Advances in Biosensing Technology for Medical Diagnosis

Essentials of Biosensors

Jacky Fong-Chuen Loo¹, ² , *, Tza-Huei Wang³, ⁴, Yi-Ping Ho¹, ⁵

¹ Department of Biomedical Engineering, The Chinese University of Hong Kong, Hong Kong, SAR

² Department of Neuroscience and Biomedical Engineering, Aalto University School of Science, Aalto, Finland

³ Department of Biomedical Engineering, Johns Hopkins University, Baltimore, Maryland, USA

⁴ Department of Mechanical Engineering, Johns Hopkins University, Baltimore, Maryland, USA

⁵ Centre for Novel Biomaterials, The Chinese University of Hong Kong, Hong Kong, SAR

Abstract

The primary objective of medical diagnosis is to precisely detect the disease onset in a timely manner for effective treatments. The rising demand in medical diagnosis, given the rapidly aging populations, increasing population mobility and complex healthcare needs, has called for support of effective diagnostic approaches, where biosensors have provided well-suited solutions. To this end, biosensors are typically examined by the yardstick of specificity, sensitivity, dynamic range, and reliability or robustness. The advancements in biomarker discoveries and signal transduction schemes have led to biosensors of improved performances; however, challenges remain particularly for translating the biosensors into clinical uses. This chapter is therefore aimed to prepare the audience with essentials of biosensors and roadblocks associated with clinical translations. Comprehension of these prerequisites is expected to accelerate the development of biosensors from lab bench to bed.

Keywords: Molecular biosensing, Medical diagnosis, Point-of-care, Signal transduction.

* Corresponding author Jacky Fong-Chuen Loo: Department of Biomedical Engineering, The Chinese University of Hong Kong, Hong Kong, SAR & Department of Neuroscience and Biomedical Engineering, Aalto University School of Science, Aalto, Finland; E-mail: [email protected]

INTRODUCTION

The ultimate goal in medical diagnosis is to pinpoint the onset of diseases precisely for timely and proper treatments. The demand has continued to escalate given the rapidly aging populations, increasing population mobility and complex healthcare needs. Detection of clinically relevant target biomarkers has long been a way for accurate medical diagnosis. Clinicians have been using an optical mic-

roscope to detect Mycobacterium tuberculosis, the bacilli responsible for tuberculosis, from samples of the patients as early as in the late 19th century [1]. Radioimmunoassay, a conventional biochemical assay relying on the interaction between the antigen and the antibody, has been developed in the 1950s [2]. These two tests utilize cell staining techniques, radioactive materials processing and operation of scintillation counters. However, in order to obtain reliable results for appropriate downstream treatment, these traditional detection methods required a centralized laboratory and well-trained technicians to handle a series of laboratory processes, including sample pre-processing, preparation of reagents, performing assays, operating equipment, and interpretation of obtained results. Take the COVID-19 pandemic in 2020 as an example, hospitals and centralized laboratories have been overloaded worldwide. Biosensors suit the so-called point-of-care (POC) setting may help to delay the outbreak by enabling self-diagnosis of viral infection or immunity without the needs of advanced equipment or trained staff [3].

Defined as a device that uses specific biochemical reactions mediated by isolated enzymes, immunosystems, tissues, organelles or whole cells to detect chemical compounds usually by electrical, thermal or optical signals by the International Union of Pure and Applied Chemistry (IUPAC), biosensors have become a natural choice for the applications in medical diagnostics, since the first invention of glucose biosensors based on electrochemical sensing in the 1970s [4]. The development has further fueled by the advanced engineering and biotechnological approaches starting from the 1980s [5, 6]. The earliest biosensors are mostly based on catalytic systems, integrating enzymes, cellular organelles, or even whole microorganisms with transducers that convert a biological response into a digital electronic signal [7]. Later advancements in the discovery of new biomarkers and exploration of novel biosensing mechanisms have led to the booming interests of molecular diagnosis based on the detection of small molecules or nucleic acids, setting the foundation of the POC diagnosis. In addition to solving the traditional medical diagnosis, the recent development on advanced technologies, such as microelectromechanical systems (MEMS), nanophotonics, or methodologies, such as CRISPR/Cas9 system for both genetic biomarker recognition and signal amplification, can be quickly adopted in the development of novel biosensors that continuously improve its performance and accuracy for medical diagnosis [8].

Based on the definition of IUPAC, a biosensor shall comprise the following three parts: a bioreceptor, a transducer and a signal readout unit, as illustrated in the schematics of Fig. (1). Bioreceptors can be regarded as recognition elements, such as antibodies, aptamers, or nucleic acid probes, that bind to the target analytes specifically [9, 10]. Binding affinity and reaction kinetics between the recognition element and the target jointly determine the sensitivity and dynamic range for the sensing of targeted biomolecules, while the stability under different conditions or prolonged storage determines the reliability of biosensing. Interpretation of the acquired data perhaps remains challenging in POC diagnosis. Optical and electrochemical are the two most commonly employed approaches to transduce the sensed events into detectable signals. Absorbance, fluorescence and luminescence sensing mechanisms are commonly used in the optical-based sensors, while potentiometric and amperometric sensing mechanisms are commonly used in the electrochemical-based sensors for signal readout [11]. For optical transducers, quantification of the detectable photons to signify the presence of targeted analyst is achieved with optoelectronic components such as photodiodes, photomultipliers and charge-coupled devices, which convert the detectable photons into electronic signal readout for downstream computation and display. For electrochemical based transducers, electrodes coated with enzymes are often applied to detect the oxidation of the targeted analyte, such as glucose, upon catalysis by an oxidoreductase enzyme, i.e. glucose oxidase [12]. The electrons liberated during this redox reaction are shuttled to the electrode through artificial electron acceptors or mediators, i.e. ferrocene, where the current generated and detected by a potentiostat is directly related to the concentration of targeted analytes.

Fig. (1))

Schematics of a biosensor. The recognition elements (i.e. antibodies, aptamers, and nucleic acid probes) are essential for the specific capture of the corresponding target analytes. The interaction between the recognition element and the target analyte is registered by a detector and subsequently transduced into a signal output in the form of for example fluorescence or electricity. The associated assessment criteria are listed on the right.

Targeting POC diagnostics, biosensors are typically examined by at least the following yardsticks: specificity, sensitivity, dynamic range, and reliability or robustness. These performance factors may be interlinked and shall be considered based upon the needs. Among these, specificity and sensitivity are equally important in most cases. Take the matrix of specificity and sensitivity as illustrated in Fig. (2) for example, high specificity and sensitivity is the best scenario. For screening of viral infection, such as the Zika virus, by detecting the viral RNA in urine, where the viral nucleic acids may be presented as low as a few copies, the diagnostic strategy would opt for high sensitivity [13]. For sensing of blood glucose and sodium, which lies in a concentration range of millimolar, high sensitivity is unnecessary, but sufficiently high specificity is critical [14]. In this chapter, we intend to give the audience a brief introduction to the essential elements of a biosensor and the evaluators for the development of a reliable biosensor particularly in the POC setting in medical diagnosis.

Fig. (2))

The matrix of sensitivity and specificity in biosensing. Possible scenarios are exemplified by the interaction between the antibody (red y-shape) and the target analyte (blue triangle). (Top left) High sensitivity and specificity: specific interaction between the two entities with high sensitivity; (Top right) High sensitivity but low specificity: non-specific interaction although the detection sensitivity is sufficiently high; (Bottom left) High specificity but low sensitivity: incapable of detecting target analytes of low concentration; (Bottom right) Low sensitivity and specificity: non-target molecules are detected.

SPECIFICITY

In the clinical setting, specificity may be broadly referred to as the specificity of both disease confirmation and biosensing of targeted biomarkers. Specificity in disease confirmation is often optimized by screening biomarkers with high clinical relevance to a particular disease, such as neurodegenerative disorders and infectious diseases. For example, 2N4R tau protein and Aβ oligomers are currently recognized as specific biomarkers for the diagnosis of Alzheimer's disease, whereas C-reactive protein, Interleukin 6 (IL-6) and microRNA-16 are biomarkers associated with sepsis [15, 16].

The specificity of biosensing hinges heavily on the recognition element, as hinted by the configuration in Fig. (1). Specific interaction has implied not only a high binding affinity between the recognition element and target analyte but also a low binding affinity between the recognition element to non-targets. Non-specific binding to non-targeted analytes resulted in the so-called false positive signals, which may lead to unnecessary treatments and a waste of clinical resources. For nucleic acids, the detection relies on the Watson-Crick base pairing. The nucleic acid probes may be designed to identify complementary strand or discriminate one-base-mismatched sequence in the targets [17]. For the case of protein markers, antibody fragments, synthetic receptors, and aptamers have been developed to specifically bind to the targeted proteins. Antibody fragments, relying on specific binding between the variable fragment (Fv) region and the target antigen, do not have the fragment crystallizable region (Fc region) to bind with non-targets [18]. Synthetic receptors and molecularly imprinted polymers, which is molded into a shape with cavities complementary to the target analyte for high affinity binding, are produced with the aid of computational chemistry and molecular imprinting for rational design of specific molecular recognition [19, 20]. Aptamer, on the other hand, typically requires an artificial selection process, i.e. systematic evolution of ligands by exponential enrichment (SELEX) method. Negative selection is also necessary for SELEX to rule out the aptamer sequences with possible binding to non-targets [21, 22]. Compared to antibody development, which requires multiple screening after an antibody with high binding affinity is raised, the aptamer SELEX process can tune its specificity to the target during the selection cycles. In addition, aptamer can be easily linked with ribozyme which acts as an on/off switch, as shown in Fig. (3A), upon binding to the target analyte, theophylline. Afterwards, the activated ribozyme triggers the downstream signal amplification process for visible signal readout Fig. (3B) [23].

Fig. (3))

Various approaches to achieve high specificity of biosensing. A. Sequence and conformational change of modified theophylline ribozyme (RS-TFU76) with an aptamer-binding domain (left, stem II), a communication module (middle, stem II), and a hammerhead ribozyme motif (right, stems I and III) from inactive to active after adhesion of theophylline (T) to the putative binding site and activates RS-TFU76 (right). B. The schematic diagram showing the whole ligand-induced ribozyme targeting and sensing system. After theophylline target (T) binds to the aptamer domain linked to a cis-acting hammerhead ribozyme (RS-TFU76) and causes the conformational change to the ON state and subsequent self-cleavage that forms two products, P1 and P3 in stage 1, downstream EXPAR amplification is triggered by the cleaved kinase-treated ribozyme (P3) and from the given templates produces ssDNA guanine quadruplex precursors (Q) in stage 2. Peroxidase activity is generated from ssDNA Q folding with hemin (H) to oxidize 3,3',5,5'-tetramethylbenzidine substrate from clear (TMBred) to blue state (TMBox) for visualization. C. The schematics showing the assay design of the electrochemical biosensor for the multiplexed detection of miRNAs with neutravidin-coated biosensing surface based on a simple one-pot assay approach where simultaneous detection of miRNAs by specific hybridization between the target miRNA and biotin-MB1- AuNP or miRNA-141/biotin-MB2-AgNP onto the neutravidin electrode followed by SSWV. The binding results in the increase in the different peak currents corresponding to AuNPs and AgNPs, which allows different signal responses from respective miRNAs to be differentiated due to the intrinsic electrochemical signature of the nanolabels. Reprinted with permission from references [23, 31].

Clinically speaking, the test specificity also concerns with the validity of targeted biomarkers among other interferential substances in the clinical samples, such as blood pigment. Take a commercially available coronavirus rapid paper-strip POC test for example, the test is based on detecting antibodies against SARS-CoV-2 for the evaluation of immunity. However, a positive signal may be acquired in the presence of SARS-CoV-2 binding antibodies, which do not provide a long-term immunity, in patients recovered from the viral infection [24]. The false positive may mislead the individual into believing that they have already been infected and developed an immunity. Apart from the detection of SARS-CoV-2 viral genome, surface protein antigens and specific antibodies, the discovery of appropriate biomarkers related to its infection is also challenging. Cytokine storm, an immune indicator of severity, is usually found in patients infected with SARS but not SARS-CoV-2 [25]. To identify highly specific human biomarkers related to SARS-CoV-2 infection, the mechanism of its infection and recovery needs further investigation.

Several optimization approaches can be adopted to enhance sensing specificity. Optimization of the reaction conditions, such as the amount of recognition elements coated on the sensing surface, pH values and ionic conditions may enhance the binding affinity between the recognition element and the target analyte, as well as reduce the background noise generated by nonspecific bindings [26, 27]. Recognition by, for example, dual- or multiple-antibodies targeting different epitopes of the same target, generating a sandwich-like 'Antibody 1-Antigen-Antibody 2' complex, functions as an AND gate. The specificity is enhanced by the signal generated from the event dual-recognized by two antibodies, such as the demonstration of whole bacteria detection [28]. Aptamer split, similar to the concept of dual antibodies, has also been used to further increase the detection specificity [29]. Additionally, the removal of the non-specific binding by charge-charge interaction is another strategy to reduce the background noise for increased specificity. Alternative recognition elements, such as peptide nucleic acid, remove the negative surface charge of nucleic acid backbone, and neutravidin minimizes the surface charge in neutral pH value in place of streptavidin during sensing [30, 31]. Fig. (3C) shows an example on the use of a neutravidin-coated surface to capture biotinylated probe-nanoparticle-microRNA complex specifically. In this configuration, there is a low chance of nonspecific charge-charge interaction between nanoparticles or microRNA and the neutravidin-coated surface. Coating of proper blockers, such as PEG, reduces hydrophobic interaction and adsorption of nonspecific protein or nucleic acid on the bare sensing surface [32].

SENSITIVITY

Sensitivity, or the limit of detection (LOD), is referred to the minimal amount of analyte that can be detected by a biosensor. The development of biosensors has been pushing the detection limit, making the early diagnostics possible by detecting extremely low amounts of the analyte. The state-of-art is already capable of analyzing single cells or even single molecules, which allows not only sensitive disease diagnosis but also further understanding of the disease progression [33]. As illustrated in Fig. (1), sensitivity of a biosensor is determined by the binding of the recognition element and the target analyte in concert with the efficiency of signal transduction. Binding of the recognition element and the target analyte is preferred in a fast kinetic and high affinity. High sensitivity in a POC test is highly desirable, so that early identification of the disease onset and accurate frequent monitoring of the disease status may become possible during regular checkup. In the COVID-19 pandemic, the sensitive viral RNA detection with a detection limit below 10⁵ copies/swab would have provided effective screening of SARS-CoV-2 infection at an early stage, when the transmission rate is high in the absence of commonly developed symptoms such as fever and coughing [34, 35]. Tests of high sensitivity would also help the SARS-CoV-2 surveillance and pandemic control by pinpointing the asymptomatic patients.

Take the nucleic acids based biosensors for example, strategies such as increasing the bound events, as shown in Fig. (4A) [36], are often used amplify the signal. For the recognition of antibodies, the affinity of rabbit polyclonal antibodies is usually 10-1000 times higher than that of a mouse monoclonal because of the high efficiency of affinity optimization during antibody production in the rabbit immune system [37]. The drawback of this polyclonal feature, i.e., multiple epitope targeting, in rabbit antibody is further tackled by the development of monoclonal antibodies with hybridoma technique [38]. On the other hand, avidin has been used to capture biotinylated molecules with a high affinity, with a Kd as low as 10-14 M between streptavidin and biotin. Also, their tight interaction is not easily affected by temperature, pH values, or osmotic pressure change, which makes it useful in the sensitive detection of biomolecules with biotin tagged in the robust screening of clinical samples.

Besides the choice of recognition elements, enzyme is widely used as a biological catalyst that increases the rate of biochemical reaction of signal generation to amplify the signal. Enzyme-linked immunosorbent assay (ELISA), an assay conjugating enzymes with recognition elements, was the first demonstrated technique to increase sensitivity in place of radioimmunoassay (RIA) in the 1950s [39, 40]. Horseradish peroxidase is widely used for colorimetric or chemil-uminescent detection via the ability of HRP of converting a substrate, such as 3,3',5,5'-Tetramethylbenzidine (TMB), into a colored product or emission of light [41, 42]. Apart from acting on the signal generation molecules, other enzymes have been employed to act on the target analyte during biosensing. Fig. (4B-C) shows two examples of signal amplification. RNA-cleaving DNAzymes have been developed for diagnostic POC tests that recognize bacterial pathogens, such as Escherichia coli [43]. RNase H has been used in repeated cycles of digesting the RNA probe in the DNA-RNA hybrid for sensitive microRNA detection, which could reveal the hidden disease for early diagnosis, as shown in Fig. (4B) [15, 44].

Fig. (4))

Various methods of signal amplification to increase sensitivity of biosensing. A. The schematic diagram showing the formation of a nanosensor assembly with fluorescence emission caused by FRET between Cy5 acceptors and a QD donor in the presence of targets. B. The schematic diagram showing the design and workflow for microRNA sensing with the MicroRNA-RNase-SPR assay, which relies on a repeated cycle of RNA probe digestion by RNase H after mature microRNA-converted cDNA (top A to D) bound on the probe to form hybridized RNA-cDNA in the gold SPR surface (bottom). It resulted in the change of the SPR signal to be measured. C. Schematic illustration of the mycobacteria biosensing system using DNA substrate complex containing scissile DNA oligonucleotide hybridized to the immobilized DNA primer captures mycobacterial TOP1A. Signal sensitivity is enhanced with rolling circle amplification (RCA) to amplify the scissile DNA circle formed by the ligation step of the mycobacterial TOP1A catalytic cycle. Reprinted with permission from references [36, 44, 48].

Moreover, enzymes, including polymerase and ligase, have been used in the nucleic acid amplification on the target nucleic acid or reporter, such as aptamers that correlate to the concentration of the target analyte. For example, Fig. (4C) shows the rolling circle amplification (RCA), an isothermal amplification technique. On the other hand, fluorescence labels using chemical dyes have been popular in the last two decades. Intercalating dyes, molecular beacons (MB), and TaqMan probes are employed to label the target nucleic acid molecule, while fluorescein-based probes that bind to ions, such as Cl-, have been coupled with Förster resonance energy transfer (FRET) to detect the presence and quantitation of these molecules [45]. Fluorescence protein, such as green fluorescence protein (GFP), are also used in cell-based biosensors because of the genetically encoded fluorescent. Split GFP fusion with surface receptors has been used to detect target molecules, where the interaction between the target and the receptor causes the split GFPs to form a functional GFP and give out fluorescence. Upconverting nanoparticles (UCNPs), on the other hand, rely on the use of non-photobleaching and non-harmful infrared spectrum, and could overcome both the limitation in photo-bleaching and background noise from excitation in the visible spectrum [46]. The increase in signal intensity, together with the decrease in noise, eventually boosts the signal-to-noise ratio for sensitivity biosensing. In addition, the use of the infrared spectrum increases the depth of excitation light to be passed, which is feasible for in vivo high-contrast imaging [47].

On the other hand, the sensitivity of biosensor is also determined by the signal-to- noise ratio during signal acquisition, where the noise could be the system noise such as background fluorescence. Although there are many labeling techniques available, the background fluorescence, procedures and time needed for labeling usually hinder their practical use in biosensing. In this regard, surface plasmon resonance (SPR) has been developed as a label-free, highly sensitive sensing method [49]. Three interrogations, i.e. angular, wavelength and phase interrogations, have been investigated for the SPR measurement. Among these, phase measurement has shown with a superior detection sensitivity. It achieves the ultra-low detection limit, compared to conventional biochemical reactions using enzymes for signal amplification [50]. With the use of molecular imprinted film, such as amoxicillin-imprinted p(HEMAGA) film for amoxicillin detection, sensitive detection of the target as low as in pg/mL level can be achieved since the SPR sensing surface is well covered by the film to maximize its ability on target recognition [51, 52]. Coupling enzymatic reactions with SPR can also further increase its sensitivity [44, 53]. Besides, SPR imaging further extends its capacity for simultaneous multiplex detection to increase the throughput without compromising its sensing sensitivity [54-56]. It is particularly useful when more than one analyst is required for measurement and result interpretation, such as olfactory sensing [57]. SPR imaging on real-time multiple targets tracking can be performed after the multiple recognition elements are printed in an array format on the SPR sensing surface [58]. SPR image scanning using a mechanical stage usually creates vibration that affects the signal detection, where the use of MEMS system, such as digital micromirror devices, could stabilize the whole sensor for stable optical measurement [59, 60]. The SPR image resolution could be further enhanced by phase measurement or nanomaterials, such as random nanodot arrays and gold nanoparticle arrays [61-63]. Unfortunately, these sensitive sensing devices are usually not portable to meet the POC need. In light of this, there are increasing research and development of the fiber-SPR sensing method, by integrating all the optical sensing in a gold-coated fiber, optical paths with fiber and junctions to connect all the components, from a light source to the spectrometer [64]. It could tackle not only the portable issue but also the instability issue for robust biosensing with conventional SPR method [65-67]. Apart from this, the use of nanoparticles, such as gold nanorods (AuNRs) that generate localized SPR (LSPR), can achieve high sensitive sensing for visible discrimination in a test-tube format [68-71]. For example, cortisol, a stress biomarker that causes mental disorders, can be quantified simply by adding single drop of saliva supernatant to the aptamer-coated nanoparticles, which has been immobilized on sensing surface, without further reagent addition or washing [72].

Concentration or localization of the target analyte is another approach to enhance sensitivity. In conventional test-tube based assays, magnetic micro-beads conjugated with recognition elements isolate the target analyte from the sample mixture, and the magnetic action assembles the microbead-analyte complexes into a tiny region [73, 74]. The alternative, created by the advancement in micro-fluidics, utilizes micro-chambers in a microfluidic to gather target molecules in a micro- or nano-scale localized zone. This also helps in the downstream capture by recognition elements and hence the sensitivity. Droplet encapsulation not only concentrates the biochemical reaction in a pico-liter droplet to increase the chance of interaction between molecules but also enables digital analysis, such as digital polymerase chain reaction (dPCR) [75]. Rapid pathogen-specific phenotypic antibiotic susceptibility testing with a high sensitivity using digital loop-mediated amplification (dLAMP) has also been developed [76]. On the other hand, optical trapping is a non-contact based concentration and defined localization of the target analyte with a laser beam or plasmonic waveguide [77, 78]. The detection can be enhanced for single-cell analysis when assisted with micro-chambers, and its ability is demonstrated in the analysis of DNA [79] and protein [80, 81]. Thermal gradient trapping is similar to optical trapping with the additional benefit of sample heating to accelerate biochemical reactions, such as DNA amplification [82, 83].

DYNAMIC RANGE

In human bodies, the physiological range of biomolecules is usually as narrow as within two orders of magnitude. Most of the conventional biochemical assays, such as ELISA, could cover the range effectively. The integrated printed biosensor was reported to have a linear response between 0 and 5 mM glucose, suitable for glucose monitoring in interstitial fluid [84]. Although the dynamic range is relatively small, the sensitivity is high and accurate enough to measure the fasting blood glucose level. On the other hand, the high dynamic range provides the addition benefit of robustness in disease screening and in-depth understanding of the pathological development of diseases, since target biomarkers show a high dynamic range in different stages. For example, the dynamic in cytokine IL-2 secretion is correlated with the pathophysiological roles in the immune system, and the high dynamic range detection support the analysis of its secretion profile for a comprehensive understanding of the role of IL-2 during a disease [81]. Also, dynamics in a single red blood cell (RBC) triggered by an environmental stimulation, such as photo-induced oxidation of the RBC, can be monitored over time [85, 86].

A POC test with a higher dynamic range could extend the biosensing ability from detecting target biomarkers in a particular type to multiple types of specimen, since the concentration of target biomarkers may vary in different body fluids. For example, the viral RNA concentration of SAR-CoV-2 is ranged from 10 ³ to 10 ⁶ copy/mL in the respiratory tract specimens such as saliva and sputum, and from 10 ² to 10 ⁴ copy/mL in human plasma sample [35, 87]. A POC test covering the said ranges would be versatile for different samples, provide additional safeguard on multiple screening of different samples from the same individual to validate their immunity, and facilitate the quantitative study of viral infection pathways and their migration inside human body by helping in the understanding of the mechanism of infection.

Various methods have been used to increase the dynamic range. A combination of several aptamers with various affinities to the target analyst was employed to detect different concentration ranges of the target [88]. The dynamic response range, adjustable by changing the environmental pH value in accordance to the isoelectric point of the target protein enables a high dynamic range of target Plasmodium falciparum lactate dehydrogenase detection from 100 pM to 10 nM [89]. In addition, duplexed aptamers, the hybridization between an aptamer and an aptamer-complementary element, i.e. a DNA oligonucleotide, was demonstrated as a ligand-responsive construct with a tunable dynamic range through base mutations on the bound DNA and screening by aptamer-complementary element scanning [90]. It was demonstrated with adenosine triphosphate (ATP) sensing with two orders of magnitude in aptamer-based electrochemical sensors [91]. Reducing the background noise to boost the detection limit, such as the use of nanomaterials UCNP@PDA@AP, supported a high linear dynamic range on cytochrome c sensing, from 50 nM to 10 μM [92]. Moreover, signal amplification with enzymes by increasing the detection limit without affecting the maximum detectable concentration helps increase the dynamic range in biosensing. For instance, immunosignal hybridization chain reaction by combining antibody-antigen interactions with hybridization chain reaction technology, has provided a broader dynamic range than the enzyme-based chemiluminescent detection method [93, 94]. The use of platinum nanocatalyst amplification on lateral-flow devices provided a broad linear dynamic range across four orders of magnitude from 1 to 10000 pg/mL on HIV diagnosis [95]. For the optical transducer for signal acquisition, the requirements for both high sensitivity and a broad dynamic range suggests that the interferometric sensor is the most fit compared to the absorbance and fluorescence measurement. SPR coupled with a continuous white light source in place of a single-wavelength laser could provide a wider coverage range of the measurable concentration, but the sensitivity will be sacrificed when using white light [96]. In this regard, the implementation of phase measurement, instead of intensity, could tackle the detection sensitivity, without compromising the dynamic range [97, 98]. This was demonstrated with a wide dynamic range for monitoring cytochrome-c leakage from 80 pM to 80 nM during cancer cell death for anti-cancer drug screening [50]. The wider coverage range of measurable concentration of the SPR method was investigated and in some recent works, such as the use of erythrocyte membrane (EM)-blanketed gold nanoparticle provide a sensitive and wide range, from 0.001–5.000 mg/mL, detection of fibrinogen in blood sample [99].

RELIABILITY AND ROBUSTNESS

Reliability and robustness of biosensors produce consistent quantitative results at the POC, which are essential to clinical decision making as valid clinical data. Fig. (5A) shows an example of an integrated biosensor platform, by merging the technical advances in printed electronics, sensing probe, and readout display for electrochemical sensing on the glucose level [84]. This portable platform in a card format supports the in-field applications, without the technical support of experts and clinical facilities. Regarding protein and nucleic acid-based biosensing, stability of recognition elements and molecular reactions are some of the key factors to generate reproducible results, but, unfortunately, the biological elements and reagents have a finite shelf life. In order to overcome this limitation, stability of the reagents is enhanced by, for example, a polymer coating and the addition of lyophilization and sugar stabilizers to reagents extend their lifetime from days to years [100]. Among various types of recognition elements, aptamers show many benefits over antibodies. The popularity of aptamer-based assays is growing among diagnostic applications, especially paper-based biosensors. Resistant to heat, aptamers can be stored in room temperature for months without losing its binding ability to the target [101]. An aptamer could also replace enzymes to convert immunoassay substrate TMB to color products in the immunoassay [23, 102]. Nevertheless, there are drawbacks of using enzymes in biological reactions to increase the signal, for instance, inconsistent reaction rates in different logs and reduction in activity over time, which hinders its practical use in stable biosensing. Enzyme-free signal generation is therefore desirable. Hybridization chain reaction is one of the enzyme-free methods for DNA amplification [93, 103]. Native DNA and even artificial DNA, e.g. locked nucleic acid (LNA) commonly used because of its resistance to nuclease, can be amplified using this method [104].

Fig. (5))

Examples of portable biosensors demonstrated with high stability and robustness. A. schematic representation of the (Upper) hybrid printed circuit design and (Lower) printed integrated system, containing the printed circuitry and battery for micro-controller programming, actuation, communication and actuation, a potentiostat chip, and display for sensing. B. Schematic diagram of a self-contained microfluidic disc (one-quarter of a full disc area) to perform sample-to-answer nucleic acid-based molecular diagnosis of specific target bacteria from clinical samples, such as blood and sputum. The operation flow of microfluidic disc in the right panel shows the fluid movement sequentially at each step, controlled by a spinning force and valves to perform sample lysis, DNA extraction, and isothermal DNA amplification (real-time loop-mediated isothermal amplification (RT-LAMP)) automatically. Reprinted with permission from references [84, 114].

The use of nanomaterials is also on the rise. In conventional fluorescence labeling, the signal output is unstable because the chemical dyes or fluorescence proteins are easily degraded by enzymes or quenched by a high-power laser input, but nanomaterials show no such limitations [36]. Fluorescence-based nanoparticles (NPs), such as gold nanorod (AuNR), quantum dot (QD), are promising alternatives [105, 106]. Conductive nanomaterials, such as graphene, could enhance electrochemical sensing stability to be more reliable in low concentration biomolecules detection by simply coating on the current sensing electrodes [107]. Graphene oxide, carbon nanotube and carbon nanosphere have been evaluated with a peroxidase-like activity, which can be applied in immunoassays in place of the unstable peroxidase [108-110].

Sample actuation is another issue in robustness of biosensors. Considering the tedious and laborious operation of sample handling and processing, modern clinical equipment in centralized laboratories in hospitals are designed in an automated action. The only procedure is to insert the sample into sample racks and assay cassettes containing the reagent for the assay into the machine and to read the result generated. To construct a miniaturized biosensor, the use of microfluidics and a MEMS device, could benefit the automatic sample actuation, instead of using traditional pipette robots. More importantly, sample-to-answer can be achieved with lab-on-a-chip, designed to complete all the actions, e.g. cell isolation and enrichment of target biomolecules, in a robust, sensitive and automated manner [111]. The use of microfluidics facilitates the accurate tem-perature control, such as thermocycling, of a limited volume of fluid in a portable and lightweight processing device, regardless of the large specific heat capacity of water. Yet, the network of external pump connections is not robust when translated into POC use. Lab-on-a-disc (LOAD) therefore provides robustness in operation because of its fully automation of liquid handling [112, 113]. In addition, the high throughput, highly parallel and quantitative analysis are enabled to support simultaneous multi-dimensional analysis, such as multiple single-cell analysis in a sample. Multiple disease diagnosis for sample-to-answer on molecular diagnosis of bacterial infection can be performed in a disc, as shown in Fig. (5B), where fluidic actuation is simply controlled with the centrifugal force only in a sequential order [114, 115]. Further combination of the advantages of label-free multiplexed SPR detection and LOAD achieve completely automated sensitive immunoassays [116]. While conventional microfluidics handle liquids continuously in microchannels, the digital microfluidic is a new type of microfluidics that manipulate fluids in discrete volumes in the form of droplets, with couplings with magnetic actuation or electrowetting-on-dielectrics (EWOD) actuation with demonstration of their potential for POC test in remote settings [117,118]. An additional benefit of digital microfluidics is that the formed droplets act as tiny micro-reactors to localize the reactant, thus increasing the sensitivity of diagnostic platforms.

Advancement of fabrication techniques, such as printing, engraving and molding, are useful in facilitating the production of biosensors and accelerating the turnaround time in development cycles. It also provides an ease of customization and supports just-in-time on-site manufacture. Integrating microfluidic biosensor development has a long history in the healthcare industry. One typical example is lateral-flow immunoassay, a biochemical test integrating recognition elements and transducers in a paper-strip format, commercially available worldwide as lateral flow dipstick for POC use, such as rapid HIV tests and rapid bacteria detection [43, 119]. Recent development shows that signal amplification, e.g. isothermal DNA amplification loop-mediated amplification (LAMP), can be accomplished in this paper-based sensor, besides protein-based immunoassays [120]. Furthermore, the evolution of affinity sensors from conventional lateral-flow paper test strips to wearable or implantable devices with soft and flexible materials, such as plastic membrane and polymers, have worked successfully in POC multifarious polymer designs that provide the base materials for sensor designs [121, 122]. Recent progress in 2D and 3D printed microfluidics biosensor has reduced the difficulties in prototyping and research development, where cycles of prototyping and optimization has been sped up [123-126].

To date, technological advancement has enabled comprehensive data analysis with high computational power. Limit of quantification (LOQ), defined as signal detection above the mean of the blank measures plus a 10× standard deviation generally, could evaluate whether the low-level signal output is valid in not only qualitatively, i.e. yes/no determination but quantitatively and reliable with a low chance of false negative signal. One major technical breakthrough in biosensor development is the image analysis technique of parallel processing of multiple optical images and effective identification of target markers in the images automatically. Machine learning, one of the categories in narrow artificial intelligence, provides more accurate image analysis, especially those based on images and large dataset, to outperform that with manual operation for disease diagnosis [127]. Machine learning is well-known for its effective and automatic cellular image analysis and therefore has a high potential to replace labor- and skill-intensive image analysis [128]. Yet, the data-processing speed becomes a key factor when handling a large sum of data, especially arrays of dataset, and only a desktop computer is capable of handling such tasks. Cloud computing may be incorporated in portable biosensors for in-field operation, unlike traditional biosensors that are linked to a personal computer. Connected to a cloud network for further analysis, cloud computing further reduces the size, electricity, and hardware requirement for data processing of biosensors, especially image-based ones [129]. As a result, cloud-based or smartphone technologies are coupled with computing algorithms for high throughput and automated quantitative analysis with a short turnaround time. Biosensors that are connected to mobile phones for data processing and signal readout have been used for rapid bacteria detection [130-132]. By leveraging machine learning algorithms and advanced optical transducers in a compact biosensor, accurate optical-based biosensing with intelligent analysis for POC use can be realized. In fact, the same advances in digital technology that are boosting the fortunes of next-generation diagnosis are also extending the lifespan of an image-based biosensor used in the current diagnosis.

CONCLUDING REMARKS

No doubt, the development of a reliable and user-friendly POC biosensor is a formidable task, which will benefit not only to the clinicians but also the public community, addressing the rapidly growing demand on frequent health check-ups, long-term continuous screening of disease biomarkers in an individual. Many technical challenges await innovative solutions, particularly considering the translation into clinical uses for medical applications, such as personalized medicine. Furthermore, clinical evaluation of the diagnostic applicability holds the key to validate a biosensor for clinical requirements. Currently, the duration from evaluation to authority approval by, for example, The Food and Drug Administration (FDA) of United States may take years. While technical innovations may improve the performance factors, revised measures shall also be included to shorten the overall idea-to-market time, to address the increasing demand in healthcare. Overall, the applications of biosensors are obvious and the perspective is optimistic. However, cautions are required for considering the performance factors for different diagnostic applications. For example, an ideal biosensor suited for POC diagnostics shall bear the essentials of cost, speed, portability and device robustness in mind. The ultimate aspiration, as a biosensor engineer, is to witness a shift in current medical intervention from reactive care to predictive or preventive care in future.

CONSENT FOR PUBLICATION

Not applicable.

CONFLICT OF INTEREST

The authors confirm that this chapter contents have no conflict of interest.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the support of the Startup Fund provided by the Chinese University of Hong Kong, the Endowment Fund Research Grant provided by the United College in the Chinese University of Hong Kong (#CA11278), and the General Research Fund provided by the Research Grants Council of the Hong Kong Special Administrative Region (Project No. CUHK 14201317). The authors would also like to acknowledge the support of the Innovative Technology Fund (ITS/061/18, GHX/004/18SZ) and the Area of Excellence scheme funding (AoE/P-0/12) provided by the Hong Kong Special Administrative Region.

REFERENCES