**1. Introduction**

Detection of pathogenic and spoilage bacteria is still a major concern to clinical, agri-food, and environmental agencies and laboratories [1]. The leading challenge is the detection speed [1]. Since the contamination level of bacteria may be relatively low and the sample matrices can significantly influence accurate and reproducible detection, extensive sample preparation steps are always required to separate the targeted bacteria from the sample matrices along with pre-enrichment [2,3]. Because the detection includes all the times starting from obtaining the samples to the signal readout, both separation and bacterial enrichment account for most of the times for bacterial detection rather than the final real detection using an instrument or a sensor [4]. For example, the conventional plating assay will take several days to confirm the growth of the targeted bacterial colony [5]. In comparison, molecular-based detection methods, such as polymerase chain reaction (PCR), requires relatively less time than the plating assay but still cannot fully avoid separation and bacterial pre-enrichment [6]. Recently, matrix assisted laser desorption ionization time-of-flight (MALDI-TOF) spectrometry has attracted considerable interest for the rapid identification of pathogens by profiling bacterial proteins from the whole cells [7]. However, this method is not suitable for characterizing a mixed sample [8] and still requires the priori cultivation and sample preparation procedure [9]. An alternative method is surface-enhanced Raman spectroscopy (SERS), an advanced Raman spectroscopic technique that enhances the vibrational modes of molecules adsorbed on or in the vicinity to the surface of metal nanoparticles. SERS provides rapid, ultra-sensitive and accurate detection with minimum requirement for sample handling and preparation.

Antibiotic resistance of pathogenic bacteria is still a leading concern to clinics as well as agri-food and veterinary medicine [10]. The key battle is to perform an accurate diagnosis of the pattern of bacterial antibiotic resistance in an early manner. Otherwise, only the broad-spectrum antibiotics can be used to treat this type of bacterial infections [11]. As aforementioned, the conventional microbiological testing, such as the determination of minimum inhibitory concentration (MIC) using the broth microdilution method, is highly time-consuming. Besides, PCR-based testing of the targeted antibiotic-resistant genes requires highly trained personnel and has a potential risk of cross contamination [12,13]. Another major limitation of this approach is that the presence of the resistance genes may not necessarily confer to the clinically relevant phenotypic resistance of bacteria [14]. Microarray offers the ability to detect a broad range of resistance genes present in the bacterial isolates with high sensitivity and specificity. However, similar to the PCR-based method, results obtained from microarrays may not always correlate to the phenotypic resistance [14]. Although MALDI-TOF mass spectroscopy can potentially differentiate the resistant and susceptible isolates based on the spectral features [7], it requires additional chemicals as the matrix for the performance of MALDI [14]. Alternative technology that can detect and characterize bacterial antibiotic resistance is therefore highly required. SERS is a powerful biochemical fingerprinting technique as it can accurately reflect the macromolecular profiles and changes that occur within the bacterial cells due to the action of the antibiotics [15].

In this mini-review paper, we will evaluate the use of SERS coupled with chemometrics as a tool to detect the trace level of antibiotic-resistant bacteria and characterize the mechanism of bacterial antibiotic resistance in an ultra-fast manner. The recent progress in this research area will be summarized and discussed mainly focusing on the following three perspectives: (1) the nanomaterials that can be used as the SERS substrates for sensing a low concentration of bacterial cells; (2) tandem-SERS technology to detect antibiotic-resistant bacteria in a sample matrix; and (3) characterizing the mechanism of bacterial antibiotic resistance and susceptibility using SERS and chemometrics.

### **2. Surface-Enhanced Raman Spectroscopy (SERS) for Sensing Trace Level of Bacteria**

### *2.1. Mechanism of SERS*

SERS is a derivative of Raman spectroscopy with the aid of nanomaterials. Numerous research studies have been conducted during the past four decades about using SERS for trace detection of the targeted analytes [16–23]. Different from the conventional Raman spectroscopic technology, SERS signal can be significantly enhanced due to both electromagnetic enhancement and chemical enhancement, with the former being the dominant contributor [24]. Electromagnetic enhancement is generated from the localized surface plasmon resonance (LSPR) in the vicinity of the nanostructured surface of noble metals, such as silver and gold [25,26]. Highly localized regions of amplified electromagnetic fields caused by LSPR are called "hot spots", which usually occurs in the gaps, crevices, or sharp vertices of supporting plasmonic materials (Figure 1a). In comparison, chemical enhancement is due to the electron transfer between the analyte molecule and the surface of the nanostructure when the energy of the incident light matches the electron transfer energy (Figure 1b) [27]. This will lead to the change of molecular polarization and subsequently enhance the Raman signal approximately 100 times. Theoretically, total SERS enhancement factors may approach to ~10<sup>14</sup> depending on the nanomaterials used. For additional details, the authors are encouraged to refer to serial publications from the Van Duyne research group [27–30].

**Figure 1.** Two mechanisms contributed to surface-enhanced Raman spectroscopy (SERS). (**a**) Electromagnetic enhancement of SERS-active silver nanoparticles. SERS "hot-spot" is generated in the gap between two close nanoparticles. (**b**) Chemical enhancement resulting from electron transfer between analytes and the surface of nanoparticles. Reproduced with permission [31]. Copyright Royal Society of Chemistry, 2014. Reproduced with permission [32]. Copyright Elsevier B.V., 2017.

### *2.2. SERS-Active Substrates for Bacterial Detection*

Because SERS can reach to single molecule detection, it has been widely applied for the detection of various analytes in an ultra-fast manner (e.g., a few seconds to less than a minute). In general, the reproducibility of the SERS signal is getting worse along with the increase of the size of the analyte [33]. For example, it is extremely challenging to harvest a reproducible SERS signal for a bacterial cell than that of a small chemical molecule, such as antibiotics and pesticides [34]. Although successful discrimination of bacteria by using SERS was reported by different research groups [15,35,36], the real world application is still extremely challenging, such as the low concentration of the targeted bacteria in the sample and a relatively large amount of interference sample components. Therefore, researchers have been developing various types of SERS-active substrates to enhance the signal intensity as well as generate more reproducible SERS signals for different biological samples, such as bacteria and viruses. Both "top-down" and "bottom-up" methods have been used for the synthesis of SERS-active substrates [37]. For the "top-down" method, large multi-dimensional materials are reduced to ideal nanoscale structures using direct fabrication process [38]. In comparison, the "bottom-up" method refers to the development of complex nanoscale structures from simple molecules or atoms [39].
