipid nanodiscs [148]. These strategies have been used to create 2- [198] to 9-plex [148] multiplexed biosensors. Several other works have employed the BioForce Nano eNabler to pattern SiP sensors with bovine serum albumin [313], ss-DNA [163], glycans [132,133], and lipid nanodiscs [149]. These works have reported the successful preparation of 2- [133] to 8-plex [163] biosensors. Angelopoulou et al. [238] spot printed antibodies and peptides on a silicon nitride MZI sensor chip with a contact printing arrayer using solid (375 μm tip, 12 nL per spot) and quill (62.5 μm tip, 0.5 nL per spot) pins. The spotting design required multiple overlapping spots to coat the waveguides with the solid pins taking 7 times as long to print despite depositing more liquid per spot compared to the quill pins. The authors found no significant difference in the sensor response between the solid versus quill pin tips.


**Table 38.** Demonstrations of bioreceptor patterning using pin and pipette spotting for the functionalization of SiP sensors.

#### *4.3. Microfluidic Patterning in Channels*

Microfluidic patterning in channels is a soft lithography technique whereby a gasket fabricated with microchannels, also called a microfluidic network (μFN), is reversibly bonded to a solid substrate and the bioreceptor solution is drawn through the microchannels (Figure 19) [33,77,321,342]. Bioreceptors are, therefore, patterned on the substrate according to the channel geometry. The μFN is usually made of molded PDMS, though laser-cut Mylar gaskets have also been used [161]. Biopatterning with μFNs was first demonstrated in 1997 by Delamarche et al. [342] for the deposition of biomolecules on solid substrates. In this work, immunoglobulins were patterned on gold, glass, and polystyrene with submicron resolution using PDMS μFNs. The channels were rendered hydrophilic with oxygen plasma and filled by capillarity to deposit the biomolecules.

**Figure 19.** Illustration of the process of bioreceptor patterning on a sensor surface using a microfluidic network (μFN). (**a**) The PDMS μFN is reversibly bonded to the sensor surface, then (**b**) the bioreceptor solution is flowed through the microchannels via capillary or pressure-driven flow. This may be followed by rinsing and blocking steps. (**c**) Lastly, the μFN is released to reveal the bioreceptorpatterned surface.

μFNs using capillary flow, like those used by Delamarche et al. [342], can achieve micron-scale pattern resolution, as the microfluidic channels can be prepared with micronscale cross section dimensions [316]. Microfluidic patterning in μFNs can also be performed using pressure-driven flow, but this requires larger channels with cross section dimensions on the order of 10 μm due to high hydraulic resistance [316,317]. Indeed, this yields poorer pattern resolution than capillary flow. However, pressure-driven flow permits the easy exchange of patterning fluids. For example, sequential surface modification steps, including crosslinker attachment, bioreceptor immobilization, rinsing, and post-processing with blocking molecules to prevent non-specific binding, can all be performed in the μFN without surface drying or removing the flow cell [32,329]. Another valuable feature of this technique is that sensing elements designed to operate in liquid media can be probed throughout the patterning process for real-time biofunctionalization monitoring [313]. Beyond biopatterning, μFNs are often used to facilitate miniaturized, simultaneous, and highly localized multi-step binding assays on functionalized sensors [321,342].

Multiplexing is typically achieved using μFNs with multiple parallel channels. Different bioreceptor solutions can be simultaneously drawn through the individually addressable channels, creating a one-dimensional array. However, this method is not well-suited to creating discrete two-dimensional patterns of bioreceptors, which would require complicated multilayer fluidics with three-dimensional flow paths [32,316]. Further, any change to the SiP sensor layout would require a redesign of the μFN [83]. Therefore, microfluidic patterning in channels has less multiplexing flexibility than pin printing, inkjet printing, and microfluidic probe-based patterning, which can localize chemical processes to arbitrary locations on a substrate [32]. Another limitation of μFNs is that reagent consumption can be high, depending on the microchannel volume, and bioreceptor molecules can be lost to microchannel walls due to nonspecific adsorption [83,316]. This is particularly undesirable when using costly bioreceptors. Similarly to μCP, materials like PDMS, which are used to fabricate the μFN, can leach uncured oligomers, which may contaminate the sensor surface and complicate functionalization and subsequent assay steps [330–332]. However, a major advantage of this technique compared to pin and inkjet printing is that bioreceptors are maintained in a controlled liquid environment throughout the patterning process. This ensures good uniformity of the biofunctionalized regions and prevents activity loss of environmentally sensitive bioreceptors due to drying [329,342]. Other advantages of this technique are that it is low cost, exceptionally simple, and unlikely to damage SiP sensing elements.

In the literature, μFNs are a popular choice for patterning SiP devices with bioreceptors (Table 39). μFNs have been used to pattern SiP sensors with antibodies [18,22,161,162,169,174,295], aptamers [174], ssDNA [174], lipid nanodiscs [150], and BSA [174,313]. This technique has been used to confine bioreceptors to select sensing structures on a single SiP device, while leaving other structures bare to control for nonspecific binding, temperature, and instrument drift [18,161]. It has also been used to compare different bioreceptor immobilization strategies. For example, Byeon et al. [295] used a 2-channel microfluidic gasket to

compare bioreceptor immobilization in the presence and absence of a chemical catalyst, while González-Guerrero et al. [313] used two microfluidic channels to compare covalent and adsorption-based bioreceptor immobilization on a single sensor. Finally, μFNs have been used to create multiplexed MRR sensors with different microrings functionalized with different bioreceptors [22,150,162,169,174].

**Table 39.** Demonstrations of bioreceptor patterning using microfluidic patterning in channels for the functionalization of SiP sensors.


#### *4.4. Inkjet Printing*

In contrast to the contact-based deposition systems discussed in Sections 4.1–4.3, which can expose the silicon waveguides and other structures to damage, non-contact inkjet systems use piezoelectric actuation for deposition without touching the sensor surface. Non-contact inkjet based printer systems were developed in the late 1990's where off-the-shelf desktop inkjet printers were repurposed to dispense controllable volumes of reagents in the ~80 pL range [343]. Initial development using home-built ink-jetting

exposed the inkjet solution to heat resulting in a loss of functionality by denaturation or decomposition of biomolecules.

Piezoelectrically actuated non-contact inkjet devices have come to the forefront for localized reagent deposition by leveraging the control provided by a piezoelectrically actuated glass capillary that is capable of depositing droplets that are on the order of one pL to a few hundred pL in size [246,318], as illustrated in Figure 20a. These systems have x-y spatial accuracies ~15–20 μm while dispensing highly accurate volumes of assay reagents without any heat source affecting the samples [344,345]. The capillary tips hover above the etched resonator area while a voltage source piezoelectrically compresses a collar surrounding the nozzle to create pressure waves within the fluid that result in expulsion of <1 nL droplets onto the sensor surface.

**Figure 20.** (**a**) Illustration of piezoelectric inkjet printing of bioreceptors on a sensor surface. (**b**) Image of antibody/antigen and BSA solutions spotted on silicon nitride photonic ring resonators using a Scienion SX microarrayer. The top (control) ring is spotted with BSA solution, and the bottom (test) ring is spotted with anti-(SARS-CoV-2 spike protein) polyclonal antibody solution (a-S1 + S2). Part (**b**) is reproduced from Ref. [167] in accordance with the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

While the high spatial and volumetric controllability of piezoelectrically actuated inkjet systems is desirable, the disadvantages must be mitigated which vary for each assay solution. Nonspecific adsorption of proteins onto the borosilicate glass capillary can cause protein loss when depositing <1 nL of low concentration (<20 μg/mL) protein solutions. Delehanty and Lingler found that both the ionic strength of the printing buffer and presence of a carrier protein greatly affected the amount of biotinylated Cy5-labeled IgG that adsorbed to the capillary surface, thus influencing the amount of IgG that was dispensed from the capillary [346]. The authors' results showed an inverse relationship between the ionic strength of the buffer (PBS) and amount of IgG protein dispensed from the capillary, which was attributed to the nonspecific adsorption of proteins to borosilicate glass. Moreover, they found that with the addition of a carrier protein (BSA), the ionic strength effect could be completely mitigated while increasing the total concentration of IgG that was dispensed up to 44-fold.

On-board cameras and positioning software allow for spot printing to be carried out in a systematic fashion by fiducial mark recognition whereby ~300 pL drops of SS-A antigen at 200 μg/ml can be spot printed using a sciFLEXARRAYER S5 (Scienion, AG, Berlin, Germany) on 128 rings with PDC-70 nozzle [293]. Kirk et al. [246] illustrated the high throughput capabilities and low assay reagent consumption by printing 10 array chips with 6 microrings per chip in 9 s, consuming a total < 25 nL of reagent. In addition to functionalizing SiP chips with multiple bioreceptors for multiplexed analyte detection, inkjet printing can be leveraged to include reference sensors. As previously discussed, the functionality of a microring resonator is based on its resonance wavelength shift and is often measured with respect to a nearby reference resonator. Positioning the reference microring resonator nearby the sensing resonator helps to eliminate shift stemming from thermal

gradients across the chip. Other sources contribute to anomalous background wavelength shift, such as non-specific binding, which may be useful to control for using reference rings. Therefore, while some reference rings may remain buried under an oxide cladding, it may be beneficial in some applications to also include protein coated reference resonators. Cognetti and Miller [167] fabricated a ring resonator set as shown in Figure 20b. One ring was functionalized with anti-SARS-CoV-2 RBD + SARS-CoV-2 RBD (a-S1 + S2) and another was functionalized with 0.1% BSA as a control for non-specific binding, illustrated by the blue and red dots, respectively. The piezoelectric inkjet process allowed for controlled deposition of the assay reagents as isolated elements and showed the relative (BSA ringsubtracted) wavelength shift in response to the SARS-CoV-2 spike protein [167].

Other types of detection mechanisms have been demonstrated through inkjet printing of assay reagents. For instance, Laplatine et al. [319] used a Scienion sciFELXARRAYER S12 to deposit an array of 64 different peptides in buffer on MZIs (spot size of ~150 μm). The MZI array was used to measure volatile organic chemicals (VOCs) with limits in the ppm range as the basis for a silicon olfactory sensor. Ness et al. [318] used a FUJIFILM Dimatix DMP-2831 materials piezoelectric inkjet printer (FUJIFILM Dimatix, Inc., Santa Clara, CA, USA) with 1 pL dispensing DMC-11601 cartridges to deposit ~30 μm diameter spots by optimizing the functional fluid to have a higher viscosity and lower surface tension which was achieved by the addition of glycerol and a surfactant, respectively. A Dimatix materials printer was also used to deposit a functional biotin-modified polymer and porous hydrogel on MZIs, whereby the functional polymer was able to sense the specific binding of protein streptavidin and the benzophenone dextran (benzo-dextran) porous hydrogel was shown to hinder the non-specific binding of BSA on the sensor surface [347]. Table 40 summarizes several demonstrations of inkjet-based bioreceptor deposition on SiP sensors.


**Table 40.** Demonstrations of bioreceptor patterning using inkjet printing for the functionalization of SiP sensors.

#### *4.5. Microfluidic Probes*

Microfluidic probes (μFPs) (Figure 21), which were first demonstrated in 2005 by Juncker et al. [348], combine the features of microfluidics and scanning probes to deliver biomolecules to surfaces. μFPs are classified as "open space microfluidics", as they confine nanoliter volumes of processing liquids on substrates without solid-walled microchannels [83,317]. This is achieved through hydrodynamic flow confinement (HFC) of the processing liquid, which is made possible by the microscale dimensions of the system and the resulting laminar flow regime [32,348].

**Figure 21.** (**a**) Illustration of simple microfluidic printing (μFP) probe used to pattern a sensor surface with a bioreceptor solution (antibody solution illustrated here as an example). (**b**) Comparison of μFP using simple hydrodynamic flow confinement (HFC), hierarchical HFC, which permits recirculation of the patterning solution in the μFP head, and radial HFC, which produces circular, rather than teardrop-shaped spots. The processing (bioreceptor) solution is shown in red, the immersion liquid is shown in light blue, and the shaping liquid used for HFC is shown in dark blue. Insets on the right show the printing footprints for each μFP type. Part (**b**) is adapted with permission from Ref. [32]. Copyright 2021 American Chemical Society.

The tip of the μFP may be fabricated from silicon [348], silicon and glass [317], or PDMS [349]. It consists of coplanar injection and aspiration microapertures and is placed 10–200 μm above the substrate [32]. An immersion liquid fills the gap between the probe tip and substrate, while processing liquid is injected from the injection aperture and collected by the aspiration aperture. The processing liquid is confined above and below by the probe tip and substrate, while it is confined laterally by hydrodynamic boundaries formed by the immersion liquid [32]. In a simple μFP configuration, the flow rate of the injected fluid, QI, must be lower than the flow rate of aspirated fluid, QA, to maintain flow confinement [348]. The ratio QA/QI can be varied, along with the distance between the probe tip and surface, to tune the shape and size of the region where the processing fluid contacts the surface [348]. Typically, this impingement area has a teardrop shape, but an alternative radial probe tip design can be used to create a circular impingement area [32]. The μFP is mobile and can scan over a substrate to create complex patterns; depending on the direction and speed of travel relative to the microfluidic flow, continuous shapes or discrete spots can be patterned [349]. Spot sizes as small as 10 × <sup>10</sup> <sup>μ</sup>m<sup>2</sup> are possible [83].

μFP-based bioreceptor patterning has not yet been demonstrated on SiP biosensors, but it may be a promising technique for future application. Firstly, μFPs are suitable for multiplexed patterning, as processing fluids can be rapidly switched using an external valve system [32]. Further, the probe can follow an arbitrary scan path, allowing for flexible and customized patterning of sensors with non-standard layouts [348]. Given that this is a non-contact technique, it is unlikely to damage fragile SiP surfaces. Unlike inkjet and pin printing, μFPs pattern surfaces in a liquid environment, which prevents uncontrolled wetting and drying effects, thus improving spot uniformity and homogeneity, while preventing aggregation or denaturation of printed biomolecules [317,348]. For example, Autebert et al. [83] demonstrated less than 6% variation in spot homogeneity for an array of 170 spots of IgG printed on polystyrene. While a simple μFP configuration

typically requires large volumes of processing fluids, a 10-fold decrease in μFP reagent consumption has been achieved using hierarchical flow confinement and recirculation, making it comparable to pin and inkjet printing (e.g., 1.6 μL to print 170 spots of IgG, each with a 50 × <sup>100</sup> <sup>μ</sup>m2 footprint).

The main challenges of this patterning technique are its low throughput and limited commercial availability [32]. Using a simple μFP configuration, only one spot can be addressed at a time, each requiring a residence time defined by the kinetics of the bioreceptor's immobilization reaction. Multiple spots could be patterned simultaneously using probe tips with microfluidic channel bifurcations to increase throughput, but this would only be suitable for SiP sensors with highly standardized layouts, as aperture spacing would need to match the spacing of SiP sensing structures [83]. The accessibility of this technique is limited, as commercial μFP-based patterning systems are not yet available. Another potential challenge is that, when applied to SiP sensor surfaces, this technique may suffer from perturbations in hydrodynamic flow confinement due to the three-dimensional topography introduced by the patterned silicon structures [317]. This, in turn, may result in reduced spot homogeneity.

#### *4.6. Summary and Future Directions*

Here, we have discussed several strategies for preparing patterns of bioreceptors on SiP sensor surfaces for multiplexed detection. In general, non-contact patterning techniques are attractive for SiP sensor biofunctionalization, as they prevent damage to the sensor surface and integrated optical and electronic components. Of the strategies discussed here, inkjet printing is a promising strategy for biopatterning multiplexed SiP sensors. Inkjet printing is a flexible, high throughput, low-waste, and multiplexable non-contact patterning strategy that can achieve sufficient resolution for the functionalization of most SiP devices [329]. However, printing protocols (e.g., actuation waveform design, environmental controls, additives to bioreceptor "ink", etc.) must be optimized for replicable deposition of uniform spots. Future studies using inkjet-based biopatterning of SiP sensors should quantify and optimize inter- and intra-spot uniformity, along with inter-spot and run-to-run replicability to validate reliable performance of this patterning technique. μFP is another flexible noncontact patterning technique, which can achieve improved spot uniformity and replicability compared to other printing methods, and may be a promising option for SiP sensors [317,348]. Nevertheless, this technique must still be validated for bioreceptor patterning on SiP surfaces.

#### **5. Critical Comparative Analysis of Solutions and Discussion of the Interplay between the Three Aspects of Biofunctionalization**

This review has provided a detailed overview of strategies that have been or can be used to functionalize SiP biosensors in terms of bioreceptor selection, immobilization chemistry, and patterning strategy. We have benchmarked potential strategies for each of these three aspects of biofunctionalization against a set of performance criteria relevant to SiP sensing. In addition to assessing the tradeoffs of individual solutions in the context of the anticipated biosensor use case, the compatibilities and incompatibilities between solutions to each of the three aspects of biofunctionalization are an essential consideration. Moreover, the interplay between bioreceptors, immobilization chemistries, and patterning techniques can affect what is considered suitable performance for a given biofunctionalization need. For example, when using a patterning technique with very low reagent consumption, bioreceptors with a greater cost per milligram may still permit very low reagent cost per sensor. This underscores the importance of considering these three aspects of biofunctionalization in concert.

The first step in designing a biofunctionalization protocol once the application of the biosensor is defined and the target(s) known, is bioreceptor selection. As discussed in Section 2, different bioreceptors are suitable for different targets. For many targets (proteins, small molecules, viruses, bacteria, etc.) antibodies, aptamers, MIPs and PCCs may be suitable. Despite being very cost-effective and stable, currently available MIPs cannot achieve sufficient binding affinity and/or selectivity to achieve detection at clinically relevant levels for many targets. Of the other three options, antibodies are the most readily available and well-characterized, but their poor stability and high cost limit their suitability for POC use. Moreover, in our group's experience, batch-to-batch variability has been a notable roadblock in the design of replicable biosensing assays using antibodies.

Synthetic antibody analogs like aptamers and PCCs, which can achieve similar affinity and specificity to antibodies, are appealing and versatile options for POC sensors. Currently, a significant roadblock in the widespread adoption of aptamers and PCCs for biosensing applications is the relatively limited availability of pre-designed products for ready use against a diverse range of targets, though this challenge can be mitigated in coming years with further research and development [30,89,234]. Additionally, aptamers often require careful sample preparation (buffering, filtering, or tight temperature control) to avoid their folding or denaturing prematurely during use. Robust aptamer formulations need to be screened with these factors in mind, given each sensing device's use case. Regardless, their low cost, good stability, and highly reproducible and scalable production are important advantages of aptamers and PCCs for POC biosensing.

For nucleic acid targets, nucleic acid probes (hybridization-based detection), HCCD, and CRISPR-dCas9-mediated detection may be suitable bioreceptor options. In applications requiring highly multiplexed nucleic acid sensing, simple hybridization-based detection with nucleic acid probes offers the greatest flexibility and assay simplicity. When exceptionally high sensitivity and selectivity are required for very low-concentration targets, HCCD or CRISPR-dCas9-mediated detection may be preferable. It should be recognized; however, these are very early- stage approaches with limited precedent for use on SiP platforms and have yet to be validated for sensing in complex samples.

Lastly, glycans, lectins, and lipid nanodiscs are valuable for the study of carbohydrateprotein and cell membrane interactions, respectively, but their often-poor affinity and selectivity limit their applications beyond such studies.

In addition to these considerations, the immobilization chemistries that are compatible with each type of bioreceptor should be kept in mind during bioreceptor selection, with particular attention paid to compatibility of the immobilization chemistries with other steps of biosensor fabrication and integration with sample fluid delivery. Broadly, passive adsorption of bioreceptors leads to poor stability of the functionalized surface and diminishes the bioreceptor's binding activity. One exception is lipid nanodiscs, which adsorb well to silicon dioxide surfaces to yield reproducible, regenerable, and stable functional layers [148,149,275,276]. For other bioreceptors, passive adsorption is not recommended, aside from in preliminary sensor validation experiments where simplicity and rapid assay design are priorities. Nevertheless, novel polymeric coating materials (e.g., PAcrAm™ and AziGrip4™ from SuSoS AG) may permit stabler and more oriented bioreceptor immobilization with similar simplicity to passive adsorption techniques, potentially comprising a valuable future research direction [309–311].

Among the various covalent and bioaffinity-based immobilization strategies explored in this review, different immobilization methods can produce very different results depending on the bioreceptor. For example, many covalent methods can readily achieve predictable and oriented aptamer and nucleic acid probe immobilization by targeting terminal functional groups incorporated into these bioreceptors during synthesis; this ensures good binding site availability for target capture. Conversely, when used for antibody immobilization, these covalent strategies typically target native functional groups that are abundant on the antibody surface, leading to random antibody orientation and reduced target-binding capacity.

When antibody binding capacity must be optimized, bioaffinity-based strategies using antibody-binding proteins, like Protein A, may be a preferable choice, though these strategies involve tradeoffs in terms of stability, regenerability, and cost. It should also be noted that Protein A-based antibody immobilization may compromise the specificity of immunoassays using amplification with a secondary antibody [350]. In our experience, using Protein A in sandwich immunoassays was correlated with a considerable nonspecific signal during the secondary antibody amplification step, which was not observed in immunoassays prepared using simple passive adsorption of the detection antibody on the SiP sensor surface [351]. This non-specific signal may be related to the unwanted capture of secondary antibodies by unoccupied Fc-binding sites on the Protein A-coated sensor surface (potentially due to incomplete functionalization with capture antibody or due to unbinding of capture antibody during the course of the experiment) [350]. One solution may be to choose secondary antibodies that do not bind well to Protein A, though this may be challenging, as Protein A and Protein G bind well with antibodies from many common host species (cow, goat, mouse, rabbit, and sheep) that are used in immunoassays [350]. Protein L, which does not bind with cow, goat and sheep antibodies and binds weakly to rabbit antibodies, may offer greater flexibility in the choice of secondary antibody, potentially making it a preferable antibody-binding protein for sandwich assays [350]. Silane-based covalent strategies have also been successfully used to immobilize capture antibodies on SiP sensors for assays using amplification with a secondary antibody [17,18,162,163]. This highlights that the assay format (label-free/labeled) and its synergy with the biofunctionalization strategy should be carefully considered.

Next, the selection of a patterning technique should take into consideration factors such as the bioreceptor cost, fluid properties of the bioreceptor solution, and changes in sensor surface hydrophilicity caused by the immobilization chemistry. For instance, patterning in microfluidic channels is a simple and popular choice in SiP biosensor functionalization protocols, but it typically has high reagent consumption. As an example, in our group's previous work, we deposited 20 μg/mL solutions of capture antibodies on SiP sensors via microfluidic channels using pressure-driven flow at 30 μL/min for 45 min [351]. This consumed a total of 27 μg of antibody, which costs roughly CAD \$135, assuming an antibody cost of ~CAD 500/100 μg. Further, bioreceptors may be lost to adsorption on the channel walls during patterning, and this inefficient reagent use is particularly undesirable for costly bioreceptors, such as antibodies. In assays using in-flow patterning followed by sample introduction using the same μFN, targets in the sample may bind to bioreceptors coating the channel walls and non-sensing regions of the SiP chip. This can deplete target molecules from the sample more rapidly than if the sensing regions, alone, were functionalized. Consequently, this may worsen the limit of detection [352]. Offline patterning of bioreceptors to ensure that they are only localized to the sensing regions is, therefore, particularly beneficial for detecting precious targets at very low concentrations.

The fluid properties of the bioreceptor solution can also dictate the success of a patterning technique. In particular, μCP, pin printing, and inkjet printing strategies are sensitive to the viscosity and surface tension of the bioreceptor solution. Required additions to bioreceptor solutions, such as glycerol to slow evaporation, must be accounted for when optimizing the patterning protocol. Surface modifications used for different bioreceptor immobilization chemistries affect the hydrophilicity of the sensor surface, which, in turn, influences the efficacy and resolution of the patterning strategy [338]. For example, silanization decreases the hydrophilicity of the SiP sensor surface [267]. In the context of μCP, this may inhibit the transfer of bioreceptor "ink" from the stamp to the sensor surface. On the other hand, this decreased surface hydrophilicity will decrease the spreading of droplets of aqueous bioreceptor solutions. This may improve the resolution of patterning techniques such as pin and inkjet printing.

While not a focus of this review, antifouling strategies must typically be integrated with biofunctionalization protocols in order to prevent non-specific adsorption of sample matrix components to the sensor surface [29]. Antifouling strategies can be included in covalent bioreceptor immobilization protocols through the use of linkers that include polyethylene glycol (PEG) chains (e.g., SM(PEG)12 [133], BS(PEG)9 [132]), which increase the hydrophilicity of the surface coating to reduce non-specific protein adsorption [29]. Other approaches include coating the surface via passive adsorption with bovine serum albumin [264,353] or commercial blockers, such as StartingBlock [109,163,196], BlockAid, and StabilCoat [167], after bioreceptor immobilization. It is important to consider how to best fit antifouling strategies into biofunctionalization workflows. For further details about antifouling strategies for SiP biosensors, readers are directed to ref. [29].

Lastly, the entire biofunctionalization procedure must be considered in the context of the overall sensor system design. While some functionalization strategies may be suitable for the SiP sensor chip itself, they may not be suitable for systems including chip-mounted electronic/photonic inputs and outputs, which can be used to translate this technology to a commercial POC platform (Figure 22) [12,13,66,354]. For example, immobilization chemistries requiring solution-phase reactions may be unsuitable for sensor designs including photonic wire bonds that connect optical inputs and outputs to the on-chip waveguides. Solvents or other chemicals used in functionalization may damage or swell the photonic wire bond or low-index photonic wire bond cladding materials, resulting in damage to the fine optical connection [355,356]. In this case, immobilization chemistries employing vapor-phase surface modifications or direct crosslinking of bioreceptors (e.g., UV crosslinking of nucleic acids or aptamers) to the unmodified surface may be preferable. Similarly, plasma or UV/ozone treatment are likely more suitable surface pre-treatment techniques than immersion in piranha solution for integrated SiP systems. For these systems, the surface topography and locations of chip-mounted components should inform the selection and design of the patterning strategy. In general, non-contact patterning techniques (e.g., inkjet printing) can permit flexible bioreceptor pattern design, while preventing damage to the system, making them preferable to techniques that require contact between the patterning tool and surface.

**Figure 22.** Integration approaches for SiP biosensors. (**a**) Multiplexed biofunctionalization of integrated SiP sensor system for POC use, which includes on-chip photonic inputs (chip-mounted fixed wavelength laser), outputs (on-chip detectors), photonic wire bonds, and microfluidics. See Ref. [13] for further information about this integration approach. (**b**) System-level integration of active SiP sensor by Laplatine et al. [66] using fan-out wafer-level packaging, showing (i–vi) schematics of the packaging process, (vii) 3D illustration of the packaged chip, and (viii) photograph of the experimental biochip setup. Part (**b**) is adapted with permission from Ref. [66]. Copyright 2018 Elsevier. (**c**) Photonic integrated circuit sensor chip presented by Mai et al. [354] using local backside release to enable integration with fluidics on one side of the chip and (i) optical coupling or (ii) optical coupling and electrical interconnects on the other. This allows for a more compact form factor than chips using front side integration only. Part (**c**) is adapted with permission from Ref. [354]. Copyright 2022 Elsevier.

In summary, it is important to consider the interplay between the three constituents of biofunctionalization as well as the silicon photonic device, fluidics, and detection assay when designing a biofunctionalization strategy. The examples above highlight the importance of considering and addressing the relationships between different bioreceptors, immobilization strategies, and patterning techniques and their suitability for different assay formats and integrated sensing architectures. This discussion aims to bring attention to the importance of considering and addressing these relationships in order to design successful biofunctionalization protocols for SiP biosensors.

#### **6. Conclusions**

When combined with carefully designed biofunctionalization strategies, SiP sensors have the potential to permit accurate and information-rich decentralized diagnostic testing for a diverse range of clinical applications. We have identified and evaluated different strategies for SiP sensor biofunctionalization in terms of bioreceptor selection, immobilization strategy, and patterning technique. Different solutions for each aspect of biofunctionalization have been benchmarked against a set of critical performance criteria relevant to multiplexed SiP biosensing and examples from the literature have been discussed and categorized. In addition to providing critical discussion about solutions for each aspect of biofunctionalization, we have also identified the interplay between these three aspects to help inform the design of SiP functionalization protocols and have highlighted additional functionalization process constraints relevant to SiP system integration for POC biosensing.

Broadly, several classes of synthetic bioreceptors (e.g., aptamers, PCCs, nucleic acids) offer excellent potential for multiplexed POC biosensing, as they can achieve high affinity and specificity, and offer scalable and cost-effective production, good stability, and regenerability. However, the availability of ready-to-use reagents remains a roadblock for the use of synthetic antibody analogs. In terms of immobilization strategies, covalent methods offer stable, scalable, and highly tailorable bioreceptor immobilization, but their success often depends highly on the reaction conditions and bioreceptor type, underscoring the potential value of developing standardized and reliable reaction protocols that are optimized for SiP surfaces. Regarding patterning, pin and inkjet-based printing are popular techniques that offer good flexibility and resolution, while inkjet printing has the additional advantages of exceptionally high throughput and being a non-contact method that will not damage the SiP surface or integrated electronic/photonic structures. μFP-based patterning is another attractive potential solution for flexible bioreceptor patterning that may achieve improved spot uniformity, though this technique has yet to be tested on SiP platforms. Overall, this review serves as a detailed overview of the biofunctionalization options available and previously tested on SiP platforms. This can help guide the design of new functionalization protocols, which must also be individually tailored for the specific target analyte(s), assay format, system architecture, and intended operating environment.

**Author Contributions:** Conceptualization, L.S.P., S.M.G., J.M.M., J.R.B., L.C., S.S. and K.C.C.; Investigation, L.S.P., S.M.G. and J.M.M.; Resources, L.C., S.S. and K.C.C.; Data Curation, L.S.P., S.M.G. and J.M.M.; Writing—Original Draft Preparation, L.S.P., S.M.G. and J.M.M.; Writing—Review and Editing, L.S.P., S.M.G., J.M.M., J.R.B., L.C., S.S. and K.C.C.; Visualization, L.S.P., S.M.G. and J.M.M.; Supervision, L.C., S.S. and K.C.C.; Project Administration, L.S.P., S.M.G., L.C., S.S. and K.C.C.; Funding Acquisition, L.C., S.S. and K.C.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was supported by Canada's Digital Technology Supercluster, a Digital Technology Supercluster Expansion project "Preventing COVID19 spread through workplace health and safety—Canada's Digital Technology Supercluster (CDTS)", the generosity of Eric and Wendy Schmidt by recommendation of the Schmidt Futures program, and the MITACS, Inc./Elevate Postdoctoral Fellowship Program, IT25501, In collaboration with Dream Photonics Inc. and Innovation, Science and Economic Development Canada.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors are grateful to work and live on land that is the traditional, ancestral, and unceded territory of the Coast Salish Peoples, including the territories of the Musqueam, Squamish, and Tsleil-Waututh First Nations. The authors would also like to acknowledge support from the Natural Sciences and Engineering Research Council of Canada (NSERC) (graduate scholarship supported by the NSERC Canada Graduate Scholarships—Master's Program) and the Silicon Electronics-Photonics Integrated Circuits Fabrication (SiEPICfab) consortium. The authors sincerely appreciate helpful discussions and feedback from Matthew B. Coppock, Avineet Randhawa, Matthew Mitchell, and Heather M. Robison.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


Iversen, P.W., et al., Eds.; Eli Lilly & Company and the National Center for Advancing Translational Sciences: Bethesda, MD, USA, 2004.


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

### *Review* **Optical Methods for Label-Free Detection of Bacteria**

**Pengcheng Wang 1, Hao Sun 2, Wei Yang <sup>2</sup> and Yimin Fang 2,\***


**Abstract:** Pathogenic bacteria are the leading causes of food-borne and water-borne infections, and one of the most serious public threats. Traditional bacterial detection techniques, including plate culture, polymerase chain reaction, and enzyme-linked immunosorbent assay are time-consuming, while hindering precise therapy initiation. Thus, rapid detection of bacteria is of vital clinical importance in reducing the misuse of antibiotics. Among the most recently developed methods, the label-free optical approach is one of the most promising methods that is able to address this challenge due to its rapidity, simplicity, and relatively low-cost. This paper reviews optical methods such as surface-enhanced Raman scattering spectroscopy, surface plasmon resonance, and darkfield microscopic imaging techniques for the rapid detection of pathogenic bacteria in a label-free manner. The advantages and disadvantages of these label-free technologies for bacterial detection are summarized in order to promote their application for rapid bacterial detection in source-limited environments and for drug resistance assessments.

**Keywords:** bacteria detection; dark-field microscopy; Raman spectroscopy; surface plasmon resonance; label-free; rapid detection

#### **1. Introduction**

Bacteria are the most abundant, widely distributed, diverse microorganisms in nature and of a special type. After a long period of natural evolution, bacteria have established complex antagonistic or symbiotic relationships with various species [1]. Although most of the bacteria are harmless, bacterial and viral infections account for approximately 70% of all human pathogenic diseases [2]. Bacterial pathogens can be obtained from food, water, animals, and even clinical settings including hospitals and other healthcare facilities. Pathogenic bacteria such as Salmonella, *Escherichia coli* (*E. coli*), Staphylococcus, etc. are the main causes of foodborne illness, which poses a constant threat to food safety. Bacterial infection is considered to be a common and costly global public health problem [3,4]. Bacteria not only cause some specific diseases in the host, but also act as opportunistic pathogens. When the host's immunity is low, the immune barrier is destroyed, flora imbalance or bacterial translocation occurs, which releases many virulent factors causing the host infection [5,6]. Treatment with antibiotics is the most effective and frequently used solution to this problem. Nevertheless, with the increasing use of antibiotics, the emergence of bacterial resistance to antibiotics is rising, which reduces the effectiveness of antibiotics for bacterial infection treatment, leading to increasing morbidity, mortality, and medical costs. According to the World Health Organization, antibiotic resistance kills 700,000 people every year, and if this problem is not addressed, the number of deaths resulting from antibiotic resistance will increase to 10 million by 2050 [7]. At present, bacterial resistance has become an increasingly serious global challenge, as well as a worldwide concern to governments and society [8]. According to the U.S. Centers for Disease Control and Prevention, about 2.8 million infections in the U.S. each year are

**Citation:** Wang, P.; Sun, H.; Yang, W.; Fang, Y. Optical Methods for Label-Free Detection of Bacteria. *Biosensors* **2022**, *12*, 1171. https:// doi.org/10.3390/bios12121171

Received: 13 November 2022 Accepted: 13 December 2022 Published: 15 December 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

related to antimicrobial resistance, implying significantly increasing treatment times and costs as well as mortality from bacterial infections [9].

The effectiveness of antibiotic treatment can be largely retained with the rational use of antibiotics. Rapid identification of pathogens is particularly important in clinical diagnosis, not only to minimize risks to patients, but also to provide a basis for physicians to prescribe pathogen-specific antibiotics rather than broad-spectrum antibiotics to reduce irrational use of antibiotics. However, rapid bacterial detection is quite a challenging task due to the large variety of bacteria and severe interference from the complex matrix in the growth environment [10]. Traditional methods, such as bacterial culture, PCR, and enzyme-linked immunosorbent assay (ELISA) are frequently used, but these methods have their own disadvantages. The bacterial culture method is the golden standard method for bacterial detection, but it is quite time-consuming, and easily contaminated by non-target bacteria. Detection of some clinically relevant pathogens by this method can take up to five days to develop an adequate culture [11]. PCR is a molecular biology technique used to amplify specific nucleic acid fragments. It replicates nucleic acid exponentially at a very low concentration [12] to a detectable amount within hours. Therefore, it has been widely used in bacteria detection. However, contamination of the test sample and erroneous DNA amplification can lead to false positive or negative results. PCR is relatively expensive and takes hours which is not rapid enough for regular use in antibiotic prescription. Immunoassays rely on the specific reaction of antigens and antibodies and are also used for the detection of bacteria [13,14] but are less sensitive and require a large amount of clinical samples.

To overcome these difficulties, more sensitive and rapid methods for bacterial detection have been extensively studied. In recent years, applications based on biosensors, which are analytical devices that convert biological responses into measurable signals, have become increasingly widespread [15]. Such an application usually consists of three parts: (1) ligands attached to the surface of the biosensor to recognize the target through specific interactions; (2) a sensor that converts biometric identification generated on the sensor surface into quantifiable physical signals such as light, electricity, heat, and voltage, etc.; (3) a signal detector. Biosensors have become an important tool for the rapid, sensitive, and selective detection of microorganisms. These methods include biosensor-based electrochemical methods [16–25], fluorescence detection methods [20–26], and spectroscopy methods [27–39]. However, most of the biosensing methods require labeling of target objects for signal reading, which significantly increases the measurement time and cost. Moreover, the presence of dyes and labels tends to interfere with the normal physiological function of bacteria, which does not reflect the true state of the bacteria, especially in the evaluation of antibiotic resistance. Therefore, label-free methods are advantageous in rapid pathogen detection and drug resistance evaluation.

Compared with the labeling methods, which generally require a long incubation time, label-free approaches are much simpler, faster, and cost-effective, making them good candidates for rapid bacterial detection in clinical application. Efforts have been made in this direction, among which the optical methods, such as Raman spectroscopy and single-particle imaging approaches, are the most promising approaches due to their high sensitivity, simplicity, and low-cost for label-free detection of bacteria [40–42]. In this review, we describe the advantages and disadvantages of optical methods such as Raman spectroscopy, SPR, and dark-field microscopy for label-free detection of bacteria and their applications in clinical detection and drug resistance evaluation.

#### **2. Surface Plasmon Resonance for Bacteria Detection**

#### *2.1. Principle of Surface Plasmon Resonance*

A typical optical system of planar SPR is mainly composed of a polarized excitation light source, a prism and a glass sensor chip coated with a thin gold film (~50 nm). The incident light passes through the prism in total internal reflection mode. The reflected light significantly decreases at a specific angle (defined as the resonance angle), while the wave vector matches the surface plasma frequency of the gold film in the propagation direction as shown in Figure 1a. The shift of the SPR angle is very sensitive to the refractive index change at the metal–liquid interface, making it a powerful tool for real-time monitoring of molecular and particle binding at the interface in a label-free manner [43]. It has been used to analyze binding specificity between molecules [44–46], the concentration of target molecules [47,48], kinetic parameters of association and dissociation [49,50], etc. More recently, with the development of SPR microscopy as shown in Figure 1b, which can directly monitor the nanoscale motion of single bacteria at the interface, SPR microscopy has become a powerful tool for rapid drug resistance evaluation [51]. In contrast to conventional SPR biosensors such as BIAcore, which provide an average signal of the designed area on the surface of the sensor chip, SPR microscopy enables the detection of areas or particles of interest on the chip surface, facilitating the detection of bacteria at the single cell level This process can be accomplished by recording an SPR image of the chip surface with a charge-coupled device (CCD) or a complementary metal oxide semiconductor camera. In addition, high spatial resolution of the perceived surface can be obtained by introducing a lens or a high numerical aperture (NA) objective into the SPR image system to replace the prism [52,53]. In addition to SPR microscopy, the use of an SPR image to detect bacteria has also been widely reported. For example, Tripathi et al. [54] proposed coating the gold surface of traditional SPR biosensors with graphene to improve the adhesion of bacteria on the surface of the sensor and applied it to the detection of *Pseudomonas* and *Pseudomason*-like bacteria. Park et al. [55,56] immobilized antibodies onto the sensor chip via EDC mediated coupling and realized the label-free and highly sensitive detection of foodborne *Salmonella* at low PH (4.6) and high antibody concentrations (up to 1000 μg/mL).

**Figure 1.** Schematic diagrams of (**a**) SPR optical system and (**b**) SPR microscopy.

#### *2.2. Method and Application of SPR Technology for Label-Free Detection of Bacteria*

The direct detection of bacteria by SPR requires specific antibodies against the target bacteria, which are immobilized on the surface of the gold film and specifically bind to the target bacteria to generate SPR signals. When the bacteria-containing solution flows to the sensor surface with specific antibody immobilization, the target bacteria bind to the gold film, which is then flushed to remove nonspecific interaction. As the SPR signal is positively correlated with the concentration of target bacteria, the number of target bacteria can be determined by setting up a calibration curve of bacterial concentration versus SPR signal intensity. The immobilization of antibodies on the sensor surface is a critical step for the detection of bacteria, which can improve the sensitivity and selectivity of bacterial SPR detection [57]. Physical adsorption and covalent binding are the main methods to fix the antibody on the sensor surface.

(i) Physical adsorption. Physical adsorption is a simple method of coating a surface that utilizes non-covalent bond interactions such as van der Waals forces, hydrogen bonds, electrostatic forces, and hydrophobic interactions to adsorb the target to the detection chip. Capturing bacteria on the surface creates a refractive index (RI) change, and RI is used to quantify the presence and quantity of the bacteria. Jarvis et al. [58] used SPR technology to track in real time the attachment of *Pseudom onas aeruginosa* bacteria to bare gold film. This study showed that the adsorption of wild-type and mutant bacteria and the concentration of bacteria in bacterial suspension could be distinguished by physical adsorption. The results of this method were compared with those of crystal violet assay for different mutant bacteria, and it was found that there was qualitative correlation between them. Another method of physical adsorption of bacteria is to first modify hydrophobic or hydrophilic compounds or biologically active molecules on the surface of the gold chip, and then incubate the bacteria with the modified surface of the gold sheet, so that it can be adsorbed to the surface of the gold chip in a non-covalent interaction. Livache et al. [59] used pyrrole co-electropolymerization to attach different types of carbohydrates to the surface of gold film. Because different carbohydrate types have different physical adsorption capacities compared to the five closely related *E. coli* strains, different types of *E. coli* were incubated and grown on the substrates modified with different carbohydrate strains. SPR imaging was used to detect their interactions with bacteria during culture. This method can detect and identify tested bacteria from an initial bacterial concentration of 102 CFU/mL.

(II) Covalent immobilization. The measurement of SPR is based on the change of refractive index. However, because the gold film itself is not selective, it is not possible to distinguish the target in the complex mixture directly on the gold chip. SPR sensors specific to an analyte can be obtained by grafting an antibody that is specifically recognized by the analyte onto the surface of the gold chip. A reasonable method of immobilization of antibodies is to chemically conjugate antibodies to the surface of the sensor; immobilization of antibodies based on self-assembled monomolecular membrane (SAM) is the most studied method at present. SAM is an ordered single molecular structure formed by the adsorption of mercapto, amine, silane, or carboxylic acid components onto the solid surface in solution [43]. SAM can help control antibody binding direction, reduce nonspecific adsorption, and provide stable and directed analyte curing [60]. Thiolate compounds with different properties can easily be prepared with monolayers of different surface properties (such as wettability). SAM can be covalently bound to the primary amine of the ligand when it contains a carboxyl group at its end. This coupling is widely used for protein fixation. During the covalent binding of ligands, the non-specific binding of ligands on gold chips hinders the active functional groups in SAMs, which reduces the specificity. Therefore, a blocking agent, such as ethanolamine, is used to block the carboxyl groups remaining on the surface. In addition, bovine serum albumin is commonly used to block the gold surface to reduce the nonspecific interaction. Srikhirin et al. [61] developed an immunosensor based on SPR imaging using specific monoclonal antibody 11E5 (MAb 11E5) for the detection of seed-borne bacterium Acidovorax avenae subsp. citrulli (Aac). Aac was detected by self-assembly of MAb 11E5 mixed with monolayers (SAM). This method can be applied to multiplex detection, and it shows good selectivity for Aac with a limit of detection (LOD) of 10<sup>6</sup> CFU/mL. Evoyet et al. [62] used cysteine labeling and mercaptan chemistry to modify a specific caudate protein (tsp) on the surface of gold film for specific capture of *Salmonella typhi* with a detection limit of 10<sup>3</sup> CFU/mL. Chen et al. applied polyclonal anti-*E. coli* O157:H7 antibody to an NHS/EDC-activated surface by activating a SAM-coated chip with a mixture of NHS and EDC to generate an NHS ester receptor capable of binding to the amino group of the antibody via an amide bond [63]. Roupioz et al. used an antigen–antibody fixation method to modify the antibodies of a series of different bacteria in different regions of the gold sheet, and then cultured the advantages of this microarray on the chip with contaminated food. The culture of the bacteria results in an increase in the concentration of the target bacteria around the specific antibody, and then surface plasma resonance imaging is used to detect the growth of the bacteria. This single-step assay method enabled multiplex testing of *Cronobacterium* and *Salmonella* in less than a day and demonstrated that both bacteria were detected in 25 g of milk powder with as few as 30 CFU cells [64].

Tao et al. modified the gold chip with a layer of PEG/PEG-COOH self-assembled monomolecular layer, and then activated PEG-COOH by NHS and EDC to generate NHS ester receptors that react with the primary amine group on the antibody by amide bonds. The polyclonal anti-*E. coli* O157: H7 IgG antibodies have been applied to NHS/EDCactivated surfaces so that bacteria can be specifically attached to the surface. By using SPR microscopy, the nanoscale-motion of bacteria can be sensitively monitored at the gold chip surface as show in Figure 2. As the nanoscale-motion of bacteria is related to their activities, Tao's group developed a culture-free antimicrobial susceptibility test (AST) by tracking the motion using SPR microscopy, facilitating rapid antimicrobial resistance testing [51].

**Figure 2.** Schematic diagram of the rapid antimicrobial susceptibility test at single bacteria level using SPR microscopy [51]. Adapted with permission from Ref. [51]. Copyright © 2015 American Chemical Society.

In addition to the above commonly used bacterial label-free detection, new methods have also been developed in recent years for SPR methods for bacterial label-free detection. Culture–Capture–Measure (CCM): The protein is covalently bound to the pyrrole monomer on the chip, and then different types of antibodies are modified on the chip in the form of microarrays. Bacteria are cultured on the surface of the chip and then combined with sensitive SPR assays, which enables rapid and specific detection of bacteria on the protein microarrays. This culture–capture–measurement method can significantly reduce the processing steps of bacterial detection and the overall analysis time of bacterial detection [64–66]. For example, Thierry et al. combined microbial incubation on chips with SPR detection to achieve rapid specific detection of *Salmonella enterica* serovar Enteritidis, *Streptococcus pneumoniae* and *E. coli* O157:H7 cultured on protein microarrays [65]. Several methods have also been proposed to further improve the sensitivity: A highly sensitive sensor based on surface material modification was constructed by modifying nanomaterials [67–69] (graphene, molybdenum disulfide, barium titanate) or organic compounds [59] (carbohydrate) on the surface of a gold chip, which can significantly improve the sensitivity of bacterial detection. Livache et al. detected their interactions with bacteria by efficiently grafting simple carbohydrates onto the surface of a gold sheet and then using surface plasma resonance imaging during the process of culturing the bacteria on the surface. It was found that each type of bacteria interacts with carbohydrate chips in different ways. Compared with the detection limit of 1.0 × <sup>10</sup><sup>4</sup> CFU/mL for other electrochemical methods, the detection limit of this method can reach 1.2 × <sup>10</sup><sup>2</sup> CFU/mL [59].

Besides the antibodies, the surface of the gold chip is modified with small molecules such as bacteriophages, polymyxin B, aptamers, etc., as a bacterial identification element [69–71]. For example, Michel Meunier et al. used l-cysteine SAM to coat a gold sheet, and then linked the T4 bacteriophage and BP14 bacteriophage to the self-assembled membrane respectively to specifically detect *E. coli* and Methicillin-resistant *S. aureus*. This method does not require the prior step of labeling or enriching bacteria and can detect concentrations of 103 CFU/mL

in less than 20 min [69]. In addition, the target bacteria were isolated and purified from complex samples by magnetic separation technology before SPR detection. Veli et al. developed a rapid and efficient magnetic separation step followed by the rapid detection of *B. melitensis* contamination in milk samples by SPR. Two aptamers with high affinity and specificity for *B. malitensis* were selected by a complete bacteria-SELEX procedure. The high-affinity aptamer (B70 aptamer) was immobilized on the surface of magnetic silica core-shell nanoparticles for the initial purification of target bacterial cells from the milk matrix. Another aptamer with high specificity for *B. melitensis* cells (B46 aptamer) was used to prepare SPR sensor chips for the sensitive determination of Brucella in magnetic purification eluted samples. This method can rapidly detect *B. melitensis* contamination in 1 mL milk samples by SPR, with LOD values as low as 27 ± 11 cells [72].

#### **3. Raman Spectroscopy for Pathogen Bacteria Detection**

#### *3.1. The Principle of Raman Spectroscopy*

Raman scattering can be defined as the inelastic scattering of photons from molecules. For every 10<sup>6</sup> photons scattered from the molecules, approximately one photon is inelastically scattered (Raman scattering). The detection of inelastic scattering photons from a molecule produces a spectrum of Raman shifts by the acquisition of energy differences from incident light. Each Raman shift corresponds to a specific vibration mode of molecular bonds, thus allowing molecular identification based on a specific vibrational fingerprint. Compared to fluorescence spectroscopy, Raman spectroscopy has higher resolution and narrower bandwidth, making it easy for the multiplex detection of different analytes. An advantage of Raman spectroscopy for bacterial detection is that Raman scattering can occur at any wavelength. This allows free choice of the excitation wavelength to meet the needs of biological Raman spectroscopy acquisition, especially in reducing the significant background from fluorescence. Raman excitation using visible wavelengths can be integrated into standard light microscopes. This shorter wavelength excitation allows higher spatial resolution compared with infrared microscopy, allowing smaller sample volumes or even the detection of individual bacteria.

#### *3.2. Label-Free Detection of Bacteria by Raman Spectroscopy*

Raman spectroscopy has been used for many years to probe the biochemistry of various biomolecules, and more recently for disease detection. Specifically, Raman spectroscopy has been used to characterize bacteria in microbial colonies to detect their presence in smaller sample sizes with rapidity. However, most bacterial detection using RS relies on microspectral identification of reference strains or clinical isolates [73–75]. Raman mic rospectroscopy can detect bacterial cells in liquid suspensions, and it can identify bacteria directly from patient body fluids without culture. Sandra et al. conducted two studies in which isolation protocols from filtration [76] and centrifugation [77] were both developed to extract bacteria from patient sputum and urine, respectively. The type of causative strain was determined by Raman spectroscopy. By combining Raman spectroscopy with hierarchical cluster analysis (HCA), Jiirgen et al. directly detected individual bacterial cells from cerebrospinal fluid samples of meningococcal patients without any sample preparation steps [78].

The major limitation is that Raman scattering is extremely weak, resulting in relatively poor sensitivity compared with other optical methods such as autofluorescence and absorption [78]. This means that collecting vibrational spectra via spontaneously generated Raman photons requires extremely sensitive detection hardware, long exposure times, and relatively high excitation power compared to other optical techniques. In recent years, surface-enhanced Raman spectroscopy (SERS) has been extensively studied in the detection of chemical and biological agents with its rapid and ultra-sensitive characteristics [79,80].

#### *3.3. Label-Free Detection of Bacteria by SERS*

SERS is a combination of Raman spectroscopy and nanotechnology. It retains the advantages of fast acquisition of RS, less sample consumption, and fingerprint spectra for specific analytes. In addition, SERS significantly enhances the sensitivity of Raman spectroscopy over several orders, thus reducing the interference from self-fluorescence. The weak Raman scattering intensity of the sample is greatly enhanced by placing the sample on the nanoscale rough noble metal surface or mixing the sample with the noble metal colloidal suspension. In SERS, the average enhancement coefficient was between 104 and 108, and it could reach 1011 in some cases [81–84].

The SERS effect can be explained by two enhancement mechanisms: electromagnetic and chemical. The former is the enhancement of electromagnetic field due to local surface plasmon resonance (LSPR) [85,86], while the latter is chemical enhancement due to the charge transfer process between metal nanoparticles and analytes [86], although the contribution of this mechanism has been shown to be much lower than that of electromagnetic enhancement. Two SERS methods have been developed, the label-based method and the label-free method. However, despite the high sensitivity, label-based methods only provide information about reporter molecules and lose the intrinsic information of bacterial cells. The accuracy of the label-based method is entirely dependent on the specificity of recognition molecules. In addition, the labeling will significantly increase the sample analysis time. Compared with the label-based SERS method, the label-free method is rapid and easy to operate without any external labeling [87]. Label-free methods can detect bacteria by measuring the SERS pattern inherent in the cell wall, allowing for direct bacteria identification. However, the sensitivity of the label-free SERS method largely depends on the SERS substrate, the bacterial species, and the sample preparation methods.

Noble metal nanoparticles such as gold and silver are the usually preferred light intensifiers in SERS. The plasmonic characteristics of these noble metal nanoparticles, namely LSPR and the electromagnetic field generated on the surface, are mainly determined by the size, shape, and mutual assembly of the metal nanoparticles and the dielectric properties of the surrounding medium [88]. In general, silver-based SERS substrates have higher SERS enhancement effects than gold. However, silver is less stable, and has a biotoxic effect on living organisms, which limits its application in living organisms. Gold is much more stable, strongly chemical inert, and less biotoxic than silver. The nanostructure of gold is stable, facilitating better control of the size and shape of particles with higher biocompatibility. In order to achieve highly sensitive and repeatable SERS detection, the size, shape, and stability of metal particles should be reasonably controlled. The aggregate of nanoparticles was found to exhibit a larger Raman-enhanced signal than individual nanoparticles due to the generation of hot spots in the gaps between nanoparticles. Additionally, the nanostructure with sharp tips can also significantly enhance the SERS intensity. The generation of hot spots is highly sensitive to the size, shape, and gap-distance of nanoparticles [87]. Therefore, top-down lithography methods and bottomup self-assembly methods have been developed to control the shapes, arrangements, and assemblies of nanoparticles [89–91].

In general, there are several strategies that have been developed for the direct label-free SERS detection of bacteria, which are summarized as follows.

#### 3.3.1. In Situ Formation of Colloidal Silver/Gold on the Surface of Bacteria

The common methods for forming colloidal silver/gold on the surface or inside of bacteria are achieved by soaking the bacteria in sodium borohydride solution, then resuspending in silver nitrate or chloroauric acid (HAuCl4). The metal ions outside the cell wall react with reducing agents released from the cell, resulting in the colloids formation on the cell wall. Tamitake et al. employed a focused near-infrared laser beam to capture individual bacteria in aqueous Ag nitrate; Ag nanoaggregates were generated on *E. coli* by an additional green laser beam stimulation. In this way, the Raman scattering signal of *E. coli* was obtained by the Raman tweezer technique at single cell level [92].

#### 3.3.2. Direct Bacteria Detection on a Planar SERS Surface

The planar SERS substrate can be gold-plated glass slides with high roughness or self-assembled SERS active substrate through rational design. The bacterial suspension is dropped on the substrate and allowed to dry for bacterial detection [93–95]. Wang et al. [96] prepared Ag/AAO SERS substrates embedding Ag nanoparticles in anodic aluminum oxide (AAO) nanochannels. This substrate possesses high reproducibility, therefore can be further analyzed by principal component analysis (PCA), linear discriminant analysis (LDA), and support vector machine (SVM) to detect Staphylococcus Aureus (Gram-positive bacterium), Klebsiella Pneumoniae (Gram-negative bacterium), and Mycobacterium Smegmatis (Mycobacterium) and other bacteria, providing a good strategy for clinical microbial detection.

Andrei et al. reported that with the modification of anti-fimbrial antibodies onto the polyethylene glycol (OEG12) molecular layer on the amorphous hydrogenated silicon (a-Si:H) film. The fimbriated *E. coli* was specifically captured onto the surface as shown in Figure 3a. The positively charged gold nanorods (Au NRs) were attracted to the negatively charged *E. coli* on the film, facilitating the reading of the SERS signals. This method has high repeatability for the detection of bacteria, due to the uniform coverage of Au NRs on the bacterial membrane [97].

**Figure 3.** Schematic detection principle of *E. coli* hydrogenated amorphous silicon a-Si:H surface modified with anti-fimbrial antibodies against the major pilin protein fimA. (**a**) Surface structures of *E. coli* expressing fimA selectively captured and positively charged Au-NRs incubated with *E. coli* for SERS sensing. (**b**) Anti-fimbriae modified array, optical imaging of spots after interaction with *E. coli* and SERS spectra after capturing bacteria [97]. Adapted with permission from Ref. [97]. Copyright © 2020 Elsevier B.V.

Lv et al. used glycidyl methacrylate and ethylene dimethacrylate to prepare a convex substrate using a concave glass mold. The surface was treated with mercaptan to capture the Au nanoparticles on the surface as shown in Figure 4. The bacterial suspension is dropped on the SERS substrate, and the SERS spectrum of *E. coli* can be obtained after the sample dries naturally, as shown in Figure 4d. This simple SERS substrate preparation method proposed in this study was able to generate homogeneous and reproducible SERS active substrates over a large area, which has significantly improved the sensitivity of Raman spectroscopy. In this experiment, propanethiol, 3-mercaptopropionic acid, and cysteamine were modified on the surface of gold nanoparticles to improve the preferential

adsorption ability of bacteria in very diluted thallus solution, while the SERS spectrum was used for the direct detection of the captured microorganisms as shown in Figure 4d [98].

**Figure 4.** Schematic and detection principle of GNP/monolith modified substrate for the capture of *E. coli*. (**a**) Cross-sectional view of *E. coli* captured on gold nanoparticles modified substrates. (**b**) SERS enhancement factor of porous substrate functionalized with 40 nm gold nanoparticles simulated by FDTD. (**c**) In the simulation, the geometry of the model is reduced to two hemispheres coated with 40 nm spherical gold nanoparticles, separated by 10 nm; the electric field intensity distributions in x-y plane and y-z plane of gold on porous monolithic substrate excited by 633 nm laser are calculated. (**d**) SERS spectra of 40 nm gold nanoparticles/substrate functionalized with cysteamine [98]. Adapted with permission from Ref. [98]. Copyright © 2015 Elsevier B.V.

#### 3.3.3. Direct Bacteria Detection in SERS Suspension

Bacteria detection can be achieved in the suspension by directly mixing the bacteria with colloid. By optimizing the volume ratio of bacterial suspension to colloidal silver. Davis et al. were able to detect *E. coli* as low as 103 CFU/mL by correcting the Raman spectrum of the wide vibrational OH band in water [99]. Jennifer developed a bacterial SERS detection platform that can detect bacteria in a controlled liquid environment that maintains the viability of bacteria in a liquid environment. Plasmon resonance nanorods with different longitudinal lengths were used to detect Gram-negative *E. coli*, *Staphylococcus epidermidis*, *Serratia marcescens,* and Gram-positive *S. aureus*. The SERS signal was much higher with the higher surface charge density of the bacteria, indicating that the higher SERS-enhanced signal comes from the electrostatic attraction between the positively charged nanorods and the negatively charged bacteria. This label-free liquid-SERS assay provides a promising strategy for bacterial identification and AST testing in living organisms [100].

#### **4. Label-Free Detection of Bacteria by Dark-Field Microscopy**

#### *4.1. Dark-Field Microscopy Imaging Principle*

Dark-field microscopy is a microscopy technique that obliquely illuminates a sample by attaching a circular opaque baffle to a condenser to prevent the incident light from directing into the camera [101]. When the incident light enters the condenser, the center part is blocked by the baffle, leaving the edge light to pass through. The annular beam formed by the incident light turns into a hollow conical beam after the light is concentrated through the condenser, and illuminates the sample, thus stimulating the scattering of sample particles. In this setting, only scattering light from objects in the medium enters the objective lens, creating a bright scattering pattern in a dark background [102]. Due to the Tyndall effect, particles far below the resolution limit of typical light microscopes can be observed using dark-field microscopes [103].

#### *4.2. Label and Label-Free Detection of Bacteria by Dark-Field Microscopy*

Dark-field microscopy is an interesting optical technique that has been successfully used to image bacteria [40,104–114] and protozoa [102,115] due to its very low background, simple construction, portability, and low cost. Since plasma nanoparticles exhibit strong scattering to visible light, dark-field microscopy is a powerful tool for imaging and localization of noble metal nanoparticles in single cell analysis [101,109,116,117]. For example, hollow gold-silver nanoparticles are used as an alternative, less invasive contrast agent to assess the uptake process of malignant lymphocytes [118]. When the nanoparticles were modified by ligand and specifically bound to the cell membrane or internalized into the organelles, bright spots of different sizes and strengths could be observed on the surface of the target bacteria or around the organelles. Bacteria can be identified or counted based on the location and intensity of the bright spots. For example, Li et al. [104] developed a simple and fast bacterial count method based on dark-field light scattering imaging of a bacteria using gold nanoparticles as reporters. Zhou et al. [119] functionalized magnetic nanoparticles (MNP) using specific antibodies, which then formed a ring structure around *E. coli*, facilitating the counting of MNP conjugated *E. coli* under a dark-field microscope, as shown in Figure 5. In a similar way, Watanabe et al. [112] used phages as biometric elements, and aggregation-induced light scattering signals from silica nanospheres assembled by gold nanoparticles as signal transducers. After mixing the samples with the phage scattering probe of *S. aureus*, the detection limit of *S. aureus* was 8 × <sup>10</sup><sup>4</sup> CFU/mL within 15–20 min.

**Figure 5.** Schematic diagram of counting *E. coli* under dark-field, using antibody functionalization of MNP to form a gold ring structure around *E. coli.* (**a**) MNP probe was obtained by culture of *E. coli* antibody onto MNP. *E. coli* samples are first mixed with MNP probes to form probe-*E. coli* complexes. (**b**)The complex of *E. coli* and MNP probes was separated by a magnet and then counted under a dark-field microscope. [119]. Adapted with permission from Ref. [119]. Copyright © 2018 The Author(s).

Shiigi et al. [117] developed a novel molecular imprinting polymer (MIP) particle coated with gold nanoparticles (AuNPs) that can act as an acceptor and an optical signal transmitter in biological systems after modifying specific antibodies on its surface. Due to the coating of AuNPs, MIP particles produce a strong scattered light signal, and the binding of MIP particles increases the light intensity of the target bacteria. This allows bacteria to be clearly visible under darkfield microscopy, allowing them to be quantified using scattered light intensity. Using this technique, they successfully quantified *E. coli* O157 cells in meat samples.

Although powerful, the above-mentioned methods require the use of nanoparticles for signal reading of bacteria via dark-field microscopy, which affects the original physiological activity state of the bacteria detected and cannot reflect the real physiological activity and quantity of the bacteria [40,117,120–122]. Therefore, it is more desirable to detect bacteria in a label-free, rapid manner as the scattering intensity of bacteria is strong enough for direct dark-field imaging. In recent years, several methods have been developed to detect bacteria label-free using dark-field microscopy. For example, Colpo et al. [40] established a sensing platform for the rapid detection of bacteria in field samples using specific antibodies as recognition elements and dark-field microscopy as detection technology. By covering a gold layer on the polished silicon wafer and covalently modifying polyclonal anti-*E. Coli* antibodies to the surface, the sensing chip can be used for the specific capture of *E. coli* on the surface. As shown in Figure 6, the circularity and size of the object were used to identify the captured bacteria by dark-field microscopy. The performance was tested and compared to the Colilert-18 test and the quantitative polymerase chain reaction (qPCR), which showed comparable results.

**Figure 6.** Schematic of detection of *E. coli* with dark-field microscopy. (**a**) Samples containing *E. coli.* (b) an anti-*E. coli* antibody functionalized gold surface. (**c**) Dark-field microscopy is used to inspect the surface of the gold sheet after 75 min incubation with the field sample and rinse with phosphate buffer solution, enlarging the image. (**d**) Statistical image analysis was used to count the bacteria captured by the antibodies [40]. Adapted with permission from Ref. [40]. Copyright © 2019 MDPI.

Creighton et al. identified Treponema Pallidum under optical microscopy with doublereflection and single-reflection dark-field condensers based on spirochetes of bacterial characteristic morphology and locomotion criteria. Ideally, this method can identify *Treponema Pallidum* using dark-field microscopy within 20 min [120].

Rapid diagnosis of bacterial infectious diseases has important clinical significance for rapid and rational use of antibiotics, so as to avoid the misuse of antibiotics. However, the detection of pathogenic bacteria generally requires molecular identification using antibodies or aptamers, which requires long incubation time, as well as complex sample pretreatment and signal amplification. To address this challenge, Fang [121] and Wang [122] used light scattering imaging methods to detect individual bacteria without labeling by the scattering intensity trajectory of particles in free solution. The scattering strength variation provides particle shape information because it is relevance to the morphological heterogeneity of the particle. The fluctuating pattern of the scattering intensity also depends on the shape and orientation of the particles in free solution, such as rod-shaped bacteria, whose scattering intensity fluctuates significantly higher than that of the spherical shape in free solution, which can be used to characterize the shape of the bacteria. Fang's group used label-free single-particle dark-field imaging for rapid and sensitive identification of bacteria in free solution by modulating the convection [121] as shown in Figure 7. Using this method, they were able to distinguish positive samples of streptococcus agalactiae from vaginal swabs within 10 min without the use of any biological reagents. In addition to the spherical shape bacteria, the optical characteristics of single bacteria with different shapes such as *E-coli* are also significantly different from the matrix, implying that the rapid detection of different types of bacteria in one clinical sample is plausible, facilitating the precise prescription of antibiotics.

**Figure 7.** Bacteria detection principle by a single-particle imaging approach. (**a**) Schematic diagram of bacteria detection by single-particle imaging. (**b**) The inhomogeneity of particle morphology is identified by tracking the fluctuations of scattering intensity in free solution. (**c**) Convection induced by an electric heater was used to screen individual bacteria in a small field of view [121]. Adapted with permission from Ref. [121]. Copyright © 2022 The Author(s).

Similarly, Wang et al. used a large-volume solution scattering imaging (LVSi) system to track the scattering intensity and movement track of individual bacteria in short videos. The machine learning algorithm was used to perform aggregation analysis on their scattering intensity and movement trajectory. The presence of *E. coli* or similar bacteria in urine could be accurately determined, and bacteria could be distinguished from other common particles in urine, as shown in Figure 8. The method can detect patients with urinary tract infection within 10 min with an accuracy of 92.3%.

**Figure 8.** The principle of tracking the rapid identification of 1 um polystyrene spheres and single cell phenotypic characteristics of *E. coli*. (**a**) *E. coli* rotation-induced scattering intensity fluctuation tracking compared with 1 μm microbeads. (**b**) SVM classification result of one representative infection negative sample. (**c**) SVM classification result of one representative infection positive sample. [122]. Adapted with permission from Ref. [122]. Copyright © 2022 American Chemical Society.

#### **5. Other Methods for Label-Free Detection of Bacteria**

Other progress in the field of label-free optical biosensors is the advent of optical fiber gratings. Smietana [123] et al. first proposed a low-cost LGPs sensor that detects specific *E. coli* without labeling by physical adsorption. To further improve the sensitivity, Saurabh [124] proposed a compact ultra-sensitive long-period fiber grating (LPFGs) detection method for high-sensitivity label-free detection of specific *E. coli*. with modification of bacteriophage as shown in Figure 9. Simona [125] developed a reflective long-period fiber grating (RT-LPG) biosensor that can rapidly detect Class C β-lactamases in simple and complex biological samples. Additionally, fiber Bragg gratings (FBGs) can be used for bacterial detection [126,127].

Alternatively, the bacteria can be detected with the preparation of SERS hot spots on a fiber tip using optical fiber technology. The fiber-optic SERS probe (SERS on-a-tip) is highly controllable and reproducible [128–130].

Similarly, Biolayer interferometry (BLI) technology has been reported for bacterial detection in recent years. BLI is a label-free optical detection technique for real-time monitoring of biomolecular interactions. When an analyte binds to a ligand immobilized on the tip surface of a glass fiber-optic biosensor, its spectrum shifts with the change in the thickness of the related molecular layer. For example, Zhang et al. [131] reported a new method for the rapid, label-free real-time detection of *Salmonella enterica* using NLI incorporating antibodies as receptors, with a detection limit of 1.6×10<sup>5</sup> CFU/mL. Gu et al. [127] used C54A mutant LysGH15 as a receptor and combined it with BLI to establish a rapid, highly specific and label-free method for real-time detection of *Staphylococcus aureus* (*S. aureus*). This method can directly detect *S. aureus*, and its detection limit is 13·CFU/mL.

**Figure 9.** Schematic illustration of the experimental arrangement. (**a**) Covalent binding of phage to SiO2 on fiber surface. (**b**) Resonance wavelength change with analyte refractive index transmission spectrum [124]. Adapted with permission from Ref. [124]. Copyright © 2012 Elsevier B.V.

#### **6. Conclusions**

In this paper, the applications of SPR, Raman spectroscopy and dark-field microscopy for the label-free detection of pathogenic bacteria are reviewed. The principle of SPR, Raman spectroscopy, dark-field microscopy as well as fiber-based methods for the labelfree detection of pathogenic bacteria are considered. These label-free optical methods possess advantages of rapidity and low-cost, and are promising candidates for the clinical use for infectious disease diagnosis, facilitating the precise prescription of antibiotics to avoid the misuse of antibiotics, which is becoming a global problem.

The SPR imaging platform has been applied for high-throughput analysis, including the simultaneous detection of different bacterial species, antibiotic and bacterial interactions, etc. However, SPR generally suffers from the problem of non-specific adsorption and the direct detection of bacteria without sample preprocessing remains a challenge. Due to the high spatial resolution, SPR microscopy is able to image the bacteria at single cell level and possibly distinguish particles by their mass, and is potentially able to differentiate nonspecific adsorption. However, SPR microscopy is not commercially available, the total internal reflection fluorescence objective used for SPR microscopy is quite expensive. SERS

is another label-free method for the rapid detection of bacteria with low cost based on the fingerprint vibration spectra. However, it is still a challenging task to detect bacteria in a label-free manner in complex biological environments. An obstacle lies in the large SERS background contributed from the complex matrix. Fortunately, recently developed machine learning methods are possibly to address this challenge. Direct detection of bacteria by darkfield microscopy on a substrate can be significant interfered with by nonspecific adsorption of other substances such as cell fragments and exosomes in the matrix, therefore, relatively few studies on the label-free detection of bacteria by this technique have been reported. However, the direct imaging of bacteria in free solution by dark-field microscopy is a unique approach reported recently which is quite promising in addressing this challenge due to it rapidity and low-cost, as it does not need any biological reagents or an incubation process. Despite the difficulties in differentiating bacteria with similar sizes and shapes, the recently developed image recognition and machine learning technologies are likely to address this challenge. Therefore, we believe that this dark-field imaging method for label-free bacterial detection in free solution will be widely used in bacterial detection, clinical diagnosis, and infectious disease control due to its high sensitivity, rapidity, simplicity, and low-cost.

Compared with the label-free optical methods, the paper based colorimetric methods have attracted increasing attention due to their simplicity and cost-effectiveness, as well as the rapid signal readout with the naked eye, making them a promising candidate for the development of point of care devices [132]. However, the colorimetric methods are largely compromised by relatively poor sensitivity. Signal amplification methods can be applied to further improve the sensitivity, but they require additional processes, which significantly increase the detection time. Therefore, we believe that the reagent-free dark-field imaging method for label-free bacterial detection in free solution is more advantageous and will be widely used in bacterial detection, clinical diagnosis, and infectious disease control due to its high sensitivity, rapidity, simplicity, and low-cost.

**Author Contributions:** Writing—original draft preparation, P.W.; writing—review and editing, H.S., W.Y. and Y.F.; funding acquisition, Y.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by National Natural Science Foundation of China (NSFC, Grant No: 21874072, 22174069); and the Key R&D Program of Jiangsu Province (Grant No. BE2021373).

**Acknowledgments:** The authors thank the Key R&D Program of Jiangsu Province (Grant No. BE2021373) and National Natural Science Foundation of China (NSFC, Grant No: 21874072, 22174069) for funding supports.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Review* **New Horizons for MXenes in Biosensing Applications**

**Decheng Lu 1, Huijuan Zhao 1,2, Xinying Zhang 1, Yingying Chen <sup>1</sup> and Lingyan Feng 1,3,\***


**Abstract:** Over the last few decades, biosensors have made significant advances in detecting noninvasive biomarkers of disease-related body fluid substances with high sensitivity, high accuracy, low cost and ease in operation. Among various two-dimensional (2D) materials, MXenes have attracted widespread interest due to their unique surface properties, as well as mechanical, optical, electrical and biocompatible properties, and have been applied in various fields, particularly in the preparation of biosensors, which play a critical role. Here, we systematically introduce the application of MXenes in electrochemical, optical and other bioanalytical methods in recent years. Finally, we summarise and discuss problems in the field of biosensing and possible future directions of MXenes. We hope to provide an outlook on MXenes applications in biosensing and to stimulate broader interests and research in MXenes across different disciplines.

**Keywords:** two-dimensional material; MXenes; biosensor; electrochemistry; optics

#### **1. Introduction**

The main two-dimensional (2D) material is a solid crystal consisting of a single or several atomic layers, a sheet thickness of 1–10 Å, and a lateral size ranging from 100 nm to several μm [1]. Two-dimensional materials with properties such as large specific surface area and unique electronics are focuses of research in many research fields [2]. Since 2004, Novoselov et al. performed exfoliation to obtain graphene nanostructures; since then, the two-dimensional material has attracted much attention [3]. In 2011, Gogotsi et al. prepared a two-dimensional Ti3C2 nanosheet named MXenes [4]. MXenes are typically a few μm laterally and 1 nm thick or less [5]. It shows superior physicochemical properties compared to other two-dimensional nanomaterials [6].

The precursor of MXenes is the MAX phase. MAX consists of Mn+1Xn units and an alternately stacked "A" element single atomic plane, expressed as Mn+1AXn. The unique crystal structure of the MAX phase combines the excellent properties of ceramics and metals [7]. Etching the "A" element of the MAX phase yields two-dimensional nanomaterial MXenes with a structural formula of Mn+1XnTx [8]. MXenes can be expressed as M2XF2, M2X(OH)2, M2XO2, etc. M is a transition metal; "A" is an element of Groups 13 and 14 of the periodic table; X is boron, carbon, or nitrogen; n includes integers from 1 to 3; Tx denotes surface groups [9] (Figure 1A,B). A list of the significant syntheses and processes in the field of MXenes research over the last decade, as well as the development of new MXenes core components and surface group control techniques, is illustrated in Figure 1C. Compared to the precursor MAX phase, derivative MXenes retain metallic and electrical conductivity benefits of MAX but also offer smaller lateral dimensions and thicknesses, as well as unique physical and chemical properties [10,11].

**Citation:** Lu, D.; Zhao, H.; Zhang, X.; Chen, Y.; Feng, L. New Horizons for MXenes in Biosensing Applications. *Biosensors* **2022**, *12*, 820. https:// doi.org/10.3390/bios12100820

Received: 7 September 2022 Accepted: 28 September 2022 Published: 2 October 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Figure 1.** (**A**) Elements represented by M, A, and X in the MAX phase. (**B**) The structure of the MAX phase and its corresponding MXenes. Reprinted with permission from Ref. [9]. Copyright 2013, Wiley-VCH. (**C**) Chronological presentations of progress in the field of MXenes. List of the main synthesis and processing breakthroughs over the first 10 years of MXenes' research and new MXenescore compositions discovered in that decade and progress in surface terminations control. <sup>δ</sup> Solid solution MXenes; - MXenesfrom non-MAX phase precursors; § out-of-plane ordered double transition metal MXene; ¤ MXenes from in-plane ordered double transition metal MAX phase analogues; \* 2D carbides and nitrides produced by bottom-up approaches; <sup>ε</sup> nitride MXenes produced by the post-synthesis treatment of carbide MXene; vacancy; <sup>ˆ</sup> mixed terminations. Reprinted with permission from Ref. [12]. Copyright 2021, Wiley-VCH.

The central area of current advanced biosensing research studies is developing biosensors for detecting biological and chemical molecules that affect disease or are damaging to the human body. The most advanced biosensors can accurately and rapidly detect the target, predict the onset of the disease in time, and receive immediate medical attention [13]. Hence, high sensitivity and selectivity are significant for the design of biosensors. Due to its unique mechanical, hydrophilic, biocompatibility, and other excellent properties, MXenes are frequently used as a new biosensing platform. Electrochemical biosensors are essential for biological, environmental, and pharmaceutical fields. It offers high sensitivity, long-term

reliability and high accuracy, rapidity, low cost, and easy miniaturisation [14]. In addition, electrochemical biosensors offer a further path for creating next-generation point-of-care testing devices [15]. With advancing nanotechnology with respect to MXene-based optical biosensors, unprecedented progress has been made in optical analysis. Optical analysis has advantages of high sensitivity, high selectivity, fast analysis, and good reproducibility. It has been widely used in biochemistry and biomedical and environmental analysis and has received increasing attention [16]. The synthesis of MXenes and their application in biosensing are reflected in Scheme 1. We will review and summarize published studies on biosensing since the development of MXenes, including those mainly classifying biosensors into electrochemical, optical biosensors and some derivative biosensors. In addition, we will also discuss the challenges of MXenes in preparing biosensors and future perspectives on applying MXenes in biosensing.

**Scheme 1.** MXenes cover both top-down and bottom-up methods of synthesis. They play an irreplaceable position in enzyme-, nucleic-acid-, immune-based electrochemical biosensing; photoluminescence; electrochemiluminescence; photoelectric effect-based optical biosensors; and other biosensors such as wearable biosensors, surface plasmon resonance, and surface-enhanced Raman spectroscopy.

#### **2. Synthesis and Structures of MXenes**

#### *2.1. Synthesis of MXenes*

There are two methods for the synthesis of MXenes. The top-down method is the most commonly used, which can be used to exfoliate multilayer materials into a few-layer or single-layer MXenes sheet. The second method is a bottom-up approach, which focuses on the growth of Mxenes from atoms or molecules [17,18].

#### 2.1.1. Top-Down Method

Selective etching disintegrates the strong covalent bonds between the MX and the A layers in the MAX phase. The primary method is etching with hydrofluoric acid (HF), molten salts, etc. In this process, oxygen (O), hydroxyl (OH), and fluorine (F) replace the M-A strong metal bond [17]. There are two main steps to gain 2D MXenes by HF: etching and exfoliation. Although the direct use of HF is straightforward and practical, it causes environmental pollution and damages to the human body [4]. In situ HF can be obtained by reacting a fluorinated salt with mild acid, which is less toxic to MXenes surfaces [19]. Researchers explored new synthetic methods (Figure 2). The typical chemical reaction equation for the synthesis of MXenes in the MAX phase is as follows [9].

$$\rm{M\_{n+1}AX\_n + 3HF \to AF\_3 + M\_{n+1}X\_n + \frac{3}{2}H\_2} \tag{1}$$

$$\rm M\_{n+1}\chi\_n + 2H\_2O \rightarrow M\_{n+1}\chi\_n(OH)\_2 + H\_2 \tag{2}$$

$$\rm M\_{n+1}\chi\_n + 2\rm HF \to M\_{n+1}\chi\_n\\F\_2 + \rm H\_2 \tag{3}$$

MXenes must undergo an exfoliation process to obtain nanosheet structures: The surface groups of MXenes result in the layers being linked by hydrogen and Van der Waals forces [3]. Exfoliation enhances the interlayer spacing by weakening interactions between layers using various molecular and ionic processes [20].

**Figure 2.** (**A**) Schematic diagram of the process of preparing MXenes by HF. Reprinted with permission from Ref. [21]. Copyright 2012, American Chemical Society. (**B**) A guide to Ti3C2 MXenes synthesis using HF. Reprinted with permission from Ref. [22]. Copyright 2017, American Chemical Society.

The molten salt method uses fluorinated molten salts, Lewis salts [23]. The synthesis does not involve fluoride, reducing the risk of synthesis [8,24]. The mechanism of MXenes formation in molten salts is similar to that of conventional HF methods: ZnCl2 and CuCl2 high-temperature molten salts strip a more comprehensive range of MAX phase materials [8] (Figure 3A). In the molten salt of Lewis acids, Zn2+, Cu2+, and Cl<sup>−</sup> are consistent with acting H+ and F<sup>−</sup> in HF. Minimally intensive layer delamination (MILD)

and electrochemical etching can also be used for MAX etching, producing high-quality, non-toxic MXenes [25,26].

#### 2.1.2. Bottom-Up Method

Bottom-up synthesis methods have been reported, such as chemical vapor deposition (CVD) [27], template [28], and plasma-enhanced pulsed laser deposition (PE-PLD) [29] (Figure 3B). MXenes produced by this method possess good crystalline quality and controllable structure and size [18].

**Figure 3.** (**A**) Preparation mechanism of Ti3C2Cl2 etched by ZnCl2. Reprinted with permission from Ref. [30]. Copyright 2019, American Chemical Society. (**B**) Bottom-up approach to obtain MXenes. Atomic layer deposition method: steps to prepare Ti3AlC2 MAX films by sputtering Ti, Al and C on a sapphire substrate (**a**), schematic diagram of Ti3C2Tx (**b**) and STEM images (**c**). CVD method: schematic diagram of the Mo2C synthesis process (**d**), AFM images of hexagonal ultra-thin Mo2C crystals (**e**) and STEM images (**f**). Reprinted with permission from Ref. [27]. Copyright 2020, American Chemical Society.

Xu et al. used CVD to synthesize high-quality Mo2C crystals [27]. The synthesis of Mo2C MXene/graphene heterostructures and Mo2C MXene-graphene hybrid films by this method has been reported [29,31]. Compared to CVD, the template method has a relatively high yield of MXenes. Two-dimensional MXenes are mainly obtained by carbonizing or nitriding two-dimensional transition metal oxide (TMO) nanosheet templates. Xia et al. prepared hexagonal-structured 2D h-MoN nanosheets using precursor MoO2 nanosheets [28]. PE-PLD is a successful method for preparing large-area ultra-thin face-centered cubic (FCC) Mo2C MXene [29].

The stability of MXenes is an important property and limits its application to a certain extent. Researchers have tried to improve its stability. High concentrations of HF accelerate the degradation of MXenes and affect its structure, so relatively mild reaction conditions are necessary [32]. Organic solvents mitigate the oxidation of MXenes. Contact with water should be avoided as much as possible to prevent oxidation [33]. The oxidation of MXenes is quicker in liquid media than in solid media, and this degradation process is exacerbated by photocatalysis and thermocatalysis [34]. The storage of MXenes in Ar-sealed vials at 4 °C exhibits high stability at room temperatures [35].

#### *2.2. Strustures of Mxenes*

The crystal structure within a 2D material can affect its properties [18]. There are six types of MXenes structures (Figure 4A): (1) single transition metal MXenes (Ti3C2 and Nb4C3); (2) solid solution MXenes ((Ti, V)3C2 and (Cr, V)3C2); (3) sequential planar internal and external bimetal MXenes with one transition metal occupying the outer layer (Cr and Mo); the central metal is another metal (Nb and Ta) [36,37]; (4) ordered doubletransition metals MXenes ((Cr2V) C2); (5) orrderly double vacancy MXenes (Mo1.33CTx) [38]; (6) random empty space MXenes (Nb1.33CTx) [39].

Computational simulation studies have been reported to identify novel stable MXenes structures, contributing to exploratory studies [40]. The properties and applications of these materials can be adapted by various parameters for composition, surface modification by heat treatment or chemical pathways, and structural adjustments [41]. MXenes have two—dimensional structures (a), one—dimensional structures (b) and (c), three—dimensional structures (d), and zero—dimensional structures (e) (Figure 4B).

**Figure 4.** (**A**) Different types of MXenes structures. Reprinted with permission from Ref. [42]. Copyright 2019, Elsevier. (**B**) 2D, 1D, 3D, and 0D structures of MXenes. Adapted with permission from Ref. [18]. Copyright 2021, Wiley-VCH.

#### **3. MXenes in Biosensing**

Several strategies involving MXenes in analytical nanoscience, biosensing, and other areas have been reported. MXenes exhibit hydrophilicity due to surface groups such as OH, O, and F. Its surface can interact with most biomolecules through hydrogen bonding, Van der Waals forces, electrostatic interactions, and ligand binding, rendering it an excellent carrier for biosensors applications [43–45]. Several different MXenes compositions have been proved to be biocompatible and non-cytotoxic [46,47].

We summarized the composition and analytical performance of some MXene-based electrochemical biosensors, optical biosensors, and other biosensors and attached them to the subsections. These cases demonstrate the broad applicability of MXenes in the fabrication of biosensors. Readers can easily extract MXene-based biosensing research and measurement data from these tables.

#### *3.1. Electrochemical Biosensing*

Due to their high electronic conductivity, MXenes can drive most electrochemical reactions, which is of great significance for the application in electrochemical biosensing [48]. The electrical properties of MXenes can be improved by changing elemental compositions and surface groups [18]. In particular, the external transition metal layer of MXenes plays a more critical role in the electronic properties than the internal layer [49]. The number and thickness of the layers of MXenes also affect electrical properties [3,50].

Biosensors based on electrical signals change the electrochemical properties of the sensor surface by binding to essential substances in the organism, such as proteins, amino acids, nucleic acids, antibodies, etc. (Figure 5). The development of MXenes for electrochemical biosensors has been intensively investigated because of their excellent properties, such as high conductivity, electrochemical activity, and large surface area. The classification of electrochemical biosensors is as follows: enzyme electrochemical, nucleic acid electrochemical, and immunoelectrochemical biosensing.

**Figure 5.** Schematic diagram of the analytical principle of an electrochemical biosensor. Reprinted with permission from Ref. [51]. Copyright 2020, Springer Nature.

#### 3.1.1. Enzyme-Based Electrochemical Biosensing

Enzyme electrochemical biosensors with higher efficiencies and substrate specificities in mild conditions have been extensively explored over the last few years (Table 1). The basic principle is the direct electron transfer (DET) process between the enzyme and the electrode. The immobilisation of enzymes on the bare electrode surface can render the enzymes biologically inactive, making it extremely difficult to perform DET on the electrode's surface [52]. MXenes can be used as a strategy for enhancing DET because of their large specific surface area, excellent electrical conductivity, and good biocompatibility.

Much of the literature has shown that MXenes or MXenes composite materials can maintain the activity of enzymes after complexing enzymes due to the various properties and unique structures of MXenes. This demonstrates that MXenes can be magnificent structures for enzyme-based biosensors. Xu et al. mixed Ti3C2 MXene and HRP enzyme directly to fabricate a biosensor for the detection of H2O2 to analyse the levels of serum samples from AMI patients before and after surgery [53]. Ma et al. fabricated a low detection limit enzyme biosensor for the detection of H2O2 using a chitosan complex of Ti3C2 MXene-loaded HRP enzyme and successfully used it to detect trace amounts of H2O2 in foods [54].


In addition, there are several reports on using MXenes in different compound types of enzymes, such as glucose oxidase [57], cholesterol oxidase [58], acetylcholinesterase [60], tyrosinase [59], etc. Wu et al. proposed a hybrid PLL/Ti3C2/glucose oxidase glucose biosensor that accelerates the breakdown of H2O2 generated during glucose oxidation by catalysing a cascade reaction [56] (Figure 6A). Xia et al. developed a Chit/cholesterol oxidase/Ti3C2Tx composite cholesterol oxidase biosensor [55]. Chit/Ti3C2Tx served as a

support matrix for immobilising the enzyme. Gold nanoparticles anchored on Ti3C2Tx MXene nanosheets enhanced the electron transfer between the enzyme and the electrode. The relative current sensitivity and LOD were 0.3–4.5 nM and 0.11 nM, respectively. Song et al. derived electrochemical etching to derive fluorine-free Nb2CTx with low cytotoxicity and constructed a Nb2CTx/acetylcholinesterase biosensor to detect sulfoxide [26] (Figure 6B). Moreover, the sensor's enzymatic activity and electron transfer are superior to the corresponding V2C and Ti3C2 MXenes biosensors. Wu et al. used Ti3C2 MXene as a new substrate to immobilise tyrosinase and facilitated the direct electron transfer process for the sensitive and rapid detection of phenol [59]. Therefore, Ti3C2 MXene can be a phenolic biosensor with high recovery and long-term stability. The biosensor exhibits good analytical performance over a wide linear range of 0.05–15.5 μM, with detection limits as low as 12 nM.

**Figure 6.** (**A**) Schematic diagram of the detection of the Ti3C2-PLL-Gox nanoreactor Reprinted with permission from Ref. [56]. Copyright 2020, Elsevier. (**B**) Schematic illustration of the enzymatic inhibition of sulfoxide detection by the HF-free Nb2CTx/AChE biosensor Reprinted with permission from Ref. [26]. Copyright 2020, Wiley-VCH.

The above examples and the contents show that it is feasible to combine enzymes directly on MXenes or with other materials to improve the performance of enzyme electrochemical biosensors.

#### 3.1.2. Nucleic Acid-Based Electrochemical Biosensing

Using nucleic acids as recognition elements allows the specific recognition of the target and the generation of some signal changes [61]. Nucleic acid is a stable and easy-tohandle biomolecule, so it has excellent detection performances [62]. Nucleic-acid-based electrochemical biosensors offer advantages of both nucleic acid probes and electrochemical detection, enabling the sensitive detection of analytes such as nucleic acid, ref. [63] proteins [64], biological molecules [65], inorganic ions [66], and cells [67] (Table 2). Nucleic acid electrochemical biosensors are based on five conformations: double-stranded, triple-stranded, quadruple-stranded, DNA nanostructures, and single-stranded DNA functionalisation (hairpin structure, aptamers, and DNAzyme) [68]. Unlike enzymes, nucleic acids possess little redox capacity. The development of nucleic acid electrochemical biosensors generally relies on molecules with redox properties, such as methylene blue (MB) and ferrocene (Fc), or through charge changes that occur during nucleic acid hybridisation [69]. The nucleic acid electrochemical biosensor has various applications in genetics, clinical medicine, and biosensing due to its rapid detection, simple experimental procedures, high sensitivity, and low cost [70]. There are two types of nucleic acid biosensors.

The first type of nucleic acid electrochemical biosensor follows the Watson–Crick pairing principle, which hybridizes a nucleic acid sequence with a complementary nucleic acid sequence through base pairing [61]. The detection principle works by immobilising nucleic acids on the electrode's surface to capture complementary nucleic acid sequences, thus obtaining an altered electrical signal for specific detection [71]. There are many reports using specific nucleic acid sequences to create biosensors for the detection of diseasepredicting miRNAs and DNA, and some electrochemical biosensors have been validated

for point-of-care detection. Duan et al. developed a Ti3C2/FePc QDs MXene nanocomposite nucleic acid biosensor with good biocompatibility [72]. The Ti3C2/FePc QDs composite material was used as a carrier to detect miRNA-155 by using a change in electrochemical impedance caused by DNA modifications. Mohammadniaei et al. used double screenprinted gold electrodes modified with MXenes and AuNPs and single-stranded DNAfunctionalised magnetic particles to detect miRNA-21 and miRNA-141 by using duplexspecific nuclease (DSN) amplification assay strategy [73] (Figure 7A). This biosensor can continue to be upgraded to quantify more analytes, forming a device for point-of-care testing (POC) cancer screening. Chen et al. fabricated a DNA electrochemical biosensor using MXene-based [74]. The surface groups were covered using ssDNA adsorbed on Ti3C2 MXene to attenuate conductivity. When target DNA and ssDNA are hybridized and desorbed from Ti3C2 MXene, the fast, simple, and sensitive detection of N-gene sequences in SARS-Cov-2 was possible (Figure 7B). The feasibility of DNA-functionalised MXenes in developing real-time monitoring diagnostic devices for clinical testing can be demonstrated.

**Table 2.** Nucleic acid-based electrochemical biosensors identify units, target, and analytical parameters.


The second nucleic acid electrochemical biosensor uses single-stranded DNA (ss-DNA) or RNA to bind to various biomolecules for analyte detection, including proteins, small biomolecules, cells, etc. [78,79]. Electrochemical biosensors made up of aptamers are easy, reliable, quick in responding, low in price, and possess acceptable repeatability [80]. Geng Xue of our research group cleverly used conformational changes of aptamers before and after capturing serotonin to construct an aptamer biosensor [81]. The interaction between aptamer and serotonin was destroyed by guanidine hydrochloride, and 98.2% of the signal was recovered, showing acceptable repeatability. Zhou et al. synthesized intercalating polypyrrole (PPy) Ti3C2Tx MXene and phosphomolybdic acid (PMo12) composites with a strong synergistic effect, promoting the anchoring of RNA aptamers on the composites [64] (Figure 7C). The G-quadruplex formed by osteopontin (OPN) and aptamer exhibits stable and high sensitivity, which proves the excellent performance of this MXene composite aptamer biosensor. Li et al. created a nuclease-driven DNA walker cascade signal amplification strategy to construct electrochemical aptamer biosensors on Au nanoparticles/MXene-modified electrodes for mucin 1 [76]. A DNA nanostructuremodified Ti3C2 MXene nanosheet biosensor was developed by Wang et al. for the detection of gliotoxins [65]. Tetrahedral DNA nanostructures were quickly immobilised on the surface of MXenes nanosheets, thus avoiding the tedious and expensive modification of DNA probes. HB5 aptamer immobilised on the MXenes layer via electrostatic interactions was highly selective for HER-2-positive cells, as reported by Vajhadin et al. Sandwich-like structures formed between magnetically captured cells, and functionalised MXenes electrodes effectively shield electron transfers, allowing quantitative cell detection with changes in the current [77] (Figure 7D).

**Figure 7.** (**A**) Schematic diagram representing the entire assay procedure for multiplex and concurrent detections of miRNA. Reprinted with permission from Ref. [73]. Copyright 2020, Elsevier. (**B**) Schematic of the ssDNA/Ti3C2Tx for the detection of the SARS-Cov-2 nucleocapsid gene. Reprinted with permission from Ref. [74]. Copyright 2022, American Chemical Society. (**C**) Schematic diagram of PPy@Ti3C2/PMo12 aptamer biosensor for OPN detection. Reprinted with permission from Ref. [64]. Copyright 2019, Elsevier. (**D**) Schematic diagram of the MXenes based cell sensor for the detection of SK-BR-3 cells: magnetic cell separation using CoFe2O4@Ag-HB5 (**a**) and electrochemical cell detection on a functionalised MXenes surface (**b**). Adapted with permission from Ref. [77]. Copyright 2021, Elsevier.

#### 3.1.3. Immunoelectrochemical Biosensing

Electrochemical immunosensors are coupled to the sensor via antigen–antibody interactions. The accessibility of antibody to a wide range of molecules and the high selectivity and sensitivity renders immunochemical methods valuable for clinical diagnosis. These electrochemical biosensors for bioanalysis have advantages of small reagent volumes, high sensitivity and specificity, and portability [82]. As observed from the contents, immunoelectrochemical biosensors offer tremendous advantages in the specific detection of biomolecules (Table 3).



In 2018, Kumar et al. fabricated the first MXene-based immunoelectrochemical sensor to detect carcinoembryonic antigens (CEAs) [84]. Aminosilane-functionalised MXenes offered more binding sites for bioreceptors than GCE, and the CEA antigen is better immobilised on Ti3C2 MXene (Figure 8A). Xu et al. synthesized a composite of 3D sodium titanate nanoribbons, anchored poly(3,4-ethylene dioxythiophene), and gold nanoparticles by oxidizing and alkalizing Ti3C2 Mxene [86]. The composites described above were used to immobilise prostate-specific substance (PSA) antibodies to create a facile electrochemical label-free immunosensor for the sensitive detection of PSA (Figure 8B).

**Figure 8.** (**A**) Schematic diagram of the detection mechanism of electrochemical CEA. Reprinted with permission from Ref. [84]. Copyright 2018, Elsevier. (**B**) The fabrication and detection steps of the immunosensor. Reprinted with permission from Ref. [86]. Copyright 2020, Elsevier. (**C**) Schematic diagram of the fabrication of the working electrode of the immunosensor. Reprinted with permission from Ref. [87]. Copyright 2020, American Chemical Society. (**D**) Illustrations for constructing the sandwich-like immunosensor of LM based on Ti3C2T*<sup>x</sup>* MXene. Reprinted with permission from Ref. [88]. Copyright 2021, Royal Society of Chemistry.

Dong et al. used a CuPtRh/NH2-Ti3C2 nanocomposite composed of trimetallic hollow CuPtRh cubic nanoboxes (CNBs) and laminated ammoniated Ti3C2 flakes to fabricate a sandwich-type immunosensor to detect cardiac troponin I (CTnI) [87]. Aminated Ti3C2 provides abundant binding sites for both CuPtRh CNBs and antibodies, while CuPtRh CNBs can prevent Ti3C2 from stacking again (Figure 8C). In addition, MXenes can serve to detect bacteria. Niu et al. constructed a sensing platform with carboxylated Ti3C2Tx MXene and rhodamine B/gold/reduced graphene oxide as the signal [88] (Figure 8D). A sandwich electrochemical immunosensing platform for detecting Listeria monocytogenes was also developed by them.

#### *3.2. Optical Biosensing*

Optical properties, including absorption, transmission, photoluminescence, scattering, and emission, are essential for applying MXenes. The surface groups, doping, and defects affect the energy band's structure [89]. A thin layer of Ti3C2Tx has been reported to absorb photons in the UV-visible region between 300 and 500 nm with a transmission of 91%. O-functionalised Ti3C2 MXene has a higher light absorption efficiency [90]. The optical properties are also affected by the thickness of the film and the distance between MXenes layers. Intercalation with hydrazine, urea, methyl ammonium hydroxide, and DMSO changes the interlayer distance of Ti3C2Tx, decreasing light transmittance [50].

MXenes have excellent hydrophilicity, biocompatibility, and optical characteristics, making them appropriate for all sorts of biosensing applications. It was discovered as a fluorescence quenching agent and a carrier for biomedical and imaging applications, contributing to high-performance optical biosensors. The interaction of light and materials is central to the optical inspection principle. It identifies samples by non-destructively monitoring changes in the intensity or spectral shift of light [91]. This section will summarize MXenes biosensing applications in photoluminescence, electrochemiluminescence, and photoelectrochemical applications.

#### 3.2.1. Photoluminescence (PL)

MXenes possess features that make MXenes excellent for fluorescence biosensors, such as larger absorption bands, higher energy levels, etc., which causes fluorescence quenches in fluorescent substances [91]. Hence, changes in fluorescence intensity can be employed as indications for biological analytes detection (Table 4). MXene quantum dots (MQDs) are luminous, extremely water-soluble, dispersible, and biocompatible [92]. As a consequence, searching for photoluminescent biosensors based on MXenes and MQDs has emerged as a widespread research issue [93,94].

**Table 4.** MXene-based photoluminescence biosensors identify units, target, and analytical parameters.


Because of MXenes' strong and broad absorption in the visible and near-infrared regions, MXenes generally act as an acceptor designed to quench the fluorescence signal emitted by sensing probes, such as metal nanoclusters, quantum dots, fluorescent dyes, etc. [94]. Due to the two-dimensional planer structure and hydrophilic surface groups of MXenes, the abundant binding sites and hydrophilic groups on MXenes provide more possibilities for biomolecular interactions [44].

Shi et al. detected glutathione by combining copper nanoclusters (Cu NCs)-functionalised MXenes [99]. MXenes quenches the fluorescence of Cu NCs through the internal filtering effect (IFE), and glutathione can analyze MXenes and Cu NCs, resulting in fluorescence recovery. Ti3C2 MXene nanosheets combined with red-emitting carbon dots (RCDs) area unit effective and selective fluorescence stimulant sensors were used for glucose detection by Zhu et al. Ti3C2 nanosheets impassively quenched the fluorescence intensity of RCDs (>96%) through IFE [94] (Figure 9A). Kalkal et al. constructed a fluorescent biosensing system based on Ag/Ti3C2 to quench the fluorescence signal on antibody/amino-graphene quantum dots [97]. The fluorescence recovered when antigen was added. It can be used to detect neuron-specific enolase with good reproducibility.

MXenes are used as fluorescence quenchers to construct optical sensors for monitoring enzyme activity and biomolecules. Similarly to the previous section, the Fc and MB of the nucleic acid biosensor can be replaced with some fluorescent materials that can be used, which are more practical for this type of biosensor. Zhu et al. reported a Ti3C2 MXene-based fluorescent biosensor to detect phospholipase D by FRET quenching of rhodamine B (RhB)-labeled phospholipids [98]. Phospholipase D cleaves phospholipids, causing RhB-labeled phospholipids to detach from Ti3C2 MXenes and re-reflorescence. Peng et al. used the affinity difference between single-stranded and double-stranded DNA on MXenes to construct fluorescent signal detection for human papillomavirus HPV-18 DNA on ultra-thin Ti3C2 MXene [95] (Figure 9B). Wang et al. presented dual-signal-labelled DNA-functionalised Ti3C2 MXene nanoprobes to achieve a dual analysis of MUC 1 and miRNA-21 at low concentrations in vitro, and the in situ imaging of MCF-7 breast cancer cells [96]. Furthermore, cell imaging can provide multiple layers of information, such as biomarkers' expression levels and spatial distribution.

**Figure 9.** (**A**) Schematic diagram of MXene-based glucose oxidase fluorescent biosensor. Reprinted with permission from Ref. [94]. Copyright 2019, Royal Society of Chemistry. (**B**) Illustration of analysis of HPV-18 type DNA using Ti3C2 MXene. Reprinted with permission from Ref. [95]. Copyright 2019, Elsevier.

When the thickness dimensions of 2D nanomaterials are less than 100 nm, MXenes can be converted into quantum dots with quantum confinement and optical properties [105]. MQDs, with an average lateral size ranging from 1.8 to 16 nm, can be obtained by hydrothermal processes [100], acidic oxidation, and chemical stripping [106]. Charge transfer is enhanced, and fluorescence is enhanced by utilizing heteroatom doping [101]. MQDs have similar properties to MXenes, such as high dispersion and good biocompatibility. The small band gaps of MXenes can expand their band gap through quantum effects, contributing to their strong fluorescence effect [107]. Some researchers synthesized MQDs that exhibited different fluorescence effects in different solvents under 365 nm UV light irradiation [93,104].

On account of their tunable size, photoluminescence, and photostability, MQDs can be applied as fluorescent probes and can also be functionalised with natural biomolecules [107]. The performance of MQDs as fluorescent agents or signals can be improved, and the application of MXenes in biosensing can be widely expanded [91,108]. MQDs have contributed enormously to detecting metal ions, biomolecules, and cellular imaging. The first MQDbased fluorescence sensor is based on the coordination of Zn2+ through hydroxyl groups on the surface of MQDs with selective quenching [100] (Figure 10B). Heteroatom-doped MQDs can be the detector for the fluorescence detection of different metal particles, such as Cu2+ [101], Ag+, and Mn2+ [109].

**Figure 10.** (**A**) Schematic diagram of fluorescence assay of alkaline phosphatase activity of MQDs. Reprinted with permission from Ref. [102]. Copyright 2018, Royal Society of Chemistry. (**B**) Schematic diagram of hydrothermal preparation of MQDs. Reprinted with permission from Ref. [100]. Copyright 2017, Wiley-VCH.

MQDs can be implemented to detect some biomolecules because they have absorption bands that overlap with the excitation and/or emission spectra of MQDs. Guo et al. designed an MQD-based fluorometric strategy for alkaline phosphatase activities and

embryonic stem cell identification [102] (Figure 10A). The effective quenching of MQD fluorescence was obtained by p-nitrophenol produced by the alkaline phosphatase-catalysed dephosphorylation of p-nitrophenyl phosphate. It can also be used as an IFE-based method to analyse ESC biomarker ALP in ESC lysates accurately. Liu et al. described a fluorescent platform for detecting cytochrome c and trypsin [103]. The fluorescence of MQDs was burst by cytochrome c through the IFE. Meanwhile, cytochrome could be degraded by trypsin, and MQDs' fluorescence could be restored. Chen et al. constructed a fluorescent sensor with the pH-dependent emission of blue fluorescence from MQDs for ratiometric MQDs probes to detect cellular pH [104].

#### 3.2.2. Electrochemiluminescence (ECL)

As a mixture of electrochemistry and optics, electrochemiluminescence is a new method for evaluations and detections. Because of its low background signal, excellent sensitivity, controllability, speed, and low cost, it is frequently employed in biochemistry for proteins, nucleic acids, enzymes, and clinical diagnostics [91]. MXenes have been proven viable as working electrodes for ECL biosensors, with improved ECL characteristics compared to glassy carbon electrodes [110]. The ECL biosensor is well suited for the analysis of nucleic acids or gene fragments, biomolecules, biomarkers, and even cells (Table 5).

In 2018, Fang et al. fabricated an ECL biosensor of Ru(bpy)3 2+ functionalised Ti3C2Tx MXene to detect unlabelled single nucleotide mismatches in human urine, using tripropylamine as a co-reactant [111]. Exposed bases in mismatched DNA bind to Ru(bpy)3 2+ on the Ti3C2Tx MXene and are more prone to electrochemical oxidation in enhancing ECL intensities. Zhuang et al. constructed ECL nanoprobes via Ti3C2Tx-mediated in situ formations of Au NPs and the anchoring of luminol and utilised the catalytic hairpin assembly (CHA) amplification of signalling to fabricate ECL biosensors for miRNA-155 detection [112] (Figure 11A). Yao et al. detected the SARS-Cov-2 gene by MXenes/PEI adsorbed Au and Ru(bpy)3 2+ DNA walkers [113]. After the DNA walker excised hairpin DNA under the action of Nb.BbvCl endonuclease, template DNA-Ag hybridized with hairpin DNA and decreased the signal of ECL (Figure 11B). Zhang et al. modified DNA probes on MXenes/PEI composites by Ru(bpy)3 2+ and AuNP and used the CRISPR-Cas12a strategy to construct an ECL signal on/off biosensor for the detection of SARS-Cov-2 (RdRp) gene [114].

**Table 5.** MXene-based electrochemiluminescence biosensors identify units, target, and analytical parameters.


Strategies for detecting biomolecules can be implemented with aptamers, resulting in higher ECL signal intensities. Sun et al. proposed PEI-functionalised MXenes and g-C3N4 composites as detection probes, and kemptide chelated with Ti in the composites after protein kinase A (PKA) phosphorylation to promote electron transfers at the electrode's interface, enhancing strategies for ECL signalling [115]. Moreover, this biosensor enables inhibitor screening and PKA activity monitoring in MCF-7 cell lysates. Mi et al. reported a method for the quantitative detection of cardiac troponin (CTnI) by electrochemical and ECL dual signals using tetrahedral DNAs (TDs) and in situ hybrid chain reaction (HPR) on Au/Ti3C2 MXene [118] (Figure 11C). Both ECL Dox-Luminol/Current Dox and Current MB/Current Dox dual signals can be used for the quantitative detection of CTnI, which is expected to be used in screen critically ill patients with COVID-19. Zhang et al. used MXenes to generate AuNPs in situ and modified aptamers and constructed an ECL biosensor to detect exosomes CD63 [117]. Zhang and colleagues developed an exosomeselective ECL biosensor using aptamer-modified Ti3C2 MXenes as probes with an LOD of 125 μL particles−<sup>1</sup> [116].

Immunochemical methods are also highly selective and sensitive in the field of ECL. Luo et al. constructed an MXene-based substrate using [Ru(bpy)2(mcpbpy)]Cl2 and Larginine as co-reactants to detect carcinoembryonic proteins (CEA) by antigen [119]. Upon antigen binding to the antibody, spatial site resistance leads to a decline in the rate of electron transfer and electrolyte diffusion at the electrode's surface, resulting in a decrease in ECL signal intensities (Figure 11D). Wei et al. constructed an ECL/SERS dual-signal biosensor to detect the causative agent of Vibrio vulnificus (VV) [120]. The pathogenic bacteria. VV is captured by Fe3O4@Ab1 as the capture unit. Ab2, R6G, and ABEI bind to AuNR as the signal unit to capture VV through Au-S and Au-N, forming a Faraday cage structure.

**Figure 11.** (**A**) Schematic diagram of the preparation of Au@Ti3C2@PEI-Ru(dcbpy)3 2+ nanocomposites (**a**); Combined unilateral DNA walker amplification strategy based on nanocomposites for ECL biosensor detection of SARS-Cov-2 RdRp gene (**b**). Reprinted with permission from Ref. [113]. Copyright 2021, American Chemistry Society. (**B**) Strategy of stable luminol-Au NPs-Ti3C2 (**a**) and construction of the proposed ECL biosensor (**b**). Reprinted with permission from Ref. [112]. Copyright 2021, Springer Nature. (**C**) Schematic representation of specific target recognition and BFP release (**a**) and the ratiometric biosensing mechanism of cTnI (**b**). Reprinted with permission from Ref. [118]. Copyright 2021, Elsevier. (**D**) Schematic representation showing the detection principle of the prepared ECL biosensor. Reprinted with permission from Ref. [119]. Copyright 2022, Royal Society of Chemistry.

#### 3.2.3. Photoelectrochemical (PEC)

Similarly to electrochemiluminescence sensing, photoelectrochemical sensing is a practical analytical method that integrates optical and electrochemical analyses. MXenes also promise photoelectrochemical sensors with their excellent optical and electronic properties [121] (Table 6).



Li et al. took advantage of Ti3C2 MXene, readily forming PN junctions with photosensitive semiconductors and, therefore, used Ti3C2/Cu2O heterostructures for the highsensitivity detection of glucose [122]. The in situ growth of Cu2O on MXenes improves photoelectrochemical performances compared to pure Cu2O. In order to improve the photocurrent conversion efficiency and detection sensitivity of glucose, a Z-type heterostructure based on TiO2/Ti3C2Tx/Cu2O was proposed in addition to the construction of a Schottkyjunction-based PEC sensor [123]. A DNA probe can also achieve the label-free PEC determination of methyltransferase (MTase) for label-free Bi2S3/Ti3C2 PN junctions [124]. For exosomes, the enzyme-induced deposition of CdS on Ti3C2 MXene forms a Ti3C2 MXene/CdS composite, creating a built-in electric field in the tight interface between CdS and Ti3C2 MXene, enabling highly accurate detection [125]. In addition to this, the use of Ti3C2@ReS2 to immobilise DNA probes and perform specific PEC detections of miRNA-141 has excellent performance [126]. For the detection of 5hmCTP on APTES/Ti3C2, the use of antibodies for PEC is also feasible [127]. Chen et al. developed a photoelectrochemical biosensor for the sensitive and selective detection of glutathione based on MQDs [128].

#### *3.3. Other Biosensing*

#### 3.3.1. Wearable Biosensing

The covalent between the M transition metal and the X element in MXenes, the terminal surface groups, and the thickness of the atomic layers resulted in excellent mechanical properties [3,18]. Numerous theoretical findings on the mechanical properties of MXenes have been reported. Kurtoglu et al. predicted a higher elasticity coefficients for various pristine and functionalised MXenes than their precursors due to the greater density of charge density in the Mn+1Xn layer [129]. The excellent mechanical properties provide favourable conditions for fabricating wearable biosensors.

Some studies have used wearable nanoelectronics to detect health-related physiological activities, such as physical or chemical stimulation, micropressure, and changes in physiological signals. Stretchable mechanical properties, high gauge factor, flexible materials integrating flexible bio-electronic interfaces, and miniaturized signalling systems need to be investigated to meet the required sensitivity of sensor devices and to improve the usability of wearable devices [130]. Recently, ultrathin MXenes comprised high-performance materials for stretchable and bendable conductive coatings [131] (Table 7). Conductive and conformal MXenes multilayers can withstand up to 2.5 mm bending and 40% tensile stretching, with recoverable electrical resistances, while maintaining a conductivity of 2000 S m−<sup>1</sup> [10].


**Table 7.** MXene-based wearable biosensors identify units, target, and analytical parameters.

Piezoresistive wearable biosensors are designed to detect weak movements of the human body using stretch changes in materials. The development of MXene-based piezoresistive biosensors has been reported to change the resistance of the biosensor by varying the MXene's layer spacing under external pressure [138]. The biosensor can monitor physical stimulation processes, such as blinking, throat swallowing, and knee bending release through electrical current. Strain sensors were also fabricated by using Ti3C2 MXenes nanocomposites with single-walled carbon nanotubes (SWCNT) obtained by layer-by-layer (LBL) spraying [139]. The multifunctional force-sensing sensor for acoustic monitoring consists of two Au electrodes on a polyethylene terephthalate (PET) substrate at the top and bottom, an intermediate MXenes layer, and a fingerprint structure on the substrate in a combined arrangement [140]. The manufactured sensor is versatile and capable of sensing sound, micro-motion, and acceleration in a single device. This biosensor can be flexibly attached to a person's throat and wrist and is used to detect a person's vocalisation and pulse. The sensor can record relevant peaks when saying "hello" and "sensor" or when detecting a steady heartbeat pulse signal. The biosensor has shown excellent sensitivity in detecting subtle human activity and other weak stresses (Figure 12). It offers a new research direction for portable and wearable sensing devices in biosensing and human behaviour analyses.

**Figure 12.** Schematic diagram of a pressure sensor as a wearable device for physiological signal detection (**a**). Corresponding signals for saying "hello" and "sensor" (**b**). Corresponding signal for the pulse on the medial wrist (**c**). Corresponding signal of the pressure sensor vibrating at different frequencies on the shaker (**d**). Corresponding signals of the pressure sensor vibrating on plate (**e**). Corresponding relationship between current values, displacement, velocity and acceleration in the vibration mode (**f**). Schematic diagram of the density of MXenes at different vibration stages (**g**). Diagram of a sound wave hitting a pressure transducer (**h**). Schematic diagram of the acoustic pressure on the sensor surface associated with sound waves and sound pulses (**i**). Corresponding detection signals and source waves for the two ringtones (**j**–**k**). Reprinted with permission from Ref. [140]. Copyright 2020, Wiley-VCH.

Wearable microfluidic biosensors were originally designed to integrate biological identifiers (enzymes, nucleic acids, enzymes, or cellular receptors) into the sensor operation. Non-invasive biomarker detection platforms via biofluids such as sweat, saliva, tears, or interstitial fluid are more practical [141]. Such wearable sensors provide real-time biochemical information about the wearer's health and offer effective disease detection and body function management [142]. A 3D electrode network electrochemical impedance immunosensor based on loaded laser-burned graphene (LBG) loaded with Ti3C2Tx was fabricated for the non-invasive monitoring of cortisol biomarkers in human sweat [132]. The sensor has a detection limit and linearity of 88 pM and 0.01–100 nM, respectively (Figure 13). In addition, microfluidic wearable biosensors can be used to detect K+, Na+ ions [135,136], glucose, lactate [134,143], pH, and other human biochemical information in biological fluids [137].

**Figure 13.** (**A**) Schematic representation of the Ti3C2Tx MXene-loaded/LBG-based cortisol biomarker assay. (**B**) Patch sensor attached to the body position (**a**), optical image of the wearable patch (**b**) and fabrication sequence of a wearable patch cortisol sensor integrated with a microfluidic system (**c**). Reprinted with permission from Ref. [132]. Copyright 2020, Elsevier.

In various wearable biosensors, sensing electrodes play an essential role in the design of wearable biosensors. MXenes offer the ability to immobilise biomolecules as a sensitive detection platform. Nevertheless, the mechanical friction and deformation of wearable devices against human skin over time leads to mechanical failure and requires a re-structuring of the device. Moreover, the attainment of signal-to-noise ratios and the stability required to achieve this device are highly challenging [133].

#### 3.3.2. Surface-Enhanced Raman Spectroscopy (SERS)

The hydrophilic nature of the MXenes surface provides a good site for Raman labelling. It serves as a potential material for SERS and provides an effective method for the ultra-sensitive determination of targets (Table 8). Sarycheva et al. showed that composites of metals and MXenes could be used as SERS subestrates for rhodamine 6G [144]. Integrating noble metal nanoparticles with MXenes exhibits an empathetic, sensitive SERS response in detecting several common dye molecules. This extends MXenes composites for visible light SERS in the sensor field. A reliable substrate for MXenes/AuNR composites was prepared by Xie et al. with high sensitivities for determining common organic dyes, such as Rh6G, crystalline violet, and peacock green [145]. It can detect organic contaminants and shows high sensitivities for more complex organic pesticides and contaminants. A ratiometric SERS aptamer sensor for ochratoxin A was developed by Zhao et al. 2-Mercaptobenzimidazole-5-carboxylic acid ligands and Au-Ag Janus nanoparticles were used as Raman signal molecules to amplify the SERS signal efficiently [146]. Liu et al. used the Ti3C2Tx-PDDA-Ag NPs hybrid platform as a sensitive and homogeneous biosensor for

the label-free quantification of the biomolecule adenine based on the SERS method [147]. These studies demonstrate that MXenes can be well-suited for SERS.

**Table 8.** MXene-based other biosensors identify units, target, and analytical parameters.


#### 3.3.3. Surface Plasmon Resonance (SPR)

SPR sensing is a non-destructive, label-free, real-time detection method. Nanomaterials modify the sensor's surface to enhance the signal and demonstrate high sensitivities to low concentration targets [148] (Table 9). MXenes-based SPR biosensors for the ultrasensitive detection of carcinoembryonic antigens (CEAs) were fabricated by Wu et al. These SPR biosensors demonstrate good reproducibility and high selectivity in human serum samples, providing a potential method for early clinical diagnoses and cancer surveillance [149]. Their group also constructed SPR biosensors for CEA using amino-functionalised MXenes [150]. MXenes were assembled on Au films to immobilise monoclonal anti-CEA antibody sensing materials covalently. MXenes were used as a substrate for binding hollow gold nanoparticles (HGNPs), and they were modified with SPA. MXenes/HGNPs/SPA/Ab2 nanocomplexes act as signal enhancers for SPR-sensing components. The sensor offers a wide linear detection range, ultra-low detection limits, and good selectivity for CEA in human serums. Chen et al. anchored targeting aptamers with thiol-modified niobium carbide MXene quantum dots and could specifically bind the SARS-Cov-2 N gene, resulting in a change in the SPR signal for laser irradiation at a wavelength of 633 nm [151]. The above studies indicate the development of MXene-based biomarker sensing chips or devices in the field of SPR to broaden the area of MXenes biosensing applications.

**Table 9.** MXene-based SPR biosensors identify units, target, and analytical parameters.


#### **4. Conclusions and Outlook**

From its discovery to the present moment, MXenes, an emerging two-dimensional biosensing material, has achieved unprecedented rapid development. Although the development time is relatively short, synthesis methods of MXenes are constantly innovating and developing towards the direction of green environmental protection, simplicity and convenience, and controllable surface groups. MXenes are known for their tunable surface groups; good hydrophilicity and biocompatibility; excellent mechanical, electrical, and optical properties; and specific morphological structures. They have become a focus of research in electrochemical and optical biosensing.

HF synthesis is undoubtedly convenient and quick, giving access to abundant hydrophilic surface groups. However, using F− can cause a certain amount of pollution to humans, the environment, and even the ecosystem. For the synthesis of MXenes, synthetic research has always progressed in the direction of less or no fluorine. Different synthetic methods and raw materials can adjust different surface groups. The structure determines the properties, and mechanical, hydrophilic, biocompatible, electrical, and optical properties can be adjusted according to different research needs using different synthetic routes.

From its large specific surface area, high electrical conductivity, large surface groups, and fast electron transfer properties, MXenes are often used as modification and immobilisation materials in electrochemical biosensors to improve electrocatalytic performances and detection sensitivities and to reduce redox potentials in order to obtain high-performance composites. However, most of the current MXene-based electrochemical biosensors use composite materials adsorbed on MXene's surface. There are relatively few reports on surface groups on MXenes as the adsorption media for the adsorption of biomolecules to prepare electrochemical biosensors. With the development of the tunable functionalisation of MXenes surface groups, we believe there is more significant potential for the direct binding of MXenes surfaces to biomolecules, such as proteins and nucleic acids. Although optical biosensors have been less studied than electrochemical biosensors, optical sensing is also a vital detection strategy in biosensing. By benefitting from the fluorescence quenching effect of MXenes and MQD's fluorescence enhancement effect, photoluminescence has been reported relatively more often in optical biosensing work. From the above studies, it can be concluded that nucleic acid-based biosensors are less selective than enzyme and immunebased biosensors. However, nucleic acids are smaller than antibodies and enzymes, so nucleic-acid-based biosensors have a higher density of surface modifications and have a higher sensitivity and reproducibility.

In contrast, SPR, SERS, and other optical sensing have been reported less often. Even so, the sparseness and singularity of functional groups on MXene's surface led to the reduced binding of biomolecules or other organic molecules. Meanwhile, the study of MQDs with high fluorescence efficiency, quantum yield, and different luminescence wavelengths has become an urgent problem.

About the main challenges mentioned above, we believe that several aspects can be studied for future developments and trends in MXenes biosensing.


with different wavelengths. MQDs can be extended to cell imaging, photothermal therapy, and other biomedical tissue applications.

**Author Contributions:** Conceptualization, D.L. and H.Z.; writing—original draft preparation, D.L.; writing—review and editing, D.L., X.Z. and H.Z.; project administration, Y.C. and L.F.; supervision, L.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (No. 21705106) and the National Science Foundation of China (No. 22177067); the Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning (No. TP2016023); and the Shanghai Sailing Program (No. 20YF1413000).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### MDPI

St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Biosensors* Editorial Office E-mail: biosensors@mdpi.com www.mdpi.com/journal/biosensors

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-7875-0