Advances in Antimicrobial Resistance Monitoring Using Sensors and Biosensors: A Review

: The indiscriminate use and mismanagement of antibiotics over the last eight decades have led to one of the main challenges humanity will have to face in the next twenty years in terms of public health and economy, i.e., antimicrobial resistance. One of the key approaches to tackling antimicrobial resistance is clinical, livestock, and environmental surveillance applying methods capable of effectively identifying antimicrobial non-susceptibility as well as genes that promote resistance. Current clinical laboratory practices involve conventional culture-based antibiotic susceptibility testing (AST) methods, taking over 24 h to ﬁnd out which medication should be prescribed to treat the infection. Although there are techniques that provide rapid resistance detection, it is necessary to have new tools that are easy to operate, are robust, sensitive, speciﬁc, and inexpensive. Chemical sensors and biosensors are devices that could have the necessary characteristics for the rapid diagnosis of resistant microorganisms and could provide crucial information on the choice of antibiotic (or other antimicrobial medicines) to be administered. This review provides an overview on novel biosensing strategies for the phenotypic and genotypic determination of antimicrobial resistance and a perspective on the use of these tools in modern health-care and environmental surveillance.


Antimicrobial Resistance
In recent years, antimicrobial resistance (AMR) has received a great deal of attention because of the impact resistant microorganisms have on public health. Consequently, it has been declared as a global public health concern [1][2][3][4][5][6][7]. According to the World Health Organization (WHO), antimicrobial resistance occurs when bacteria adapt and grow in the presence of an antimicrobial (AM) agent [8]. It can also be described as a phenomenon when microorganisms (bacteria, viruses, fungi, and parasites) are no longer affected by an AM to which it was previously sensitive, as a result of mutation of the microorganism to evade the effect of the antimicrobial or of the acquisition of the resistance gene [9,10]. It is estimated that by 2050, human deaths related to AMR will amount to 10 million, resulting in an economic cost of about 100 billion dollars [11]. Although, there is compelling economic justification for the development of new generations of antibiotics by 2030 [12], the acquisition rate of biological resistance to new drugs is now a major consideration in preventing their introduction [7,13,14].
The WHO has established a list of antimicrobials (limited to antibiotics) that are important in human medicine and in veterinary practice. The top priority antimicrobials are cephalosporins (3rd, 4th, and 5th generation), glycopeptides, macrolides and ketolides, The WHO has established a list of antimicrobials (limited to antibiotics) that are important in human medicine and in veterinary practice. The top priority antimicrobials are cephalosporins (3rd, 4th, and 5th generation), glycopeptides, macrolides and ketolides, polymyxins and quinolones [4,15]. In 2017, the WHO published a list of priority pathogens resistant to antibiotics; on this list, 12 families of bacteria are selected for the research and development of new antibiotics due to the growing need for innovative treatments against multi-resistant bacteria in hospital environments ( Figure 1) [16,17]. One of the leading causes contributing to the global increase in AMR is the empirical use of broad-spectrum AMs in clinical and veterinary practices [18]. For the early detection of microorganisms, for the prevention of diseases and spread of AMR, it is essential to have well-equipped laboratories specialized in the diagnosis of infections [19]. It is difficult to meet this requirement in developing countries because of the lack of funds, shortage of trained personnel, and lack of the required infrastructures. According to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines, reliable antimicrobial resistance tests require phenotypic evaluation, i.e., experimental procedures to monitor the microbial growth in presence of AM. The routine procedure to determine if a pathogen is resistant is through isolation, including enrichment culture (i.e., urine culture) and plate cultivation, identification and finally antimicrobial susceptibility testing (AST) using different AM loads. The results can be expressed as the lowest concentration of the antibiotic that prevents visible growth, i.e., the minimal inhibitory concentration (MIC) [20]. ASTs based on culturing methods require from 24 to 72 h. This time lapse can be crucial for the patient [21,22]. Another disadvantage of culturing-based techniques is that trained personnel is required [23]. The disk diffusion test, introduced, in the 1960s by Bauer and Kirby to determine the susceptibility to antibiotics in Staphylococcus strains, is still one of the most frequently used phenotypic method for AST [24,25]. Broth microdilution assays is another commonly used technique [26,27], based on visually identifiable growth. Automated commercially available AST instruments have helped in reducing the analysis times to less than a day [28]. Examples of automated instruments include the Vitek systems (BioMèrieux), the BD Phoenix instruments, the Alfred 60AST system (Alifax), and the MicroScan WalkAway system (Beckman Coulter) [18,25,[29][30][31].
Other candidates for AST are flow cytometry (FC) and isothermal microcalorimetry (IMC). FC is an optical technique used for cell counting and the detection of biomarkers One of the leading causes contributing to the global increase in AMR is the empirical use of broad-spectrum AMs in clinical and veterinary practices [18]. For the early detection of microorganisms, for the prevention of diseases and spread of AMR, it is essential to have well-equipped laboratories specialized in the diagnosis of infections [19]. It is difficult to meet this requirement in developing countries because of the lack of funds, shortage of trained personnel, and lack of the required infrastructures. According to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines, reliable antimicrobial resistance tests require phenotypic evaluation, i.e., experimental procedures to monitor the microbial growth in presence of AM. The routine procedure to determine if a pathogen is resistant is through isolation, including enrichment culture (i.e., urine culture) and plate cultivation, identification and finally antimicrobial susceptibility testing (AST) using different AM loads. The results can be expressed as the lowest concentration of the antibiotic that prevents visible growth, i.e., the minimal inhibitory concentration (MIC) [20]. ASTs based on culturing methods require from 24 to 72 h. This time lapse can be crucial for the patient [21,22]. Another disadvantage of culturing-based techniques is that trained personnel is required [23]. The disk diffusion test, introduced, in the 1960s by Bauer and Kirby to determine the susceptibility to antibiotics in Staphylococcus strains, is still one of the most frequently used phenotypic method for AST [24,25]. Broth microdilution assays is another commonly used technique [26,27], based on visually identifiable growth. Automated commercially available AST instruments have helped in reducing the analysis times to less than a day [28]. Examples of automated instruments include the Vitek systems (BioMèrieux), the BD Phoenix instruments, the Alfred 60AST system (Alifax), and the MicroScan WalkAway system (Beckman Coulter) [18,25,[29][30][31].
Other candidates for AST are flow cytometry (FC) and isothermal microcalorimetry (IMC). FC is an optical technique used for cell counting and the detection of biomarkers [32]. Since the mechanism of action of antibiotics induces modifications in some morpho-functional and physiological characteristics of the cell, it is possible to determine

Sensors and Biosensors for Phenotypic AMR Detection
The previous decade has seen a notable increase in the development of chemical sensors and biosensors for phenotypic AMR detection [18,[74][75][76][77]. Herein, sensors and biosensors for AMR testing have been classified according to the transducer used, and thus are discussed as magnetic, mass, mechanical, thermal, optical, and electrochemical. Mechanical sensors and biosensors are devices sensitive to physical changes in mechanical properties and capable to convert interactions and processes occurring at their surface into measurable mechanical properties [76]. Optical sensors are based on the measurement of reflectance, chemiluminescence, absorbance, fluorescence, Raman scattering, and surface plasmon resonance (SPR) [78][79][80][81]. SPR is one of the most widely used optical label-free platforms [24,82]. Electrochemical (bio)sensors can measure potentiometric, amperometric, voltametric, or impedimetric signals [51,75,[83][84][85][86].

AST Magnetic, Mass, and Mechanical (Bio)Sensors
The magnetic properties of some materials have been explored for AMR monitoring. Asynchronous magnetic bead rotation (AMBR) is one of these. In this technique, the rotational rate of magnetic beads in a rotating magnetic field is monitored. The measurement is based on the changes in the rotation frequency given by the physical properties of the bead itself (shape and volume) and its environment (viscosity and bacterial load) [87]. The number of bacterial cells present in a sample affects the rotation frequency. In presence of an antibiotic, the MIC of the antibiotic can be estimated [88]. Kinnunen et al. measured the bacterial growth of uropathogenic E. coli isolate and its response to the presence of streptomycin and gentamicin by coating magnetic beads with specific antibodies for the bacteria. Bacterial growth alters the drag of the rotating magnetic bead cluster due to the change in viscosity (Figure 2A) [89]. With this device, the MICs of streptomycin (8 µg/mL) and gentamicin (2 µg/mL) were obtained in 2 h.
Other devices that also use beads are those based on the measurement of the Brownian motion. Wang et al. developed a self-powered sensor based on vancomycin-modified microbeads and the use of their Brownian motion for rapid AST and in situ monitoring of microorganisms [90]. The proposed approach is based on the fact that vancomycin is a β-lactam antibiotic that possesses a similar structure to the enzyme dd-transpeptidase that catalyzes the synthesis of bacterial cell walls. Cell walls of almost all bacteria contain d-Alad-Ala residues. Gram-negative bacteria are covered with an extra layer of outer membrane, however d-Ala-d-Ala ligases still are exposed to the outside environment through some defects [91]. Thus, the vancomycin-modified microbeads can selectively capture bacteria instead of normal cells or proteins and inhibit their wall synthesis. When the functionalized microbeads capture the bacteria, giving rise to a change in diffusivity, the Brownian motion is weaker; the images of the analysis were then obtained by means a fluorescent microscope. In this study, E. coli and S. aureus strains were used, and gentamicin was added for AST. It was determined that concentrations below 2 µg/mL were insufficient to inhibit the bacteria; the determined MIC was 4 µg/mL with a detection limit of 10 5 CFU/mL in an analysis time of 2 h and a volume of 5 µL.
Quartz crystal microbalance (QCM) allows piezoelectric/acoustic dynamic measurements. This technique measures the resonant frequency change from mass changes on the transducer surface. The surface of the transducer is a piezoelectric material; quartz is commonly used for its low cost and stability against thermal, chemical, and mechanical stresses [92]. The quartz crystal is coated with a gold electrode; antibodies, DNA or RNA capture probes, enzymes or nanostructures that are specific and highly sensitive to the substance to be detected can be immobilized on the gold surface [93,94]. Reyes et al. reported a rapid and highly sensitive biosensor for monitoring the effects of antibiotics on resistant or susceptible strains of E. coli and S. cerevisiae [95]. A magnesium zinc oxide (MZO) nanostructure-modified quartz crystal microbalance (MZOnano-QCM) was used ( Figure 2B). The biosensor was applied to measure the sensitivity of the strains to ampicillin and tetracycline for E. coli, and amphotericin-B and miconazole for S. cerevisiae through the device's time-dependent frequency shift and motional resistance [95]. An innovative alternative for the determination of AST was proposed by Guliy et al. [96]. The principle of operation of the sensor is based on the fact that the attenuation of an acoustic wave propagating in a piezoelectric plate and its velocity depend on the electrical boundary conditions on the surface of the plate [97]. In the proposed approach the lower part of a lithium niobate plate was excited with interdigital transducers (IDT), the signal generated was determined in the form of an acoustic wave with horizontal polarization; the suspension of E. coli cells, placed on top of the lithium niobate plate, led to the appearance of pronounced resonance peaks. Subsequently, the bacterial suspension was exposed to ampicillin and measurements of the depth and frequency of the resonant peaks were made; at 2 µg/mL of antibiotic, a change in the depth and frequency of the resonance peaks can be observed because ampicillin modifies the properties of the cytoplasmic membrane, favoring the release of species to the suspension by modifying its conductivity; this increase in conductivity decreases the depth and frequency of the resonant peaks of the sensor. This device has the advantage of short analysis times (10-15 min). However, it is affected by temperature changes, causing fluctuations in the resonance frequency.
Mechanical sensors have been proposed for AST [76,98]. The forces produced by the interaction of the target species with the receptors immobilized on the cantilever determine a change in the cantilever resonance frequency that behaves as a harmonic oscillator. For the determination of bacteria and their susceptibility or resistance, the cells adhere to the cantilever, and according to the added mass and position, the change in resonance frequency is monitored [74,99,100]. Currently, the atomic force microscope (AFM) is one of the main nanomechanical techniques used in AST due to its ability to detect displacements in the range of 0.1 Å and a resolution of microseconds [98,101]. Using an AFM system, Stupar et al. developed a methodology capable of detecting nanomotion of E. coli bacteria and observing their behavior when exposed to ciprofloxacin, ampicillin, and ceftriaxone [102]. The bacteria themselves induce a fluctuation in the sensor; when exposing the bacteria to antibiotics, a change in the fluctuations is rapidly recorded ( Figure 2C).  Table 2 summarizes some of the most recent works for the determination of susceptibility through magnetic, mass, and mechanical sensors. Mainly, these sensors were tested in bacterial culture. Thus, in Table 2, detection times after the isolation and identification of the strains are detailed, as well as the MICs of the antibiotics used for susceptibility or resistance measurement and the detection limits. Single  Table 2 summarizes some of the most recent works for the determination of susceptibility through magnetic, mass, and mechanical sensors. Mainly, these sensors were tested in bacterial culture. Thus, in Table 2, detection times after the isolation and identification of the strains are detailed, as well as the MICs of the antibiotics used for susceptibility or resistance measurement and the detection limits.
Colorimetric sensors offer a number of advantages including: simple procedure, low cost, and the possibility of observing color changes with the naked eye or with simple instrumentation in the case of quantitative analysis [70,118]. The use of chemical species that changes color with respect to bacterial growth, such as Resazurin [19,119], or the use of enzymes, such as Horseradish Peroxidase (HRP), that modifies substrates according to the number of microbial cells, can help to measure susceptibility in real-time [70].
In the approach of Sun et al. [120], p-benzoquinone was used to monitor the bacterial concentration and specifically distinguish E. coli from four other common bacteria, namely, Enterococcus faecalis (E. faecalis), Staphylococcus aureus (S. aureus), Streptococcus mutans (S. mutans), and Salmonella pullorum (S. pullorum). A visible color change, evaluated without any complex instruments but with a simple smartphone, is used for the determination of the concentration of the bacteria.
Many (bio)sensors are based on fluorescent labels including nanomaterials [70,121,122]. Nazemi et al. developed a rapid approach to evaluate antibiotic susceptibility of E. coli based on photoluminescence emission of GaAs/AlGaAs quantum well (QW) biochips [123]. Bacterial cells were captured on the surface of the QW, delaying the photoluminescence maximum. The biochips were functionalized with polyclonal antibodies for E. Coli. After 4 h of incubation (LB broth, penicillin, or ciprofloxacin), the position of photoluminescence maximum was measured. The photoluminescence maximum depends on the photocorrosion process due to the immobilized bacteria on the surface of the GaAs/AlGaAs biochips ( Figure 3A). When bacterial cells are affected by antibiotic exposure, peak photoluminescence occurs faster than in presence of bacteria in LB.
Tang et al. [124] studied the change in pH due to the rapid accumulation of metabolic products during cell growth. A microfluidic pH sensor manufactured by integrating pHsensitive chitosan hydrogel with poly(dimethylsiloxane) (PDMS) microfluidic channels was used in their study. To observe the change in pH, Fourier transform reflectometric interference spectroscopy (FT-RIFS) was used. The suspension of the bacterial cells (E. coli, 1 × 10 6 CFU/mL) with different antibiotics (azithromycin, amikacin, ciprofloxacin, and tetracycline) was injected into the device measuring the reflectance spectra. The increase in EOT was monitored in real-time by determining the possible inhibition of bacterial growth by different antibiotics ( Figure 3B). The technique was adequate to monitor microbial growth and MIC of antibiotics in real-time (azithromycin 1 µg/mL, amikacin 16 µg/mL, ciprofloxacin 0.25 µg/mL, and tetracycline 0.5 µg/mL) in 60 min.
Rapid analysis of antibiotics susceptibility (ceftazidime, cefotaxime, ampicillin, amoxicillin, levofloxacin, and doxycycline) was demonstrated in P. aeruginosa and E. coli bacteria by Nag et al. [125] by using localized SPR phenomenon and gold nanoparticles. The analysis consisted of capturing the bacteria (10 5 CFU/mL) on a gold nanoparticle-coated optical fiber sensor directly from urine ( Figure 3C). Subsequently, the bacteria were exposed and induced to cell lysis with different antibiotics for 2 h, observing the change in absorbance with time and degree of susceptibility or resistance of the bacteria to the different antibiotics ( Figure 3C). The sensor was also tested in tap water samples [125].
Raman spectroscopy (RS), is commonly used to measure molecular vibrational, rotational, and other low-frequency modes providing characteristic information about the effects of AMs on microbial cells [126]. Intensity and wavenumber of anelastic-scattered light of RS can be used to analyze living cells without labeling or fixation procedures. Surfaceenhanced Raman scattering (SERS) can be used to amplify the response [28,74,127,128]. Bi et al. developed an AST procedure based on SERS-active Au@Ag core-shell nanorods (Au@AgNR) [129]. The manufacture of the SERS nanotag system consisted of the functionalization of the Au@AgNR with streptavidin (SA) and their interaction with biotinylated antibodies specific for E. coli to form the complete SERS nanotags (Au@AgNR@SA@Ab) ( Figure 3D). Bacterial cells were trapped by the nanotags ( Figure 3D) and measured at different CFU concentrations to check the biosensor sensitivity. Finally, the bacterial cells were exposed to ampicillin to observe the susceptibility of the bacteria. The technique showed a detection limit of 10 2 CFU/mL, an estimated analysis time lower than 4 h and the possibility to analyze complex biological matrices such as blood. Some of the most recent publications related to optical (bio)sensors are summarized in Table 3. bacteria. The technique showed a detection limit of 10 2 CFU/mL, an estimated analysis time lower than 4 h and the possibility to analyze complex biological matrices such as blood. Some of the most recent publications related to optical (bio)sensors are summarized in Table 3.

AST Electrochemical (Bio)Sensors
Electrochemical sensors have also been proposed in AST tests [43,75,142]. Bacteria can be electroactive and carry out a great variety of redox reactions. Thus, electrochemical sensors can determine the electrical response of microbial cells with great sensitivity, in some cases, eliminating the prolonged isolation and culturing times and allowing the measurement directly in biological and environmental matrices [75,143]. Depending on the interaction of the microbial cell with the recognition element, the response can be measured indirectly or directly [144]. In direct detection, the binding of the cell itself with an antibody or a bacteriophage immobilized on the electrode, produces a reduction or increase in the transfer of electrons due to the presence of membrane enzymes [145]. In the indirect detection, the cell is stimulated to secrete some electroactive species that can be measured [144]. Screen-printed electrodes (SPE) have a significant advantage over traditional electrochemical cell methods since the three electrodes (working, counter, and reference electrode) are integrated into a single chip, facilitating the analysis and reducing manufacturing costs and time [146][147][148]. Hannah et al. implemented an SPE to monitor bacterial growth and select the correct antibiotic to administer to treat infections caused by E. coli at a low-cost diagnosis ( Figure 4A) [149]. A gold electrode was modified with an antibiotic-seeded hydrogel to measure the resistance to current flow presented by E. coli strains by electrochemical impedance spectroscopy (EIS). The bacteria were exposed to streptomycin (MIC of 4 µg/mL) for 2.5 h showing a sensitivity of ≥10 5 CFU/mL. Furthermore, electroanalytical techniques can be used for pathogen identification significantly reducing the susceptibility determination times [66,150,151]. Altobelli et al. [150] developed an assay for phenotypic antimicrobial susceptibility testing and rapid uropathogen identification. The quantitative determination of bacterial 16S rRNA is based on the use of an electrochemical array functionalized with a panel of complementary DNA capture probes. Pathogen ID was determined using pathogen-specific probes, whereas phenotypic AST with ciprofloxacin MIC was determined using an Enterobacteriaceae probe to measure 16S rRNA levels as a function of bacterial growth. Results are obtained in very short time ( ∼ =6 h) in comparison to 2-3 days required for the standard diagnosis.
Mach et al. developed an electrochemical, point of care device for determining the antibiotic susceptibility of pathogens of urinary tract directly in urine samples [152]. In this case, each of the working electrodes of a sensor array was modified with a capture probe specific for the bacterial 16S rRNAs of clinically relevant bacterial urinary pathogens. Detection of the target-probe hybrids was achieved through binding of HRP-conjugated anti-fluorescein antibody to the detector probe. Uropathogenic E. coli was cultured, and samples taken at time intervals of 20 min for the analysis by standard quantitative plating as well as the biosensor assay. Proportional increases in signal strength were observed over a 4-log unit range. A similar correlation between increased cell number and biosensor signal was observed with other common uropathogens including E. faecalis and P. aeruginosa. The specific probes for E. coli were immobilized on the microfabricated sensor with a thin, optical-grade layer of gold electrodes deposited on plastic. Subsequently, the strains were cultured for 2.5 h in different antibiotics (ampicillin, ciprofloxacin, trimethoprim/sulfamethoxazole gentamicin, ceftriaxone, and cefepime). After this time, cell lysis was performed, and amperometric current versus time was measured using a multichannel potentiostat to observe the susceptibility or resistance of the microbiological species [152]. In the analysis of samples directly from urine, it was determined that the identified E. coli bacteria were resistant to ampicillin, ciprofloxacin, trimethoprim/sulfamethoxazole, and gentamicin and susceptible to ceftriaxone and cefepime. A similar approach was reported in [153]. Genefluidics Inc. (Duarte, CA, USA) commercializes a system based on this approach, as also described in the Section 3.
Complementary to oligonucleotide-based capture probes, antibodies are an excellent tool for detecting whole cells due to their high specificity and potential for ID [154]. Recently, Shi et al. implemented a biosensor based on the capture antibody-bacteriadetection antibody sandwich immune complex to demonstrate a culture-free approach and to achieve a rapid diagnosis of bloodstream infections ( Figure 4B) [155]. Here, redox electroactive enzymes attached to antibodies are used as sensing elements; the device consists of a three-electrode cell modified with insulated gating electrodes (GE) for applying a gating voltage (VG) between the gating electrode and the working electrode. VG modifies the interfacial charge distribution by inducing an electric field at the solution-enzymeelectrode interface to reduce the tunnel barrier for electrons. A first antibody (specific for E. coli) is immobilized on the surface of the working electrode; the sample is incubated with the biosensor; finally, another antibody marked with HRP is added. Cyclic voltammetry (CV) was used to measure the peak reduction of HRP. The E. coli bacteria were incubated in the presence of ampicillin to measure whether the strains were susceptible or resistant to the antibiotic. The system allows the determination of bacterial cells in the order of 8 CFU/mL in 1-2 h.
Aptamers have also been proposed for bacterial cell determination [156][157][158][159]. To measure, in real-time, the growth and susceptibility of E. coli and S. aureus to the antibiotics gentamicin, ampicillin, and tetracycline, Jo et al. reported an aptamer-functionalized capacitance sensor array [160]. The specific DNA aptamers for each of the different strains were immobilized on the surface of the Au-working electrode. Bacteria can act as capacitors connected between electrodes in parallel, providing valuable information on bacterial activity. When bacteria are treated with antibiotics at inhibitory concentrations, the capacitance decreases. When the bacteria is resistant to the antibiotic, the capacitance increases in the same way as if the strains were not exposed to the antibiotic ( Figure 4C); the analysis can be performed in 1 h with a sensitivity of 10 CFU/mL [160].
functionalized capacitance sensor array [160]. The specific DNA aptamers for each of the different strains were immobilized on the surface of the Au-working electrode. Bacteria can act as capacitors connected between electrodes in parallel, providing valuable information on bacterial activity. When bacteria are treated with antibiotics at inhibitory concentrations, the capacitance decreases. When the bacteria is resistant to the antibiotic, the capacitance increases in the same way as if the strains were not exposed to the antibiotic ( Figure 4C); the analysis can be performed in 1 h with a sensitivity of 10 CFU/mL [160].  Graphene field effect transistors (G-FETs) have been gaining attention due to their high sensitivity in the detection of biomarkers and DNA, scalability, biocompatibility, and ease of incorporation of many kind of substrates [161,162]. However, the use of G-FET sensors for bacterial detection counts only a small number of papers describing the detection of a lab strain of E. coli, but no reports on the sensing of clinically relevant pathogenic bacteria, nor on antibiotic resistant strains [163,164]. A rapid, selective, and single cell electrical system for the detection of antibiotic-resistant bacteria based on G-FET was recently developed by Kumar at al. [165]. In their approach pyrene-conjugated selected peptides immobilized on G-FETs were used for detecting pathogenic S. aureus. A similar device was also proposed to discriminate between antibiotic resistant and sensitive strains of Acinetobacter baumannii, suggesting that these devices can also be used for detecting antibiotic resistive pathogens. In order to enhance the bacteria attachment, electric-field-assisted binding was used, i.e., bacteria were pushed into the graphene transducer by using electrical pulses. This strategy allowed reducing the detection limit to 10 4 cells/mL and the detection time to below 5 min. Table 4 shows some electrochemical sensors developed in recent years and different recognition elements used to determine antimicrobial susceptibility. * Antibiotic and MIC correspond to E. coli; ** Initial concentration; 3-APBA, 3-aminophenylboronic acid; AuNPs-gold nanoparticles; CV, cyclic voltammetry; DPV, differential pulse voltammetry; EIS, electrochemical impedance spectroscopy; FD, frequency dependence; G-FET, graphene field effect transistors; MRSA, methicillin-resistant Staphylococcus aureus; MWCNTs, multiwalled carbon nanotubes; Not measured; Not reported; POA, pyrazinoic acid; SPCE, screen-printed carbon electrodes; SPE, screen-printed electrodes.

Sensors and Biosensors for Genotypic Antimicrobial Resistance Detection
Microorganisms have developed different biological mechanisms to adapt and survive AM attack; these mechanisms are conserved and transmitted to other microorganisms through resistance genes. These genes information can be transferred vertically to the offspring or horizontally through mobile genetic elements (plasmids, prophages, integrons) [176,177].
The continuous mutations in the genome of microorganisms make it challenging to detect resistance genes. Short sequences of oligonucleotides (single-stranded DNA, PNA, LNA) immobilized onto transducer surfaces are the most common techniques in detecting ARGs by using electrochemical platforms, [47,86,178,179], SPR chip [180,181], or thin film transistor [182].
A direct gene-circuit/electrode interface for the electrochemical detection of colistin antibiotic resistance genes has been proposed by Mousavi et al. [183]. The colistin resistance genes (i.e., mcr-1, mcr-2, mcr-3 and mcr-4) have recently been identified in livestock globally. The interesting strategy described in ref. [183] is based on the production of restrictionenzyme-based reporters to catalyze the release of a reporter DNA (single-stranded DNA, ssDNA, labeled with methylene blue), which hybridizes with a DNA capture probe immobilized onto the electrode surface. The approach begins with the synthesis of toehold switches complementary to the 24 top-ranked binding sites within each mcr gene and the ligation of a unique reporter enzyme gene to each set of switches. Restriction enzymes cleave annealed reporter DNA, which is free floating solution, releasing reporter DNA labeled with methylene blue. The capture probes then recruit the redox-active reporter DNA to the surface of nanostructured microelectrodes, generating an electrochemical signal. The approach was validated against the mcr-4 gene. Thus, the mcr-4 gene was expressed in E. coli and then total RNA was collected from the resulting culture. Selected RNA fragments were then amplified by isothermal amplification reaction (1 h) at a concentration of 30 ng/µL, using mcr-4 specific primers. The amplified mcr-4 mix was added without purification to solution containing the specific switch, followed by incubation off-and on-electrochemical chip for 45 min at 37 • C, prior to taking the first electrochemical measurement. The measurement was specific to mcr-4 RNA in the presence of high background off-target RNA sequences and provided a strong, distinct signal against negative controls.
Korri-Youssoufi group [178,179] demonstrated the detection of rpoB gene resistance in Mycobacterium tuberculosis in real samples in a PCR-free approach. The sensitivity of the proposed assay was based on the association of magnetic nanoparticles and polypyrrole on a gold electrode surface. The catalytic properties of iron oxide nanoparticles combined with conducting properties of polypyrrole platform, allowed the detection of 1 fM of DNA target in a 50 µL drop corresponding to 3 × 10 4 copies of DNA. The naphthoquinone redox signal decreased after the hybridization between a selected probe and the complementary DNA target without cross-hybridization with single nucleotide mismatch DNA target. The biosensors were applied to detect and discriminate wild type DNA from mutated DNA strands of M. tuberculosis, in both PCR amplified samples and samples without PCR amplification.
An electrochemical biosensor based on isothermal strand-displacement polymerization reaction for detection of mecA gene in methicillin resistance-Staphylococcus aureus (MRSA) was proposed in [184]. Methylene blue (MB)-labeled hairpin capture probes were immobilized by self-assembling on a gold electrode, in order to have the MB molecules confined close to the electrode surface for efficient electron transfer. After hybridization with the complementary target DNA, the hairpin probes undergo conformational changes, which led to a decreased electrochemical response. Furthermore, the primers annealed with the opened stems of the hairpin probes were extended by DNA polymerase, which in turn released the target DNA to trigger the next polymerase cycle. Therefore, a significant amplified current suppression for mecA gene detection is generated by the mass of MB molecules moved away from the electrode surface since many rounds can be performed.
Huang et al. developed an electrochemical biosensor based on a specific PNA probe to detect bla NDM , the gene encoding the New Delhi metallo-beta-lactamase, using labelfree EIS [185]. The capture probe for the single-stranded DNA of the resistance gene was immobilized on the surface of a gold screen-printed electrode. After extraction and digestion of the plasmatic DNA, hybridization with the biosensor was performed. The detection limit of 100 pM for PCR amplification products of the bla NDM gene was obtained, with excellent results observed for amplification-free from a bla NDM -harboring plasmid.
Rolling-circle amplification (RCA) was coupled to a simple electrochemical impedimetric sensor for the detection of resistance genes of antibiotic belonging to the diverse group of β-lactamases [186].
Using a bimodal waveguide interferometer (BiMW) technique, Maldonado et al. [187] demonstrated a direct detection, without amplification, of blaCTX-M-15 gene and blaNDM-5 gene, two clinically relevant and frequent antimicrobial resistance encoding sequences. Selected DNA capture probes were immobilized on the surface of the BiMW sensor chip. After DNA extraction, fragmentation, and denaturation process, the samples were carried out for alignment with the sensor ( Figure 5). The BiMW biosensor showed high sensitivity (20-30 aM) and short analysis times (≤40 min, including sample pretreatment) for resistance gene analysis.   Hu et al. developed a system based on multiple cross displacement amplification (MCDA) assay and label-based lateral flow biosensor (LFB) for the amplification and detection of nucleic acids labeled with different chain displacement biomarkers in the presence of Bst polymerase under isothermal conditions [188]. In this work, the presence of the bla OXA-23-lik (carbapenem resistance) gene was determined in A. baumannii. The reading of the results using the MCDA-LFB technique consists of capturing the amplification products on the test lines based on the combination of antibodies embedded in the LFB test lines and labeled antigens in the amplification products. The method shows a detection limit of 100 fg per reaction with pure culture and high selectivity for the carbapenem resistance gene. Different types of genotypic sensors to determine resistance genes are shown in Table 5.  [195] Optic-fluorescence DNA probe rpoB Rifampicin resistance 1 nM ssDNA in 1 mL sample volume Yes [196] Optic-fluorescence DNA probe rpoB Rifampicin resistance 100 nM Yes [197] Optic-SPR DNA probe rpoB Rifampicin resistance NR No [198] Electrochemical-DPV PNA probe rpoB Rifampicin resistance 1 CFU/ml No [199] Optic

Conclusions and Perspectives
The main challenges in AMR detection is to obtain reliable results in minutes or hours (instead of days as with standard techniques) and with easy-to-use and cost-effective tools.
Automated systems have helped in the pursuit to decrease analysis time in the deployment of AST and various methods based on optical imaging and microscopy-based techniques have been developed to measure bacterial growth [204][205][206].
Chemosensors and biosensors are one of the alternative techniques with the most significant projection in the coming years. The use of chemosensors and biosensors will result in the development of fast point-of-care devices in multiplexing formats. Progress in the different transducing techniques will allow AMR monitoring in low resource settings without the need for trained personnel.
However, despite the pressing market demands, commercially available tools are still at the development stage or do not meet validation requirements. There are a number of challenges that have to be surmounted in the validation of the developed methods. Key of which is the detection of low concentrations of analytes in the presence of high concentrations potential interferents. Enrichment and amplification are still necessary for the analysis of real samples due to the need to detect target species at low concentrations (i.e., initial concentration of bacteria in blood is around 1 to 100 CFU/mL) and eventually in the presence of a large excess of other bacterial species.
Integration of many different technologies is needed to solve problems related to low initial pathogen number and the presence of contaminating sample matrices. The problem of contamination can be avoided by the use of magnetic beads or nanoparticles coated with ligands-targeting bacteria. Furthermore, the development of novel isothermal amplification procedures will benefit (bio)sensor field. Among these techniques we can mention the nucleic acid sequence-based amplification (NASBA), the loop-mediated isothermal amplification (LAMP), the recombinase polymerase amplification (RPA), and others. These amplification methods have simplified nucleic acid amplification, often allowing a minimum sample treatment, cut down the costs of instrumentation and are suitable for miniaturization. Moreover, some of these techniques are less sensitive to inhibitors than standard PCR, allowing a minimal treatment of biological fluids. By contrast, multiplexing approaches are less developed for these isothermal amplification techniques than for PCR. Recently, clustered regularly interspaced short palindromic repeats (CRISPR)associated methods have been studied for the detection of nucleic acids. This and other novel biochemical approaches represent innovative and powerful tools for rapid sensitive and selective monitoring of resistance genes.
Miniaturization of the transducers and the coupling with microfluidic platforms will enable the deployment of multiplexed, compact, and easy-to-use systems that will result in greater acceptance of chemosensors and biosensors by end-users and industrial stakeholders. Indeed, the progress in microfluidics, nucleic acid isothermal amplification, and (bio)sensor technology has recently provided several interesting devices that may eventually be used in AST, especially for low-resource field settings. An example is the Alveo platform (Alveo Technologies Inc, Alameda, CA, USA), an innovative device designed to multiplex 100 or more simultaneous tests from a single sample. The device consists of a fluidic cartridge coupled with electrochemical detection to measure the change in electrical conductivity that occurs during nucleic acid amplification: actually, during biochemical synthesis of DNA from nucleotides, the number and the mobility of electrically charged molecules are altered. Thus, the detection strategy is based on processing the sample by performing RPA and, optionally, a second isothermal amplification reaction. The second isothermal amplification reaction, which includes LAMP, can be carried out with amplification products of the RPA itself. The presence of the amplified target nucleic acid is then evaluated by measuring the impedimetric signal of the amplification reaction solution compared to a control. The analysis time can be of few minutes as well as many hours, depending on the desired degree of amplification that is necessary to achieve.
However, miniaturization requires a high level of standardization of the conditions, as the samples should represent similar growth states and culture densities. Automation can help with accurate reagent dispensing. The UtiMax Lab Automation System of Genefluidics Inc. (Duarte, CA, USA) is an automated diagnostic platform for the identification of uropathogens directly from urine samples, consisting of a robotic liquid handling systems with associated reagent kits and disposable sensor array chips. As already mentioned the chip allows the electrochemical-measurement of species-specific ribosomal 16S rRNA. Each sample is lysed chemically. A built-in multi-channel potentiostat reads the electrical current from the steady-state enzymatic cycling amplification: signal is proportional to the bound 16S rRNA content from lysate and reported in ranges of CFU per milliliter through an established calibration curve.
Research in the field of chemo-and biomimetic receptors, including peptides, imprinted polymers and aptamers, will help in improving selectivity and thus bacterial identification that is another important issue for improving the reliability of the (bio)sensing approach to AMR monitoring. For instance, recently, antimicrobial peptides are considered as interesting biorecognition elements for bacteria ID and detection. These peptides can interact with bacteria, causing their lysis and inhibition of their growth. Some antimicrobial peptides have been described as specific to certain bacterial species, while other peptides can recognize any bacteria. This diversity can be employed both to design highly specific ligands and wide-spectrum probes able to interact with many different bacteria species. In the second case, profiling fingerprints and statistical analysis of data are necessary for the species-specific identification of bacteria.
Thus, in a context of growing global bacterial resistance, the development of better and more rapid detection of pathogenic threats based on biosensors is for sure one of the main interesting targets. This of course implies the development of more efficient sensing devices and also adequate biorecognition elements in the future to allow the specific identification of the various pathogens, particularly in medical settings where the nature of the threat is often not clearly identifiable.