**1. Introduction**

Point-of-care diagnostics (POCD) for health monitoring involves the evaluation of indices from human health through tests performed outside of the clinical laboratory, typically right at the site of patient care [1–4]. While the advantages of POCD such as rapid diagnosis, operational efficiency, and costs have been well established [2], widespread implementation of this technology has not ye<sup>t</sup> been achieved [5]. One of the main limitations of POCD is the requirement of unprocessed test samples drawn from patients to provide the desired sensitivity and specificity for conducting reliable assays in situ. However, most of the currently validated clinical assays are carried out using plasma, prepared from whole blood samples inside sufficiently equipped laboratory conditions. Here, we develop a biosensor platform that could be reliably used for instantaneous detection of target molecules from non-invasively obtained bodily fluid samples for real-time point-of-care testing applications.

Although whole blood assays are generally considered as the gold standard for bioanalysis, such an approach may not be convenient to implement under all circumstances. In general, high-risk

patients under intensive care or long-term hospitalization su ffering from extremely sick conditions may not always be able to bleed enough quantity of blood. Possible nerve damage during venipuncture [6] could render the procedure extremely painful, traumatic and psychologically disturbing for the patients, such that even skilled phlebotomists could face limitations to extract enough quantity of blood under such circumstances.

One strategy to overcome the trauma associated with such painful venous blood draw procedures could be to adopt a technique that is less invasive and utilizes smaller volumes, possibly through a finger prick as implemented with some of the commercially available handheld blood glucose monitors. However, a limitation with this approach is that reduced sample volumes call for greater sensitivity of the measurement setup and may also not always contain the necessary distribution of the target molecules under all circumstances. This could lead to inconsistent readings and false negative results [7]. Furthermore, common clinical problems with whole blood assays such as hemolysis or coagulation could also result in an inaccurate representation of the analyte concentration [8].

An alternate strategy would be to eliminate the necessity for whole blood, and use other test samples for carrying out the required bioanalysis. Among the bodily fluids generally used in clinical assays (shown in Figure 1), those commonly available through non-invasive extraction procedures include urine, saliva [9] and sweat [10]. While inducing sweat out of hospitalized patients is practically di fficult, urinary extraction may also necessitate additional diuretics which could directly interfere with the measurands, aside from inducing other significant side e ffects [11]. Considering these factors, saliva presents a viable alternate to the traditional biofluids used in biochemical analyses.


**Figure 1.** List of commonly available bodily fluids showing their nature of the extraction procedure and general feasibility of potassium detection from those fluids.

Saliva is considered as a filtrate of the blood representing the physiological state of the body. Given the ability of molecules from blood to di ffuse into saliva, salivary diagnostics is increasingly being recognized as an equivalent to serum analysis [12]. Research studies using saliva as a bio-diagnostic fluid includes works such as evaluation towards dialysis needs in renal failure patients [9] and on-site analysis for biochemical factors [13]. The incorporation of microfluidic methods in bioassays of saliva

has further reduced sample and reagen<sup>t</sup> consumption, and has decreased the overall assay times [13]. Research studies in salivary diagnostics involving microfluidics include measurement of C-reactive protein using an fluorometric immunoassay [14], on-chip polymerase chain reaction (PCR) system for rapid fluorometric detection of genetic deletion [15], paper-based device for quantification of the nitrate concentration [16], detection of thiocyanate through surface-enhanced Raman scattering (SERS) [17], and spectrometric absorbance detection of NH3 and CO2 in saliva as a biomarker for stomach cancer [18]. In addition, most of the reported microfluidic systems for salivary diagnostics included integrated optical sensing for their bio-detection scheme [19].

In this work, we present the development of a simple, hybrid integrated optical microfluidic biosensor for rapid analysis of saliva, and we have demonstrated the application of the proposed biosensor format by detecting the presence of potassium from whole, untreated salivary samples. Inspired by the age-old concept of colorimetry published as early as the first half of the 20th century [20,21], we hereby show that potassium can be detected from human saliva samples through the principle of optical absorption. A commercially available potassium detection reagen<sup>t</sup> mixed with saliva samples creates turbidity based on the potassium concentration, which could then be measured through the absorption of light passing through this turbid path. We have implemented this optical sensing principle on a lab-on-a-chip platform by integrating colorimetric detection within a microfluidic system that facilitates transportation and handling of the analytes and the reagents. Homogeneous mixing of fluids inside the microfluidic system has been achieved through the integration of piezo-actuated acoustic micromixing. The results of this work show that the proposed hybrid integrated device can be applied to real-time optical biosensing using whole, unprocessed samples, which could be extremely useful for automated point-of-care testing applications.

#### **2. Materials and Methods**

#### *2.1. Development of the Integrated Optical Biosensor*

This section provides the details of the design, fabrication, integration and packaging of our hybrid integrated optical microfluidic setup. The schematic illustration of the processes is shown in Figure 2.

#### 2.1.1. Device Design and Fabrication

The biosensor (schematically shown in Figure 2A) consists of a microfluidic system with two inlets which accommodates two different types of fluids. The fluids are transported to a piezo-actuated (acoustofluidic) micromixer unit, and upon mixing are subsequently transported to the optical detection unit. The configuration of the bulk acoustofluidic micromixer unit was adopted from [22,23]. The optical detection unit consists of SU8 waveguides core integrated onto poly dimethylsiloxane (PDMS) (cladding) through the inlet and outlet waveguide channels (as indicated in Figure 2A). Convex curvatures at the end of the optical channel enable convergence of light into the fluidic channel. Two different types of optical waveguide assembly were designed. In the configuration used in the present work, the output waveguide was fabricated co-axially with the input waveguide. This configuration allowed direct coupling of the input light onto the output waveguide, as required in optical absorption measurements.

To fabricate the devices on PDMS (fabrication and integration processes schematically shown in Figure 2B), silicon master molds with patterned SU8 (100 μm thickness) were fabricated using standard soft lithography process. Poly dimethylsiloxane (Sylgard 184, Dow Corning, Midland, MI, USA) with prepolymer to curing agen<sup>t</sup> volumetric ratio of 10:1 was mixed, degassed and poured onto the silicon mold. The setup was left to cure at 70 ◦C for four hours, after which PDMS was peeled off the mold and diced to create the individual devices. To create optical waveguides, the PDMS devices were exposed to oxygen plasma (2 min, 200 mTorr pressure, 20 cubic centimeter (ccm) flow rate, and 60 W Radio Frequency (RF) power), and thereafter SU8-5 (Microchem, Westborough, MA, USA) with viscosity 290 cSt, and refractive index ~1.6 was allowed to fill in the optical channels through capillary flow. The photoresist was cured by flood exposure to ultraviolet (UV) light (12 mW/cm2) for 60 s, and the devices were diced thereafter to enable fiber attachment with the SU8 waveguides.

**Figure 2.** (**A**) Schematic working principle of the acoustofluidic micromixer integrated optical sensing for colorimetric detections. The liquid test specimen (fluid 1) and its corresponding chromogenic agen<sup>t</sup> (fluid 2) are mixed using acoustic waves generated by a piezoacuator, and the mixture is transported into the optical detection unit, wherein the extent of color change induced due to the presence of target analyte in the sample is sensed by the absorption of light passing through the sample. (**B**) Schematic illustrayion of the fabrication and packaging of the hybrid integrated biosensor (**C**) Scanning electron microscope (SEM) images showing the microfluidic channel and the optical channels of the poly dimethylsiloxane (PDMS) device (i) before and (ii) after integration with SU8 (Scale bars represent 500 m) (**D**) Micro-positioning setup for precise alignment and coupling of fibers with the SU8 waveguides. (**E**) Images of the functional integrated optical microfluidic biosensor (i) with one fluid inlet and (ii) two fluid inlets.

#### 2.1.2. Integration and Packaging

Coupling of the input and the output fibers were carried out under a stereo microscope. FC connectorized tapered lens-ended fiber (OZ Optics, Ottawa, ON, Canada) which gives a spot size of 5 μm, was connected to a broadband laser light source (Ocean Optics, Largo, FL, USA) for optical input. The output collector fiber was SMA end connectorized and coupled to a handheld spectrometer (USB 2000, Ocean Optics, Largo, FL, USA). Fiber strippers and precision cleavers (Newport, RI, USA) were used to remove the buffer layer and cladding around the fiber. The input and the output fibers were positioned in separate five axis fiber positioners on supporting V-grooves, so that each of the fibers and the waveguides can be maneuvered and coupled appropriately for acquiring the maximum optical signal. Maximum optical light coupling into the fiber was ensured by fine-tuning the position of the fiber and by observing the maximum signal from the output fiber coupled into the spectrometer. Thereafter, UV index matching gel (NOA60, Norland International Inc., Lincoln, NE, USA) was applied at the tip of the fiber and the setup was exposed to UV for 60 s to bond the fiber with the SU8 waveguides.

The device was then treated with atmospheric plasma using a handheld plasma cleaner (Plasmaetch, Carson City, NV, USA) and the channels were sealed using a 100 μm thick glass coverslip. Piezoceramic discs (T216-A4NO-173X, Piezo Systems Inc., Cambridge, MA, USA) used for generating the acoustic waves were attached to this glass diaphragm using silver conductive epoxy (also used as the bottom

electrode). The top electrode was soldered to the piezoceramic. For ease of handling, the device was attached with a polycarbonate support using double-sided adhesive tape, without damaging the optical fibers. A through hole drilled on the polycarbonate prior to the attachment of the optical-microfluidic chip accomomodates the piezo actuator inside the cavity.

#### *2.2. Chemicals and Reagents*

Integrated optical-microfluidic characterization experiments were carried out with de-ionized (DI) water and ethanol (Sigma Aldrich, St. Louis, MO, USA). Prior to introducing fluid samples into the system, the microfluidic channels were flushed with ethanol in order to remove any contaminants or air bubbles. Initial micromixing characterization experiments were conducted by staining the working fluids (DI water or ethanol) with standard food coloring dyes. Preliminary experiments for the characterization of optical absorption with integrated micromixing were conducted using glucose enriched RPMI 1640 cell culture media (A1049101, Thermofisher, Waltham, MA, USA) as the working fluid. Herein, the glucose present in cell culture media was tested to produce an expected colorimetric reaction with Benedict's reagen<sup>t</sup> prepared according to the protocol described by Cochran et al. [24]. For colorimetric salivary analysis experiments, potassium colorimetric assay kit (E-BC-K279) was procured from Cedarlane Labs (Burlington, NC, USA). Whole samples of unstimulated saliva were collected from volunteers in centrifuge tubes by a simple spitting method, and the specimens were vortexed thoroughly. Prior to the experiments, protein precipitant and chromogenic agents were prepared as described in the protocol prescribed by the manufacturer. 20 μL of the saliva samples were mixed with 180 μL of the protein precipitant and centrifuged at 1100 g for 10 min. Thereafter, the supernatant was used in the experiments, mixed with a volumetrically consistent chromogenic agen<sup>t</sup> for the colorimetric optical absorbance measurements.

#### *2.3. Integrated Testing and Measurement*

The piezoelectric system was driven by an external signal generator (33120A, Hewlett Packard, Palo Alto, CA, USA) and an amplifier (PCB Piezotronics, Depew, NY, USA) where a sinusoidal signal from the signal generator was amplified 20-fold by the amplifier. Typically, the operating voltage of 100 Vp-p was used for piezo-actuation. For continuous flow experiments, liquids were injected at the same flow rates inside the microchannel using syringe pump (KDS-210, KD Scientific, Holliston, MA, USA). In all other cases, liquids were manually pipetted into the microfluidic channels.

The mixing e fficiency was calculated based on the change in the pixel values of the respective color fluids.

$$\text{Mixing efficiency } (\%) = \text{(Initial pixel count-final pixel count)} (\text{[initial pixel count]} \times 100 \qquad (1)$$

Optical signals were collected using a handheld spectrometer (USB2000, Ocean Optics, USA). The optical absorbance of the specimens measured by the spectrometer is given by the formula

$$A\_{\lambda} = \ -\log g\_{10}\left(\frac{S\_{\lambda} - D\_{\lambda}}{R\_{\lambda} - D\_{\lambda}}\right) \tag{2}$$

where,

λ—Wavelength of light used

*A*λ—Absorbance

*S*λ—Intensity of light passing through the sample

*D*λ—Dark intensity

*R*λ—Intensity of light passing through a reference medium.

The dark intensity was recorded by measuring the optical signal intensity when the light source was turned o ff. For potassium measurement experiments, the chromogenic agen<sup>t</sup> was used as the reference medium to measure the reference intensity.

#### *2.4. Imaging and Statistical Analysis*

Images were recorded using an OFV-A-534-Cax video camera inbuilt with the single point Laser Doppler Vibrometer (LDV, Polytec, Detroit, MI, USA). Scanning electron microscopy (SEM, SU3500, Hitachi Hi-Technologies, Tokyo, Japan) was conducted using standard protocols for variable-pressure imaging mode (3.0 kV, 30 Pa,) allowed SEM observations of the PDMS devices without the need for additional sample manipulation or conductive coating. The images were processed using ImageJ (National Institute of Health, Bethesda, MD, USA), following standard protocols. All statistical comparisons were made using one- or two-way analyses of variance (ANOVA) with Tukey post-hoc comparisons (Prism; GraphPad Software, La Jolla, CA) with *p*-values < 0.05 considered significant, and graphical data reported as means ± standard error for at least n = 3 experiments.
