Visible-light and near-infrared fluorescence and surface-enhanced Raman scattering point-of-care sensing and bio-imaging: a review

Abstract

This review article deals with the concepts, principles and applications of visible-light and near-infrared (NIR) fluorescence and surface-enhanced Raman scattering (SERS) in in vitro point-of-care testing (POCT) and in vivo bio-imaging. It has discussed how to utilize the biological transparency windows to improve the penetration depth and signal-to-noise ratio, and how to use surface plasmon resonance (SPR) to amplify fluorescence and SERS signals. This article has highlighted some plasmonic fluorescence and SERS probes. It has also reviewed the design strategies of fluorescent and SERS sensors in the detection of metal ions, small molecules, proteins and nucleic acids. Particularly, it has provided perspectives on the integration of fluorescent and SERS sensors into microfluidic chips as lab-on-chips to realize point-of-care testing. It has also discussed the design of active microfluidic devices and non-paper- or paper-based lateral flow assays for in vitro diagnostics. In addition, this article has discussed the strategies to design in vivo NIR fluorescence and SERS bio-imaging platforms for monitoring physiological processes and disease progression in live cells and tissues. Moreover, it has highlighted the applications of POCT and bio-imaging in testing toxins, heavy metals, illicit drugs, cancers, traumatic brain injuries, and infectious diseases such as COVID-19, influenza, HIV and sepsis.

---

Yingjie Hang

Yingjie Hang is a graduate student under the supervision of Prof. Wu at the University of Massachusettes Amherst. She received her BS degree from Soochow University in 2016. Her research interest lies in point-of-care testing, plasmon-enhanced fluorescence and surface-enhanced Raman scattering (SERS) for protein and nucleic acid detection.

Jennifer Boryczka

Jennifer Boryczka received her MS degree in materials science and engineering from West Virginia University, USA. Currently she is a graduate student under the supervision of Prof. Wu at the University of Massachusetts Amherst. She has co-authored about 5 journal articles. Her research interest lies in plasmon-enhanced fluorescent sensing, imaging and photodynamic therapy.

Nianqiang Wu

Dr Nianqiang (Nick) Wu received his PhD degree in materials science & engineering from Zhejiang University, China. Currently he is an Armstrong-Siadat Endowed Chair Professor in materials science at the University of Massachusetts Amherst. He is Fellow of the Electrochemical Society (FECS) and The Royal Society of Chemistry (FRSC). He received several honors and awards such as Highly Cited Researcher (Clarivate Analytics, Thomson Reuters) and ECS Sensor Division Outstanding Achievement Award. He has authored or co-authored about 200 journal articles, 3 book chapters and 1 book entitled “Biosensors Based on Nanomaterials and Nanodevices”.

---

1. Introduction

Chemical compounds in human fluids and tissues reflect physiological or pathological processes, and can be used for diagnosis of diseases and assessment of medical intervention efficacy. It is essential to perform in vitro analysis of analytes such as metal ions, small molecules, proteins and nucleic acids in human fluids (blood, serum, plasma, urine, saliva and cerebrospinal spinal fluid).^1–5 In most cases, in vitro analysis is performed using large-scale analytical techniques such as spectrophotometry, mass spectrometry, immunoassays, and electrophoresis. For example, enzyme-linked immunosorbent assay (ELISA) and polymerase chain reaction (PCR) techniques are considered the gold standards for quantitation of soluble proteins and nucleic acids, respectively. Although ELISA and PCR are accurate methods, they are expensive, tedious, time-consuming, and require professionals to operate in a central laboratory. These shortcomings have motivated the development of point-of-care testing (POCT) tools to meet the needs of inexpensive, rapid, high-throughput, field-deployable, in vitro analyses by laypersons.^6–8 The emergence of pandemics (such as swine flu (HIN1), severe acute respiratory syndrome (SARS), and COVID-19) and prevalence of epidemic diseases (such as malaria, dengue, chikungunya, yellow fever and Zika) have aroused an increasing demand for POCT tools to rapidly test for these infectious diseases at home, clinics, schools, employer sites, communities or hospitals.

To enable portability, numerous sensors have been developed for in vitro analysis. According to the signal transduction mechanism, sensors can be categorized into different types such as electric, electrochemical, acoustic, magnetic, colorimetric, fluorescent, Raman devices, etc. Among these types of sensors, fluorescence and surface-enhanced Raman scattering (SERS) devices show great potential in in vitro analysis of human fluids due to their high sensitivity and high resistance to interference from sample matrices.^9–11 For sensing applications, visible-light fluorescence and SERS probes are widely used because fluorescence dyes have large fluorescence quantum yields and Raman reporters have large scattering cross-sections, leading to strong signals in buffer sample matrices. Although most NIR-fluorescence dyes exhibit lower quantum yields than visible-light counterparts, and NIR Raman dyes have smaller scattering cross-sections, they are subject to less interference from the sample matrices of blood and plasma due to the reduced light absorption and less auto-fluorescence. On the other hand, it is worth noting that not all sensors are POCT tools because these sensors still require multi-step operations performed by well-trained personnel in a field or even in a laboratory. One of the effective means to enable POCT is the integration of sensors into microfluidic chips to create lab-on-chips, which are characteristic of militarization, field-deployment, automation, swiftness, minimal or non-invasive detection, and easy operation by laypersons.^12,13 The availability of POCT tools will extend in vitro diagnosis (IVD) from hospitals to clinics and communities.

Besides in vitro testing, NIR fluorescence and SERS are finding increasing applications in in vivo imaging due to their minimal or non-invasive detection and relatively deep penetration into human tissues compared to other ultraviolet and visible-light analytical techniques. Current common bio-imaging technologies include magnetic resonance imaging (MRI), computed tomography (CT), ultrasound, X-ray radiography and positron-emission tomography scanning (PETS).¹⁴ However, these techniques are quite time-consuming and require minutes to hours to achieve imaging results. Besides, they are unable to obtain target-specific results and usually need professionals to distinguish the target area from the background. Also, long-time exposure to radiation under X-ray radiography, CT and PETS may cause harm to human health. Thus, it is difficult to obtain long-term and real-time imaging.^15,16 In contrast to these imaging modalities, NIR fluorescence and SERS enable real-time monitoring of expressed genes, pathogens, metabolites, and drug compounds in living cells and tissues. Moreover, the target-specific results enhance diagnostic accuracy.

This article will first clarify the importance of biological transparency windows for selecting NIR fluorescence and SERS for POCT and bio-imaging, and then describe how to use plasmon to enhance the intensity of fluorescence emission and SERS. Next, it will give a summary of the design strategies of fluorescence and SERS sensors in the detection of metal ions, small molecules, proteins and nucleic acids, placing an emphasis on signal transduction and signal amplification. Subsequently, this article will describe a combination of sensors and sample handling modules in single microfluidic devices to create lab-on-chips toward POCT. Lastly, it will give an overview of bio-imaging based on NIR fluorescence and SERS, including instruments, probes and applications, and highlight how the principles of signal transduction and amplification are used in imaging living cells and tissues.

2. Optical properties of SERS and fluorescence

2.1 Biological transparency windows

For optical detection and imaging, an incident light beam is directed towards the sample to excite the light–matter interaction, and a detector is used to collect the light emitted or scattered from the sample. Biological samples may strongly absorb both the incident and the emitting light. Also, auto-fluorescence emission can occur under the excitation of the incident light, leading to interference. These effects may attenuate the incident light quickly with an increase in the penetration depth, reduce the intensity of emitting/scattering light, and increase the signal-to-noise ratio. Such side effects become deteriorated in the ultraviolet (UV) and visible-light spectral ranges, which make it difficult or even impossible for deep-tissue imaging and trace analyte detection. Fortunately, these side effects are minimized in three NIR spectral ranges, which are called “biological transparency windows”. The first biological window covers the wavelength from 650 nm to 950 nm (NIR-I), the second biological window ranges from 1000 nm to 1350 nm (NIR-II), and the third biological window spans from 1550 nm to 1870 nm (NIR-III).^16–18 In short, this interprets why NIR fluorescence sensing or SERS imaging systems should be designed within the biological transparency windows. For example, traditional organic fluorescent dyes exhibit fluorescence in the visible-light or NIR-I window, such as indocyanine green (ICG) and methylene blue (MB) NIR-I probes.¹⁵ Recently, many newly developed organic dyes show the extended emission wavelengths at the NIR-II window.¹⁹ The fluorescence of quantum dots can be tuned from visible-light, NIR-I to NIR-II or NIR-III windows by tailoring the size or/and chemical composition. Semiconducting single-walled carbon nanotubes exhibit fluorescence emission in the NIR spectral range mainly between 900 and 1600 nm, corresponding to NIR-II or NIR-III windows.¹⁶

Fig. 1 shows the light absorption spectra obtained from human skin, which guides the selection of biological transparency windows for in vivo optical imaging (IVI).¹⁶Fig. 2 reveals the molecules that cause strong light absorption and scattering in human tissues.²⁰ The major light-extinction molecules in human tissues are, but not limited to, melanin, collagen, fat, water and blood (e.g., Ox-Hgb and D-Hgb). Light below 650 nm is strongly absorbed by oxy-hemoglobin (Ox-Hgb) and deoxy-hemoglobin (D-Hgb), while water is the main absorber in the spectral range over 1350 nm (Fig. 2).²⁰ It can be seen from Fig. 1 that both absorption and scattering of light are reduced dramatically when the wavelength increases from the UV and visible-light regions to the NIR regions. As the wavelength falls into the biological transparency windows, light scattering decreases significantly, and light absorption is maintained at a low level in the spectral range of 500 nm to 1300 nm with a small increase in the NIR-III region. Therefore, optical loss in human tissues is minimal in the three biological transparency windows.

Fig. 1 Optical loss of human skin. Reproduced from ref. 16, copyright 2016, The Royal Society of Chemistry.

Fig. 2 Light absorption spectra with X-axis on the natural logarithm base for different chromophores present in human tissues. Spectra for oxy-hemoglobin (Ox-Hgb) and deoxy-hemoglobin (D-Hgb), proteins, water, collagen, fat and cytochrome oxidase (CtOx) in the region from 100 nm to 3000 nm with respect to the specific concentration in mM are shown. This figure is further modified from ref. 20, copyright 2014, Elsevier.

It is worth noting that light scattering is evident in a narrow spectral region (1350–1550 nm) between the NIR-II and NIR-III windows. Also, a narrow spectral region (950–1000 nm) between the NIR-I and NIR-II windows should be avoided for bio-imaging and detection although light scattering is negligible. Light is absorbed in this spectral region by water in biological samples, as shown in Fig. 3, and the absorbed light can heat biological samples.²¹ This is why NIR upconversion rare-earth nanoparticles such as NaYF₄:Yb,Er, which are typically excited with a 980 nm laser, may encounter a problem when they are used for bio-imaging and detection.

Fig. 3 Light absorption and autofluorescence species in human blood as well as excitation and emission wavelengths of common fluorophores and SERS probes. Rectangles represent the auto-fluorescence spectral bands of plasma/blood proteins including pyridoxine (Pyr), retinol A (Ret), collagen (Col), cholecalciferol (Chol), folic acid (FA), and porphyrin (PP); diamonds represent the absorbent peaks of oxy-hemoglobin (OxHe) and deoxy-hemoglobin (D-He); hexagons represent the excitation wavelength of SERS probes; circles represent the excitation and emission of NIR fluorescence probes including indocyanine green (IGG), Ag₂S quantum dots (Ag₂S QDs), single-walled carbon nanotubes (SWCNTs), infrared fluorescent proteins (IFPs) and lead halide perovskites (LHPs). The light spectra in this figure were acquired in Wu's lab. Some data in this figure were taken from ref. 21 with permission from the Oxford University Press.

In buffer sample matrices, a better limit of detection can be achieved using visible-light fluorescence probes because of their higher fluorescence quantum yields than their NIR counterparts. However, the visible-light fluorescence intensity can be reduced largely in blood or plasma matrices due to the strong light absorption and auto-fluorescence of the sample matrix. In contrast, the intensity of NIR fluorescence (or SERS) probes may be reduced slightly or even not changed when the sample matrix is switched from buffer to blood/plasma. Hence, NIR fluorescence and SERS probes are attractive for in vitro diagnosis (IVD), especially for blood or plasma samples. Fig. 3 shows the light absorption spectra obtained from human blood and plasma, which guides the selection of excitation light sources, dyes and fluorophores for in vitro optical detection in blood or plasma sample matrices. This figure reveals the major molecules in blood that cause strong light absorption and auto-fluorescence.²¹ The light-absorption molecules include oxy-hemoglobin (Ox-Hgb), deoxy-hemoglobin (D-Hgb) and bilirubin. D-Hgb exhibits the strongest light absorption at around 420 nm and the second light absorption peak at around 580 nm. Ox-Hgb displays its strongest light absorption at around 410 nm, and two secondary peaks at 550 nm and 600 nm. Overall, light absorption by Ox-Hgb and D-Hgb decreases with an increase in the wavelength, and becomes much weaker in the spectral range above 770 nm. Auto-fluorescence leads to a low signal-to-background ratio. The typical auto-fluorescence molecules in blood are amino acids and proteins, pyridoxine (Pyr), retinol A (Ret), collagen (Col), cholecalciferol (Chol), folic acid (FA), and porphyrin (PP).^22–25 These molecules may exhibit strong auto-fluorescence in the visible-light regions, as shown in Fig. 3. Interestingly, auto-fluorescence becomes low in both the NIR-I and NIR-II windows in a blood sample matrix. The authors' group has compared the behavior of visible-light and NIR fluorescence probes in buffer and plasma, respectively. It has been found that the fluorescence intensity of the visible-light fluorescence dye (rhodamine 6G) was reduced by more than 50% when the sample matrix was switched from buffer to blood or plasma. In stark contrast, the NIR fluorescent dye (Cyanine 7) exhibited a negligible change in the fluorescence intensity.

The prevailing colorimetric paper-based lateral flow assays employ gold nanoparticles as colorimetric probes. These gold nanoparticles typically exhibit a light absorption maximum at around 520–560 nm, showing a red wine-like color on the test line of the assay, which is visible to human eyes. When these colorimetric paper-based lateral flow assays are applied to blood, serum and plasma samples, they suffer from severe interference of biomolecules in sample matrices, leading to poor sensitivity and repeatability. In contrast, NIR fluorophores and SERS probes that fall into the biological transparency windows (NIR-I and NIR-II) have advantages when testing plasma, serum and whole blood samples. Typically, the 785 nm or 1064 nm laser is used to excite SERS probes. And the typical examples of NIR fluorophores are organic dyes (e.g., indocyanine green, IGG), quantum dots (QDs, e.g., Ag₂S), carbon nanomaterials (e.g., single-walled carbon nanotubes, SWCNTs), infrared fluorescent proteins (IFPs) and halide perovskite nanocrystals (e.g., lead halide perovskites, LHPs).^{14,19,26–29}

2.2 Surface plasmon resonance

Although NIR fluorescence or SERS detection systems in the biological transparency windows are characteristic of the reduced light absorption and the minimized autofluorescence from biological samples, their signals are relatively low compared to their visible-light counterparts. NIR fluorophores typically exhibit lower quantum efficiency than visible-light counterparts. The quantum yield of visible-light fluorophores typically ranges from 20% to 80%, and is reduced to ∼10% for NIR-I fluorophores, and only ∼0.01–1.4% for NIR-II counterparts except for some nanomaterials such as PbS quantum dots.³⁰ On the other hand, Raman scattering of a molecule is a dipole-like interaction with the excitation light, which shows that Raman scattering strength is proportional to the frequency of the excitation light to the fourth power. Consequently, the Raman scattering signal is greatly reduced when shifting from the visible-light to the NIR spectral regions. Therefore, surface plasmon resonance (SPR) is typically tuned into the NIR spectral region to enhance NIR fluorescence and SERS. Thus, we introduce the SPR concept and the plasmon-enhancement principles before we talk about NIR fluorescence and SERS sensing/imaging. Understanding the underlying mechanism will enable “devices-by-design” and “materials-by-design”.

SPR is a phenomenon that numerous conduction electrons collectively oscillate on the surface of a metallic nanostructure upon the resonant excitation of incident light.³¹ It can generate an intense local electromagnetic (EM) field surrounding the nanostructure, which can modulate optical processes including photon absorption, Raman scattering and fluorescence.³² SPR has two typical forms, namely, propagating surface plasmon polariton (SPP) and localized SPR (LSPR). SPP refers to electron oscillation on the surface of a thin metal film when the frequency of incident light is smaller than the plasma frequency (Fig. 4a).³² The SPP frequency is determined by the plasma frequency of the bulk metal (ω_p) and the dielectric constant of the surrounding medium (ε_diel), which is expressed as^33,34

Here, ω_p is the plasma frequency of the bulk metal and is expressed aswhere ω_p is determined by the charge carrier density (n) and charge (e) of electrons, the effective mass (m_eff) and the permittivity of free space (ε₀).³⁵ The generated EM field can extend into the dielectric environment, which decays in ∼100–200 nm.³⁶ When it comes to nanoparticles (NPs), plasmon will occur in the form of LSPR (Fig. 4b)³² if the size of the metallic NPs is comparable or smaller than the wavelength of the incident light. The electron density of the nanoparticles shows a uniform displacement.³² A strong restoring force is then formed due to the positive ionic core background, which ultimately results in a characteristic oscillation frequency in the metal electrons like a simple harmonic oscillator.^37,38 In the Mie theory or a simple harmonic oscillator model, the extinction cross-section of light absorption and scattering induced by LSPR in a nanosphere is given by³⁷in which ω is the frequency of light; V is the volume of the spherical nanoparticle; ε_diel is the dielectric constant of the surrounding medium; and are the real and imaginary parts of the dielectric function, respectively.³⁸ The denominator can vanish when is negative and is small at the required wavelength range such as the near-infrared range. Strong plasmon resonance could be achieved at^39,40In addition, the factor of 2 in front of ε_diel is a variation dependent on the shape of the nanoparticles.³² Therefore, the LSPR frequency is dependent upon the nanoparticle shape in addition to other two factors (metal types and surrounding environment). The EM field generated by LSPR shows a shorter decay (∼10–30 nm) compared to the EM field from SPP (∼100–200 nm).^41,42 Specifically, nanoparticles (NPs) with sharp edges or a small gap (<10 nm) between neighboring plasmonic NPs, “hot spots” are formed where the EM field is concentrated with further enhanced field intensity (Fig. 4c and d).^43,44 In addition, after LSPR is excited by incident light, both energy absorption and re-radiation occur along with light absorption and scattering in plasmon. Generally, no obvious Stokes shift was observed between the absorption and scattering spectra of the NPs, since the radiative decay of LSPR is often earlier than the non-radiative process.⁴⁵ However, the occurrence of absorption and scattering depends on the particle size. For metal nanoparticles smaller than 15 nm, light absorption dominates, then plasmons absorb energy and quickly transfer it into heat.^46,47 For nanoparticles larger than 15 nm, plasmons tend to re-radiate energy, inducing light scattering with a large scattering cross-section.^46,47

Fig. 4 (a) The SPP-induced EM field and the decay. (b) The LSPR-induced EM field and the decay. (c) Finite-difference time-domain (FDTD) simulated EM field distributions for NPs with small gaps or sharp edges. (d) FDTD simulated EM field distributions for nanostars. Reproduced with permission from (a) ref. 32, copyright 2015, The Royal Society of Chemistry; (b) ref. 32, copyright 2015, The Royal Society of Chemistry; (c) ref. 44, copyright 2011, American Chemical Society; (d) ref. 45, copyright 2012, IOP Publishing Ltd.

2.3 SERS-enhancement mechanisms and SERS probes

Mechanisms. Raman scattering is extremely inefficient. The cross-section of Raman scattering (10⁻²⁸–10⁻³⁰ cm² per molecule) is ∼14 orders of magnitude weaker than the fluorescence cross section of fluorophores (10⁻¹⁶ cm²).^48,49 Fortunately, this problem is mitigated by SERS, in which Raman scattering is enhanced by molecules adsorbed on metallic nanostructures or rough surfaces, or semiconductors.⁵⁰ SERS signals are typically enhanced by chemical enhancement or/and electromagnetic enhancement (EM) mechanisms.^51,52 The chemical enhancement mechanism relies on charge transfer between the metal substrate and Raman molecules, which is estimated in two cases.^53–55 For the first case, charge transfer forms a new energy state of the metal substrate–molecules system. Such charge transfer itself could increase Raman scattering, and the resonance of the charge transfer system further enhances the SERS signal after it is excited by incident light. For the second case, a new electron transfer pathway is formed either from the highest occupied molecular orbital (HOMO) of the molecule to the unoccupied state above the Fermi level of the metal substrate, or from the lowest unoccupied molecular orbital (LUMO) of the molecule to the occupied state below the Fermi level of the metal substrate, assuming that the Fermi level of the metal substrate could match with the HOMO or LUMO of the molecule.

The EM enhancement mechanism is based on the enhanced localized EM field around a plasmonic metal nanostructure.^11,56 In the case of the influence of both chemical enhancement and EM enhancement, the SERS signal intensity (P_SERS) can be estimated by⁴²

P_SERS ∝ chemical enhancement × EM enhancement (5)

P_SERS ∝ N·α^R_abs·|A(ν_e)|²·|A(ν_s)|²·I(ν_e) (6)

where N is the number of molecules absorbed on the metal substrate; α^R_abs is the Raman probability of the absorbed molecules; A(ν_e) and A(ν_s) represent the enhancement factors of excitation and Raman scattering light, respectively; I(ν_e) shows the intensity of excitation light. In this equation, α^R_abs is the main contributor to the CE mechanism since I(ν_e) influenced by CE is negligible, while A(ν_e) and A(ν_s) are the main contributors to EM enhancement. By comparison, the enhancement factor from EM enhancement (between 10⁴ and 10⁸) is much larger than that from CE mechanism (between 10 and 100), thus EM enhancement is dominant in SERS enhancement.⁵⁷ By just considering EM enhancement, the SERS enhancement factor (EF) can be determined as⁴²where P_0,SERS represents the SERS signal intensity without EM enhancement. E^l and E^LOC are the amplitudes of the excitation and local electric fields, respectively. ω_e is the frequency of excitation light and ω_s is the frequency of Raman scattering light. Because the Raman shift between the excitation and Raman scattered frequency is smaller than the width of the LSPR range, the Raman shift can be ignored (that is, ω_e is approximate to ω_s), which gives |A(ν_e)|² ∼ |A(ν_s)|². Therefore, the SERS enhancement factor by the EM field (EF_EM) can be approximated by⁵⁴In other words, when molecules are in the enhanced localized EM field, both the intensity of the incident light and the Raman scattering light of the molecules are enhanced. The SERS signal is maximal with an |E⁴| enhancement when both the excitation and Raman scattered frequency are in resonance with the LSPR frequency. Following the EM enhancement mechanism, the Raman reporter molecules should be placed as close to the plasmonic metal as possible when designing SERS probes. This is because the plasmon-included EM field usually decays exponentially away from the plasmonic nanoparticles. The design of SERS probes is different from that of plasmonic fluorescence probes where the fluorophores are typically kept at a distance of 10–30 nm away from the plasmonic nanoparticles.

Apart from these two traditional SERS enhancement mechanisms, a new mechanism named “molecular hot spots” was proposed recently.⁵⁸ It is found that the polarizability of aromatic molecules conjugated on a plasmonic metal nanoparticle interacts strongly with SPR, leading to a spatially extended enhanced EM field into the conjugated aromatic molecule assembly. This creates “molecular hot spots” surrounding the metal nanoparticle. The enhanced EM field is delocalized and spreads throughout the aromatic molecule layer on the surface of the metal nanoparticle. The “molecular hot spots” can be generated in aromatic molecules even on a small metal nanoparticle (<15 nm) with a weak LSPR. The “molecular hot spots” can result in a large enhancement of SERS signals of other molecules that are co-adsorbed with the aromatic molecules on the metal substrate. Even linear-chain molecules, which typically have extremely low polarizability, can exhibit strong SERS signals when they co-exist with aromatic molecules near the metal nanoparticle.

SERS probes. Based on the EM enhancement mechanism, SERS probes have been designed typically with three types. The first type is “Raman dyes”.^59–61 To enable high sensitivity, Raman reporter dyes require relatively large scattering cross sections, such as molecules with high polarizability moieties, e.g., conjugated π-systems.^61,62 For spectral multiplexing properties, it is necessary for molecules to have a minimal number of vibrational Raman bands and no spectral overlap. Because Raman dyes have no SERS effect themselves, Raman dyes must be coupled to a rough metal surface or a plasmonic metal nanostructure to enable EM enhancement during SERS detection. Raman dyes may be modified with additional functional groups to allow for covalent bonding to the surface of metals.^61,63 Based on these criteria, many “Raman dyes” have been selected or developed, including benzotriazole derivatives (e.g., rhodamine 6G, R6G), malachite green isothiocyanate (MGITC), and olefin- or alkyne-based molecules (e.g., 5-ethylnyldeoxyuridine, EdU).⁵⁹

The second type is “metal@Raman reporter”, in which Raman reporter molecules are attached to a metal nanoparticle. Because the Raman scattering cross-section of Raman reporter molecules is quite small, only 10⁻²⁸–10⁻³⁰ cm² per molecule, they alone cannot be used as SERS probes.⁴⁹ Typically, plasmonic metal nanoparticles such as Au or Ag are utilized to generate LSPR to amplify Raman signals. Because LSPR-induced EM fields are decayed to a very low level beyond a distance of about 30 nm away from the particle surface, Raman reporters are required to stay in close proximity to the metal nanoparticles. Thus, metal@Raman reporters are formed through surface functionalization of plasmonic nanoparticles with Raman reporters.⁴² SH- or NH₂-containing molecules with great affinity to the Au or Ag substrate are usually chosen as Raman reporters, such as malachite green isothiocyanate (MGITC), 4-aminothiophenol (4-ATP), 4-mercaptobenzoic acid (4-MBA), 4-mercaptopyridine. (4-MPY), and 5,5-dithiobis-2-nitrobenzoic acid (DTNB).^64–66 To further enhance the SERS signal of Raman reporters, many researchers have engineered metal nanoparticle shapes from nanospheres to nanorods and nanostars. It was reported that more sharp tips in nanoparticles generate stronger EM fields, thus nanostars showed the highest SERS enhancement, nanorods took the second place, while nanospheres exhibited the lowest enhancement.⁴⁴ The “metal@Raman reporter” particles may aggregate in an aqueous solution with high ionic strength or get detached from the metal particle surface during sensing/imaging, which leads to signal loss and poor repeatability.^44,67 On the other hand, it is difficult to conjugate additional antibodies or DNA into the “metal@Raman reporter” particles due to the occupancy of Raman dyes on the metal surface.

The third type is sandwich-structured “metal core@Raman dye@shell” particles,⁴⁴ in which the Raman reporter molecules are sandwiched between a metal core and an outer shell. This prevents from leaching out of Raman reporter molecules during detection. The plasmonic metal core is used to amplify SERS signals. The thin outer shell is typically made from silica (SiO₂) or polymers such as polyethylene glycol (PEG).^44,68–70 The silica shell not only provides a clean surface for bioconjugation with antibodies and DNA, but also renders excellent water-solidity, which make the sandwich-structured SERS probe stable in an aqueous solution with high ionic strength such as blood. In short, sandwich-structured SERS probes are characteristic of high sensitivity, robustness, water-solubility, and biocompatibility and repeatability.⁴⁴

In general, both the second and the third types of SERS probes can be used in buffer sample matrices for sensing applications. Also, the sandwich-structured “metal core@Raman dye@shell” SERS probes are stable and robust in blood and plasma matrices. However, the “metal@Raman reporter” SERS probes may aggregate in blood or plasma due to the high ionic strength of sample matrices.

2.4 Plasmon-enhanced fluorescence mechanisms and NIR fluorophores

Mechanisms. SPR is the most common means for enhancing fluorescence. Gold and silver are the most popular plasmonic materials, hence, plasmon-enhanced fluorescence (PEF) is sometimes called “metal-enhanced fluorescence (MEF)” or “surface-enhanced fluorescence (SEF)”. The mechanisms of plasmon-enhanced fluorescence are described in detail in the literature.^32,71 Briefly, when a fluorophore is located near a plasmonic nanostructure, energy transfer can occur between the plasmon and the fluorophore via the Förster resonance energy transfer (FRET) process or/and the Purcell effect.^32,71 For metal nanostructures which primarily scatter, enhancement will be possible through both FRET at short distances (∼10 nm) and the Purcell effect at long distances (10–50 nm). At short distances (∼1–10 nm), FRET can take place from the donor to the acceptor via the dipole–dipole interaction.^32,71 It occurs when the emission band of the donor overlaps with the absorption band of the acceptor, and the fluorescence of the donor is then quenched. The FRET efficiency (Eff_FRET) depends on the dipole-to-dipole separation distance (R) and the Förster distance (R₀) (usually ranges from 3 nm to 8 nm) of this pair of dipoles, as expressed by⁷²The efficiency of FRET decays at 1/R⁶, while the Purcell effect scales with 1/R³, and becomes dominant in longer distances (∼10–50 nm). The Purcell effect originates from a resonant cavity that modulates the local density of optical states (LDOS) that is expressed as⁷³

ρ_LDOS(ω) ∼ |E_loc(ω)|² (10)

where |E_loc(ω)|² represents the normalized local electric field of the cavity according to the incident field intensity. When the radiative dipole is in the resonant cavity of plasmonic nanostructures and the plasmon's absorption or scattering band is overlapped with the fluorescence emission band, the LDOS increases, leading to a higher rate of fluorescence emission into the cavity as compared to the dipole in air. The transferred energy can be absorbed or re-radiated as scattering by the cavity, leading to the quenching or enhancement of fluorescence emission.

Considering the spectral overlap of the plasmon and the fluorescence, and the gap, quenching or enhancement of fluorescence emission can be determined (Fig. 5a and b).³² First, when the plasmon band is overlapped with the fluorophore absorption band in Fig. 5a, fluorescence emission can be enhanced.³² It can be illustrated as follows. For metal nanoparticles smaller than ∼15 nm, the particles mainly absorb and generate an enhanced EM field, leading to excitation enhancement by FRET, followed by emission enhancement. It should be noticed that plasmon's radiation enhanced by the Purcell effect is negligible, thus excitation is only enhanced by FRET within a few nanometers of distance in this case. For metal nanoparticles larger than ∼15 nm, the particles mainly scatter, and both FRET at short distances and the Purcell effect at long distances can result in enhanced fluorescence emission.^32,46,74 Second, when the plasmon band is overlapped with the fluorophore emission band in Fig. 5b, quenching or enhancement of fluorophore emission can be achieved.³² At a short distance (∼10 nm) between the plasmon and the fluorophore, fluorescence emission can be quenched through FRET. At a distance beyond FRET (>10 nm), strong Purcell enhancement is induced because of the large local density of states (LDOS) of the local plasmonic field, which increases the radiative rate of the fluorophore, and if the plasmon can scatter efficiently, strong emission enhancement can occur.^32,46,74 Typically, the maximum enhancement takes place at an optimal gap of 10–30 nm between the plasmon and the fluorophore. Besides the overlap of the plasmon and absorption/emission band, it is important to choose a suitable laser line. For the overlap of the plasmon and absorption band, a signal laser line is enough to excite both the plasmon and fluorescence probe. When the plasmon is overlapped with the emission band, two laser lines are required. One laser line with a shorter wavelength is used to excite the fluorescence probe, and another with a longer wavelength can excite the plasmon.

Fig. 5 Mechanisms of plasmon-enhanced fluorescence. (a) When the plasmon overlaps with the fluorophore absorption spectrum, an enhanced fluorophore emission can be achieved. (b) When the plasmon overlaps with the fluorophore emission spectrum, quenching or enhancement of fluorophore emission can be achieved. (c) When the spectrum peak of the plasmon is longer than the absorption of the fluorophore, fluorescent intensity can be enhanced through the PIRET process if there is an overlap between the plasmon spectrum and fluorophore absorption. Reproduced with permission from (a) ref. 32, copyright 2015, The Royal Society of Chemistry; (b) ref. 32, copyright 2015, The Royal Society of Chemistry; (c) ref. 45, copyright 2016, American Chemical Society.

Recently, a new mechanism, plasmon-induced resonance energy transfer (PIRET), is proposed for plasmon enhanced fluorescence.⁷⁵ It is also a non-radiative energy transfer via the dipole–dipole interaction. Different from FRET which only occurs from the donor at a shorter wavelength to the acceptor at a longer wavelength, PIRET can realize the blue-shift coherent energy transfer from the longer-wavelength donor to the shorter-wavelength acceptor. For a plasmon–fluorophore pair, when the SPR band of the plasmon is overlapped with the absorption band of the fluorophore, the emission intensity of the fluorophore can be enhanced through PIRET. The schematic mode of energy transfer is shown in Fig. 5c.⁴⁵ The plasmon spectrum often overlaps simultaneously with the absorption and emission spectrum of the fluorophore, such that FRET occurs to quench fluorescence and PIRET takes place to enhance fluorescence simultaneously. The net influence of FRET and PIRET is determined by the dephasing time (T₂), which is expressed by^45,75

where T₁ refers to the recombination time of relaxation from the excited state to the ground state of the plasmon (or the fluorophore); refers to the pure dephasing time of the plasmon (or fluorophore). Given (around fs range) is much shorter than T₁ (around ps range), and T₂ is related to . If the fluorophore's dephasing time is faster than that of the plasmon, PIRET shows higher efficiency than FRET and the net influence is enhanced fluorescence. Conversely, FRET dominates in the whole process with the result of quenched fluorescence.

NIR fluorophores. Various NIR fluorophores are commercially available, such as organic dyes, inorganic quantum dots (QDs), carbon nanomaterials, halide perovskite nanocrystals and fluorescent proteins.^{14,19,26–28} Plasmonic fluorescence probes are typically formed by the placement of the above-mentioned fluorophores near a plasmonic nanoparticle with an optimal gap of 10–30 nm. It is necessary to have a spacer to keep an optimal gap to prevent the quenching of fluorescence. The spacer can be a thin SiO₂ layer, a polymer, a double-stranded DNA, an aptamer, etc.^76–81 For example, an Au nanostar@SiO₂@double strand DNA-labeled Cy5.5 probe was created, showing the enhanced NIR fluorescence.⁸² A 22 nm-thick SiO₂ shell has optimized the fluorescence enhancement to 21 folds. Besides silica, polymers can also be used as a shell or spacer to separate NIR fluorophores from plasmonic nanoparticles.^79,83 In addition, the biotin–avidin complex has also been used as the spacer.⁸³ The biotinylated bovine serum albumin was self-assembled on the surface of the Au nanostar, followed by the incorporation of streptavidin-labeled fluorophores via the biotin–avidin interaction. Because the biotinylated bovine serum albumin and the streptavidin are around 8 nm and 4 nm in size, respectively, a 12 nm gap was created between the fluorophores and the Au nanostar, which activated the Purcell effect for fluorescence enhancement.

Plasmonic fluorescence probes based on PIRET show advantages in bio-sensing and bio-imaging.^45,75,84,85 Because it is a blue-shifted energy transfer from the longer-wavelength end to the shorter-wavelength end, excitation can reduce photodamage to biological samples. The signal-to-noise ratio can also be improved. For example, Wang et al. developed Au nanorods covered by a merocyanine-doped poly(methyl methacrylate) film.⁴⁵ It was observed that PIRET occurred when there was a significant overlap between the LSPR band and the fluorophore absorption band, which resulted in an average enhancement factor of up to 1854. As PIRET is a newly discovered mechanism for fluorescence enhancement, plasmonic fluorescence probes based on PIRET have just emerged.

In a special case, when Raman scattering and fluorescence spectra are overlapped for a fluorescent molecule at a given excitation wavelength, that is, both SERS and fluorescence are enhanced by the plasmonic electric field, the total cross-section (σ_(total)) of the spectrum is theoretically the sum of the Raman scattering cross-section spectrum (σ_(RS)) and the fluorescence cross-section spectrum (σ_(FL))⁸⁶

σ_(total) = σ_(RS) + σ_(FL) (12)

Owing to enhancement by the EM field, the total cross-section becomes

σ_{(enhanced total)} = σ_(ERS) + σ_(EFL) (13)

σ_(ERS) = σ_(RS)·|A(ν_e)|²·|A(ν_s)|². (14)

It is noted that |A(ν_e)|²·|A(ν_s)|² is an enhancement factor of electric field intensity. The σ_(EFL) item in eqn (13) includes the product of EM enhancement of absorption and fluorescence. The EM enhancement of absorption (|A(ν_e)|²) is equivalent to an increase in the effective absorption cross-section. The EM enhancement (|A(ν_s)|²) is due to the emission enhanced by the coupling of the plasmon resonance with fluorescence light. Simultaneously, fluorescence may be quenched by a factor of q through a resonance energy transfer process. Thus, these synergistic effects lead to

σ_(EFL) = q·σ_(FL)·|A(ν_e)|²·|A(ν_s)|² (15)

Combining eqn (13), (14) and (15),

σ_{(enhanced total)} = [σ_(RS) + qσ_(FL)]·|A(ν_e)|²·|A(ν_s)|² (16)

The enhanced light intensity is proportional to the enhanced total cross-section, and dependent on the fourth power of the electric field.

3. Design of fluorescence and SERS sensors

In this section, we review the design of fluorescence and SERS sensors for the detection of typical analytes in healthcare and biomedical research. We discuss how to transduce and amplify sensing signals. In addition, we highlight the application examples of fluorescence and SERS sensors.

Signals of both fluorescence and SERS can be amplified dramatically by surface plasmon resonance, achieving high sensitivity of detection, even reaching a single molecule level. It is worth noting that SERS signals increase with a decrease in the distance between the Raman reporter and the plasmonic nanostructure while the fluorescence signals of a fluorophore reach a maximum at an optimal distance between the fluorophore and the plasmonic nanostructure. On the other hand, both NIR fluorescence and SERS have strong resistance to interference from complex sample matrices (blood and plasma) and have deep penetration capability in bio-imaging. When designing sensors, keep in mind that few analyte molecules have signature or characteristic fluorescence peaks, while many analytes can generate fingerprinting or signature SERS spectra. Therefore, SERS can be used to conduct label-free detection of many trace analyte molecules, and is good at real-time monitoring metabolites from live cells. Low-cost fluorescence readers are well developed, and a lot of commercial products are available. In contrast, today Raman readers are more expensive than their fluorescence counterparts. But rapid technology progress will increase portability and reduce the costs of Raman readers in the future. Indeed, this trend was expedited in the last decade. Palm-sized Raman readers, which can be operated with cell phones via Bluetooth, are available commercially now. As for in vitro imaging, both modern fluorescence and Raman microscopes are able to real-time image live cells quickly. As for in vivo imaging, commercial fluorescent probes and imaging instruments are also well developed and available commercially. In vivo collection of Raman signals from deep tissues remains a challenge.

3.1 Protein sensors

Significance. In humans, proteins play important roles in multiple facets of life: growth and maintenance of cells and tissues, delivery of messages, retention of mechanical integrity of cells and tissues, regulation of pH values, transport and storage of nutrients, supply of energy, performing reactions, and fighting infection. The state-of-the-art ELISA can achieve a LOD down to ∼10 pM toward protein measurement in serum. But the levels of proteins range from 10⁻¹⁶ to 10⁻¹² M in serum in some cases of diseases.^87,88 Hence, it is essential to develop highly sensitive sensors to meet such needs.

Signal transduction. In sensor design, proteins are selectively recognized by specific ligands, such as antibodies, aptamers, peptides and polymers through affinity, and electrostatic and hydrophobic interactions. Several signal transduction modes have been developed for protein detection. The first one is the formation of a sandwiched antibody–antigen–antibody structure (Fig. 6a). The capture antibody is initially immobilized on a solid substrate, and the detection antibody labelled with a fluorescence or SERS probe is suspended in a solution. In the presence of an analyte (antigen), the antigen will bind to both the capture and the detection antibodies, forming a sandwich structure and immobilizing the optical probes close to the solid substrate. After washing non-specific biomolecules away, the number of immobilized optical probes reflects the analyte concentration in a sample. The second mode is a competitive immunoassay (Fig. 6b), in which a detection antigen labelled with an optical probe is initially bound to the capture antibody fixed on a solid substrate. When an analyte (antigen) is present, it will bind to the capture antibody because the analyte has a higher affinity to the capture antibody than the detection antigen. In the meanwhile, the detection antigen labelled with an optical probe will go away due to loss of binding competition. Besides antibodies, aptamers (single-stranded DNA or RNA) are alternative candidates of molecular recognition elements to proteins. They can replace antibodies in a sandwiched assay (Fig. 6c). An aptamer can also serve as a molecular beacon to transduce fluorescence or SERS signals (Fig. 6d). Initially, a fluorophore and a gold nanoparticle are linked to two ends of a hairpin-structured aptamer, respectively, and the emission of the fluorophore is quenched by the gold nanoparticle via FRET. When an analyte (antigen) is present in the assay, the analyte will bind to the aptamer and open the loop of the hairpin, pushing the gold nanoparticle away from the fluorophore. This turns on the emission of the fluorophore. If the fluorophore is substituted by a Raman dye, SERS signals will scatter from the Raman dye due to EM-enhancement by the nearby plasmonic gold nanoparticle. When an analyte is present in the assay, SERS signals will be turned off due to the moving of the gold nanoparticle away from the Raman dye.

Fig. 6 Signal transduction modes/principles for protein detection. (a) Sandwiched antibody–protein–antibody assay; (b) competitive assay; (c) sandwiched aptamer–protein–aptamer assay; (d) aptamer beacon that can change its conformation.

Applications. Following the above signal transduction modes, various NIR fluorescence or SERS sensors are built. Wu's group developed a sandwich-structured SERS immunosensor, where the Au nanostar@Raman dye@SiO₂ nanoparticles acted as SERS probes and the Au triangle nanoarray served as the solid substrate (Fig. 7a).⁸⁷ Compared to traditional Au nanosphere particles and a planar Au film, a high density of “hot spots” in a 3D hierarchical architecture was generated by coupling the Au nanostars to the Au triangle nanoarray in the presence of an antigen analyte, amplifying the SERS signals remarkably. It showed a wide dynamic linear range from 0.1 pg mL⁻¹ to 10 ng mL⁻¹ at a LOD of 7 fg mL⁻¹ toward human immunoglobulin G (IgG) protein in a buffer solution. It eliminated multiple washing steps and directly detected the vascular endothelial growth factor (VEGF) in the blood plasma samples of clinical breast cancer patients. These antibody/antigen/antibody sandwich immuno-sensors are most common for protein detection.^89–96 For example, the graphene quantum dots (QDs)-capture antibody/CA-125 antigen/detection antibody labelled with a horseradish peroxidase (HRP) (Ab-HRP) sandwich immuno-sensor was designed for the detection of the CA-125 ovarian cancer biomarker.⁹⁷ When the antigen (CA-125) was present, chemiluminescence catalyzed by HRP was quenched by the graphene QDs via a resonance energy transfer process. In addition, many efforts have been made to develop optical probes in the sandwich immuno-sensor. Apoferritin with a cavity structure is an intriguing example.^97,98 It not only exhibits biocompatibility but also entraps tens of fluorophores or quantum dots into the cavity for signal amplification.⁹⁷

Fig. 7 (a) ELISA-like SERS immunosensor using the antibody–antigen interaction for cancer biomarkers in blood plasma. (b) ELISA-like fluorescence immunosensor using the aptamer–antigen interaction. (c) Competitive assay based on FRET mechanism. Reproduced with permission from (a) ref. 87, copyright 2013, American Chemical Society; (b) ref. 100, copyright 2011, American Chemical Society; (c) ref. 102, copyright 2014, MDPI.

Antibodies are expensive and produced by a complicated synthesis process, and require low temperature for storage. Hence, the substitution of antibodies with aptamers are receiving more attention because aptamers possess tailored specificity and affinity, relatively high thermostability, and can be produced massively to almost any epitope in a programable route.⁹⁹Fig. 7b shows an aptamer-based sandwich assay for the detection of a cancer biomarker in buffer and diluted human serum.¹⁰⁰ A pair of aptamers was immobilized on the surface of recombinant fluorescent ferritin nanoparticles as an optical probe, and the targeted cancer biomarker was sandwiched by the pair of aptamers. Such an aptamer sensor achieved a LOD down to 100 fM in a buffer.

Competitive immunoassays simplify the procedure and require the reduced steps of operation compared with sandwich-structured assays.^91,101–103Fig. 7c reveals a FRET-based competitive sensor for the detection of the cyclic adenosine 3′,5′-monophosphate (cAMP) protein that was used for the evaluation of G protein-coupled receptor activation.¹⁰² Prior to detection, the fluorescence of the donor dye was quenched, but the fluorescence was emitted from the acceptor dye due to the FRET effect. When the analyte (cAMP) appeared in the assay, the FRET process was interrupted, and the donor emitted fluorescence. Competitive assays can also employ aptamers as molecular recognition elements.^104–107 Xing et al. designed an aptamer-based competitive sensor for the determination of human epidermal growth factor receptor 2 (HER2).¹⁰⁵ Initially, the HER2-specific aptamer linked with MnCuInS/ZnS QDs was hybridized with a complementary single-stranded DNA (ssDNA) linked to gold nanoparticles. The double-stranded DNA (dsDNA) was rigid and held the Au nanoparticles close to the QDs, and enabled FRET, leading to quenching of fluorescence of the QDs. During the detection process, the binding of the aptamer to HER2 led to the dissociation of dsDNA due to the higher affinity of the aptamer with HER2. As a result, the ssDNA became free-standing, and the gold nanoparticles were kept away from the QDs, which turned on the fluorescence of the QDs. This aptamer-based FRET sensor achieved a LOD of 1 ng mL⁻¹ with a detection range from 2 to 100 ng mL⁻¹, and successfully distinguished the healthy individuals (2–15 ng mL⁻¹) from the individuals with the disease (15–75 ng mL⁻¹).

3.2 Nucleic acid sensors

Significance. Nucleic acid sequences are usually used to identify a particular species or subspecies of organisms.^108,109 Either deoxyribonucleic acid (DNA) or ribonucleic acid (RNA) acts as a carrier of genetic information for viruses and bacteria. DNA expression is an important index in evaluating the number of bacteria, such as S. aureus, S. typhimurium and Escherichia coli.^110,111 Typical DNA-based viruses include hepatitis B virus (HBV), herpes virus and papilloma virus, which seriously threaten human health. Early detection of viruses is essential for the prevention and treatment of diseases.¹¹² Different from the double-helix structure of DNA, RNA is often found as a single-stranded molecule in nature. It is the main genetic material in most viruses, such as SARS, COVID-19, Ebola, human immunodeficiency virus (HIV), influenza virus and tobacco mosaic virus.^6,113,114 RNA also plays significant roles in coding, decoding and expression of genes, among which miRNAs (20–24 nucleotides) are a group of endogenous and noncoding RNAs, which are essential in post-transcription of gene expression. They are closely associated with proliferation, differentiation, development and metabolism in cells. Hence, they act as valuable biomarkers in various physiological disorders like tumors, neural diseases, cardiovascular diseases and diabetes.^115,116 Detection of nucleic acids is significant for early diagnosis of many diseases because nucleic acids are expressed at a very early stage of diseases while antigens or/and antibodies appear in human fluids typically later or even much later than nucleic acids.

Nucleic acid amplification. In many cases, the expressed nucleic acids are at a very low level (10–1000 copies per mL). Therefore, amplification is a necessary step prior to detection. Taking advantage of the chain reaction of nucleic acids, a variety of methods have been explored to make copies of nucleic acids, such as the polymerase chain reaction (PCR), loop-mediated isothermal amplification (LAMP), recombinase polymerase amplification (RPA), strand displacement amplification (SDA), rolling circle amplification (RCA), ligase chain reaction (LCR) and sustained sequence replication (3SR).^117–122Fig. 8 exemplifies the four mechanisms of nucleic acid amplification that are most commonly used in analytical devices.

Fig. 8 Schematic illustration of nucleic acid amplification. (a) Polymerase chain reaction (PCR). (b) Loop-mediated isothermal amplification (LAMP). (c) Recombinase polymerase amplification (RPA). (d) Strand displacement amplification (SDA). Reproduced with permission from (a) ref. 118, copyright 2013, Elsevier; (b) ref. 123, copyright 2019, PLOS; (c) ref. 125, copyright 2020, MDPI; (d) ref. 120, copyright 2014, American Chemical Society.

(i) PCR is carried out by temperature cycling (Fig. 8a).¹¹⁸ At a high temperature of the denaturation step, two strands of duplex DNA are fully separated. The temperature is then reduced to allow primers to anneal to each ssDNA. In the presence of deoxynucleoside triphosphates (dNTPs) and heat-stable polymerase, primers are extended to form new double helical DNA at around 72 °C. Theoretically, the number of double-stranded DNA pieces is doubled in each cycle, achieving 2ⁿ copies of DNA after n cycles in the PCR process.

(ii) LAMP is performed isothermally at 60–65 °C for 45–60 min with the aid of the specific DNA polymerase, dNTPs and primers (Fig. 8b).¹²³ The LAMP process starts when the F2 region of the forward inner primer (FIP) hybridizes to the F2c region of the target DNA and extends the strand, followed by the F3 primer which is complementary to the F3c region of the target elongating the strand. The FIP-linked strand is then displaced and forms a loop at the 5′-terminus, which also acts as a template for the backward inner primer (BIP). The B2 region of the BIP undergoes a similar strand synthesis as the F2 region and opens the 5′-terminus loop. The synthesized strand is then displaced by the B3 primer extended strand, forming a dumbbell-shaped DNA, which is an initiator of LAMP recycling. As a result, a mixture of stem-loop DNA structures with varied stem lengths and various cauliflower-shaped structures with multiple loops is obtained assisted by these primers.

(iii) RPA was developed by Niall Armes et al. in 2006 using three proteins (recombinase protein, single-stranded binding protein and a strand displacing DNA polymerase).¹²⁴ A typical example is shown in Fig. 8c.¹²⁵ The recombinase protein (T4 UvsX protein) binds to primers in the presence of a loading factor (T4 UvsY) and adenosine triphosphate (ATP) to form a recombinase-primer complex that seeks homologous sequences in duplex DNA and inserts the strand by primers at the cognate site. A single-stranded binding protein (SSB, T4 gp32 protein) is used to prevent the ejection of the invaded primer by branch migration. Subsequently, the recombinase protein disassembles from the complex and a strand displacing DNA polymerase (Bsu or Sau) binds to the 3′-end of the primer and initiates the elongation assisted by dNTPs, ultimately resulting in the exponential accumulation of the target double helical DNA.

(iv) SDA is also carried out under in vitro isothermal conditions. Fig. 8d gives an example of SDA for miRNA amplification.¹²⁰ Primers used in SDA consist of a particular sequence (marked in red in Fig. 8d) as the recognition site of the nicking enzyme. SDA starts with the binding of the target miRNA to Primer 1 at the 3′-end, followed by the extension of Primer1 and miRNA in the presence of Klenow fragment polymerase. The nicking enzyme is added to cleave Primer 1 at the recognition site. Next, the elongation happens again at the nicking site, thus displacing the replicated strand that serves as a new template in the following step and the left duplex strand enters the nicking step to initiate the first cycle. The replicated strand bonds with Primer 2 to trigger the second cycle through extension, nicking, polymerization and displacement steps and ultimately achieves the new target miRNA, which could also enter the first cycle.

Signal transduction. In biosensors, nucleic acids are usually recognized specifically by complementary base-pairing. Fig. 9 shows the five signal transduction modes that are usually used in optical sensors. Fig. 9a reveals a classical “three-strand” mode, in which the capture strand is grafted on a solid substrate and the detection strand functionalized with an optical probe is suspended in the liquid. The analyte (single-stranded DNA or RNA) will be hybridized to the complementary capture and detection strands to form a rigid double-stranded DNA (or RNA). In Fig. 9b, a fluorophore-labeled detection single strand is initially immobilized onto a graphene oxide sheet via π–π stacking and hydrogen bonding between the single-stranded DNA and graphene oxide. In this case, the emission of the fluorophore is quenched by graphene oxide via FRET. After the analyte (ssDNA) and detection ssDNA form a double-stranded DNA (dsDNA), the interaction between dsDNA and graphene oxide is too weak to overcome the interfacial electrostatic repulsion. Consequently, the dsDNA is pushed away from graphene oxide and disables FRET, turning on fluorescence emission. Fig. 9c demonstrates the signal transduction using a molecular beacon (MB), which is a single-stranded oligonucleotide containing 25–35 nucleotides.¹²⁶ In the absence of the analyte, it is folded into a stem-loop (hairpin) structure, where the loop part consists of 18–30 nucleotides, and the stem part comprises 5–8 base pairs labeled with a fluorophore; the other end is immobilized onto a graphene oxide sheet or a plasmonic metal nanoparticle. In this case, the emission of the fluorophore is quenched by graphene oxide or a plasmonic metal nanoparticle via FRET. When the loop part is hybridized to the complementary analyte (ssDNA), the loop will be opened, forming a rigid dsDNA, which can accurately control the gap between the fluorophore and the graphene oxide sheet (or plasmonic metal nanoparticle). This is because the inter-distance between two adjacent base-pairs in rigid dsDNA is 0.3 nm; the gap between the two ends of dsDNA can thus be accurately tailored by the number of base-pairs with an increment of 0.3 nm. If the gap between the two ends of dsDNA is large enough beyond the FRET effect, the fluorophore at one end will emit. Fig. 9d illustrates the signal transduction triggered by polymerase enzyme, where the target sequence is first hybridized with a probe-labeled strand and a primer which is located at one end of the analyte. Polymerase enzymes are added to elongate the primer and digest the probe-labeled strand, leading to the release of the optical probe. After the collection and detection of the released optical probes, the concentration of the analyte is derived eventually. Fig. 9e presents a special enzyme-assistant assay using a CRISPR/Cas12a system. With the guidance of CRISPR RNA (crRNA), CRISPR–Cas12a can recognize the analyte through a T nucleotide-rich protospacer-adjacent motif (PAM) site on the target dsDNA.^110,127 The formed Cas12a–crRNA–analyte dsDNA ternary complex will activate the trans-cleavage activity of Cas12a to arbitrarily cut the fluorophore quencher-labeled ssDNA reporters, leading to the recovery of fluorescence.

Fig. 9 Signal transduction modes for nucleic acid detection. (a) “Three-strand” hybridization mode; (b) “two-strand” hybridization mode-based competitive assay; (c) “molecular beacon” mode; (d) enzyme-assisted assay using polymerase; (e) enzyme-assisted assay using CRISPR/Cas12a.

Applications. Fig. 10a illustrates a classical “three-strand” mode-based DNA sensor.¹¹² The analyte, hepatitis B virus (HBV) DNA, brought the SERS probes (Ag nanorice@Raman label@SiO₂ nanoparticles) close to a gold triangle nanoarray pattern, which created a 3D plasmonic field, and resulted in a SERS enhancement factor (|E|⁴) of ∼40 000, achieving a LOD of 50 aM in a buffer solution with specific discrimination of a single-base mutant of DNA. It is desirable to combine nucleic acid amplification and detection in a single platform to perform the nucleic acid amplification test (NAAT).^128–131 Guven et al. incorporated RCA with the “three-strand” detection scheme to detect the 35S promoter from cauliflower mosaic virus (CaMV35S) for tracking gene-modified organisms.¹²⁹ After the analyte sequence was sandwiched by the primer and the capture strand on a gold substrate, the primer was catalyzed by DNA ligase and hybridized with an RCA template. With the assistance of Phi29 DNA polymerase and dNTPs, long nucleic acid products were generated via the RCA reaction, followed by hybridization with the detection strand-modified gold nanorods. This NAAT device reached a LOD of 6.3 fM with a wide detection range from 100 fM to100 nM.

Fig. 10 Nucleic acid detection with SERS or fluorescent sensors. (a) “Three-strand” hybridization assay using SERS probes for the detection of hepatitis B virus DNA. (b) “Two-strand” hybridization mode-based competitive interaction between single-stranded and double-stranded complex graphene oxide. (c) “Molecular beacon” mode for the detection of influenza virus RNA. (d) Nucleic acid amplification test (NAAT) using self-circulation of chain reaction for signal amplification and “molecule beacon” mode for detection. (e) Taq polymerase-assisted assay. (f) CRISPR/Cas12a-assisted assay with fluorescence probes determined by the size of metal nanoparticles and their distance. Reproduced with permission from (a) ref. 112, copyright 2013, American Chemical Society; (b) ref. 132, copyright 2013, American Chemical Society; (c) ref. 136, copyright 2016, Elsevier; (d) ref. 135, copyright 2019, The Royal Society of Chemistry; (e) ref. 137, copyright 2012, The Royal Society of Chemistry; (f) ref. 138, copyright 2020, American Chemical Society.

Fig. 10b shows a sensor based on the “two-strand” mode as described in Fig. 9b.¹³² It has achieved a LOD of 1 pM toward miRNA in a buffer solution. The “two-strand” mode has also been used in NAAT.^133–135 For instance, a biosensor was constructed via target cycling amplification to detect highly pathogenic avian influenza H7N9 virus.¹³³ The number of analyte DNA after recycling was dependent on the released capture strands which were immobilized onto the surface of Ag nanorods. The detection strands labeled with the fluorescent dye Cyanine-5 (Cy5) were finally hybridized with the capture strands to record the analyte concentration. This NAAT assay exhibited a linear range from 1 fM to 100 pM at a LOD of 31 aM for H7 and 44 aM for N9 in buffer solutions, respectively.

Molecule beacons have been widely used in SERS and fluorescence sensors for nucleic acid detection.^{126,134–136,139}Fig. 10c reveals a NIR fluorescent sensor base on a molecule beacon labeled with CdZnSeTeS QDs, showing a LOD of 2 copies per mL towards influenza virus H1N1 RNA in both buffer and human serum.¹³⁶ In addition, the molecular beacon mode-based detection scheme has been incorporated with nucleic acid amplification methods. Fig. 10d presents a NAAT assay using self-circulation of chain reaction between the target miRNA and a molecular beacon (A) together with other three hairpin structures without probes (B, C, and D).¹³⁵ In the presence of the analyte miRNA-34a, a breast cancer biomarker, the molecular beacon was unfolded via hybridization with miRNA-34a, forming a duplex with B (A–B) spontaneously. This released the target to activate the first cycle steps 1–2. The A–B duplex entered the next cycles by the formation of the C–D duplex, which catalyzed A and B to generate C–D in turn. This sensor detected 2.5 pM miRNA-34a in a buffer.

The enzymatic cleavage of phosphodiester linkages of DNA or RNA is also common in sensor design (Fig. 9d).^140–142 Harper et al. utilized this method in a SERS sensor for the determination of mecA gene sequence of methicillin-resistant S. aureus (MRSA) bacterium (Fig. 10e).¹³⁷ In this design, a sequence complementary to a specific region of mecA gene was modified with a 5′-biotin residue and a TAMRA dye-labeled 10 adenine bases at the 3′-end, respectively, which was termed as the TaqSERS probe. When the TaqSERS probe and a primer were hybridized to the target sequence, the Taq polymerase enzyme extended the primer with the addition of deoxynucleoside triphosphate (dNTP) and digested the TaqSERS probe simultaneously. The TAMRA dye-labeled DNA was then separated in the magnetic field after any free biotin and undigested TaqSERS probes were absorbed by streptavidin coated magnetic beads. The separated dye labeled DNA was aggregated by the negatively charged citrate surface of the Ag nanoparticles. In turn, these DNA sequences neutralized the charge of Ag nanoparticles, facilitating the formation of a nanoparticle cluster. Compared with the fluorescence intensity of the dye alone, the incorporation of the amplification method resulted in the LOD being an order of magnitude lower. Moreover, nucleic acid amplification by LAMP was combined with the enzyme-assisted detection scheme.¹⁴¹ Briefly, after amplification by LAMP, the products of the analyte sequence of DNA were complementary to a capture ssDNA strand labeled with Cy5 and Au nanoparticles. The Cy5 and Au nanoparticles were linked by oligonucleotides in close proximity (<2 nm), enabling remarkable enhancement of the SERS signal from Cy5. Upon adding the S1 nuclease, the SERS signal remained at high intensity in the presence of the target DNA strands. Otherwise, the S1 nuclease digested ssDNA into free dNTPs, and separated Cy5 away from the nanoparticles, leading to low SERS signals. Apart from the amplification feature, nuclease was used to digest any non-specific DNA or potential contamination to minimize the interference to SERS signals. This sensor was used to detect Salmonella enterica, a major foodborne pathogen, achieving a LOD of 66 CFU mL⁻¹ in a buffer.

CRISPR/Cas systems are adaptive immunity found in bacteria and archaea to degrade intruding nucleic acids, which have been wildly developed as gene-editing tools because of their recognition and degradation ability toward the target nucleic acids induced by Cas protein under the direction of a guide RNA. CRISPR/Cas systems have been extended to sensor design for nucleic acid testing.^{110,127,138,143,144}Fig. 10f shows the use of a CRISPR/Cas12a system to design a sensor for the detection of breast cancer gene-1 (BRCA-1).¹³⁸ The sensor probe was designed as 60 nm sized Au nanoparticles (60-AuNP), and 20 nm sized Au nanoparticles (20-AuNP) were linked to a long fluorophore-labeled dsDNA (7 nm) and a short ssDNA (2 nm). Because the fluorescence probe was close to 60-AuNP, it was quenched by 60-AuNP at the initial state. When the analyte BRCA-1 gene was present, it was complementary to crRNA and formed a Cas12a–crRNA–BRCA-1 complex, activating the CRISPR/Cas system to cleave ssDNA in the surrounding environment. The ssDNA in the probe was then cut to dissociate the fluorophore from 60-AuNP. As a result, a fluorescence signal was emitted and enhanced by the 20-AuNP via FRET. The combination of CRISPR/Cas with FRET achieved a femtomolar level of LOD both in buffer and under human serum conditions within a short assay time (<30 min).

3.3 Small molecule sensors

Significance. Several types of small molecules affect human life and the surroundings, including inherent functional molecules, exogenous nutrients and prescribed medicine, as well as toxins. Inherent functional molecules, such as adenosine triphosphate (ATP), glucose, bilirubin, and non-essential amino acids (such as glycine, cysteine, alanine and tyrosine), play vital roles in regulating physiological and pathological processes.^4,145–147 For example, ATP transports chemical energy for cellular metabolism and is involved in various life activities, such as liver glycogen metabolism, cell proliferation, muscle contraction and nerve conduction.^64,145,148 The expression of ATP out of the intracellular concentrations (1–10 mM) is the sign of diseases including cardiovascular and Parkinson's disease.¹⁴⁸ Glucose exists as a main energy resource in human adult blood with a normal level less than 100 mg dL⁻¹ (5.6 mmol L⁻¹) after an overnight fast.¹⁴⁹ Diabetes is confirmed if the blood glucose level reaches 126 mg dL⁻¹ (7 mmol L⁻¹) or higher on two separate tests. Diabetes is a global metabolic disease diagnosed by over 463 million people worldwide.¹⁴⁹ Bilirubin is created from biliverdin catalyzed by enzyme biliverdin reductase during heme moiety catabolism and has a potent antioxidant capacity to avoid reverting to biliverdin.^146,150 It is normally excreted in urine and bile to maintain a safe range (0.2–1.2 mg dL⁻¹) of total serum bilirubin in a human adult; otherwise, excess accumulation of bilirubin is responsible for hyperbilirubinemia and jaundice.¹⁵⁰

Nutrients such as essential amino acids and vitamins cannot be synthesized to a sufficient amount in the body and must be obtained from food, and are indispensable for human health with limited amount.^147,151 Specifically, in the family of essential amino acid, methionine is associated with greying of hair, depression and atherosclerosis; lysine plays a crucial role in proteinogenesis and epigenetic regulation; valine takes part in blood sugar regulation.^152–157 In addition, adequate vitamin intake is important for the enzyme–substrate reaction, growth and development. The family of B complex vitamins forms coenzymes to facilitate catalysis processes. For instance, thiamine (B1) can function as thiamine pyrophosphate (TPP) in the body and this coenzyme is a vital player in carbohydrate metabolism.¹⁵⁸ B9 participates in the synthesis of nucleic acids, metabolism of amino acids and cell proliferation.¹⁵⁹ It is reported that long-term supplementation of B9 can lessen the risks of stroke, heart disease and prostate cancer.¹⁶⁰ The deficiency of these amino acids and vitamins causes functional disorders, for example, insufficiency of vitamin E can lead to myopathies, spinocerebellar ataxia, peripheral neuropathy and retinopathy, etc.^161,162

Illicit drugs, pesticides, herbicides, and food pollutants pose a threat to human health and the environment. Abuse of illicit drugs is a global problem with 271 million people (1 out of 18 people) from 15 to 64 years old.¹⁶³ Drug abuse is one of the sources for hepatitis C and a quarter with human immunodeficiency virus (HIV).¹⁶⁴ Therefore, drug testing is essential for drug control and prevention. Nitrite, the main species in nutrient pollution, is of concern to the ecological system.⁷⁰ It is also widely present in meat products as food preservatives and excessive consumption causes diseases such as diabetes, cancer and neurodegenerative diseases.^42,165 The United States Environmental Protection Agency (EPA) and the Legislation of the European Union have enacted the maximum contaminant level (MCL) of nitrite as 1.0 μg mL⁻¹ in drinking water and the maximum allowable level as 50 mg kg⁻¹ in meat products.^70,165 Release of pesticides and herbicides into surface waters can result in pollution of the environment, accumulation in the food chain, uptake by human eventually, leading to adverse health effects, etc.^166,167

Signal transduction. In Fig. 11, several signal transduction modes have been developed for the detection of small molecules. Like the detection of proteins, the capture antibody/analyte/detection antibody sandwich mode (Fig. 11a), the molecular beacon-based mode (Fig. 11b) and the competitive assay (Fig. 11c) are also commonly used for signal transaction in small molecule sensors. Details of these three modes have been described above. Taking advantage of SERS, two unique signal transduction modes have been employed for small molecule detection. As shown in Fig. 11d, aromatic compounds, such as amphetamine-type stimulants, dyes and phenolic pollutants, are SERS-active due to their high polarizability. They can be brought close to a plasmonic substrate for SERS detection. For the SERS-inactive analyte molecules, they may in situ react with specific reagents in an assay to produce SERS-active molecules. The SERS-active products can be brought close to a plasmonic substrate for SERS detection (Fig. 11e).

Fig. 11 Signal transduction modes for small molecule detection. (a) Sandwiched antibody–molecule–antibody assay; (b) molecular beacon assay; (c) competitive assay; (d) direct detection using a substrate with “hot spots” to enhance Raman scattering; (e) target detection through monitoring their chemical reaction product using a SERS sensor.

Applications. Antibodies are widely used as molecular recognition elements in small molecule detection.^168–171 And signal transduction is typically realized by a sandwich-structured assay of the capture antibody/molecule/detection antibody-labeled with an optical probe.¹⁷¹ Such a type of assay with CdSe/ZnS QDs as fluorescence probes exhibited a LOD of 0.3 and 0.4 ng mL⁻¹ toward clothianidin and thiacloprid in a sodium borate buffer, respectively. This sensor can be employed in different applications, such as river water, soil, cabbage, rice and tomato. In the last decade, more and more aptamers appeared to substitute antibodies as molecular recognition elements for small molecule detection.^{164,172–174} Guler et al. decorated QDs on the surface of μ-well plates and further grafted the cocaine-specific aptamer labeled with Au nanoparticles on QDs.¹⁷² When cocaine was present, it was bound to the aptamer and made the aptamer folded to reduce the distance between the QDs and the Au nanoparticles, quenching the fluorescence of the QDs. It showed a LOD of 42 pg mL⁻¹ in a PBS buffer, which was much lower than that of traditional liquid chromatography-tandem mass spectrometry (LC-MS, 0.162 ng mL⁻¹) and capillary electrophoresis-electrospray ionization-mass spectrometry (CE-ESI-MS, 100 ng mL⁻¹). This sensor was also able to distinguish cocaine even from its metabolite benzoylecgonine. With the rapid development of specific ligands, a competitive assay using different affinities is well established according to the signal transduction mode mentioned in Fig. 11d. A competitive biosensor in Fig. 12a was constructed using high affinity between the morphine and Au nanoparticles because of the overall effects of the phenolic hydroxy groups as well as the tertiary nitrogen ring atom in the morphine.¹⁷⁵ In this design, fluorescein was first quenched by colloidal Au nanoparticles via a nanometal surface energy transfer (NSET) process. After adding morphine, it coordinated with Au nanoparticles and introduced the aggregation of nanoparticles since it had two phenolic hydroxy groups, thus fluorescein was departed from the nanoparticles and turned on its fluorescence. This competitive assay utilized the specific interaction of morphine with Au nanoparticles, and exhibited high selectivity and sensitivity without interference from metal ions to other illicit drugs, and achieving a LOD of 53 pM, 67 pM, 62 pM and 78 pM in buffer, serum, urine and urine, respectively. Competitive assays recently became more popular for small molecule detection due to the availability of more commercial aptamers.⁶⁴ In Fig. 12b, a gold film was initially modified with ssDNA that was hybridized to an aptamer labeled with a SERS probe (Au nanostar@malachite green isothiocyanate (MGITC)@SiO₂ nanoparticles).⁶⁴ Upon adding ATP molecules, the noncanonical G/A base pairs were disrupted in the duplex DNA due to the higher affinity of ATP with the aptamer, leading to the detachment of the aptamers from the gold surface. This sensor showed a very low LOD of 12.4 pM toward ATP detection in a buffer solution because strong SERS signals originated from the enhanced electromagnetic field around Au nanostars as well as “hot spots” generated between Au nanostars and the Au substrate.

Fig. 12 Detection of small molecules with fluorescence or SERS sensors. (a) A competitive assay based on the different affinities of fluorescein and morphine to Au nanoparticles. (b) Molecular beacon assay using SERS probe-linked aptamer for ATP detection. (c) SERS sensors with a “built-in” electrostatic preconcentration unit for the capture of the antibiotic analyte. (d) In situ generation and detection of SERS-active H₂O₂ for glucose detection. (e) In situ generation of SERS-active azo-moieties for nitrite detection. Reproduced with permission from (a) ref. 175, copyright 2018, Springer Nature; (b) ref. 64, copyright 2012, American Chemical Society; (c) ref. 166, copyright 2017, Wiley-VCH; (d) ref. 149, copyright 2014, Elsevier; (e) ref. 70, copyright 2018, Elsevier.

Many small molecules contain benzene rings, which are usually SERS-active. Such molecules can be trapped on a plasmonic substrate to trigger SERS signals. Following this principle, various SERS sensors have been constructed.^42,176,177 Huang et al. fabricated an Au@Ag nanopillar array with a 10 nm of gap between adjacent pillar tips.¹⁷⁶ Due to its enhancement factor of up to 10⁷, this SERS sensor was able to detect a 5 × 10⁻⁶ M 2,3,3′-trichloribiphenyl (PCB20) pesticide. To create stronger “hot spots”, Zhu et al. developed a hierarchically ordered array of Ag-nanorod bundles with about 2 nm of gaps between adjacent nanorods.¹⁷⁸ This increased the SERS enhancement factor to 10⁸. Furthermore, this sensor was also able to distinguish multiple analytes simultaneously, which benefited from the narrow band of each peak in SERS spectra. Using a Ag nanorod bundle array, methyl parathion had characteristic bands at 852, 1109, 1260, and 1328 cm⁻¹ while 2,4-dichlorophenoxyacetic acid (2,4-D) exhibited specific peaks at 436, 915, 1298, 1373, and 1620 cm⁻¹. The strong EM field enhancement (E/E₀ ∼ 3 × 10⁴) of the Ag nanorod bundle array not only achieved high sensitivity toward methyl parathion (a LOD of 21.5 nM) and 2,4-dichlorophenoxyacetic acid (2,4-D, a LOD of 61.9 nM) in water, but also discriminated two pollutants in a mixture. To trap and preconcentrate the charged or polar analytes, electrophoresis can be integrated into SERS sensors because charged or polar analytes are driven to the oppositely charged electrode under the electric field. This process was employed for the selective detection of polar molecules by altering potential polarity. Recently, a copper microgrid, which was modified with germanium nanowires and further decorated with Ag nanoparticles, served as both an electrode and a SERS substrate (Fig. 12c).¹⁶⁶ After applying a voltage, the evenly distributed analytes in the solution were forced to move along the electric field to the SERS substrate, leading to a high concentration near the substrate. In this case, polar antibiotics, 2.4 nM 6-aminopenicillanic acid and 0.9 nM penicillin G, were detected, which were approximately 5.7 and 2.8 folds of improvement by electrostatic preconcentration.

In situ detection of chemical reaction products of an analyte can result in low interference and a high signal-to-noise ratio. Fig. 12d shows a SERS sensor for the detection of glucose in saliva.¹⁴⁹ The Au nanostar@SiO₂ core–shell nanoparticles were functionalized with glucose oxidase enzyme. The sensor catalyzed glucose to produce H₂O₂ that exhibited a finger-print SERS peak. The produced H₂O₂ molecules were close to the surface of core–shell nanoparticles, which were subject to a strong electromagnetic field generated by the “hot spot” near the Au nanostar tips, amplifying SERS signals remarkably. This sensor response was not affected by the co-existing ascorbic acid and uric acid, and it showed a dynamic range of concentration from 25 μM to 25 mM toward glucose detection in saliva. An additional intriguing example is that the product from the in situ chemical reaction served as a linker, which was covalently bound to two adjacent SERS substrates.⁷⁰ In their design, the Ag nanopyramid array and the Au nanostars were modified with 1-naphthylamine (1-NA) and 4-aminothiophenol (4-ATP), respectively (Fig. 12e).⁷⁰ After an analyte (nitrite) was present in the assay, it reacted with 1-NA and 4-ATP to form the azo group that was sandwiched between the Au nanostars and the Ag nanopyramid array. Owing to a small gap, the “hot spots” between the Ag nanopyramid and the Au nanostars significantly enhanced the SERS signals of the azo group, and showed a LOD of 0.6 pg mL⁻¹ toward nitrite detection in water with a wide linear detection range from 1 pg mL⁻¹ to 10 μg mL⁻¹, which covered the MCL of 1.0 μg mL⁻¹ of nitrite in drinking water regulated by the U.S. EPA.

3.4 Metal ion sensors

Significance. Metal ions include essential metals in living organisms (potassium, sodium, copper, zinc, iron ions, etc.) and toxic ions (mercury, silver, lead, chromium, arsenic ions, etc.). The former ions play an important role in various physiological processes such as enzyme catalysis, cell growth and metabolism, while the toxic ions pose a threat to human health and the ecological environment.^11,179,180 For example, potassium (K⁺) and sodium (Na⁺) ions are critical in many biological functions, including management of blood pressure and pH, nerve, muscle and enzyme activity.^181,182 The interaction of heavy metals, such as mercury (Hg²⁺) and silver (Ag⁺), with biological systems threatens human health, and even a low level of intake makes it difficult to metabolize them.^69,76,183 Metal ions are usually measured by atomic absorption spectroscopy and inductively coupled plasma mass spectrometry (ICP-MS) in a central laboratory, which are expensive, tedious, time-consuming and require professional personnel to operate.¹⁸⁴ Therefore, portable sensors show a great potential in the on-site measurement of metal ions.

Signal transduction. Metal ions alone are neither SERS-active nor fluorescence-active. To enable signal transduction in SERS or fluorescence sensors, they need to be specifically coordinated with small molecules or aptamers to form complexes prior to detection. Fig. 13 illustrates several signal transduction modes for metal ion detection in SERS or fluorescence sensors. For example, a fluorescent molecule is initially quenched via a photo-induced electron transfer (PET) process (Fig. 13a). After the molecule is bound to the metal ion, the PET process is interrupted, triggering fluorescence. In Fig. 13b, plasmonic metal nanoparticles, which are functionalized with SERS-active ligands with selective affinity to specific metal ions, are initially dispersed in a solution. When the metal ion analytes are present in the assay, the open end of the ligand will bind to the metal ions, which induces aggregation of plasmonic nanoparticles given that another end of the ligand is covalently linked to the plasmonic nanoparticles. As a result, the SERS-active ligand is located between the close-contact plasmonic nanoparticles, leading to strong SERS enhancement. Fig. 13c shows signal transduction via DNA hybridization. Initially, two ssDNA cannot be hybridized owing to their mismatched sequences deliberately designed. In the presence of the target metal ions, the metal ions will bind to ssDNA on the mismatched sites, leading to the hybridization of DNA. In Fig. 13d, an aptamer labelled with a fluorophore is initially immobilized on a graphene oxide sheet due to the π–π interaction between the aptamer and the graphene oxide sheet. At this moment, the emission of the fluorophore is quenched by graphene oxide via FRET. When an analyte is present, it will coordinate with the aptamer to form a G-quadruplex or a hairpin structure, breaking the π–π interaction. This G-quadruplex or hairpin structure will be repelled away from the graphene oxide sheet, turning on fluorescence emission.

Fig. 13 Signal transduction modes/principles in SERS or fluorescence sensors for metal ion detection. (a) “One-to-one” model using single coordinate site between the optical probe and target ion; (b) aggregation of optical probes induced by multiple coordinate sites; (c) DNA hybridization assay; (d) molecular beacon-based competitive assay.

Applications. The photo-induced electron transfer (PET) process in a single molecular entity has been used to design fluorescent sensors for metal ion detection (Fig. 13a).^{179,185–187}Fig. 14a shows a molecule that consists of a fluorophore (rhodamine analog, Rh) and a triazacryptand (TAC) moiety that selectively binds to potassium ion, which forms an electron donor–accept structure.¹⁸¹ In the absence of K⁺ ions, PET led to weak fluorescence. After the formation of a coordination complex with a K⁺ ion, the PET process was inhibited, leading to strong fluorescence. Since TAC not only acted as an electron donor in sensing K⁺ but also had the ability to target mitochondrion, a sensor was employed to detect intracellular K⁺. A wide detection range of 16 to 400 mM was successfully measured, showing a linear range of 100–300 mM. Taking advantage of the large coordination number of metal ions, complexes could form with two or more molecules.^185,188 For example, gold nanoparticles were grafted with 4-mercaptobenzoic acid (MBA) for Hg²⁺ detection (Fig. 14b).¹⁸⁸ In the absence of Hg²⁺, the functionalized gold nanoparticles were dispersed and repelled from each other due to the negative charging of the carboxylic terminal in MBA. At this moment, the SERS signals of the MBA molecule were very weak. When Hg²⁺ ions were present in the assay, Hg²⁺ ions were bound to the carboxylic moieties, leading to a linkage between the nanoparticles. This caused the aggregation of Au nanoparticles, creating strong “hot spots” between neighboring nanoparticles. The MBA molecules were thus exposed to strong electromagnetic field in the “hot spots”, amplifying the SERS signals of MBA remarkably. This simple SERS sensor showed a LOD of 5 ppt toward Hg²⁺ in water.

Fig. 14 Detection of metal ions by the formation of metal ion–ligand complexes. (a) Coordination complex formed with one metal ion and one fluorescence probe, inhibiting photo-induced electron transfer (PET) to enhance the intensity of fluorescence. (b) Coordinate interaction induced aggregation of gold nanoparticles for SERS enhancement. (c) SERS probes coupled with an ordered Au nanohole array. (d) Competitive assay by the formation of a G-quadruplex/Pb²⁺ complex. (e) Catalytic nucleic acid (DNAzymes) activity activated by specific metal ions. Reproduced with permission from (a) ref. 181, copyright 2020, The Royal Society of Chemistry; (b) ref. 188, copyright 2017, IOP; (c) ref. 69, copyright 2015, The Royal Society of Chemistry; (d) ref. 189, copyright 2013, Elsevier; (e) ref. 190, copyright 2018, American Chemical Society.

Fig. 14c shows a SERS sensor based on the metal ion-induced DNA hybridization (Fig. 13c).⁶⁹ A single-stranded DNA (ssDNA) was initially immobilized on a gold nanohole array substrate, and another ssDNA labelled with a SERS probe (gold nanostar@Raman-reporter@silica nanoparticle) was suspended in the solution. Because Hg²⁺ and Ag⁺ showed strong affinity to thymine nucleobase (T) and cytosine nucleobase (C) to form the T–Hg²⁺–T and C–Ag⁺–C structures, respectively, the presence of analytes resulted in the formation of double-stranded DNA and moved the SERS probes close to the gold nanohole array substrate. The “hot spots” were formed between the gold nanostars and the Au nanohole array, amplifying the SERS signals. The sensor exhibited a LOD of 2.3 pM for Hg²⁺ and 0.17 nM for Ag⁺ in human saliva. This sensor can be used to monitor the release of metal ions from the silver–mercury amalgams in dental fillings.

Because aptamers can be programmed to selectively bind to different analytes, there is plenty of room for the application of aptamers in sensors.^76,182,189 Wu's group used an aptamer to construct a fluorescent sensor for lead(II) ion detection (Fig. 14d).¹⁸⁹ Initially, an aptamer conjugated with QDs was immobilized on the graphene oxide sheet through π–π stacking interactions. Upon binding to Pb²⁺ ions, the aptamer was transformed to a G-quadruplex/Pb²⁺ complex, turning on the fluorescence of the QDs. This sensor reached a LOD as low as 90 pM and was successfully applied to river water. Additionally, a special aptamer with catalytic nucleic acids (DNAzymes) activity, which can be activated by metal ions, provided an alternative method for detection.^191,192 Lu's group demonstrated this method by conjugating Zn²⁺-specific DNAzyme with a quencher (Q2) at the end to lanthanide-doped upconversion nanoparticles, NaYF4:Yb,Tm@NaYF4 UCNP (Fig. 14e).¹⁹⁰ The substrate strand of DNAzyme was linked with a fluorophore (F) and another quencher (Q1) at each end, respectively. The synergistic effect of Q1 and Q2 allowed complete quenching of the fluorophore in the initial state. The Zn²⁺ ions activated the catalytic ability of DNAzymes, leading to the cleavage of the scissile ribonucleotide adenosine (rA) and the dehybridization of the substrate strand. This triggered the emission of the fluorophore. In this design, the scissile ribonucleotide adenosine (rA) in the middle of the substrate strand was specifically protected by a 2′-nitrobenzyl photocage group (PG) that was photo-dissociated when the upconversion nanoparticles converted 980 nm NIR light into 365 nm emission. This showed a precise control of the location and time in detection by regulating NIR light. Similarly, Wang et al. introduced it into a SERS sensor where the gold-coated substrate was functionalized by DNAzyme. A cleavage substrate was used as a bridge with the bottom part hybridizing with the DNAzyme and the upper part hybridizing with the Au nanoparticles.¹⁹³ Upon the addition of Pb²⁺, the SERS signal was reduced due to the detachment of Au nanoparticles from the gold surface.

4. Optofluidic devices toward point-of-care testing

Biosensors, which have been described in Section 3, have provided convenience for in vitro diagnosis. However, in many cases, these sensors have to be operated manually by professionals in a central laboratory, and thus cannot be used in POC settings. An effective solution is to incorporate SERS or fluorescent sensors into a microfluidic platform to form “all-in-one” optofluidic devices, which can pretreat samples, transport liquid, and detect analytes in a single chip. To enable POCT, optofluidic devices are desired to be miniaturized, portable, automotive, non- or minimally invasive, can be read out by a handheld or a portable reader, and operated by a layperson with minimized or no training. In this section, we highlight some important features and research progress in optofluidic devices that are relevant to POCT. In POC settings, devices are applied to various sample matrices such as water, buffer, saliva, urine and blood. Blood testing is most challenging among all these sample matrices in POC settings. So far, few commercial portable devices are available for direct and quantitative measurement of analytes in finger-prick whole blood samples in POC settings.

4.1 Criteria and features of point-of-care-technologies

Many people call biosensors and microfluidic devices POC technologies, which is misleading. Not all biosensors and microfluidic devices can be identified as “point-of-care (POC)” technologies unless they meet the following criteria.^194,195

Portable. The whole detection system can be carried with a portable carry-on bag, and does not need installation in field except for “plug-in” actions. In addition, all reagents and devices can be stored at room temperature or in a refrigerator with a long shelf life.

Field-deployable. A POC device can be operated in field by end users without professional training at any place such as at a hospital bedside, the emergency room, clinics, home, in communities and other settings. It should be automatic or just needs several simple steps from sample collection to result readouts. For blood testing, it is desirable to directly measure finger-prick blood samples in field.

Accurate. A POC device should achieve the required analytical precision, accuracy and range with minimal false negatives and minimal false positives to meet sensitivity and specificity for a specific analyte in a real-world sample.

Rapid. For POC testing of non-nucleic acid samples, the detection duration is typically 5 to 30 minutes. For POC testing of nucleic acid samples, the detection duration is desired to be less than 45–60 minutes.

Affordable. The overall POCT cost should be much lower than that of current laboratory-based test kits. For example, testing a protein biomarker in blood with ELISA typically costs $100–150 in the USA. Correspondingly, the POCT testing may lead to a reduced cost down to $5–50.

4.2 Active optofluidic devices

4.2.1 Design of operational components in microfluidic devices. Microfluidic devices can be categorized into active and passive microfluidics according to flow control processes. In active microfluidics, liquid flow is driven and/or controlled by active pumps (syringe pumps or pressure pumps).¹³ It provides great flexibility in controlling velocity, volume and distance of fluids via regulation of external pressure. Here we introduce some operational components in active microfluidics that are important for POCT.
Pumps. For active microfluidic devices, traditional external pumping systems include syringe pumps, vacuum pumps, electro-osmotic pumps, electrowetting devices, etc.^13,196–198 However, these pumps are large, expensive and some of them require training for operation. Thus, small and easy-operating pumps are desirable for POCT.¹⁹⁹ For example, a commercially available latex balloon served as a pressure pump to drive liquid flow in microfluidic channels, and sodium polyacrylate with a high swelling ratio acted as a suction pump to accelerate flow by absorbing water.^200,201 Additionally, some hand-operated systems including syringes and pipettes are used to inject liquid into microfluidic channels.^199,202,203 Similarly, one can push a deformable elastomer chamber with a finger to change pressure to deliver fluid into the target zones.¹⁹⁹
Valves. A valve is another important component that stops fluid flow, changes flow direction and sequentially delivers multiple preloaded reagents.^13,199,204 Various types of valves have been developed to meet different needs in flow control. For instance, hydrophobic valves are introduced to stop or regulate liquid flow through generating a capillary barrier. Pneumatic valves operate flow based on a change in the pressure. Screwing, twisting, aligning holes or inducting magnetism can also realize fast switching and on-demand flow regulation.^205–207 Compared with these manually operating valves, a series of automatic valves have been designed to better satisfy automation requirements in POCT. For example, a meltable valve made of wax can open in response to heating to separate assay steps.²⁰⁸ And an integrated pH-responsive hydrogel valve can sense the surrounding pH to display different extents of swelling or contracting to control flow.²⁰⁹ In addition, retention burst valves with predetermined heights and widths can encode capillary difference changes, enabling automatic sequential liquid delivery.²¹⁰ Moreover, elastic reversible valves on a centrifugal microfluidic platform are used to release liquid on demand. Furthermore, a circular chip incorporated with peptide microarrays can automatically separate, collect and detect five different proteins from a single serum sample.²¹¹
Mixing modules. Typically, mixing is performed by confluence of fluids from different inlets to induce molecule diffusion. Multiple S-shaped channels are exploited to increase the diffusion efficiency.²¹² Apart from these passive micromixers with patterned channels, active stir mixing units can tune the mixing processes. For example, a stir mixing was powered by an electrical field using two aqueous electrolyte solutions,²¹³ in which fluctuating waves were generated and propagated toward the downstream. Alternatively, flow instability conditions can be achieved by acoustic-driven micromixers, centrifugal mixers or magnetic mixers under magnetic fields.^197,214,215
Purification/separation modules. In the case of handling complex sample matrices such as blood, urine and silva, separation may be required to extract desirable analytes, or to remove non-specific molecules prior to detection. Classical separation methods, such as nanoporous membrane filtration, cross-flow filtration, electrophoresis and isotachophoresis, rely on the size and/or charge difference of molecules in samples.^{13,199,216–219} For example, in a three-stage spiral channel device, multiplex cross-flow channels were patterned along the spiraling direction. As a result, tumor cells and blood cells were separated by optimizing the flow rate, achieving 1100-fold of analyte preconcentration.²¹⁷ In order to increase specificity, specific ligand–analyte interaction, or additional surface modification with ovine serum albumin (BSA) is used for sample separation. For example, a porous hierarchical graphene foam served as the channel substrate, in which the electrospun titanium dioxide nanofibers were modified with the antibody of epidermal growth factor receptor 2.²²⁰ The surface-modified substrate in microchannels can capture the protein analyte and remove non-specific species by washing. Additional simpler purification methods include modulation of steric hindrance on the microchannel substrate and immobilization of the capture ligand on magnetic beads. The targets captured on magnetic beads can be isolated under the magnetic field.^221,222
On-chip blood plasma separation unit. Testing blood in POC settings is much more challenging than saliva, urine and buffer. Whole blood consists of ∼55% plasma as well as ∼45% blood cells and platelets.²²³ Blood cells may cause severe interference during testing. Therefore, plasma is typically separated from whole blood prior to testing. In a central laboratory, plasma separation is processed by centrifugation and other lab-based equipment. However, such large-scale equipment is unavailable in POC settings, and is not a choice for laypersons. Hence, plasma separation is typically performed in situ by an on-chip unit in microfluidic devices using active or passive methods.^196,224,225 Active separation methods include centrifugation, dielectrophoresis and acoustic waves (Fig. 15). Centrifugal separation, which utilizes difference in mass density between blood cells and plasma, was used in a compact-disk (CD) microfluidic device (Fig. 15a).²²⁶ By varying the disk rotation speed, the separated plasma was subsequently divided into smaller samples of equal volume to perform multiple testing simultaneously. Fig. 15b shows plasma separation by acoustic standing waves, in which blood cells were driven to the pressure nodes in the standing wave field. Acoustic microstreaming is usually observed, and air bubbles oscillate under the acoustic field. As shown in Fig. 15b, blood cells were trapped in the microstreaming vortex at a certain frequency while plasma flowed downstream.²²⁷Fig. 15c demonstrates a dielectrophoresis process, in which the polarization forces were formed in a non-uniform electric field to separate the suspended particles from fluid. Blood cells have dielectric properties different from proteins and small molecules, and can be forced to move in the microchannel. As shown in Fig. 15c, blood cells were trapped in the dead-end branches and left blood plasma in the middle of the channel.²²⁸

Fig. 15 On-chip plasma separation methods including active (a–c) and passive (d and e) methods: (a) centrifugal separation in a CD-shaped disk. (b) Acoustic microstreaming vortex to trap blood cells. (c) Dielectrophoresis-induced plasma separation. (d) Silicon-based microfilter including (i) weir-type filter, (ii) pillar-type filter, (iii) cross-flow filter, and (iv) membrane filter. (e) Dual separation by combination of sedimentation and membrane filtration. Reproduced with permission from (a) ref. 226, copyright 2006, The Royal Society of Chemistry; (b) ref. 227, copyright 2021, Elsevier; (c) ref. 228, copyright 2015, Springer Nature; (d) ref. 229, copyright 2008, Springer Nature; (e) ref. 238, copyright 2013, American Chemical Society.

Filtration is a common passive separation method, which is based on the size exclusion effect to impede cell movement. Pillar-type, grating-type, weir-type microchannels, packed-bead column, and filter membranes have been explored for plasma separation in microfluidic devices.^{225,229–231} Ji et al. compared four types of filtration structures, including weir-type, pillar-type, cross-flow and membrane filter (Fig. 15d).²²⁹ It was found that the cross-flow filter showed the highest efficiency in separating white blood cells from red blood cells in whole blood. However, blood cells could block the microchannels in these filters, limiting continuous operation. Hence, other passive separation methods such as hydrodynamic flow and capillary pumps have been exploited. Hydrodynamic flow is based on the Zweifach–Fung bifurcation law that cells prefer channels at a higher flow rate, leaving plasma in the channels at a lower flow rate.^232,233 Both blood cells and plasma keep flowing in the microchannel, thus such devices can work for a longer time without blocking. For example, Yang et al. fabricated a “Y”-shaped channel with an asymmetric flow rate ratio to separate plasma.²³⁴ Heath's group designed a comb-liked channel to direct plasma skimming into high-resistance branches (Fig. 16b).²³⁵ By utilizing the Zweifach–Fung bifurcation effect, Heath's group further increased the efficiency of plasma separation with inertial forces.²³⁶ This method permitted direct separation in the same channel without skimming to other channels, leading to a larger proportion of plasma. Natural sedimentation is quite slow. The separation process will stop once the sediment gap is filled with cells. To accelerate this process, Maria et al. combined hydrodynamic effects and sedimentation to separate 2 μL of plasma from 10 μL of whole blood with 99.9% separation efficiency within 15 min.²³⁷ In Fig. 15e, Liu et al. combined sedimentation and membrane filtration into a plasma separator.²³⁸ It took less than 7 min to extract about 275 μL of plasma from 1.8 mL of whole blood. As a result, HIV virus was measured in the plasma matrix.

Fig. 16 A typical structure of active optofluidic device for protein detection. (a) SERS-based microdroplet device for detection of prostate-specific antigen (PSA) cancer markers. It has 5 compartments with different functions: (i) generating microdroplet and mixing reagents; (ii) forming immunocomplexes on magnetic beads; (iii) separating immunocomplexes by a magnetic bar; (iv) moving the supernatant that contained unbound SERS probes; (v) separating magnetic immunocomplexes through droplet splitting. (b) In vitro fluorescence microfluidic device for multiplexed detection of proteins in finger-prick whole blood. Reproduced with permission from (a) ref. 218, copyright 2016, The Royal Society of Chemistry; (b) ref. 235, copyright 2008, Springer Nature.

Digital microfluidic modules. Digital microfluidics (DMF) has been developed to precisely manipulate small quantities of liquid reagents in microfluidics.²³⁹ Following the electrowetting-on-dielectric principle, continuous liquid can form picoliter-to-microliter droplets due to the change in surface wettability (hydrophobic vs. hydrophilic) of discrete insulated electrodes laid out close to each other.^240–242 Under the electric field, each droplet with different inclusions (targets, detection and capture compounds) can individually move, split, merge and dispense within a short period of time, realizing sample transport, mixing and separation.^240,243 Digital microfluidics can also reduce sample consumption and avoid cross contamination besides their own merit of low power consumption.²³⁹ For instance, in a digital microfluidic device, the bottom of microfluidic channels was made of glass. An array of discrete ITO electrodes was deposited onto the glass substrate; the ITO electrodes were then completely covered with a dielectric material (SU-8/Al₂O₃). Subsequently, a hydrophobic layer was coated onto the surface of the dielectric material. The dielectric material served as a spacer between the top hydrophobic layer and the underlying ITO electrode. Without an applied electric field, the top surface layer was hydrophobic. When applying an electric field onto an ITO electrode, the top surface layer became polarized, showing a hydrophilic feature.²⁴⁴ Droplets were consequently created on the parallel substrates and transported to the target area to function as a fluorescence immunoassay. In this digital microfluidic device, the magnetic beads linked with the capture antibody were mixed with the sample droplets containing the protein analytes under electrowetting force, followed by incubation for capturing analytes on the magnetic beads. These magnetic beads can be extracted under the magnetic field and washed with buffer prior to fluorescence detection. Compared with conventional ELISA that requires at least 50 μL of a sample volume, this digital microfluidic device required only 520 nL, which made it possible to sensitively test a rare sample volume (5–10 μL) from fertilization in vitro. The total duration was shortened to less than 40 min for multiplexed detection of protein biomarkers.

4.2.2 Integration of fluorescence and SERS sensors into microfluidic platforms. Microfluidic devices can pretreat, transport, and control liquid flow, which provide attractive platforms for the integration of fluorescence and SERS sensors to achieve POCT.^235,245,246 Such a technology combines the advantages of fluidics and optics, thus is called optofluidics. Signal transduction in optofluidic devices is similar to that in sensors as illustrated in Section 3. The major difference is that optofluidic devices just require simple operation or can realize automation in mixing of sample and reagents, incubation, washing, separation, preconcentration and detection while the sensors, which are described Section 3, typically require manual operation of these steps by trained personnel. Herein, we highlight the design and applications of fluorescence and SERS optofluidic devices in protein and nucleic acid detection.
Detection of proteins with optofluidic devices. Fig. 16a shows a typical example of optofluidic immunoassay where the liquid sample, the Raman dye-labelled Au nanoparticles with detection antibodies, and the magnetic beads with capture antibodies along with the carrier oil were externally injected into three microchannels, respectively.²¹⁸ All liquids were fully mixed as the microdroplets passed through the winding channel, followed by the formation of sandwich magnetic immunocomplexes. These complexes were then separated and concentrated by a magnet. At the end of the device, the Au nanoparticles linked to magnetic beads were aggregated, leading to enhanced SERS signals. This device has achieved a LOD (0.1 ng mL⁻¹) toward prostate-specific antigen (PSA) detection in serum. Like this device, aggregation of plasmonic nanoparticles in a microfluidic channel was commonly used for SERS transduction in optofluidic devices, which can be realized with other approaches such as micro–nano channel junctions and nanoporous silica microspheres besides the magnetic field.²⁴⁷ However, aggregation methods involve a lot of reagents in the device, which bears a burden on storage, shipping and operation. A different method is capture of analytes into a detection substrate on the bottom of the microfluidic chamber. In this way, the SERS or fluorescence probes are in situ immobilized onto the detection substrate.^248–250 Heath's group has reported an integrated microfluidic system patterned with antibody barcode arrays as the multiplexed detection substrate (Fig. 16b).²³⁵ After a finger-prick blood sample was loaded onto the optofluidic device, plasma was quickly separated via the Zweifach–Fung effect. Different antigen biomarkers were then captured onto different barcode detection zones by forming the sandwiched configuration of the capture antibody–antigen–detection antibody labeled with a fluorescence probe. Such in situ immunoassays can detect multiple proteins simultaneously.
Detection of nucleic acids with optofluidic devices. When the level of nucleic acids is high, nucleic acids can be directly detected using optofluidic devices; the structure of devices is similar to that of optofluidic immunoassays. For example, Qi et al. reported an “in situ” measurement of nucleic acids using molecular beacons immobilized on a nanoporous gold disk inside the PDMS microchannel.²⁴⁸ When the targets flowed through the gold disk substrate, they were hybridized with the molecular beacons via base pairing, generating fluorescence signals on the detection zone. However, in most cases, the level of nucleic acid is very low in the samples. For example, the serum level of DNA in HIV patients within the first week of infection is as low as 50–1000 copies per mL. It is very difficult to directly measure such a level of nucleic acids. Hence, a nucleic acid amplification unit needs to be included into an optofluidic POCT device besides a detection unit.

A variety of nucleic acid amplification methods, such as polymerase chain reaction (PCR) and isothermal amplification, have been exploited in microfluidic devices to amplify signals.²⁵¹ Particularly, digital PCR (dPCR) is a promising high-throughput approach.²⁵² Amplicons are generated exclusively from a single DNA template in large-scale multiple diluted samples, converting fluorescent signals into linear and digital signals. Digital PCR is generally integrated into microfluidic systems in the form of droplets or integrated flow circuit chips. For example, a SlipChip was fabricated with two plates etched with wells and ducts (Fig. 17a).²⁵³ In the assembly of two plates, the elongated wells allowed for a continuous flow path to transport reagents into each reaction chamber. When the plates slipped, the misaligned channels and ridges broke off the flow path to obtain droplets. This chip with 1280 wells can achieve 1280 droplets and each of them contained ∼2.6 nL of the reaction solution, thus is able to perform single-copy PCR assay. Although PCR or digital PCR technology shows high sensitivity and accuracy in detection, 2–3 hours of thermal cycles cannot meet the POCT need. In contrast, isothermal amplification is relatively simpler and has shorter duration, which is suitable for microfluidic systems.^254,255 For example, a centrifugal microfluidic chip was integrated with the recombinase (RAA)-aided amplification unit and the CRISPR/Cas12a detection unit.²⁵⁶ After centrifuging, the loaded sample was mixed with preloaded RAA–Cas12a reagents, followed with isothermal nucleic acid amplification at 37 °C.

Fig. 17 Integrated microfluidic devices for nucleic acid testing. (a) Schematic digital PCR process with a SlipChip: (1) assembled SlipChip after splitting; (2) and (3) the top plate and the bottom plate, respectively; (4–6) overlapping of the top and bottom plates to form a continuous fluidic path; (7–9) the aqueous reagent (red) filling the SlipChip through the connected elongated wells; (10–12) aqueous droplets formed in each compartment after splitting. (b) A self-contained microfluidic cassette integrated separation, amplification, and detection of nucleic acids. Reproduced with permission from (a) ref. 253, copyright 2010, The Royal Society of Chemistry; (b) ref. 267, copyright 2010, Springer Nature.

Lysis and/or extraction of nucleic acids is sometimes required prior to detection. In this case, a lysis/extraction unit needs to be added into optofluidic devices so that nucleic acid tests can be done in POC settings. One of the common methods is chemical lysis, where lytic agents are mixed with the sample liquid in the microchannel to disrupt the cell membrane.²⁵⁷ For example, ammonium chloride is capable of red blood cell lysis; guanidinium thiocyanate and guanidinium chloride are used to disrupt membrane proteins; lysozyme works on the bacteria wall.^257,258 Sethu et al. performed lysis by sandwiching whole blood with a lysis buffer on both sides.²⁵⁹ The mixture passed through a long serpentine channel to achieve full lysis. Besides chemical lysis, physical lysis is also common, including mechanical lysis, electroporation lysis, thermal lysis, laser lysis and sonication lysis.^260,261 Kim et al. performed mechanical lysis with ZnO nanowires patterned in a microfluidics device.²⁶¹ The ZnO nanowires were characteristic of high aspect ratios and sharp tips, which teared the plasma membrane and released nucleic acids but held cells. After lysis, nucleic acids may need to be extracted and separated for downstream amplification/detection. This can be done via micro-solid phase extraction, electrostatic interactions, nanoporous membrane filtration, or magnetic bead separation.^262,263 Micro-solid phase extraction and magnetic bead separation technologies are commercially available for nucleic acid extraction.²⁶⁴ In microfluidic devices, silica micro-pillars or silica beads were built in the microchannel to perform micro-solid phase extraction.²⁶⁵ The negatively charged nucleic acids can bind to silica in a high ionic strength buffer, followed by washing with a non-polar solvent to remove other biological components. These nucleic acids were then eluted with a low ionic strength solution.²⁶⁶

“All-in-one” microfluidic devices are desirable for nucleic acid testing in POC settings, that is, combing lysis, extraction, amplification and detection modules in a single microfluidic chip.^265,268Fig. 17b shows a typical example, where the self-contained microfluidic cassette can perform separation, amplification, and detection of nucleic acids. It employed the miniaturized linear motors to control fluid flow and valves.²⁶⁷ All the control units such as motors, heater controllers, micro-vacuum pump, and a liquid trap were packaged into a portable box as an analyzer. Before introducing the sample into the mixing chamber, the sealing layer was removed for the engagement of the connecting film. The cassette was then inserted into the analyzer to perform the preprogrammed process operation, including lysis with a prestored lysis buffer, extraction through the silica membrane, collection of pure nucleic acids with an elution buffer, amplification in the PCR chamber and detection in a lateral flow strip. This integrated system was successfully used for monitoring the DNA of bacteria (B. cereus) and the RNA of HIV, and it detected 1000 pathogen particles in the sample. In addition, centrifugal microfluidic CDs are well-known “all-in-one” devices for nucleic acid testing. Stumpf et al. designed a LabDisk with the pre-stored reagents.²⁴⁶ By controlling the rotational frequencies of the chip and using a magnet, a sample containing influenza H3N2 virus was sequentially transferred to each chamber and mixed with specific reagents. After the sample was mixed with the lysis buffer, nucleic acids were released and bound with the functionalized magnetic beads using a binding buffer. These magnetic beads were subsequently mixed with an elution buffer under the magnetic field to extract nucleic acids. Subsequently, nucleic acids were centrifuged into specific compartments to conduct reverse-transcript PCR, and to read fluorescence signals.

Moreover, the electric field gradients were used for preconcentration, purification, mixing, and acceleration of reactions among the sample and reagents.^269,270 For example, the electric field gradients were formed using isotachophoresis (ITP), which utilized a low-mobility trailing electrolyte (TE) buffer and a high-mobility leading electrolyte (LE).²⁷¹ Under an electric field, TE and LE ions can ensure that the target sample ions focus with the electro-migrating LE-to-TE interface. In this design, SARS-CoV-2 viral RNA and host DNA from the lysed samples of COVID-19 patients were extracted and preconcentrated in ITP, leaving impurities, proteins and other inhibitors. After loop-mediated isothermal amplification (LAMP), the electric field gradients in ITP were used to control and confocus the Cas12–gRNA enzymatic reaction. Such ITP-based nucleic acid device exhibited the test time down to 30 minutes.

4.3 Passive microfluidics

Active microfluidics usually uses external syringe pumps or pressure pumps, which provide great flexibility in designing the structure of microfluidic devices, selecting different liquids, and controlling liquid flow in a well-tuned manner. However, active microfluidics systems are relatively expensive and large, and require a lot of steps to operate. Therefore, there is an incentive to develop passive microfluidic devices, in which the “built-in” micropumps are used to drive liquid flow to realize self-powered fluidic systems, eliminating the use of external pumps and the associated electronic controllers. Although there are some constraints on regulating liquid flow and realizing sequential operation, passive microfluidic devices are attractive for POCT because they are relatively inexpensive, compact, and have a relatively simple structure. Lateral flow is the dominant form in passive microfluidics. Therefore, this section is focused on lateral flow devices that are divided into two different types: (i) plastic-/glass-based lateral flow devices and (ii) paper-based lateral flow assays.

4.3.1 Plastic-/glass-based lateral flow devices.
Basic structure of lateral flow devices. In lateral flow devices, fluids flow almost horizontally along the relatively long microchannels that are divided into different functional zones.²⁷⁵ It can consecutively or programmably perform sample handling and analyte detection as fluids flow to different functional zones. Fig. 18a shows the basic structure of a lateral flow device.²⁷² It contains four main functional zones, including a pretreatment unit, a detection chamber, a fluid resistor and a “built-in” pump. A liquid sample is initially loaded into the pretreatment unit, and fills the pretreatment unit, the detection chamber and the fluid resistor sequentially, and finally reaches the “built-in” micropump. The resistor is used to adjust the flow rate and to optimize the incubation duration for fully mixing or reacting the reagents and analytes.^{221,273,276,277} In addition, a sample pretreatment unit may need to be placed on the upstream of the detection zone to perform separation or purification. To better regulate the separation and purification processes, an additional “built-in” micropump may be added into the pretreatment unit.

Fig. 18 (a) Basic structure of plastic-/glass-based lateral flow device. It is composed of an interfacing layer (cover layer) and a fluid layer. The fluid layer contains four main functional zones, including a pretreatment unit (plasma separation and extraction), a detection chamber, a fluid resistor and a “built-in” micropump. (b) and (c) Examples of “built-in” micropumps: (b) capillary pumps such as “vertical lines”, “parallel lines”, “rounded hexagons” and “tree lines”. (c) A pneumatic pump fabricated by multiple vacuum pillars. Reproduced with permission from (a) ref. 272, copyright 2017, MDPI; (b) ref. 273, copyright 2007, The Royal Society of Chemistry; (c) ref. 274, copyright 2017, The Royal Society of Chemistry.

“Built-in” micropumps. “Built-in” micropumps are used to drive liquid flow in passive microfluidic devices, and designed based on different mechanisms such as capillary forces, air transfer (solubility or permeability), diffusion and osmosis pressure, magnetic field, electrophoresis, electroosmosis, electrowetting, etc.^{273,278–280} The first two mechanisms are most popular. The capillary pumps drive flow via capillary forces, which are generated by interfacial tension. Once the device is dull of liquid, capillary pumping will stop. The flow rate can be regulated by the surface properties, size and shape of microstructures in a micropump, the distance between different pumps, and fluid viscosity.^281,282Fig. 18b illustrates various microstructures in capillary pumps. Line-shaped microstructures are fabricated with a round periphery to reduce the occurrence of fluid pining.²⁷³ The fluids flow in both vertical and horizontal directions, and are modulated by the dimension of lines and the gap between neighboring lines. Zimmermann et al. fabricated a vertical line-shaped capillary pump, and found that 15 μm-sized lines exhibited the highest flow rate.²⁷³ However, such a microstructure showed a random filling front, which can be altered into a straight filling front by rotating the microstructure by 90°, namely, “parallel line”. In addition, this pump was designed with gradually increasing lateral dimension from the entrance of the pump, which can prevent the capillary pressure difference in the connection part to make the fluid efficiently flow to the capillary pump. Some capillary pumps exhibit an interlocked microstructure. For example, rounded hexagon-shaped pumps can achieve uniform flow rates due to the inhibited fluid pinning. Furthermore, a tree line-shaped capillary pump has been designed with branched filling paths separated from the main flow region (Fig. 18b).²⁷³ Such a structure can program sequential filling flow through generating different capillary pressures of the branched paths. This “tree-line” pump may suffer from a decrease in flow rate as fluids flow along each branch. Gradient surface wettability can solve this problem by moving spontaneously toward a more hydrophilic area owing to contact angle hysteresis.²⁸³

Pneumatic micropumps are also commonly used in passive microfluidic devices, in which pumping is realized through gas generation or storage of vacuum pillars inside microchannels. Zhu et al. fabricated a hydrogen generator using platinum-black to catalyze ammonia borane solution in a microchannel.²⁸⁰ The generated hydrogen bubbles pushed the liquid flow along the channel, and the velocity of the liquid was regulated by the gas removal rate at the gas outlet.²⁷⁹ Although this technology could achieve self-circulation by collecting hydrogen bubbles to add fresh reactants, it is quite difficult to ensure stable fluid flow, and the chemical reaction may be disturbed when loading the sample liquid or detection probes. Vacuum-driven microfluidics provides an alternative method utilizing gas solubility and permeability of PDMS.^274,278,284 When it is degassed, the formed negative pressure forces the fluid to flow from the inlet to the dead-end microchannel. Compared with capillary pumps, surface treatment is typically not required in vacuum-driven microfluidics. However, it is challenging to maintain a constant flow rate due to a gradual decrease in the pressure difference with prolonging time. In this case, prolonging vacuuming time or changing channel patterns was proposed to mitigate this problem. Specifically, multiple vacuum pillars were introduced into the microchannel for continuous flow control (Fig. 18c).²⁷⁴

Purification/separation modules. Filtration, capillary pumping and hydrodynamic flow are commonly used for purification/separation in passive microfluidic devices. Hydrodynamic flow may be combined with filtration and capillary pumping to improve the separation efficacy. Maria's group designed a vertical cylinder with gradient hydrophilicity as a hydrophobic patch for blood plasma separation.²³⁷ The central portion of the vertical cylinder was hydrophobic while both sides of the well as well as their connected two parallel channels were hydrophilic. As blood flowed through the bottom hydrophilic microchannel and was elevated along the vertical cylinder by capillary action, the capillary force decreased due to the hydrophobic patch near the central part, slowing down the flow. The blood cells in the sample subsequently tended to be deposited because of their higher density while the plasma was lifted up at a higher velocity because of its lower viscosity. Ultimately, the separated plasma flowed into the upper channel for analyte detection. Similarly, a cyclic olefin copolymer was coated with a super-hydrophilic surface by the layer-by-layer assembly technique. Thus, asymmetric capillary forces were used for blood plasma separation.²⁸⁵ In addition, vacuum pillars were designed for automatic blood plasma separation without showing hemolysis. A slant capillary valve led the blood flow to the vacuum pillars. As a result, the blood cells were trapped while the plasma flowed into the separation channel.²⁸⁴
Integration of fluorescence or SERS units. Integration of SERS or NIR fluorescence sensing units into passive microfluidic devices is appealing for POCT. Gao et al. incorporated a SERS sensor into a polydimethylsiloxane (PDMS) lateral flow chip for monitoring prostate-specific antigen biomarkers (Fig. 19a).²²¹ Driven by capillary forces, the liquid sample, the SERS probes conjugated detection antibodies and the magnetic beads linked with the capture antibodies were loaded from three different inlets and fully mixed in the winding-shaped channel due to the chaotic advection effect. The sandwiched structure was formed via the antigen–antibody interaction, and then purified by applying a magnetic field onto the magnetic beads. The SERS probes linked to the targeted antigen was related to the level of the analyte. The whole process was finished within 5 min. This integrated device achieved a LOD of 0.01 ng mL⁻¹ in a human serum. Additionally, a blood plasma separation unit and a fluorescent hairpin molecular beacon were integrated into a vacuum-driven microfluidic chip for the detection of thrombin in whole blood (Fig. 19b).²⁷⁴ In this device, blood cells flowed into the vacuum microchambers while the plasma was driven to a plasma chamber, where the molecular beacons with the quenched fluorophores were stored. In the presence of thrombin, the molecular beacon opened the loop of the hairpin, pushing the quencher away from the fluorophore, and turning on the fluorescence.

Fig. 19 (a) A passive microfluidic device with an integrated capillary micropump and SERS probes, consisting of three compartments for (i) mixing the sample and reagents; (ii) preloading reagents and detecting; (iii) the capillary pump. (b) A passive microfluidic device with an integrated pneumatic micropump and fluorescence SERS probes. Reproduced with permission from (a) ref. 221, copyright 2019, American Chemical Society; (b) ref. 274, copyright 2017, Elsevier.

4.3.2 Paper-based lateral flow assays. In a typical paper-based lateral flow assay (PLFA), which is also called a paper-based lateral flow strip (PLFS), a porous nitrocellulose paper membrane provides the main microfluidic channels in a test strip, and a microstructured polymer, sintered polymer, glass fibers and other materials, more or less, are incorporated into the test strip to realize different functions. In such devices, liquid flow is driven by capillary forces.^286,287 The PLFS is of particular interest to POCT because of its low-cost, easy operation on-site, and relatively simple manufacturing processes. The home pregnancy test strip, which detects hormones in urine, is a typical example of such devices. Owing to the success of the home pregnancy test, this type of device is sometimes called a lateral flow test (LFT) strip. The rapid development of this technology has expanded its detection scope to a wide range of analytes such as proteins, nucleic acids, small molecules, metal ions, etc. If antibodies are employed as molecular recognition elements, it can also be called a lateral flow immunochromatographic assay (LFIA).
Basic structure and principle. Fig. 20a shows the basic structure of a PLFS.⁶⁷ It mainly consists of a sample pad, a conjugated pad, a nitrocellulose microfluidic membrane, an absorbent/wick pad, a control line and one to three test lines/zones. These pads/membranes are connected to each other with small overlaps to ensure continuous liquid flow from the sample pad all way to the absorbent pad. The sample pad is usually made of cellulose, woven meshes or filtration matrices to retain high tensile strength after wetting. The sample pad may be treated with a dry pH regulator to standardize the buffer condition of the sample liquid.^286,288 The conjugate pad should prevent non-specific binding so that the detection probes can be released upon wetting by the liquid sample, and the analytes would not be held there. Hence, glass fibers, cellulose and polyester with high bed volume and low non-specific binding are suitable for the conjugate pad to store a large number of probes and release them quickly as liquid flows through.²⁸⁹ The nitrocellulose membrane is enriched by the amine group, which exhibits strong binding to antibodies or other types of bioreceptors, and is surface-modified with a surfactant or other reagents for prevention from non-specific binding. It also has tunable wicking properties to regulate capillary flow. This absorbent pad should have a sufficient volume so that it is not full even after the whole process is complete. Complete filling or overloading of the absorbent pad will stop the lateral flow of the liquid in the whole strip, and even cause the back flow or liquid accumulation on the nitrocellulose membrane, inducing the background noise on the test line.

Fig. 20 Structure of PLFS with integration SERS probes. (a) SERS-PLFS for the detection of traumatic brain injury (TBI) protein biomarker in diluted blood plasma samples. (i) Basic constructure of PLFS; (ii) side view of prestored reagents in the PLFS before detection; (iii) side view after detection. (b) Gold nanostar@Raman reporter@silica sandwiched nanoparticles coupled with a Au pyramid array for the detection of TBI biomarker S-100β in blood plasma: (i) structure; (ii) before detection; (iii) after detection. (c) SERS-PLFS with a built-in plasma separation unit for whole blood detection: (i) Structure; (ii) before detection; (iii) after detection. Reproduced with permission from (a) ref. 67, copyright 2017, American Chemical Society; (b) ref. 291, copyright 2021, Elsevier; (c) ref. 292, copyright 2020, American Chemical Society.

Integration of SERS into PLFS. A traditional PLFS usually employs gold nanoparticles as colorimetric transducers and has achieved huge commercial success in home pregnancy testing in POC settings. A colorimetric PLFS can be read out by human eyes without the need for any instrument, and used for the qualitative measurement of a high level of analytes in water, buffer, saliva and urine. But they are unable to quantitatively measure a low level of analytes due to their low sensitivity. Particularly, when a PLFS is applied to plasma, serum or whole blood, it suffers from severe interference from sample matrices. One of the major reasons is that the strong red wine color of hemoglobin in such samples is overlapped with the color of gold nanoparticles on the test line. As described in Section 2, NIR fluorescence and SERS fall into the biological transparency windows, and are read out by portable readers. Thus, NIR fluorescence or SERS-based PLFS have great potential in the quantitative measurement of trace analytes even in complex sample matrices such as plasma, serum or whole blood.^1,67,290

Many SERS probes are made by directly attaching Raman dye molecules to bare metal nanoparticles. Such SERS probes are unstable and easily aggregate in liquid with high ionic strength.⁴⁴ Hence, Wu's group designed a sandwich-structured SERS probe, in which Raman dye molecules are sandwiched between a plasmonic metal core and a thin SiO₂ outer shell, which exhibits excellent water solubility and stability in the blood sample matrix.^44,64 His team also compared the SERS enhancement factors of Au nanospheres, nanorods and nanostars and found that Au nanostars contributed a lot of “hot spots”, achieving the highest SERS enhancement.⁴⁴ The gold nanostar@Raman reporter@silica sandwiched SERS probes were incorporated into a PLFS to measure the neuron-specific enolase (NSE), a traumatic brain injury (TBI) biomarker, in clinical blood plasma, achieving a LOD of 0.86 ng/mL, which was three orders of magnitude lower than that of the colorimetric PLFS (Fig. 20a).⁶⁷ In order to further enhance SERS signals, a gold nano-pyramid array chip was integrated into the detection zone of the PLFS (Fig. 20b).²⁹¹ By covering the inorganic Au nano-array substrate with a surfactant bridge, the conjugate of SERS probes and analytes can flow smoothly throughout the chip. The “hot spots” formed between the gold nanostars in SERS probes and the gold nano-pyramid array created a hierarchical 3D plasmonic field, which greatly enhanced the SERS signals and achieved a LOD down to 5.0 pg mL⁻¹ toward the TBI S-100β biomarker in blood plasma. This portable SERS-PLFS system showed a comparable LOD with ultrasensitive ELISA assays but in a much shorter assay time (∼30 min for PLFS vs. 2 h for ELISA). Furthermore, Wu's group integrated a blood plasma separation unit into a SERS-PLFS following the principle of salt-induced aggregation of blood cells (Fig. 20c).²⁹² Such a SERS-PLFS has directly detected carcinoembryonic antigen (CEA), a cancer protein biomarker, in a drop of whole blood (30 μL) at a LOD of 1.0 ng mL⁻¹. This portable “all-in-one” SERS-PLFS system can serve as a POCT tool for testing finger-prick blood samples, assisting in vitro diagnostics (IVD) in resource-limited settings. This SERS-PLFS can be adapted easily to detect other protein biomarkers in blood by replacing the antibodies.

Integration of fluorescence into PLFS. Fluorescent organic dyes and semiconducting quantum dots (QDs) have been used as fluorescent probes in the PLFS.^{288,293–295} Organic dyes usually suffer from low quantum yield, photobleaching and blinking/quenching. To avoid such problems, QDs have been explored in the PLFS. A PLFS integrated with dsDNA-functionalized QDs was developed to measure HIV-DNA in human serum.²⁹⁶ However, the small size of the QDs limits their conjugation with antibodies, especially for proteins over 100 kDa due to the steric effect. In addition, binding with large antibodies would possibly change the surface properties of the QDs and even cause instability in aqueous solutions. Macromolecule-coated nanobeads were thus introduced to not only stabilize the QDs during detection but also increase the fluorescence intensity in each entity. For instance, the dual-color CdSe/ZnS–COOH QDs were immobilized onto SiO₂ nanoparticles via polyethyleneimine (PEI)-mediated electrostatic adsorption, and conjugated with antibodies.²⁹³ Such a fluorescent PLFS was used to quantify two sepsis biomarkers in human serum simultaneously, namely, procalcitonin (PCT) and C-reactive protein (CRP) (Fig. 21a).

Fig. 21 Fluorescent PLFS. (a) PLFS using dual QD nanobeads for the simultaneous detection of two sepsis biomarkers. (b) Competitive assay based on the “turn-on” fluorescence signal in PLFS. Reproduced with permission from (a) ref. 293, copyright 2020, Springer Nature; (b) ref. 295, copyright 2018, Springer Nature.

FRET is of particular interest to PLFS because it can reduce the background signals. A quencher can either “turn-on” or “turn-off” fluorescence.^295,297 For example, an immunoassay PLFS utilized FRET to detect carcino embryonic antigen where fluorescein isothiocyanate (FITC)-labeled antibodies were initially pre-immobilized onto the test zone of the PLFS.²⁹⁷ When analytes were present, the Au NPs functionalized with detection antibodies were brought in proximity to the fluorescent dye via the formation of a classic sandwich structure. Thus, the fluorescence of FITC was quenched by Au NPs via FRET. In addition, the “turn-on” FRET PLFS was developed for the measurement of fumonisin mycotoxins based on the competitive mechanism (Fig. 21b).²⁹⁵ In this design, the CdSe/ZnS QDs functionalized with antigen molecules were coated on the detection line of the PLFS where the metal NPs with antibodies stayed close to the QDs via the antigen–antibody interaction, keeping the quenching of QD fluorescence prior to detection. When analytes were present on the test line, they formed a complex with the metal NPs, which pushed the metal NPs from the QDs as the fluid flows, turning on the fluorescence of the QDs.

Engineering PLFS toward POCT – “on-strip” blood plasma separation. It is challenging to enable the testing of finger-prick blood samples with PLFS in POC settings. A promising route is to combine “on-strip” plasma separation and NIR fluorescence or SERS into a single paper-based microfluidic strip to create an “all-in-one” device. Filtration is a common approach for “on-strip” plasma separation in PLFS. Filter membranes with a pore size smaller than 2.5 μm have been placed in the sample pad to block blood cells.^224,298,299 However, such small pores hinder the flow of plasma through the membrane. Hence, various agglutination reagents, such as agglutination antibody, salt and chitosan, have been used for the surface-modification of large pore membranes to induce aggregation of blood cells.^291,300,301 The aggregated blood cells are too large to pass through the membrane while plasma can do.²³⁰ For example, an assembly of three-layer filtration membranes was treated with salts and then integrated into a SERS-PLFS.²⁹¹ The first layer was functionalized with salts, which impaired the electric double-layer of the blood cell surface, causing deformation and aggregation of blood cells due to the osmotic pressure. The surfactant coated on the second layer accelerated the liquid flow to the third layer where salts held the residual blood cells. Such an “on-strip” plasma separation unit enabled the PLFS to directly measure whole blood samples.
Engineering PLFS toward POCT – tunning fluid. The flow rate is an important factor governing the performance of PLFS, which can be regulated by designing different path patterns and lengths.^289,302 Computational simulation was combined with experimental analysis to investigate the transport processes in simple two-dimensional geometries and assisted optimization for sequential reagent delivery for sample pretreatment and analysis.³⁰³ In addition, valves were constructed for flow control and reagent loading.^{204,299,304,305} Paraffin wax, PDMS droplets or dissolvable reagents (e.g., sugar) were deposited on the paper surface as hydrophobic barriers to control the flow direction. Surfactant treatment can accelerate the flow.^289,305 To slow down the flow, a cellulose pad was placed on the upstream of the test line as a shunt for time delay, tuning the flow rate by varying its size and number.³⁰⁵ Apart from these passive methods, active valves were built into PLFS by folding, stacking or cutting papers to initiate or stop the flow.^{304,306–309} For example, a paper-based device contained three parts and each part was loaded with different reagents. The loaded liquid can move to other parts by folding the specific part.³¹⁰ Inspired by electrowetting on dielectrics, the “on-strip” electrical valves were fabricated by inkjet-printing and spraying technology on a paper-based device. The dielectric surface in the valve can be switched from hydrophobicity to hydrophilicity under an applied bias. Thus, the ionized sample solution can stop or flow by the external bias.
Engineering PLFS toward POCT – reshaping the overall structure. Besides the basic/classic structure shown in Fig. 20a, a lot of varieties have been developed to either improve the performance or realize new functions. An intriguing example is the origami paper-based devices that transfer and mix sample and reagents by folding papers.^2,310,311 As shown in Fig. 22a,³¹⁰ an origami device was composed of three layers where the target DNA, a short ssDNA with a quencher (Q1) and a long ssDNA with a fluorophore (F1) were loaded on each layer, respectively. Because Q1 was shorter than the target DNA, the target showed priority to hybridize with F1 compared with Q1, and the remaining F1 strands were complementary to Q1. The amount of the target DNA was proportional to the intensity of fluorescence.

Fig. 22 Advanced structures of PLFS. (a) An origami paper-based device for nucleic acid detection. (b) Hybrid device nucleic acid extraction, amplification and detection by the combination of an origami paper for the extraction of nucleic acid, a plastic microfluidic device for the LAMP reaction and a paper-based strip for the detection of malaria DNA. Reproduced with permission from (a) ref. 310, copyright 2013, American Chemical Society; (b) ref. 2, copyright 2019, National Academy of Sciences.

Additionally, nucleic acid amplification sometimes needs to be conducted at a relative high temperature. For example, loop-mediated isothermal amplification (LAMP) of nucleic acids is typically performed at around 72 °C, which is difficult to be carried directly on the paper surface. Hence, a paper-based strip was combined with microfluidic PDMS to create hybrid structures, which not only reduced sample consumption but also enabled more functions.^{308,309,312–314} Phillips et al. modified the sample pad of PLFS to make a LAMP unit for the detection of HIV-1 virus in blood.³¹⁵ The HIV-1 viral particles were separated by a size-selective filtration membrane, and migrated in a buffer solution to a polyether sulfone membrane preloaded with the LAMP reagents to perform LAMP at room temperature. The amplicons generated by fluorescence-labeled primers flowed along the downstream paper strip, finally captured on the test line for fluorescence recording. The whole process was finished within 90 minutes, achieving a LOD of 2.3 × 10⁷ copies per mL in whole blood. Recently, Cooper's group developed a multiplex microfluidic diagnostic platform using hybrid devices.² An additional example of a hybrid device is shown in Fig. 22b.² The origami structure, plastic microfluidic modules and paper-based lateral flow strips were combined to create a hybrid device for the detection of malaria in blood. The liquid sample was pretreated with cell lysis. The released DNA was captured onto magnetic beads, and then loaded onto the sample zone of the origami paper part where excessive buffer and other residues were removed by the capillary force of the waste panel. The DNA bounded with the magnetic beads were distributed into three spots through distributing channels and transferred into different LAMP chambers. In each chamber, the LAMP reaction proceeded independently by adding different primers labeled with fluorophores. After that, finger pumps were used to push the liquid to flow to the paper test strips. As a result, it can detect two different types of malaria DNA together with the positive and the negative controls simultaneously.

5. NIR fluorescence and SERS imaging of live cells and tissues

NIR fluorescence and SERS take advantage of biological transparency windows, feature the extended penetration depth and the reduced interference from cells and tissues, and they are non-invasive techniques. Therefore, they find increasing applications in the real-time imaging of internal processes of living cells, tissues and organisms for diagnostic or biomedical research. This section gives an overview of the instruments, optical probes, imaging activation, and applications.

5.1 Instruments

Fluorescence microscopes for in vitro imaging. Conventional optical microscopes mainly comprise a light source, a condenser, an objective and an eyepiece.^316,317 The light from the light source is concentrated by the condenser, leading to a cone of light with uniform intensity to pass through the specimen. After that, a primary magnified image is produced by the objective as well as a secondary magnified image by the eyepiece. The total magnification is equal to the objective magnification times’ eyepiece magnification. A conventional optical microscope has been developed for a large family of commercial products, including the brightfield, differential interference contrast, phase contrast, Hoffman modulation contrast and polarized light microscope.³¹⁸ Different from these conventional optical microscopes, fluorescence microscopes rely on the condensed light from the light source passing through the objective to the specimen, and the emitted light by the specimen will then pass through the same objective to the eyepiece.³¹⁹ Given the fact that most fluorophores only absorb a small range of excitation light and their fluorescence emission is quite a lot weaker than the incident light, the main research work on the fluorescence microscope is to effectively reduce the influence of background light and collect the emission light. Hence, filters are the major means to (i) segregate a defined bandpass of excitation light with an excitation filter; (ii) segregate a defined bandpass of emission light with an emission filter; (iii) segregate the excitation light from the emission light by fluorophores using a dichromatic mirror.³¹⁹ By the combination of these three filters, the condensed light from the source is purified using an excitation filter with most of the nonessential light reflected back to the source, and then is reflected down by the dichromatic mirror to the objective and reaches the specimen. Non-coherent fluorescence emission passes through the same objective and the dichromatic mirror, followed by filtration with the emission filter to remove the contaminating light including the unabsorbed excitation light and the light not fully removed by the previous filters, and finally reaches the eyepiece. Apart from the filters, light sources and objectives should be carefully chosen to provide sufficient excitation energy in the required spectrum ranges and gather a large amount of emission light, respectively. Light sources include mercury, LEDs and metal-halide lamps.³²⁰ Mercury lamps are the line-emitting arc discharge lamps whose 405 and 546 nanometer lines can overlap Alexa Fluors 405 and 546, while LEDs and metal-halide lamps are suitable for Alexa Fluor 488.^321,322 The objectives are fabricated with various lens elements such as high light transmission values over large spectrum ranges, and large numerical apertures. It should be noted that water immersion objectives used in live-cell imaging need color correction for the compensation of coverslip and temperature variations, while oil immersion objectives used in fixed specimens can reduce light loss due to the reflections at the glass–air interface of the objectives. Additionally, imaging experiments should be performed in a relatively dark room to reduce light disturbance.

Developed from a basic fluorescence microscope, a confocal laser scanning microscope (CLSM) features controllable depth scanning, serial optical sectioning, spatial filtering, and adjustable magnification.^323,324 The key to success is the application of pinholes located between the laser source and excitation filter (called the light source pinhole), and between the emission filter and eyepiece or detector (called the detector pinhole), respectively. With the addition of a light source pinhole, a point-liked laser is formed and reaches the specimen. Then, the emission fluorescence from the in-focus specimen passes through the detector pinhole to the detector. Otherwise, the fluorescence from the out-of-focus specimen is reflected. Thus, the use of pinhole not only realizes the high signal-to-noise ratio but also acquires multiple two-dimensional (2D) images at different depths in a sample, followed by reconstruction to form a three-dimensional (3D) image. Particularly, for uneven samples, depth scanning can collect and integrate serial optical sections into a single image, which help eliminate out-of-focus imaging. In addition, the confocal laser scanning microscope has a zoom function that adjusts magnification by altering the scanning rate and scanning area of the laser without changing the objectives. Given that a high zoom factor would lead to photobleaching, it is advisable to achieve a balance between the zoom factor and imaging. In terms of thick samples, such as embryos and brain slices, confocal microscopy with single-photon excitation is weak in serially scanning in deeper planes, thus multiphoton microscopy has emerged.³²⁵ A two-photon excitation fluorescence microscope depends on the simultaneous absorption of two photons, followed by non-radiative relaxation and fluorescence emission.³²⁶ It shows that the excitation wavelength is longer than the emission wavelength, which is opposite to the traditional fluorescence microscope. More importantly, NIR excitation light is commonly used in a two-photon microscope, leading to deep tissue penetration, low background signals and reduced photobleaching. Despite these advantages and commercial applications of a confocal microscope, the theoretical resolution limit is around 200 nm within the visible light spectrum because of the far-field diffraction limit.³²⁷ Fortunately, a series of super resolution techniques and microscopies has been introduced, such as 4Pi and I⁵M, stimulated emission depletion (STED), saturated structure illumination (SSIM), single-molecule localization and far-field microscopy.^328–331 It is desirable that the nanometer scale can be resolved using a commercial fluorescence microscope in the near future.

Fluorescence microscopes for in vivo imaging. The main problems faced by in vivo imaging are light absorption and autofluorescence of tissues, which can be partially overcome by NIR fluorescence probes. In commercial in vivo imaging systems, the basic light path is similar to that in fluorescence microscopes. However, specimens are living creatures, and the optical pathlength in a specimen is much longer.³³² The core of an in vivo imaging system is a charge coupled device (CCD) camera.³³³ Upon illumination by light, each pixel creates a charge which is proportional to the illumination intensity. The important parameters of a CCD include the quantum efficiency and dark current level. Quantum efficiency can be improved by generating back-thinned and back-CCD to increase photon absorption. The dark current level is referred to the charge electrons that leak into a pixel within an exposure process in the absence of light, which limits the minimal light intensity of CCD detection. The low dark current can be achieved by cooling the CCD camera or introducing a cooled intensifier tube into the system.

In addition, an optical fiber-based imaging system is one of the in vivo imaging tools, which is referred to fluorescence microendoscopy (FME).³³⁴ The optical fiber in FME can be inserted into a tissue in a minimally invasive way, and flexibly maneuvered in the hollow tissue cavities. Various types of optical fibers are used in FME, such as fiber bundle, single-mode fiber, double-clad fiber, multi-mode fiber and graded index microlens.³³⁵ For example, a fiber bundle generally consists of ∼100 000 individual step index fibers, which are closely packed with a total diameter, ranging from hundreds of micrometers to a few millimeters. Each core is of high-refractive index and serves as a pixel of an image. Because the arrangement of individual fibers in a fiber bundle is relatively maintained at both ends, the intensity of the image can be transmitted in a pixelated form. The fiber bundle can be introduced into basic fluorescence and confocal laser scanning as well as a two-photon excited fluorescence microscope. Specifically, each core of fiber bundles could also act as a pinhole in a confocal instrument. In a typical FME, the excitation light from the light source (e.g. lasers, LEDs and lamps) is transmitted to the distal end of the fiber bundle, illuminating the sample. The fluorescence emission light from the sample is then gathered by the fiber bundle, followed by sequentially passing through the objective, the dichroic mirror and the emission filter, and finally reaches a CCD camera.

Raman microscopes. Raman signals are in the form of light scattering and are commonly built under the confocal model, which consists of a laser source, band pass filters, pinholes, dichroic mirrors, an objective and a CCD camera.³³⁶ Like laser scanning confocal microscopy (LSCM) of fluorescence, out-of-focus Raman scattering light is minimized to increase the signal-to-light ratio with the assistance of point-like scanning. Depth scanning is also applicable in a confocal Raman microscope to realize 3D imaging. Taking advantage of the fingerprinting spectrum of a specific Raman-active molecule, Raman microscopes can be used for imaging living cells and tissues even without labeling. And the distribution of chemical species in images can be further analyzed using statistical methods such as principal component analysis (PCA).³³⁷ In addition, the common excitation lasers, such as 785 nm and 830 nm, fall in a biological transparent window. Thus, confocal Raman microscopes could also be used for in vivo imaging, but the penetration depth is limited because an objective lens is typically needed.³³⁸ To achieve deeper imaging capability, various types of Raman optical fibers are introduced in a confocal Raman microscope for endoscopic use.^337,339

Considering that the spatial resolution of a confocal Raman microscope is limited to the submicron scale, hyperspectral imaging is combined with Raman scattering into a single microscope to perform dual functions. Hyperspectral imaging can enhance the spatial resolution to tens of nanometers. In addition, Raman scattering is incorporated with a scanning near-field optical microscopy (SNOM) because SNOM can reach the nanometer scale resolution. In a single instrument, SNOM can be used to map the morphology of a sample and Raman scattering can be used to acquire chemical information.

Recent Raman microscopes are mainly used for in vitro imaging for cultured cells and biopsy tissues. The reason is that a sample under Raman microscopes needs to be focused within a specific length to the object. It is difficult to collect Raman scattering with a controlled distance under in vivo conditions. Also, to obtain a high-resolution image in a large area, a Raman microscope should scan the target area sequentially, which is a quite time-consuming process. These drawbacks impede its development in in vivo imaging. Different from Raman imaging, fluorescence emission from a sample is much easier to collect from a large-area or a large sized sample.

5.2 Materials for NIR fluorescence probes

The general design and materials of NIR fluorescence probes have been described in Section 2.4. Compared to SERS imaging, there are much more varieties of materials for NIR fluorescence bio-imaging. Hence, this sub-section is focused on materials for NIR fluorescence probes. For imaging of live cells and tissues, NIR fluorescence probes require a high molar extinction coefficient, high quantum yield, low- or even non-toxicity and good photostability in an aqueous environment. Particularly, in vivo imaging requires deep penetration and low background noise. In the past few decades, numerous NIR probes have been exploited, which can be classified into three types, namely, organic dyes, nanomaterials and proteins.

Organic dyes. The basic structures of organic dyes contain cyanine, squaraine, rhodamine, 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene (BODIPYs), or porphyrin.^19,340,341 Among them, the emission wavelengths of classical rhodamine, BODIPYs and porphyrin dyes are not in the NIR region. Hence, NIR derivatives have been developed by chemical modification to regulate their photophysical properties. For example, NIR rhodamine derivatives can be generated by extending the conjugation of the xanthene moiety or replacing the central oxygen atom with other atoms including N, C, S, Si, P, Se, and Tn.³⁴² Although typical cyanine (Cy5 and Cy7) and squaraine dyes emit in the NIR region, chemical modification is still required to improve their photophysical properties. For example, the emission peak of cyanine dyes can be tuned from the NIR-I window to NIR II window through extending the conjugated structure length or regulating their chemical substitutions.³⁴³ Moreover, traditional dyes suffer from poor water solubility and self-aggregation due to their hydrophobic moieties. When organic dyes aggregate at high concentrations, their fluorescence intensity will decrease because of the aggregation-caused quenching (ACQ) effect. To mitigate these problems, small hydrophilic groups including sulfonate, carboxyl, carboxylate, amine, and maleimide are added to these dyes to increase water solubility. Besides, fluorogens with aggregation-induced emission (AIE) features were exploited by Tang et al.³⁴⁴ AIE fluorogens showed enhanced and stable fluorescence intensity upon aggregation, as well as a high photobleaching threshold. Recently, many AIE fluorogens with NIR emission have been developed by introducing a hydrophobic π-conjugated bridge into AIE initial fluorogens.^345,346 Overall, organic dyes exhibit easy processability and good biocompatibility. There are some issues with organic dyes, such as small Stoke's shift, self-aggregation, and poor photostability in an aqueous medium.

Nanomaterials. Nanomaterials show great potential in NIR fluorescence imaging, such as quantum dots (QDs), graphene-based nanomaterials, carbon nanotubes and halide perovskite nanocrystals.^14,15 At the nanoscale, they can be designed to overcome biological barriers and penetrate cells and tissues. In addition, the optical properties of nanomaterials can be modulated by quantum effects. The emission wavelength depends on the size and shape of nanomaterials. Fluorescent QDs are a class of zero-dimensional (0D) nanomaterials with a typical diameter of 2–20 nm. They exhibit greater photostability and can achieve higher quantum yields than traditional organic dyes, thus enabling long-term imaging with high signal intensity. In addition, they have broad absorption spectra and their emission peaks are tunable and narrow, facilitating multi-component imaging with minimal overlapping.^347,348 At the early stage, NIR QDs contained toxic elements such as Cd, Hg, Pb and As, which lead to cytotoxicity, limiting their biological applications. To mitigate this problem, core/shell nanostructures are designed by coating the toxic core with ZnS, ZnS, SiO₂ or C.³⁴⁹ Alternatively, some materials with less cytotoxicity are used to synthesize NIR QDs, including Ag₂S and Ag₂Se, silicon and carbon.^350,351 Si exists in human body in the form of orthosilicic acid and could be efficiently excreted out of the body through urine. Thus, Si QDs display relatively good biocompatibility and biodegradability features. Fluorescence emission of Si QDs can be tuned to the NIR-I region by varying the particle size; they can become water soluble by surface modification. Because the bulk band gap (1.12 eV) of Si is larger than that of PbS and AgS, it is not practical to increase their size for reaching the NIR-II emission region. In this case, boron (B) and phosphorus (P) co-doping makes it possible to generate NIR-II Si QDs.

Similarly, biocompatible and biodegradable carbon quantum dots (CQDs) are alternatives to traditional metal-based QDs. Their absorption and NIR emission spectra can also be tuned via size control, surface modification and heteroatom doping. Furthermore, carbon quantum dots display excitation-wavelength-dependent fluorescence due to the “giant red-edge effect”.³⁵² Hence, they are able to present different fluorescence emission bands in imaging under different excitation wavelengths. Other carbon materials such as graphene oxides and carbon nanotubes (CNTs) have been explored for bio-imaging. Pristine graphene with solely sp²-carbon atoms do not show fluorescence. Graphene oxides and graphene quantum dots (GQDs) consist of a mixture of sp² and sp³ carbon atoms, thus they have an opened band gap; their surface possess hydrophilic groups such as hydroxyl, carboxyl and carbonyl groups that make themselves water soluble. Like carbon quantum dots, free-standing graphene oxides and graphene quantum dots exhibit excitation-wavelength-dependent fluorescence, and they are photostable and biocompatible.

Recently halide perovskite nanocrystals have emerged for NIR fluorescence imaging.^28,353 Among the vast perovskite families, lead-based halide perovskite (LHP) nanocrystals have attracted great interest due to their high quantum yield. It has been reported that lead-based halide perovskite nanocrystals achieved a photoluminescence quantum yield of up to 90% without any post-treatment.³⁵⁴ Lead-based halide perovskites have a very narrow fluorescence emission band that can be tuned by either their composition or/and size. However, halide perovskite nanocrystals typically contain toxic elements, and may exhibit poor stability in an aqueous solution.

Proteins. Fluorescence proteins (FPs) were initially discovered from jellyfish and corals. Fluorescence proteins and their derivatives have revolutionized many fields of molecular and cell biology. However, most of them exhibit the excitation band in the spectral range less than 610 nm. It is worth noting that Tsien's group reported monomeric NIR fluorescence protein with the excitation/emission maxima at 684/708 nm by engineering the bacterial phytochromes (BphPs) of Deinococcus radiodurans.³⁵⁵ Similarly, other NIR fluorescence proteins have been exploited from Rhodopseudomonas palustris.²⁷ The chromophore of NIR fluorescence protein is converted from haeme, which is influenced by haeme oxygenase 1. The fluorescence intensity of these proteins in cells could be enhanced via engineered cofactor biosynthesis.³⁵⁶ In contrast to other NIR probes, fluorescence proteins feature exceptional biocompatibility and biodegradation; they are widely used in monitoring gene expression, tumor growth, and bacterial or viral infections in vitro and in in vivo imaging.

5.3 Signal transduction

For fluorescence or SERS imaging of live cells and tissues, there are three common signal transduction modes/principles: (i) direct imaging with non-targeting optical probes, (ii) imaging with targeting optical probes, and (iii) activatable imaging with triggering modes.

Direct imaging with non-targeting optical probes. Direct imaging is independent on specific binding events to the target/analyte, and a signal is under the “always on” mode. Generally, live cells or tissues are stained with fluorescence probes to directly visualize the structures of cells or tissues (e.g., tumor vessels) and to monitor their dynamic changes.^357–359 For example, Dai's group fabricated NIR-II PbS/CdS QDs for noninvasive in vivo imaging of tumor tissues (Fig. 23a).³⁶⁰ After the tail vein injection of such QDs into a mouse with xenograft MC38 tumors, the PbS/CdS QDs were circulated over time in vessels. Because of the increasing structural abnormalities and the permeability of tumor vasculatures (enhanced permeability and retention (EPR) effect), these QDs were gradually leaked from vessels and accumulated at the tumor sites, brightening up the whole tumor. Moreover, these QDs could be excreted over 1 month, providing a tool for localizing tumors and real-time monitoring their dynamic changes. Because inorganic quantum dots exhibit long-term photostability and capability of real-time localization, they are suitable for imaging-guide surgery, especially for the surgery resection of brain tumors. Brain tumor margins are irregular and indistinct, and extensive tumor-induced angiogenesis increases the risk of irreversible damage. Li et al. reported a dual-modality nanoprobe (Gd–DOTA–Ag₂S QDs) together with both MRI and fluorescence imaging functions for highly efficient and precise physical resection of brain tumors.³⁵⁷ After nanoprobes were accumulated at the tumor sites based on the permeability and retention effect, the deep tissue penetration capability of MRI enables the identification of tumor locations. Simultaneously, NIR fluorescence imaging guided the intraoperative resection precisely.

Fig. 23 Direct imaging of tissues with stained NIR fluorescence or SERS probes. (a) In vivo fluorescence imaging of tumor tissues through the tail vein injection of PbS/CdS QDs into a mouse with xenograft MC38 tumors. (1) Schematic structure of PbS/CdS QDs probe. (2) TEM image of QDs. (3) Schematic illustration of non-invasive in vivo confocal imaging of the mouse through the skin; (4) in vivo wild-field fluorescence imaging of tumor tissues after the tail vein injection of QDs. Scale bar: 1 mm; (5) confocal fluorescence imaging of tumor vessels at a depth of ∼180 μm. Scale bar: 200 μm. (b) SERS imaging of characteristic biomolecules in cells. (1) SERS signals from absorbed molecules and changes with their distance to gold nanoparticles. (2) A photograph of a macrophage cell using a dark-field microscope and the gold nanoparticles shown as a small white spot (white arrow) under the dark-field microscope. (3) SERS spectra obtained from (a) the three characteristic peaks of phosphate at 997 cm⁻¹, CH₂ and CH₃ at 1457 cm⁻¹, and Amid II at 1541 cm⁻¹. (4) Trajectory of gold nanoparticles under the dark-field microscope. (5) False-color mapping of the molecular distribution based on the trajectory of the nanoparticles (4). Reproduced with permission (a) ref. 360, copyright 2018, National Academy of Sciences; (b) ref. 361, copyright 2011, American Chemical Society.

SERS probes without specific targeting moieties can also be used in tumor imaging by utilizing the permeability and retention effect. For example, PEGlyated sandwiched SERS nanostars with a LOD down to 1.5 fM were successfully used for the imaging of premalignant lesions of pancreatic and prostatic neoplasias.³⁶² In addition, based on the “finger-printing” features of SERS spectra, SERS imaging can be directly used for the identification of cells and monitoring of molecular evolution in cells or tissues. After the Raman dye-free plasmonic metal nanoparticles entered the cells or tissues, the SERS-active molecules in the living body, such as proteins with chromophore groups (e.g. hemoglobin and myoglobin), nucleic acids, and metabolites (e.g. lactic acid and glucose), can be detected when they were in proximity with the plasmonic nanoparticles. In the meanwhile, a dark-field microscope was employed to map the position of plasmonic metal nanoparticles in living cells, tracking the SERS signals of characteristic biomolecules at different sites of cells, which can continuously monitor intracellular information (Fig. 23b).³⁶¹ In order to mitigate the uncertainty and uncontrollability of the distribution of nanoparticles in living cells, researchers deposited metal nanoparticles on a solid substrate to form a SERS-active substrate.³⁶³ After seeding cells on such a SERS-active substrate, the biomolecules on the cell membrane get close to the substrate, thus obtaining the SERS signals of these biomolecules. The SERS images, which are obtained from the cells on the plasmonic substrate, can monitor the cell state because these biomolecules on the cell membrane play a vital role in regulating the signal pathway, cell behavior and in identifying the cell type and condition. The large amount of data generated in a dynamic system can be analyzed by statistical methods such as principal component analysis (PCA) and multivariate curve resolution (MCR).

Imaging with targeting optical probes. In this mode, the signals of probes are under the “always on” mode like the direct imaging mode, but the probe is conjugated with the specific ligands (antibodies or aptamers) as the molecular recognition elements that specifically bind to the targets/analytes through the antibody–antigen interaction or the aptamer–nucleic acid interaction. Compared with direct imaging, this mode can provide more specific information about biological events.^364–366 Furthermore, ligand-assisted imaging could concentrate optical probes at the targeted sites, leading to higher sensitivity. Ghosh et al. compared the signal intensity of single-walled carbon nanotubes (SWNTs) with the conjugated ovarian tumor-targeting ligands and without ligands.³⁶⁷ It was found that those with the conjugated ligands showed 10 times higher signal-to background ratio than that of the un-conjugated counterparts. The NIR photoluminescence of single-walled carbon nanotubes falls into the biological transparent window, which endows them with higher sensitivity for imaging-guided surgery, especially for submillimeter tumors. SERS probes are also characteristic of high sensitivity for in vivo imaging. For example, a sandwich SERS probe with a polyethylene glycol (PEG) shell was used for the imaging of human cancer cells and xenograft tumors in mice (Fig. 24).³⁶⁸ These PEGlyated SERS probes showed not only great stability over a wide range of pH solutions but also high signal intensity. It was observed that the SERS intensity of the adsorbed molecules increased as much as 10¹⁴ to 10¹⁵ folds when the SERS probes were conjugated with the tumor-targeting ligands.

Fig. 24 In vivo cancer targeting and SERS detection using the PEGlyated SERS probes conjugated with the single-chain variable fragment (ScFv) antibodies, which can recognize the tumor biomarker epidermal growth factor receptor (EGFR). (a) Schematic illustration of the structure of the PEGlyated SERS probes. (b) In vivo SERS spectra obtained from the targeting and non-targeting SERS probes. (c) Photographs illustrating a 785 nm laser beam focusing on the tumor site or on the anatomical location of the liver. Reproduced with permission ref. 368, copyright 2008, Springer Nature.

Imaging with triggering modes. Fluorescence or SERS imaging can be conducted on either the “turn on” or “turn off” mode in response to targets, bio-events or their surroundings. Compared with the “always on” mode, activatable imaging is more convenient, accurate and sensitive in imaging with minimal interference by the background signal. In activatable fluorescence imaging, the signal transduction principles are based on various physicochemical processes such as electron transfer (ET), Förster resonance energy transfer (FRET), aggregate-induced emission (AIE), etc. Signal transduction via ET and FRET is based on the donor–receptor interaction. ET can be categorized into intracellular charge transfer (ICT) and photoinduced-electron transfer (PET), which often take place when the distance between the donor and the acceptor is less than 3 nm, such as chemical bounds and physical absorption. Different from ET, FRET is a relatively long-range energy transfer process up to 10 nm. In addition, the AIE process is often used in signal transduction, in which the initial “fluorescence-off” probes are solute or dispersive in the aqueous environment. In the presence of a hydrophobic target (e.g., proteins), the “fluorescence-off” probes bind with the target specifically, which induces the aggregation of probes, turning on fluorescence emission.

In activable SERS imaging, the signal trigger is the environment-sensitive Raman reporter. The signal variation of Raman reporters reflects the physicochemical changes in living cells and tissues. For example, 4-mercaptobenzoic acid (MBA), 2-aminothiophenol and 4-mercaptopyridine are pH-sensitive Raman reporters, while indocyanine green (ICG) is a thermosensitive reporter.^369–371

Fluorescence or SERS imaging can be triggered by various operating parameters (e.g. pH, viscosity, temperature) or biomarkers (e.g., small molecules, protein, nucleic acids) in cells and tissues. For example, the pH value plays a vital role in biological processes, such as ion transport, endocytosis and cell growth. Under normal physiological conditions, the pH value is around 7.2 and 7.4 in the cytosolic and extracellular environment, respectively; the pH value in lysosomes and mitochondria is approximately 5.0 and 8.0, respectively. Abnormal pH values may reflect pathological conditions, including cancer and Alzheimer's diseases. For instance, the pH value of lysosomes in the tumor cells becomes more acidic (pH_ly 3.8–4.7) than the normal one. Fig. 25a illustrates the effect of pH on the ET process resulting from protonation/deprotonation-induced electron distribution.³⁷² Briefly, a pH-responsive probe was constructed using a heptamethine cyanine dye, which had good biocompatibility and a relatively high quantum yield. The NIR fluorophore (heptamethine cyanine derivatives IR783) served as the electron acceptor, and the amine moiety on one indole ring was the electron donor in neutral pH. When the probe was delivered into lysosomes through an organic anion transporting peptide, the acidic environment protonated the amine groups, triggering fluorescence emission. Visualization of a pH change in cancer cells can localize cancer cells specifically, and guide tumor ablation. In addition, Tang et al. developed a pH-triggered imaging method using the PET process, which strongly depends on a pH change from 6.70 to 7.90.³⁷³ It was used successfully for the real-time imaging of cellular pH in living HepG2 and HL-7702 cells. However, it is difficult to achieve a broad range of pH detection in these single-electron transfer modes, especially from acidity to alkalinity, to show the overall pH change in cells. A combinational electron transfer-energy transfer method was thus exploited to create an ICT-PET-FRET probe with dual pH-responsive sites.³⁷⁴ Therefore, the probe responded to a wide pH range from 4.0 to 8.0 with high sensitivity. It can be utilized not only in fluorescence imaging of lysosomal pH values with one of the pH sites targeting the lysosome place, but also in therapeutic effect evaluation based on the intracellular pH changes in cells after chloroquine treatment. In addition, taking advantage of the pH change from normal to tumor tissues, researchers have developed pH-sensitive imaging and drug delivery to realize imaging-guided surgery. For example, Song et al. labeled a hydrophilic polymer brush, a pH-sensitive hydrophobic brush and a Raman reporter on the surface of gold nanoparticles to form SERS-encoded amphiophilic nanoparticles (Fig. 25b).³⁷⁵ These nanoparticles can be further self-assembled into vesicles with the anticancer drugs embedded in the hollow cavity. After being tagged with the cancer-targeting ligands on the outer surface of SERS vesicles, these vesicles were successfully bound to the target cancer cells with strong SERS signals due to their assembly structure. Afterwards, the SERS vesicles were transferred into cytosol through endocytosis and the acidic intracellular compartments triggered the pH-responsive disassembly of SERS vesicles to form single gold particles. Compared with the SERS vesicles, these dispersed single gold nanoparticles showed much weaker signal intensity. Simultaneously, the embedded drugs were released with the disassembly of the SERS vesicles.

Fig. 25 Biomarker-triggered imaging. (a) pH activated electron transfer (ET) process in an organic NIR dye for cancer cell visualization. (b) Drug-loaded and targeted plasmonic vesicles for cancer cell targeting, imaging and therapy via pH-sensitive drug release. (c) Viscosity activated in vivo imaging of tumor tissues. Reproduced with permission (a) ref. 372, copyright 2016, The Royal Society of Chemistry; (b) ref. 375, copyright 2012, American Chemical Society; (c) ref. 376, copyright, American Chemical Society.

Viscosity is one of the critical microenvironment parameters, which is associated with the transportation and reactions of biomolecules as well as chemical signals within cells. Viscosity perturbances are caused by diseases and pathologies, such as diabetes, atherosclerosis, cancers and Alzheimer's diseases. As it is an important indicator in disease diagnostics, a series of molecular rotors has been developed to map the cellular viscosity. For example, a molecular rotor, malononitrile moiety, was linked to 4H-chromene groups modified triphenylamine as the NIR fluorophore to form viscosity-sensitive probes. As shown in Fig. 25c, in low-viscosity media, after the fluorophore was excited by a proton, the excited electron was transferred into a twisted excited state due to the rotation of the malononitrile group, which reached a non-emissive decay.³⁷⁶ In contrast, in a high-viscosity media, molecular rotation was restricted, and the ET process was then inhibited, leading to the emissive decay of NIR fluorescence. This method was used to identify cancer, which showed higher viscosity than normal tissues, and to evaluate the efficacy of anticancer drugs.

Imaging can be triggered by the unique features of biomolecules or reactions, such as hydrophobic property of proteins, chain reactions of nucleic acids, or specific chemical reaction of small molecules. Fig. 26a shows the images of nucleic acids using FRET signal transduction and nucleic acid amplification to overcome the problems of intrinsic low-abundance analytes.³⁷⁷ First, the target microRNA-21 (miR-21), a crucial tumor-associated biomarker, was recycled via an Exo-III-assisted assay, and the exposed DNAzyme was further partially hybridized with a fluorophore/quencher pair. After experiencing biocatalysis by Mg²⁺ in cells, the fluorophore/quencher pair was cleaved, thus the fluorescence of the separated fluorophore was restored. In addition, the base-pairing complementarity can also be used in the intermolecular FRET process. For example, Hwang et al. developed a fluorophore-labeled nucleic acid as an imaging probe whose fluorescence was initially quenched by graphene oxide (GO) via FRET.³⁷⁸ After the probes were internalized into cancer cells through CD44 receptor-mediated endocytosis, the nucleic acid of the probe was specifically hybridized with the target when encountering the target miRNA, leading to the detachment of the probes from graphene oxide and recovery of fluorescence. Fig. 26b reveals an AIE imaging process activated by amyloid-β (Aβ) plaques, an important pathological hallmark for the early diagnosis of Alzheimer's disease.³⁷⁹ Initially, the AIE probe was under the “fluorescence-off” state due to the hydrophilicity of the sulfonate group of the probe. The fluorescence of the probe was turned on when binding to the Aβ aggregates via the N,N-dimethylamino group in the probe.

Fig. 26 (a) MicroRNA-21 (miR-21)-activated imaging through a FRET process together with Exo-III-assisted nucleic acid amplification. (i) Schematic illustration of the imaging process; (ii) fluorescence imaging of miR-21 in breast cancer cells (MCF-7) with high miR-21 expression, HeLa cells with low miR-21expression and MRC-5 cells with no miR-21expression. The scale bar is 20 μm. (b) Amyloid-β (Aβ) plaque-activated AIE process for in situ imaging of Alzheimer's disease (AD). (i) Schematic illustration of synthesis of AIE probe and imaging process; (ii) in vivo imaging of Aβ deposition in the AD model (APP/PS1transgenic) mice in comparison to normal wild-type mice; (iii) ex vivo fluorescence imaging of brain slices from APP/PS1 mice after the injection of AIE probes (QM-FN-SO3) with the colocalization of Aβ plaques stained with anti-Aβ antibody-2454. Reproduced with permission from (a) ref. 377, copyright 2020, American Chemical Society; (b) ref. 379, copyright 2019, American Chemical Society.

In cells, there exist a lot of chemically active molecules, such as reactive oxygen species (ROS), reactive nitrogen species (RNS), enzymes and metal ions. Thus, activable imaging can also be triggered by chemical reactions. Fig. 27 shows a NIR fluorescence imaging process activated by an alkaline phosphatase (ALP) reaction.³⁸⁰ Initially, fluorescence of merocyanine was quenched by the modified phosphate group via the electron transfer process. When it comes to ALP-positive tumor tissues, the membrane-bound ALP dephosphorylated the phosphate group, leading to the emission of NIR fluorescence. In this design, fluorescence imaging was combined with magnetic resonance imaging (MRI) through the modification of paramagnetic DOTA-Gd chelate to the probe with a hydrophobic dipeptide liner. After being dephosphorylated, the hydrophobic interaction of dipeptide promoted the molecular self-assembly into magnetic nanoparticles.

Fig. 27 In vivo bimodal imaging for tumor tissue visualization: alkaline phosphatase (ALP) reaction-triggered ET imaging combined with magnetic resonance imaging (MRI). (a) Chemical mechanism of the ALP reaction-triggered NIR fluorescence imaging and the self-assembly for further signal amplification; (b) schematic illustration of in vivo bimodal imaging of ALP-positive tumor cells; (c) ALP reaction-triggered NIR fluorescence imaging (left) and MRI imaging (right). Reproduced with permission from ref. 380, copyright 2019, American Chemical Society.

6. Remarks and perspective

In summary, further efforts are needed to design and fabricate plasmonic nanostructures to achieve the effective coupling of plasmon with fluorescence or SERS to amplify signals. These efforts will include the commercialization of plasmonic NIR fluorescence and SERS probes. These probes should be inexpensive, water-soluble, “bright” (strong intensity), stable in storage, shipment and operation, and repeatable from batch to batch.⁴⁴ Commercial availability of these plasmonic probes is the key step in boosting the application of NIR fluorescence and SERS in POCT and bio-imaging.

The pregnancy test strip is the most successful example of a commercial POCT device. Like pregnancy test strips, most commercial lateral flow assays are based on colorimetric transduction. Such devices are inexpensive and can be used for rapidly and qualitatively detecting a relatively high level of analytes. However, they suffer from low sensitivity and severe interference from complex sample matrices such as blood. In contrast, NIR fluorescence and SERS microfluidic devices show great potential in sensitive and quantitative measurements of a low level of analytes in complex sample matrices. However, such POCT devices typically require the integration of a complex plasmonic nanostructure into a microfluidic chip in order to amplify NIR fluorescence or SERS signals. Although significant progress has been made in such a direction, there remain large technical barriers in fabricating and integrating 1D, 2D or 3D plasmonic nanostructures into microfluidic POCT devices.²⁹¹

Today there are a lot of commercial POCT devices based on immunoassays with antibodies as molecular recognition elements. These immunoassays have been successfully used for detecting illicit drugs, HIV antibodies, biological warfare agents, pathogens and so on. They are typically used for the qualitative measurement of analytes in relatively simple sample matrices such as tap water, river water, saliva and urine. So far, few POCT devices are commercially available for the direct and quantitative measurement of analytes in finger-prick blood samples because whole blood is a much more challenging sample matrix as compared to buffer solutions, water, saliva and urine. Integration of a plasma separation unit and a NIR fluorescence or SERS detection platform into a single microfluidic or paper lateral flow strip is an effective solution to blood analysis in POC settings.²⁹²

There are few commercial POCT devices for nucleic acid testing. For example, no commercial POCT device is now available for reliable testing of SARS-COV-2 viral RNA in saliva or nasal swab samples in POC settings. One of the reasons is that the levels of nucleic acid biomarkers of interest, such as DNA, RNA and micro-RNA, are typically too low to be detected directly with a sensor. To enable nucleic acid testing in POC settings, both the amplification and detection of nucleic acid should be performed with a single lab-on-chip device. In some cases, an on-chip lysis and nucleic acid extraction module also need to be integrated into a POCT device. In short, nucleic acid POCT devices are complicated and quite challenging. Many researchers are tackling such challenges now. It is believed that more and more POCT devices will appear in the market for rapidly testing nucleic acids.

Besides test chips/strips/cartridges, readers are critical for POCT testing, and many inexpensive handheld fluorescence readers are commercially available for directly reading fluorescence signals from test chips/strips/cartridges. However, inexpensive handheld SERS readers are not commercially available to fit into test chips/strips/cartridges, which has hindered the commercialization of SERS-based POCT systems.

Smartphones show great potential in recording signals from fluorescence and SERS POCT test chips/strips/cartridges.³⁸¹ One option is the integration of an optical detector into an existing smartphone, enabling the direct reading of signals with the smartphone. Another option is the connection of a smartphone with an existing reader via Bluetooth, enabling the wireless operation of the reader with the smartphone using an app, allowing for the wireless transmission of data to a doctor's office or a central station via a wireless 4G or 5G network. Such readers, which are wirelessly connected to smartphones, could be operated by someone at a remote central station.

Incorporation of numerous POCT devices, ambulances, clinics and hospitals into a wireless sensor network or a cloud-based network is an important direction in the future. Such a cloud-based network will play important roles in multi-facets of public health and security in smart-cities, -towns, or -countries. For example, it will allow for the real-time monitoring for the reoccurrence or of the treatment process of chronic diseases. And it can be used to track acute, infected, or asymptotic COVID-19 patients. Furthermore, it can be employed to monitor the occurrence and progression of pandemic or epidemic diseases.

Besides in vitro POCT, NIR fluorescence and SERS imaging can achieve fast screening and real-time tracking and monitoring of specific targets in living cells and tissues. They can greatly overcome the drawbacks of current large diagnostic instruments, such as high cost, time consumption and requirement of professional training. It is worth noting that longer-wavelength probes exhibit weaker signals than shorter-wavelength counterparts. However, the reduced light absorption and scattering by tissues and the depressed tissue autofluorescence not only enable deeper penetration but also allow for longer acquisition time with a greater signal-to-background ratio. On the other hand, in the case of rapid screening demand such as image-guided surgical applications, use of shorter NIR wavelengths and tolerance of higher background noise may be necessary for achieving stronger signals.

Some of the technical barriers for bio-imaging are the biocompatibility and biodegradability of NIR fluorescence and SERS probes. It is highly recommended to not only construct biocompatible and biodegradable probes, but also to reveal their mechanisms of action and metabolism, which can provide clear guidelines for choosing appropriate materials and structures to design probes. In addition, improvement in spatial resolution of imaging is an important research direction. This will benefit the site-specific detection or mapping of physiological or pathological processes in live cells, tissues and organs. Moreover, theranostics, which combines therapy with imaging functions into a single entity to form therapeutic agents, shows great potential. It will allow for time-solved and space-resolved tracking and adjustment of medical intervention.

It is estimated that the global POCT market will increase quickly from $29.5 billion in 2020 to $50.6 billion by 2025.³⁸² The market will drive the advancement in POCT techniques. With the increasing need for sensitive and quantitative measurements of analytes in POC settings, the market share will shift from colorimetric and electrochemical devices to NIR fluorescence and SERS devices. To the best of our knowledge, we do not have quantitative data about the market value of NIR fluorescence and SERS bio-imaging. The technological development in NIR fluorescence and SERS bio-imaging will boost the growth of the market in these areas in the future.

Conflicts of interest

The authors declare no conflict of interests.

Acknowledgements

This work is partially supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (1U01NS119647). Wu is partially supported by the Armstrong-Siadat Professorship Endowment. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

References

1. D. Wang, S. He, X. Wang, Y. Yan, J. Liu, S. Wu, S. Liu, Y. Lei, M. Chen, L. Li, J. Zhang, L. Zhang, X. Hu, X. Zheng, J. Bai, Y. Zhang, Y. Zhang, M. Song and Y. Tang, Nat. Biomed. Eng., 2020, 4, 1150–1158.
2. J. Reboud, G. Xu, A. Garrett, M. Adriko, Z. Yang, E. M. Tukahebwa, C. Rowell and J. M. Cooper, Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 4834–4842.
3. M. K. Yuzon, J.-H. Kim and S. Kim, BioChip J., 2019, 13, 277–287.
4. S. Chaiyo, E. Mehmeti, W. Siangproh, T. L. Hoang, H. P. Nguyen, O. Chailapakul and K. Kalcher, Biosens. Bioelectron., 2018, 102, 113–120.
5. L. Wu and X. Qu, Chem. Soc. Rev., 2015, 44, 2963–2997.
6. B. Ning, T. Yu, S. Zhang, Z. Huang, D. Tian, Z. Lin, A. Niu, N. Golden, K. Hensley, B. Threeton, C. J. Lyon, X.-M. Yin, C. J. Roy, N. S. Saba, J. Rappaport, Q. Wei and T. Y. Hu, Sci. Adv., 2021, 7, eabe3703.
7. J. Meng, X. Tang, B. Zhou, Q. Xie and L. Yang, Talanta, 2017, 164, 693–699.
8. Y. Song, B. Lin, T. Tian, X. Xu, W. Wang, Q. Ruan, J. Guo, Z. Zhu and C. Yang, Anal. Chem., 2019, 91, 388–404.
9. X. Gao, H. Xu, M. Baloda, A. S. Gurung, L.-P. Xu, T. Wang, X. Zhang and G. Liu, Biosens. Bioelectron., 2014, 54, 578–584.
10. A. W. Martinez, S. T. Phillips, G. M. Whitesides and E. Carrilho, Anal. Chem., 2010, 82, 3–10.
11. M. Li, H. Gou, I. Al-Ogaidi and N. Wu, ACS Sustainable Chem. Eng., 2013, 1, 713–723.
12. Y. Song, Y. Y. Huang, X. Liu, X. Zhang, M. Ferrari and L. Qin, Trends Biotechnol., 2014, 32, 132–139.
13. S. F. Berlanda, M. Breitfeld, C. L. Dietsche and P. S. Dittrich, Anal. Chem., 2021, 93, 311–331.
14. C. H. Quek and K. W. Leong, Nanomaterials, 2012, 2, 92–112.
15. F. Yang, Q. Zhang, S. Huang and D. Ma, J. Mater. Chem. B, 2020, 8, 7856–7879.
16. E. Hemmer, A. Benayas, F. Legare and F. Vetrone, Nanoscale Horiz., 2016, 1, 168–184.
17. A. M. Smith, M. C. Mancini and S. Nie, Nat. Nanotechnol., 2009, 4, 710–711.
18. L. Sordillo, Y. Pu, S. Pratavieira, Y. Budansky and R. Alfano, J. Biomed. Opt., 2014, 19, 056004.
19. S. Zhu, R. Tian, A. L. Antaris, X. Chen and H. Dai, Adv. Mater., 2019, 31, e1900321.
20. F. Scholkmann, S. Kleiser, A. J. Metz, R. Zimmermann, J. Mata Pavia, U. Wolf and M. Wolf, NeuroImage, 2014, 85(part 1), 6–27.
21. C. Swanson and A. D'Andrea, Clin. Chem., 2013, 59, 641–648.
22. Y. Haga, H. David Kay, M. A. Tempero and R. K. Zetterman, Clin. Biochem., 1992, 25, 277–283.
23. O. O. Abugo, R. Nair and J. R. Lakowicz, Anal. Biochem., 2000, 279, 142–150.
24. A. N. Semenov, B. P. Yakimov, A. A. Rubekina, D. A. Gorin, V. P. Drachev, M. P. Zarubin, A. N. Velikanov, J. Lademann, V. V. Fadeev, A. V. Priezzhev, M. E. Darvin and E. A. Shirshin, Molecules, 2020, 25, 1863.
25. H. Zeng, C. MacAulay, D. I. McLean and B. Palcic, J. Photochem. Photobiol., B, 1997, 38, 234–240.
26. R. A. Álvarez-Puebla, J. Phys. Chem. Lett., 2012, 3, 857–866.
27. A. Miyawaki, Nat. Methods, 2016, 13, 729–730.
28. I. Lignos, V. Morad, Y. Shynkarenko, C. Bernasconi, R. M. Maceiczyk, L. Protesescu, F. Bertolotti, S. Kumar, S. T. Ochsenbein, N. Masciocchi, A. Guagliardi, C. J. Shih, M. I. Bodnarchuk, A. J. deMello and M. V. Kovalenko, ACS Nano, 2018, 12, 5504–5517.
29. X.-H. Chang, J. Zhang, L.-H. Wu, Y.-K. Peng, X.-Y. Yang, X.-L. Li, A.-J. Ma, J.-C. Ma and G.-Q. Chen, Micromachines, 2019, 10, 422.
30. A. L. Antaris, H. Chen, S. Diao, Z. Ma, Z. Zhang, S. Zhu, J. Wang, A. X. Lozano, Q. Fan, L. Chew, M. Zhu, K. Cheng, X. Hong, H. Dai and Z. Cheng, Nat. Commun., 2017, 8, 15269.
31. M. R. Jones, K. D. Osberg, R. J. Macfarlane, M. R. Langille and C. A. Mirkin, Chem. Rev., 2011, 111, 3736–3827.
32. M. Li, S. K. Cushing and N. Wu, Analyst, 2015, 140, 386–406.
33. J. Dostálek and W. Knoll, Biointerphases, 2008, 3, Fd12–Fd22.
34. J. R. Lakowicz, Plasmonics, 2006, 1, 5–33.
35. N. W. Ashcroft and D. N. Mermin, Solid State Physics, Cengage Learning, 1976.
36. V. M. Shalaev and S. Kawata, Nanophotonics with Surface Plasmon, Elsevier, 2007.
37. K. M. Mayer and J. H. Hafner, Chem. Rev., 2011, 111, 3828–3857.
38. P. Mulvaney, Langmuir, 1996, 12, 788–800.
39. P. L. Stiles, J. A. Dieringer, N. C. Shah and R. P. Van Duyne, Annu. Rev. Anal. Chem., 2008, 1, 601–626.
40. M. Moskovits, Rev. Mod. Phys., 1985, 57, 783–826.
41. A. M. Schwartzberg and J. Z. Zhang, J. Phys. Chem. C, 2008, 112, 10323–10337.
42. H. Tang, C. Zhu, G. Meng and N. Wu, J. Electrochem. Soc., 2018, 165, B3098–B3118.
43. J. M. McMahon, S. Li, L. K. Ausman and G. C. Schatz, J. Phys. Chem. C, 2011, 116, 1627–1637.
44. M. Li, S. K. Cushing, J. Zhang, J. Lankford, Z. P. Aguilar, D. Ma and N. Wu, Nanotechnology, 2012, 23, 115501.
45. M. Wang, B. B. Rajeeva, L. Scarabelli, E. P. Perillo, A. K. Dunn, L. M. Liz-Marzan and Y. Zheng, J. Phys. Chem. C, 2016, 120, 14820–14827.
46. J. R. Lakowicz, Anal. Biochem., 2005, 337, 171–194.
47. I. Gryczynski, J. Malicka, Z. Gryczynski and J. R. Lakowicz, Anal. Biochem., 2004, 324, 170–182.
48. M. Käll, H. Xu and P. Johansson, J. Raman Spectrosc., 2005, 36, 510–514.
49. S. D. Christesen, Appl. Spectrosc., 1988, 42, 318–321.
50. M. Fleischmann, P. J. Hendra and A. J. McQuillan, Chem. Phys. Lett., 1974, 26, 163–166.
51. D. L. Jeanmaire and R. P. Van Duyne, J. Electroanal. Chem. Interfacial Electrochem., 1977, 84, 1–20.
52. M. G. Albrecht and J. A. Creighton, J. Am. Chem. Soc., 1977, 99, 5215–5217.
53. W. E. Doering and S. Nie, J. Phys. Chem. B, 2002, 106, 311–317.
54. K. Kneipp, H. Kneipp, I. Itzkan, R. R. Dasari and M. S. Feld, Chem. Rev., 1999, 99, 2957–2976.
55. J. Kim, Y. Jang, N.-J. Kim, H. Kim, G.-C. Yi, Y. Shin, M. H. Kim and S. Yoon, Front. Chem., 2019, 7, 582.
56. G. C. Schatz, M. A. Young and R. P. Van Duyne, Surface-Enhanced Raman Scattering: Physics and Applications, Springer, Berlin Heidelberg, 2006 DOI:10.1007/3-540-33567-6_2.
57. E. C. Le Ru, E. Blackie, M. Meyer and P. G. Etchegoin, J. Phys. Chem. C, 2007, 111, 13794–13803.
58. M. Li, S. K. Cushing, G. Zhou and N. Wu, Nanoscale, 2020, 12, 22036–22041.
59. M. Schutz, C. I. Muller, M. Salehi, C. Lambert and S. Schlucker, J. Biophotonics, 2011, 4, 453–463.
60. D. Graham, K. Faulds and W. E. Smith, Chem. Commun., 2006, 4363–4371.
61. T. Keller, S. Brem, V. Tran, O. Sritharan, D. Schäfer and S. Schlücker, J. Biophotonics, 2020, 13, e201960126.
62. X. Mu, Y. Guo, Y. Li, Z. Wang, Y. Li and S. Xu, J. Raman Spectrosc., 2017, 48, 1196–1200.
63. L. Lin, X. Tian, S. Hong, P. Dai, Q. You, R. Wang, L. Feng, C. Xie, Z.-Q. Tian and X. Chen, Angew. Chem., Int. Ed., 2013, 52, 7266–7271.
64. M. Li, J. Zhang, S. Suri, L. J. Sooter, D. Ma and N. Wu, Anal. Chem., 2012, 84, 2837–2842.
65. A. M. Gabudean, D. Biro and S. Astilean, J. Mol. Struct., 2011, 993, 420–424.
66. A. Michota and J. Bukowska, J. Raman Spectrosc., 2003, 34, 21–25.
67. X. Gao, P. Zheng, S. Kasani, S. Wu, F. Yang, S. Lewis, S. Nayeem, E. B. Engler-Chiurazzi, J. G. Wigginton, J. W. Simpkins and N. Wu, Anal. Chem., 2017, 89, 10104–10110.
68. X. Tan, Z. Wang, J. Yang, C. Song, R. Zhang and Y. Cui, Nanotechnology, 2009, 20, 445102.
69. P. Zheng, M. Li, R. Jurevic, S. K. Cushing, Y. Liu and N. Wu, Nanoscale, 2015, 7, 11005–11012.
70. P. Zheng, S. Kasani, X. Shi, A. E. Boryczka, F. Yang, H. Tang, M. Li, W. Zheng, D. E. Elswick and N. Wu, Anal. Chim. Acta, 2018, 1040, 158–165.
71. J. E. Park, J. Kim and J. M. Nam, Chem. Sci., 2017, 8, 4696–4704.
72. J. R. Lakowicz, Principles of Fluorescence Spectroscopy, Springer Academic, 2006.
73. T. Itoh, Y. S. Yamamoto and Y. Ozaki, Chem. Soc. Rev., 2017, 46, 3904–3921.
74. J. B. Khurgin and G. Sun, J. Opt. Soc. Am. B, 2009, 26, B83–B95.
75. J. Li, S. K. Cushing, F. Meng, T. R. Senty, A. D. Bristow and N. Wu, Nat. Photonics, 2015, 9, 601–607.
76. M. Li, X. Zhou, W. Ding, S. Guo and N. Wu, Biosens. Bioelectron., 2013, 41, 889–893.
77. J. Yang, F. Zhang, Y. Chen, S. Qian, P. Hu, W. Li, Y. Deng, Y. Fang, L. Han, M. Luqman and D. Zhao, Chem. Commun., 2011, 47, 11618–11620.
78. L. Trotsiuk, A. Muravitskaya, O. Kulakovich, D. Guzatov, A. Ramanenka, Y. Kelestemur, H. V. Demir and S. Gaponenko, Nanotechnology, 2020, 31, 105201.
79. F. Tang, F. He, H. Cheng and L. Li, Langmuir, 2010, 26, 11774–11778.
80. D. Gontero, A. V. Veglia, D. Boudreau and A. G. Bracamonte, J. Nanophotonics, 2017, 12, 012505.
81. X. Wang, L. Zhang, A. Hao, Z. Shi, C. Dai, Y. Yang and H. Huang, ACS Appl. Nano. Mater., 2020, 3, 9796–9803.
82. Y. Gao, J. Wang, W. Wang, T. Zhao, Y. Cui, P. Liu, S. Xu and X. Luo, Anal. Chem., 2021, 93, 2480–2489.
83. I. G. Theodorou, Z. A. R. Jawad, Q. Jiang, E. O. Aboagye, A. E. Porter, M. P. Ryan and F. Xie, Chem. Mater., 2017, 29, 6916–6926.
84. H. Ren, J.-L. Yang, W.-M. Yang, H.-L. Zhong, J.-S. Lin, P. M. Radjenovic, L. Sun, H. Zhang, J. Xu, Z.-Q. Tian and J.-F. Li, ACS Mater. Lett., 2020, 3, 69–76.
85. M. Yang, P. Moroz, Z. Jin, D. S. Budkina, N. Sundrani, D. Porotnikov, J. Cassidy, Y. Sugiyama, A. N. Tarnovsky, H. Mattoussi and M. Zamkov, J. Am. Chem. Soc., 2019, 141, 11286–11297.
86. T. Itoh, K.-i. Yoshida, H. Tamaru, V. Biju and M. Ishikawa, J. Photochem. Photobiol., A, 2011, 219, 167–179.
87. M. Li, S. K. Cushing, J. Zhang, S. Suri, R. Evans, W. P. Petros, L. F. Gibson, D. Ma, Y. Liu and N. Wu, ACS Nano, 2013, 7, 4967–4976.
88. D. M. Rissin, C. W. Kan, T. G. Campbell, S. C. Howes, D. R. Fournier, L. Song, T. Piech, P. P. Patel, L. Chang, A. J. Rivnak, E. P. Ferrell, J. D. Randall, G. K. Provuncher, D. R. Walt and D. C. Duffy, Nat. Biotechnol., 2010, 28, 595–599.
89. X. Fu, L. Chen and J. Choo, Anal. Chem., 2017, 89, 124–137.
90. Z. Yang, H. Liu, Y. Tian, Y. Chen, Z. Niu, C. Zhou, F. Wang, C. Gu, S. Tang, T. Jiang and J. Zhou, J. Mater. Chem. C, 2020, 8, 2142–2154.
91. T. Jiang, X. Wang, J. Zhou and H. Jin, Sens. Actuators, B, 2018, 258, 105–114.
92. M. Lee, H. Kim, E. Kim, S. Y. Yi, S. G. Hwang, S. Yang, E. K. Lim, B. Kim, J. Jung and T. Kang, ACS Appl. Mater. Interfaces, 2018, 10, 37829–37834.
93. Z. Farka, M. J. Mickert, A. Hlavacek, P. Skladal and H. H. Gorris, Anal. Chem., 2017, 89, 11825–11830.
94. Z. Cheng, N. Choi, R. Wang, S. Lee, K. C. Moon, S. Y. Yoon, L. Chen and J. Choo, ACS Nano, 2017, 11, 4926–4933.
95. Z. Chen, S. M. Tabakman, A. P. Goodwin, M. G. Kattah, D. Daranciang, X. Wang, G. Zhang, X. Li, Z. Liu, P. J. Utz, K. Jiang, S. Fan and H. Dai, Nat. Biotechnol., 2008, 26, 1285–1292.
96. I. Al-Ogaidi, H. Gou, Z. P. Aguilar, S. Guo, A. K. Melconian, A. K. Al-Kazaz, F. Meng and N. Wu, Chem. Commun., 2014, 50, 1344–1346.
97. Y. Luo, X. Wang, D. Du and Y. Lin, Biomater. Sci., 2015, 3, 1386–1394.
98. M. Uchida, H. Kosuge, M. Terashima, D. A. Willits, L. O. Liepold, M. J. Young, M. V. McConnell and T. Douglas, ACS Nano, 2011, 5, 2493–2502.
99. H. Qian, Y. Huang, X. Duan, X. Wei, Y. Fan, D. Gan, S. Yue, W. Cheng and T. Chen, Biosens. Bioelectron., 2019, 140, 111350.
100. S.-E. Kim, K.-Y. Ahn, J.-S. Park, K. R. Kim, K. E. Lee, S.-S. Han and J. Lee, Anal. Chem., 2011, 83, 5834–5843.
101. F. Degorce, A. Card, S. Soh, E. Trinquet, G. P. Knapik and B. Xie, Curr. Chem. Genomics, 2009, 3, 22–32.
102. L. Norskov-Lauritsen, A. R. Thomsen and H. Brauner-Osborne, Int. J. Mol. Sci., 2014, 15, 2554–2572.
103. S. Hepojoki, J. Rusanen, J. Hepojoki, V. Nurmi, A. Vaheri, A. Lundkvist, K. Hedman and O. Vapalahti, J. Clin. Microbiol., 2015, 53, 2292–2297.
104. K. Shao, L. Wang, Y. Wen, T. Wang, Y. Teng, Z. Shen and Z. Pan, Anal. Chim. Acta, 2019, 1068, 52–59.
105. H. Xing, T. Wei, X. Lin and Z. Dai, Anal. Chim. Acta, 2018, 1042, 71–78.
106. L. Zhao, M. Cheng, G. Liu, H. Lu, Y. Gao, X. Yan, F. Liu, P. Sun and G. Lu, Sens. Actuators, B, 2018, 273, 185–190.
107. J. Yang, Z. Zhang and G. Yan, Sens. Actuators, B, 2018, 255, 2339–2346.
108. E. Pyrak, J. Krajczewski, A. Kowalik, A. Kudelski and A. Jaworska, Molecules, 2019, 24, 4423.
109. B. C. Strachan, H. S. Sloane, J. C. Lee, D. C. Leslie and J. P. Landers, Analyst, 2015, 140, 2008–2015.
110. L. Peng, J. Zhou, L. Yin, S. Man and L. Ma, Anal. Chim. Acta, 2020, 1125, 162–168.
111. X. Liu, F. Wang, R. Aizen, O. Yehezkeli and I. Willner, J. Am. Chem. Soc., 2013, 135, 11832–11839.
112. M. Li, S. K. Cushing, H. Liang, S. Suri, D. Ma and N. Wu, Anal. Chem., 2013, 85, 2072–2078.
113. S. Wang, M. A. Lifson, F. Inci, L. G. Liang, Y. F. Sheng and U. Demirci, Expert Rev. Mol. Diagn., 2016, 16, 449–459.
114. L. Huang, L. Ding, J. Zhou, S. Chen, F. Chen, C. Zhao, J. Xu, W. Hu, J. Ji, H. Xu and G. L. Liu, Biosens. Bioelectron., 2021, 171, 112685.
115. Y. Sun and T. Li, Anal. Chem., 2018, 90, 11614–11621.
116. Z. Liang, J. Zhou, L. Petti, L. Shao, T. Jiang, Y. Qing, S. Xie, G. Wu and P. Mormile, Analyst, 2019, 144, 1741–1750.
117. Y. P. Wong, S. Othman, Y. L. Lau, S. Radu and H. Y. Chee, J. Appl. Microbiol., 2018, 124, 626–643.
118. L. Garibyan and N. Avashia, J. Invest. Dermatol. Symp. Proc., 2013, 133, 1–4.
119. J. Li, J. Macdonald and F. von Stetten, Analyst, 2018, 144, 31–67.
120. C. Shi, Q. Liu, C. Ma and W. Zhong, Anal. Chem., 2014, 86, 336–339.
121. Y. Zhou, B. Li, M. Wang, J. Wang, H. Yin and S. Ai, Mikrochim. Acta, 2017, 184, 4359–4365.
122. M. Fakruddin, K. S. Mannan, A. Chowdhury, R. M. Mazumdar, M. N. Hossain, S. Islam and M. A. Chowdhury, J. Pharm. BioAllied Sci., 2013, 5, 245–252.
123. C. O. Nzelu, H. Kato and N. C. Peters, PLoS Neglected Trop. Dis., 2019, 13, e0007698.
124. O. Piepenburg, C. H. Williams, D. L. Stemple and N. A. Armes, PLoS Biol., 2006, 4, e204.
125. A. S. James and J. I. Alawneh, Diagnostics, 2020, 10, 399.
126. N. Bidar, F. Oroojalian, B. Baradaran, S. Eyvazi, M. Amini, A. Jebelli, S. S. Hosseini, P. Pashazadeh-Panahi, A. Mokhtarzadeh and M. de la Guardia, TrAC, Trend Anal. Chem., 2020, 131, 116021.
127. J. S. Chen, E. Ma, L. B. Harrington, M. Da Costa, X. Tian, J. M. Palefsky and J. A. Doudna, Science, 2018, 360, 436–439.
128. I. Khalil, W. A. Yehye, N. M. Julkapli, S. Rahmati, A. A. Sina, W. J. Basirun and M. R. Johan, Biosens. Bioelectron., 2019, 131, 214–223.
129. B. Guven, I. H. Boyaci, U. Tamer, E. Acar-Soykut and U. Dogan, Talanta, 2015, 136, 68–74.
130. L. Ge, X. Sun, Q. Hong and F. Li, ACS Appl. Mater. Interfaces, 2017, 9, 32089–32096.
131. J. Zhang, Y. Yang, X. Jiang, C. Dong, C. Song, C. Han and L. Wang, Biosens. Bioelectron., 2019, 141, 111402.
132. S.-R. Ryoo, J. Lee, J. Yeo, H.-K. Na, Y.-K. Kim, H. Jang, J. H. Lee, S. W. Han, Y. Lee, V. N. Kim and D.-H. Min, ACS Nano, 2013, 7, 5882–5891.
133. C. Song, Y. Liu, X. Jiang, J. Zhang, C. Dong, J. Li and L. Wang, Talanta, 2019, 205, 120137.
134. J. Huang, X. Su and Z. Li, Anal. Chem., 2012, 84, 5939–5943.
135. K. Guk, S. G. Hwang, J. Lim, H. Y. Son, Y. Choi, Y. M. Huh, T. Kang, J. Jung and E. K. Lim, Chem. Commun., 2019, 55, 3457–3460.
136. O. Adegoke, T. Kato and E. Y. Park, Biosens. Bioelectron., 2016, 80, 483–490.
137. M. M. Harper, B. Robertson, A. Ricketts and K. Faulds, Chem. Commun., 2012, 48, 9412–9414.
138. J. H. Choi, J. Lim, M. Shin, S. H. Paek and J. W. Choi, Nano Lett., 2021, 21, 693–699.
139. Y. Liu, A. Kannegulla, B. Wu and L.-J. Cheng, ACS Appl. Mater. Interfaces, 2018, 10, 18524–18531.
140. J. A. Dougan, D. MacRae, D. Graham and K. Faulds, Chem. Commun., 2011, 47, 4649.
141. M. S. Draz and X. Lu, Theranostics, 2016, 6, 522–532.
142. G. A. Obande and K. K. Banga Singh, Infect. Drug Resist., 2020, 13, 455–483.
143. X. Zhao, S. Li, G. Liu, Z. Wang, Z. Yang, Q. Zhang, M. Liang, J. Liu, Z. Li, Y. Tong, G. Zhu, X. Wang, L. Jiang, W. Wang, G.-Y. Tan and L. Zhang, Sci. Bull., 2020, 66, 69–77.
144. Y. Liu, S. Li, L. Zhang, Q. Zhao, N. Li and Y. Wu, RSC Adv., 2020, 10, 28037–28040.
145. M. Yu, Y. Yao, B. Cui, C. Sun, X. Zhao, Y. Wang, G. Liu, H. Cui and Z. Zeng, ACS Appl. Nano Mater., 2018, 2, 48–57.
146. W. Tan, L. Zhang, J. C. G. Doery and W. Shen, Lab Chip, 2020, 20, 394–404.
147. Y. Zhou and J. Yoon, Chem. Soc. Rev., 2012, 41, 52–67.
148. R. Feng, Y. Xu, H. Zhao, X. Duan and S. Sun, Analyst, 2016, 141, 3219–3223.
149. I. Al-Ogaidi, H. Gou, A. K. Al-Kazaz, Z. P. Aguilar, A. K. Melconian, P. Zheng and N. Wu, Anal. Chim. Acta, 2014, 811, 76–80.
150. W. Tan, L. Zhang, J. C. G. Doery and W. Shen, Sens. Actuators, B, 2020, 305, 127448.
151. J. Yin, Y. Kwon, D. Kim, D. Lee, G. Kim, Y. Hu, J. H. Ryu and J. Yoon, J. Am. Chem. Soc., 2014, 136, 5351–5358.
152. R. M. Wood, J. K. Rilling, A. G. Sanfey, Z. Bhagwagar and R. D. Rogers, Neuropsychopharmacology, 2006, 31, 1075–1084.
153. M. Rondanelli, A. Opizzi, N. Antoniello, F. Boschi, P. Iadarola, E. Pasini, R. Aquilani and F. S. Dioguardi, Clin. Nutr., 2011, 30, 571–577.
154. F. Xiao, J. Yu, Y. Guo, J. Deng, K. Li, Y. Du, S. Chen, J. Zhu, H. Sheng and F. Guo, Metabolism, 2014, 63, 841–850.
155. C. Martin and Y. Zhang, Nat. Rev. Mol. Cell Biol., 2005, 6, 838–849.
156. J. M. Wood, H. Decker, H. Hartmann, B. Chavan, H. Rokos, J. D. Spencer, S. Hasse, M. J. Thornton, M. Shalbaf, R. Paus and K. U. Schallreuter, FASEB J., 2009, 23, 2065–2075.
157. I. C. G. Weaver, F. A. Champagne, S. E. Brown, S. Dymov, S. Sharma, M. J. Meaney and M. Szyf, J. Neurosci., 2005, 25, 11045–11054.
158. G. Mkrtchyan, V. Aleshin, Y. Parkhomenko, T. Kaehne, M. Luigi Di Salvo, A. Parroni, R. Contestabile, A. Vovk, L. Bettendorff and V. Bunik, Sci. Rep., 2015, 5, 12583.
159. N. Naderi and J. D. House, Advances in Food and Nutrition Research, Academic Press, 2018, vol. 83, pp. 195–213.
160. M. Visentin, N. Diop-Bove, R. Zhao and I. D. Goldman, Annu. Rev. Physiol., 2014, 76, 251–274.
161. Institute of Medicine, Dietary Reference Intakes for Vitamin C, Vitamin E, Selenium, and Carotenoids, The National Academies Press, 2000.
162. R. Brigelius-Flohé and M. G. Traber, FASEB J., 1999, 13, 1145–1155.
163. U. Nations, World Drug Report 2017, United Nations, 2017 DOI:10.18356/bdc264f4-en.
164. E. De Rycke, C. Stove, P. Dubruel, S. De Saeger and N. Beloglazova, Biosens. Bioelectron., 2020, 169, 112579.
165. D. Li, Y. Ma, H. Duan, W. Deng and D. Li, Biosens. Bioelectron., 2018, 99, 389–398.
166. Q. Zhou, G. Meng, J. Liu, Z. Huang, F. Han, C. Zhu, D.-J. Kim, T. Kim and N. Wu, Adv. Mater. Technol., 2017, 2, 1700028.
167. X. Hu, P. Zheng, G. Meng, Q. Huang, C. Zhu, F. Han, Z. Huang, Z. Li, Z. Wang and N. Wu, Nanotechnology, 2016, 27, 384001.
168. W. Xue, X. Tan, M. K. Khaing Oo, G. Kulkarni, M. A. Ilgen and X. Fan, Analyst, 2020, 145, 1346–1354.
169. W. Yu, C. Jiang, B. Xie, S. Wang, X. Yu, K. Wen, J. Lin, J. Wang, Z. Wang and J. Shen, Anal. Chim. Acta, 2020, 1102, 91–98.
170. X. Hua, Y. Ding, J. Yang, M. Ma, H. Shi and M. Wang, Sci. Total Environ., 2017, 583, 222–227.
171. M. Li, M. Ma, X. Hua, H. Shi, Q. Wang and M. Wang, RSC Adv., 2015, 5, 3039–3044.
172. E. Guler, G. Bozokalfa, B. Demir, Z. P. Gumus, B. Guler, E. Aldemir, S. Timur and H. Coskunol, Drug Test. Anal., 2017, 9, 578–587.
173. S. R. Ahmed, R. Chand, S. Kumar, N. Mittal, S. Srinivasan and A. R. Rajabzadeh, TrAC, Trend Anal. Chem., 2020, 131, 116006.
174. L. Wu, G. Li, X. Xu, L. Zhu, R. Huang and X. Chen, TrAC, Trend Anal. Chem., 2019, 113, 140–156.
175. J. Nebu, J. S. Anjali Devi, R. S. Aparna, B. Aswathy, A. O. Aswathy and G. Sony, Mikrochim. Acta, 2018, 185, 532.
176. Z. Huang, G. Meng, Q. Huang, Y. Yang, C. Zhu and C. Tang, Adv. Mater., 2010, 22, 4136–4139.
177. C. Qian, Q. Guo, M. Xu, Y. Yuan and J. Yao, RSC Adv., 2015, 5, 53306–53312.
178. C. Zhu, G. Meng, P. Zheng, Q. Huang, Z. Li, X. Hu, X. Wang, Z. Huang, F. Li and N. Wu, Adv. Mater., 2016, 28, 4871–4876.
179. Z. Guo, S. Park, J. Yoon and I. Shin, Chem. Soc. Rev., 2014, 43, 16–29.
180. R. A. Alvarez-Puebla and L. M. Liz-Marzan, Angew. Chem., Int. Ed., 2012, 51, 11214–11223.
181. G. Song, D. Jiang, L. Wang, J. Ning, X. Sun, F. Su, M. Chen and Y. Tian, Chem. Commun., 2020, 56, 5405–5408.
182. C. Li, D. Yao, X. Jiang, A. Liang and Z. Jiang, J. Mater. Chem. C, 2020, 8, 11088–11101.
183. M. Li, Q. Wang, X. Shi, L. A. Hornak and N. Wu, Anal. Chem., 2011, 83, 7061–7065.
184. P. Chen, X. Jiang, K. Huang, P. Hu, X. Li, L. Wei, W. Liu, L. Wei, C. Tao, B. Ying, X. Wei and J. Geng, ACS Appl. Mater. Interfaces, 2019, 11, 36476–36484.
185. D. Li, Y. Ma, H. Duan, F. Jiang, W. Deng and X. Ren, Anal. Chim. Acta, 2018, 1038, 148–156.
186. X. Chen, Y. Zhou, X. Peng and J. Yoon, Chem. Soc. Rev., 2010, 39, 2120–2135.
187. J. Cao, C. Zhao, X. Wang, Y. Zhang and W. Zhu, Chem. Commun., 2012, 48, 9897–9899.
188. P. Zheng, X. Shi, K. Curtin, F. Yang and N. Wu, Mater. Res. Express, 2017, 4, 055017.
189. M. Li, X. Zhou, S. Guo and N. Wu, Biosens. Bioelectron., 2013, 43, 69–74.
190. Z. Yang, K. Y. Loh, Y.-T. Chu, R. Feng, N. S. R. Satyavolu, M. Xiong, S. M. Nakamata Huynh, K. Hwang, L. Li, H. Xing, X. Zhang, Y. R. Chemla, M. Gruebele and Y. Lu, J. Am. Chem. Soc., 2018, 140, 17656–17665.
191. R. J. Lake, Z. Yang, J. Zhang and Y. Lu, Acc. Chem. Res., 2019, 52, 3275–3286.
192. Y. Lin, Z. Yang, R. J. Lake, C. Zheng and Y. Lu, Angew. Chem., Int. Ed., 2019, 58, 17061–17067.
193. Y. Wang and J. Irudayaraj, Chem. Commun., 2011, 47, 4394–4396.
194. P. K. Drain, E. P. Hyle, F. Noubary, K. A. Freedberg, D. Wilson, W. R. Bishai, W. Rodriguez and I. V. Bassett, Lancet Infect. Dis., 2014, 14, 239–249.
195. H. Kettler, K. White and S. Hawkes, Mapping the landscape of diagnostics for sexually transmitted infections, World Health Organization, Geneva, 2004.
196. F. Yang, Y. Zhang, X. Cui, Y. Fan, Y. Xue, H. Miao and G. Li, Biotechnol. J., 2019, 14, 1800181.
197. R. Gorkin, J. Park, J. Siegrist, M. Amasia, B. S. Lee, J. M. Park, J. Kim, H. Kim, M. Madou and Y. K. Cho, Lab Chip, 2010, 10, 1758–1773.
198. P. N. Nge, C. I. Rogers and A. T. Woolley, Chem. Rev., 2013, 113, 2550–2583.
199. J. Park, D. H. Han and J.-K. Park, Lab Chip, 2020, 20, 1191–1203.
200. P. Thurgood, J. Y. Zhu, N. Nguyen, S. Nahavandi, A. R. Jex, E. Pirogova, S. Baratchi and K. Khoshmanesh, Lab Chip, 2018, 18, 2730–2740.
201. J. Seo, C. Wang, S. Chang, J. Park and W. Kim, Lab Chip, 2019, 19, 1790–1796.
202. B. Kim, S. Oh, D. You and S. Choi, Anal. Chem., 2017, 89, 1439–1444.
203. L. Gervais, N. de Rooij and E. Delamarche, Adv. Mater., 2011, 23, H151–H176.
204. J. Han, A. Qi, J. Zhou, G. Wang, B. Li and L. Chen, ACS Sens., 2018, 3, 1789–1794.
205. M. T. Guler, P. Beyazkilic and C. Elbuken, Sens. Actuators, A, 2017, 265, 224–230.
206. Y.-D. Ma, Y.-S. Chen and G.-B. Lee, Sens. Actuators, B, 2019, 296, 126647.
207. J. Park and J.-K. Park, Lab Chip, 2019, 19, 2973–2977.
208. M. Díaz-González, G. Boix, C. Fernández-Sánchez and A. Baldi, SPIE, 2017.
209. B. Jo, L. M. V. Lerberghe, K. M. Motsegood and D. J. Beebe, J. Microelectromech. Syst., 2000, 9, 76–81.
210. A. O. Olanrewaju, A. Robillard, M. Dagher and D. Juncker, Lab Chip, 2016, 16, 3804–3814.
211. M. M. Aeinehvand, L. Weber, M. Jiménez, A. Palermo, M. Bauer, F. F. Loeffler, F. Ibrahim, F. Breitling, J. Korvink, M. Madou, D. Mager and S. O. Martínez-Chapa, Lab Chip, 2019, 19, 1090–1100.
212. A. Enders, I. G. Siller, K. Urmann, M. R. Hoffmann and J. Bahnemann, Small, 2019, 15, e1804326.
213. W. Luo, K. Yarn, M. Shih, Y. Wu, K. Chang and M. Weng, 3rd IEEE International Conference on NEMS, 2008, pp. 801–806.
214. Q. Xiong, C. Y. Lim, J. Ren, J. Zhou, K. Pu, M. B. Chan-Park, H. Mao, Y. C. Lam and H. Duan, Nat. Commun., 2018, 9, 1743.
215. M. R. Rasouli and M. Tabrizian, Lab Chip, 2019, 19, 3316–3325.
216. A. T. Bender, M. D. Borysiak, A. M. Levenson, L. Lillis, D. S. Boyle and J. D. Posner, Anal. Chem., 2018, 90, 7221–7229.
217. N. Xiang, Q. Li and Z. Ni, Anal. Chem., 2020, 92, 6770–6776.
218. R. Gao, Z. Cheng, A. J. deMello and J. Choo, Lab Chip, 2016, 16, 1022–1029.
219. C. Szydzik, K. Khoshmanesh, A. Mitchell and C. Karnutsch, Biomicrofluidics, 2015, 9, 064120.
220. M. A. Ali, K. Mondal, Y. Jiao, S. Oren, Z. Xu, A. Sharma and L. Dong, ACS Appl. Mater. Interfaces, 2016, 8, 20570–20582.
221. R. Gao, Z. Lv, Y. Mao, L. Yu, X. Bi, S. Xu, J. Cui and Y. Wu, ACS Sens., 2019, 4, 938–943.
222. M. Sharafeldin and J. J. Davis, Anal. Chem., 2021, 93, 184–197.
223. S. Tripathi, Y. V. B. Varun Kumar, A. Prabhakar, S. S. Joshi and A. Agrawal, J. Micromech. Microeng., 2015, 25, 084004.
224. M. S. Maria, T. S. Chandra and A. K. Sen, Microfluid. Nanofluid., 2017, 21, 72.
225. S. Tripathi, Y. V. B. Varun Kumar, A. Prabhakar, S. S. Joshi and A. Agrawal, J. Micromech. Microeng., 2015, 25, 083001.
226. S. Haeberle, T. Brenner, R. Zengerle and J. Ducrée, Lab Chip, 2006, 6, 776–781.
227. S. C. Liu, P. B. Yoo, N. Garg, A. P. Lee and S. Rasheed, Sens. Actuators, A, 2021, 317, 112482.
228. M. Mohammadi, H. Madadi, J. Casals-Terre and J. Sellares, Anal. Bioanal. Chem., 2015, 407, 4733–4744.
229. H. M. Ji, V. Samper, Y. Chen, C. K. Heng, T. M. Lim and L. Yobas, Biomed. Microdevices, 2008, 10, 251–257.
230. X. Yang, O. Forouzan, T. P. Brown and S. S. Shevkoplyas, Lab Chip, 2012, 12, 274–280.
231. H. Shimizu, M. Kumagai, E. Mori, K. Mawatari and T. Kitamori, Anal. Methods, 2016, 8, 7597–7602.
232. K. Svanes and B. W. Zweifach, Microvasc. Res., 1968, 1, 210–220.
233. Y.-C. Fung, Microvasc. Res., 1973, 5, 34–48.
234. S. Yang, A. Undar and J. D. Zahn, Lab Chip, 2006, 6, 871–880.
235. R. Fan, O. Vermesh, A. Srivastava, B. K. Yen, L. Qin, H. Ahmad, G. A. Kwong, C. C. Liu, J. Gould, L. Hood and J. R. Heath, Nat. Biotechnol., 2008, 26, 1373–1378.
236. J. Wang, H. Ahmad, C. Ma, Q. Shi, O. Vermesh, U. Vermesh and J. Heath, Lab Chip, 2010, 10, 3157–3162.
237. M. S. Maria, P. E. Rakesh, T. S. Chandra and A. K. Sen, Sci. Rep., 2017, 7, 43457.
238. C. Liu, M. Mauk, R. Gross, F. D. Bushman, P. H. Edelstein, R. G. Collman and H. H. Bau, Anal. Chem., 2013, 85, 10463–10470.
239. G. Luka, A. Ahmadi, H. Najjaran, E. Alocilja, M. DeRosa, K. Wolthers, A. Malki, H. Aziz, A. Althani and M. Hoorfar, Sensors, 2015, 15, 30011–30031.
240. K. Choi, A. H. Ng, R. Fobel and A. R. Wheeler, Annu. Rev. Anal. Chem., 2012, 5, 413–440.
241. B. Coelho, B. Veigas, H. Águas, E. Fortunato, R. Martins, P. Baptista and R. Igreja, Sensors, 2017, 17, 2616.
242. F. Mugele and J.-C. Baret, J. Phys.: Condens. Matter, 2005, 17, R705–R774.
243. A. M. Foudeh, D. Brassard, M. Tabrizian and T. Veres, Lab Chip, 2015, 15, 1609–1618.
244. M. S. Lee, W. Hsu, H. Y. Huang, H. Y. Tseng, C. T. Lee, C. Y. Hsu, Y. C. Shieh, S. H. Wang, D. J. Yao and C. H. Liu, Biosens. Bioelectron., 2020, 150, 111851.
245. V. Srinivasan, V. K. Pamula and R. B. Fair, Lab Chip, 2004, 4, 310–315.
246. F. Stumpf, F. Schwemmer, T. Hutzenlaub, D. Baumann, O. Strohmeier, G. Dingemanns, G. Simons, C. Sager, L. Plobner, F. von Stetten, R. Zengerle and D. Mark, Lab Chip, 2016, 16, 199–207.
247. S. H. Yazdi, K. L. Giles and I. M. White, Anal. Chem., 2013, 85, 10605–10611.
248. J. Qi, J. Zeng, F. Zhao, S. H. Lin, B. Raja, U. Strych, R. C. Willson and W. C. Shih, Nanoscale, 2014, 6, 8521–8526.
249. A. Kaminska, E. Witkowska, K. Winkler, I. Dziecielewski, J. L. Weyher and J. Waluk, Biosens. Bioelectron., 2015, 66, 461–467.
250. C. H. Sang, S. J. Chou, F. M. Pan and J. T. Sheu, Biosens. Bioelectron., 2016, 75, 285–292.
251. C. M. Chang, W. H. Chang, C. H. Wang, J. H. Wang, J. D. Mai and G. B. Lee, Lab Chip, 2013, 13, 1225–1242.
252. P. J. Sykes, S. H. Neoh, M. J. Brisco, E. Hughes, J. Condon and A. A. Morley, Biotechniques, 1992, 13, 444–449.
253. F. Shen, W. Du, J. E. Kreutz, A. Fok and R. F. Ismagilov, Lab Chip, 2010, 10, 2666–2672.
254. M. Safavieh, M. K. Kanakasabapathy, F. Tarlan, M. U. Ahmed, M. Zourob, W. Asghar and H. Shafiee, ACS Biomater. Sci. Eng., 2016, 2, 278–294.
255. H. Yuan, Y. Chao and H. C. Shum, Small, 2020, 16, 1904469.
256. Y. Chen, Y. Mei, X. Zhao and X. Jiang, Anal. Chem., 2020, 92, 14846–14852.
257. J. Kim, M. Johnson, P. Hill and B. K. Gale, Integr. Biol., 2009, 1, 574–586.
258. Z. Pang, A. Al-Mahrouki, M. Berezovski and S. N. Krylov, Electrophoresis, 2006, 27, 1489–1494.
259. P. Sethu, M. Anahtar, L. L. Moldawer, R. G. Tompkins and M. Toner, Anal. Chem., 2004, 76, 6247–6253.
260. L. Nan, Z. Jiang and X. Wei, Lab Chip, 2014, 14, 1060–1073.
261. J. Kim, J. W. Hong, D. P. Kim, J. H. Shin and I. Park, Lab Chip, 2012, 12, 2914–2921.
262. Y. Shin, A. P. Perera, C. C. Wong and M. K. Park, Lab Chip, 2014, 14, 359–368.
263. Y. K. Cho, J. G. Lee, J. M. Park, B. S. Lee, Y. Lee and C. Ko, Lab Chip, 2007, 7, 565–573.
264. W.-P. Hu, Y.-C. Chen and W.-Y. Chen, Sci. Rep., 2020, 10, 21132.
265. J. Yin, Y. Suo, Z. Zou, J. Sun, S. Zhang, B. Wang, Y. Xu, D. Darland, J. X. Zhao and Y. Mu, Lab Chip, 2019, 19, 2769–2785.
266. S. Petralia, E. L. Sciuto and S. Conoci, Analyst, 2017, 142, 140–146.
267. D. Chen, M. Mauk, X. Qiu, C. Liu, J. Kim, S. Ramprasad, S. Ongagna, W. R. Abrams, D. Malamud, P. L. Corstjens and H. H. Bau, Biomed. Microdevices, 2010, 12, 705–719.
268. C. Zhu, A. Hu, J. Cui, K. Yang, X. Zhu, Y. Liu, G. Deng and L. Zhu, Micromachines, 2019, 10, 537.
269. M. Solsona, E. Y. Westerbeek, J. G. Bomer, W. Olthuis and A. van den Berg, Lab Chip, 2019, 19, 1054–1059.
270. C. Liu, J. Sun, J. Li, C. Xiang, L. Che, Z. Wang and X. Zhou, Sci. Rep., 2017, 7, 7552.
271. A. Ramachandran, D. A. Huyke, E. Sharma, M. K. Sahoo, C. Huang, N. Banaei, B. A. Pinsky and J. G. Santiago, Proc. Natl. Acad. Sci. U. S. A., 2020, 117, 29518–29525.
272. V. Betancur, J. Sun, N. Wu and Y. Liu, Micromachines, 2017, 8, 376.
273. M. Zimmermann, H. Schmid, P. Hunziker and E. Delamarche, Lab Chip, 2007, 7, 119–125.
274. E.-C. Yeh, C.-C. Fu, L. Hu, R. Thakur, J. Feng and L. P. Lee, Sci. Adv., 2017, 3, e1501645.
275. W. C. Mak, V. Beni and A. P. F. Turner, TrAC, Trends Anal. Chem., 2016, 79, 297–305.
276. L. Gervais and E. Delamarche, Lab Chip, 2009, 9, 3330.
277. S. Ghosh and C. H. Ahn, Analyst, 2019, 144, 2109–2119.
278. L. Xu, H. Lee, D. Jetta and K. W. Oh, Lab Chip, 2015, 15, 3962–3979.
279. L. Qin, O. Vermesh, Q. Shi and J. R. Heath, Lab Chip, 2009, 9, 2016–2020.
280. L. Zhu, N. Kroodsma, J. Yeom, J. L. Haan, M. A. Shannon and D. D. Meng, Microfluid. Nanofluid., 2011, 11, 569–578.
281. L. Gervais, M. Hitzbleck and E. Delamarche, Biosens. Bioelectron., 2011, 27, 64–70.
282. L. Xu, A. Wang, X. Li and K. W. Oh, Biomicrofluidics, 2020, 14, 031503.
283. J. M. Morrissette, P. S. Mahapatra, A. Ghosh, R. Ganguly and C. M. Megaridis, Sci. Rep., 2017, 7, 1800.
284. S. Shin, B. Kim, Y. J. Kim and S. Choi, Biosens. Bioelectron., 2019, 133, 169–176.
285. K. K. Lee, M. O. Kim and S. Choi, J. Pharm. Biomed. Anal., 2019, 162, 28–33.
286. C. Parolo, A. Sena-Torralba, J. F. Bergua, E. Calucho, C. Fuentes-Chust, L. Hu, L. Rivas, R. Alvarez-Diduk, E. P. Nguyen, S. Cinti, D. Quesada-Gonzalez and A. Merkoci, Nat. Protoc., 2020, 15, 3788–3816.
287. J. Hu, S. Wang, L. Wang, F. Li, B. Pingguan-Murphy, T. J. Lu and F. Xu, Biosens. Bioelectron., 2014, 54, 585–597.
288. H. Wei, Y. Peng, Z. Bai, Z. Rong and S. Wang, Analyst, 2021, 146, 558–564.
289. S. Kasetsirikul, M. J. A. Shiddiky and N.-T. Nguyen, Microfluid. Nanofluid., 2020, 24, 17.
290. R. Wang, W. Zhang, P. Wang and X. Su, Mikrochim. Acta, 2018, 185, 191.
291. X. Gao, J. Boryczka, P. Zheng, S. Kasani, F. Yang, E. B. Engler-Chiurazzi, J. W. Simpkins, J. G. Wigginton and N. Wu, Biosens. Bioelectron., 2021, 177, 112967.
292. X. Gao, J. Boryczka, S. Kasani and N. Wu, Anal. Chem., 2021, 93(3), 1326–1332.
293. X. Yang, X. Liu, B. Gu, H. Liu, R. Xiao, C. Wang and S. Wang, Mikrochim. Acta, 2020, 187, 570.
294. X. Weng and S. Neethirajan, Mikrochim. Acta, 2017, 184, 4545–4552.
295. L. Anfossi, F. Di Nardo, S. Cavalera, C. Giovannoli, G. Spano, E. S. Speranskaya, I. Y. Goryacheva and C. Baggiani, Mikrochim. Acta, 2018, 185, 94.
296. X. Deng, C. Wang, Y. Gao, J. Li, W. Wen, X. Zhang and S. Wang, Biosens. Bioelectron., 2018, 105, 211–217.
297. J. Wang, F. Cao, S. He, Y. Xia, X. Liu, W. Jiang, Y. Yu, H. Zhang and W. Chen, Talanta, 2018, 176, 444–449.
298. Z. Lu, E. Rey, S. Vemulapati, B. Srinivasan, S. Mehta and D. Erickson, Lab Chip, 2018, 18, 3865–3871.
299. A. K. Yetisen, M. S. Akram and C. R. Lowe, Lab Chip, 2013, 13, 2210.
300. W. Guo, J. Hansson and W. van der Wijngaart, Anal. Chem., 2020, 92, 6194–6199.
301. H. Li and A. J. Steckl, Anal. Chem., 2019, 91, 352–371.
302. D. M. Cate, J. A. Adkins, J. Mettakoonpitak and C. S. Henry, Anal. Chem., 2015, 87, 19–41.
303. E. Fu, S. A. Ramsey, P. Kauffman, B. Lutz and P. Yager, Microfluid. Nanofluid., 2011, 10, 29–35.
304. A. Ainla, M. M. Hamedi, F. Güder and G. M. Whitesides, Adv. Mater., 2017, 29, 1702894.
305. J. R. Choi, Z. Liu, J. Hu, R. Tang, Y. Gong, S. Feng, H. Ren, T. Wen, H. Yang, Z. Qu, B. Pingguan-Murphy and F. Xu, Anal. Chem., 2016, 88, 6254–6264.
306. C. K. Koo, F. He and S. R. Nugen, Analyst, 2013, 138, 4998–5004.
307. J. R. Choi, J. Hu, R. Tang, Y. Gong, S. Feng, H. Ren, T. Wen, X. Li, W. A. Wan Abas, B. Pingguan-Murphy and F. Xu, Lab Chip, 2016, 16, 611–621.
308. J. T. Connelly, J. P. Rolland and G. M. Whitesides, Anal. Chem., 2015, 87, 7595–7601.
309. P. Chen, C. Chen, Y. Liu, W. Du, X. Feng and B.-F. Liu, Sens. Actuators, B, 2019, 283, 472–477.
310. K. Scida, B. Li, A. D. Ellington and R. M. Crooks, Anal. Chem., 2013, 85, 9713–9720.
311. Z. Yang, G. Xu, J. Reboud, S. A. Ali, G. Kaur, J. McGiven, N. Boby, P. K. Gupta, P. Chaudhuri and J. M. Cooper, ACS Sens., 2018, 3, 403–409.
312. H. Ahn, B. S. Batule, Y. Seok and M. G. Kim, Anal. Chem., 2018, 90, 10211–10216.
313. T. Komatsu, M. Maeki, A. Ishida, H. Tani and M. Tokeshi, ACS Sens., 2020, 5, 1287–1294.
314. B. A. Rohrman and R. R. Richards-Kortum, Lab Chip, 2012, 12, 3082–3088.
315. E. A. Phillips, T. J. Moehling, K. F. K. Ejendal, O. S. Hoilett, K. M. Byers, L. A. Basing, L. A. Jankowski, J. B. Bennett, L. K. Lin, L. A. Stanciu and J. C. Linnes, Lab Chip, 2019, 19, 3375–3386.
316. S. Ebnesajjad, Surface Treatment of Materials for Adhesive Bonding, William Andrew Publishing, Oxford, 2nd edn, 2014, pp. 39–75 DOI:10.1016/B978-0-323-26435-8.00004-6.
317. A. Di Gianfrancesco, Materials for Ultra-Supercritical and Advanced Ultra-Supercritical Power Plants, Woodhead Publishing, 2017, pp. 197–245 DOI:10.1016/B978-0-08-100552-1.00008-7.
318. A. S. Holik, Encyclopedia of Materials: Science and Technology, Elsevier, Oxford, 2001, pp. 6458–6463 DOI:10.1016/B0-08-043152-6/01142-6.
319. J. W. Lichtman and J.-A. Conchello, Nat. Methods, 2005, 2, 910–919.
320. B. Huang, M. Bates and X. Zhuang, Annu. Rev. Biochem., 2009, 78, 993–1016.
321. B. G. de Grooth, M. van Dam, N. C. Swart, A. Willemsen and J. Greve, Cytometry, 1987, 8, 445–452.
322. D. J. Webb and C. M. Brown, Cell Imaging Techniques: Methods and Protocols, Humana Press, Totowa, NJ, 2013, pp. 29–59 DOI:10.1007/978-1-62703-056-4_2.
323. S. W. Paddock, Mol. Biotechnol., 2000, 16, 127–149.
324. A. S. Maddox and P. S. Maddox, Methods in Cell Biology, Academic Press, 2012, vol. 107, pp. 1–34.
325. A. M. Larson, Nat. Photonics, 2011, 5, 1.
326. P. T. C. So, C. Y. Dong, B. R. Masters and K. M. Berland, Annu. Rev. Biomed. Eng., 2000, 2, 399–429.
327. L. Li, W. Guo, Y. Yan, S. Lee and T. Wang, Light: Sci. Appl., 2013, 2, e104.
328. C. G. Galbraith and J. A. Galbraith, J. Cell Sci., 2011, 124, 1607–1611.
329. G. Perinetti, T. Müller, A. Spaar, R. Polishchuk, A. Luini and A. Egner, Traffic, 2009, 10, 379–391.
330. J. M. Pullman, J. Nylk, E. C. Campbell, F. J. Gunn-Moore, M. B. Prystowsky and K. Dholakia, Biomed. Opt. Express, 2016, 7, 302–311.
331. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn and S. W. Hell, Nature, 2006, 440, 935–939.
332. S. K. Cool, K. Breyne, E. Meyer, S. C. De Smedt and N. N. Sanders, J. Fluoresc., 2013, 23, 909–920.
333. M. Lesser, High Performance Silicon Imaging, Woodhead Publishing, 2014, pp. 78–97 DOI:10.1533/9780857097521.1.78.
334. B. Q. Spring, A. O. Abu-Yousif, A. Palanisami, I. Rizvi, X. Zheng, Z. Mai, S. Anbil, R. B. Sears, L. B. Mensah, R. Goldschmidt, S. S. Erdem, E. Oliva and T. Hasan, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, E933–E942.
335. G. Oh, E. Chung and S. H. Yun, Opt. Fiber Technol., 2013, 19, 760–771.
336. C. C. Horgan, M. S. Bergholt, A. Nagelkerke, M. Z. Thin, I. J. Pence, U. Kauscher, T. L. Kalber, D. J. Stuckey and M. M. Stevens, Theranostics, 2021, 11, 2006–2019.
337. W. Wang, J. Zhao, M. Short and H. Zeng, J. Biophotonics, 2015, 8, 527–545.
338. S. H. Seo, B. M. Kim, A. Joe, H. W. Han, X. Chen, Z. Cheng and E. S. Jang, Biomaterials, 2014, 35, 3309–3318.
339. E. Cordero, I. Latka, C. Matthaus, I. Schie and J. Popp, J. Biomed. Opt., 2018, 23, 1–23.
340. X. Yi, F. Wang, W. Qin, X. Yang and J. Yuan, Int. J. Nanomed., 2014, 9, 1347–1365.
341. J.-B. Li, H.-W. Liu, T. Fu, R. Wang, X.-B. Zhang and W. Tan, Trends Chem., 2019, 1, 224–234.
342. L. Wang, W. Du, Z. Hu, K. Uvdal, L. Li and W. Huang, Angew. Chem., Int. Ed., 2019, 58, 14026–14043.
343. R. Weinstain, T. Slanina, D. Kand and P. Klan, Chem. Rev., 2020, 120, 13135–13272.
344. H. Wang, E. Zhao, J. W. Y. Lam and B. Z. Tang, Mater. Today, 2015, 18, 365–377.
345. W. Xu, D. Wang and B. Z. Tang, Angew. Chem., Int. Ed., 2021, 60, 7476–7487.
346. L.-J. Liu, W. Liu, G. Ji, Z.-Y. Wu, B. Xu, J. Qian and W.-J. Tian, Chin. J. Polym. Sci., 2019, 37, 401–408.
347. P. Zhao, K. He, Y. Han, Z. Zhang, M. Yu, H. Wang, Y. Huang, Z. Nie and S. Yao, Anal. Chem., 2015, 87, 9998–10005.
348. P. Zhao, Q. Xu, J. Tao, Z. Jin, Y. Pan, C. Yu and Z. Yu, Wiley Interdiscip. Rev.: Nanomed. Nanobiotechnol., 2018, 10, e1483.
349. S. Chinnathambi and N. Shirahata, Sci. Technol. Adv. Mater., 2019, 20, 337–355.
350. M. Montalti, A. Cantelli and G. Battistelli, Chem. Soc. Rev., 2015, 44, 4853–4921.
351. X. Bao, Y. Yuan, J. Chen, B. Zhang, D. Li, D. Zhou, P. Jing, G. Xu, Y. Wang, K. Hola, D. Shen, C. Wu, L. Song, C. Liu, R. Zboril and S. Qu, Light: Sci. Appl., 2018, 7, 91.
352. P. Zheng and N. Wu, Chem. – Asian J., 2017, 12, 2343–2353.
353. J. Shamsi, A. S. Urban, M. Imran, L. De Trizio and L. Manna, Chem. Rev., 2019, 119, 3296–3348.
354. L. Protesescu, S. Yakunin, M. I. Bodnarchuk, F. Krieg, R. Caputo, C. H. Hendon, R. X. Yang, A. Walsh and M. V. Kovalenko, Nano Lett., 2015, 15, 3692–3696.
355. X. Shu, A. Royant, M. Z. Lin, T. A. Aguilera, V. Lev-Ram, P. A. Steinbach and R. Y. Tsien, Science, 2009, 324, 804–807.
356. D. Yu, W. C. Gustafson, C. Han, C. Lafaye, M. Noirclerc-Savoye, W.-P. Ge, D. A. Thayer, H. Huang, T. B. Kornberg, A. Royant, L. Y. Jan, Y. N. Jan, W. A. Weiss and X. Shu, Nat. Commun., 2014, 5, 3626.
357. C. Li, L. Cao, Y. Zhang, P. Yi, M. Wang, B. Tan, Z. Deng, D. Wu and Q. Wang, Small, 2015, 11, 4517–4525.
358. L. Y. Huang, S. Zhu, R. Cui and M. Zhang, Anal. Chem., 2020, 92, 535–542.
359. D. E. Kim, D. Schellingerhout, F. A. Jaffer, R. Weissleder and C. H. Tung, J. Cereb. Blood Flow Metab., 2005, 25, 226–233.
360. M. Zhang, J. Yue, R. Cui, Z. Ma, H. Wan, F. Wang, S. Zhu, Y. Zhou, Y. Kuang, Y. Zhong, D.-W. Pang and H. Dai, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, 6590–6595.
361. J. Ando, K. Fujita, N. I. Smith and S. Kawata, Nano Lett., 2011, 11, 5344–5348.
362. S. Harmsen, R. Huang, M. A. Wall, H. Karabeber, J. M. Samii, M. Spaliviero, J. R. White, S. Monette, R. O’Connor, K. L. Pitter, S. A. Sastra, M. Saborowski, E. C. Holland, S. Singer, K. P. Olive, S. W. Lowe, R. G. Blasberg and M. F. Kircher, Nat. Biotechnol., 2015, 7, 271ra277.
363. J. Y. Huang, C. Zong, L. J. Xu, Y. Cui and B. Ren, Chem. Commun., 2011, 47, 5738–5740.
364. T. Wei, H. Xing, H. Wang, Y. Zhang, J. Wang, J. Shen and Z. Dai, Talanta, 2020, 210, 120625.
365. A. Huefner, W. L. Kuan, R. A. Barker and S. Mahajan, Nano Lett., 2013, 13, 2463–2470.
366. J. Chang, A. Zhang, Z. Huang, Y. Chen, Q. Zhang and D. Cui, Talanta, 2019, 198, 45–54.
367. D. Ghosh, A. F. Bagley, Y. J. Na, M. J. Birrer, S. N. Bhatia and A. M. Belcher, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, 13948–13953.
368. X. Qian, X.-H. Peng, D. O. Ansari, Q. Yin-Goen, G. Z. Chen, D. M. Shin, L. Yang, A. N. Young, M. D. Wang and S. Nie, Nat. Biotechnol., 2008, 26, 83–90.
369. J. Chen, Z. Sheng, P. Li, M. Wu, N. Zhang, X. F. Yu, Y. Wang, D. Hu, H. Zheng and G. P. Wang, Nanoscale, 2017, 9, 11888–11901.
370. A. Jaworska, L. E. Jamieson, K. Malek, C. J. Campbell, J. Choo, S. Chlopicki and M. Baranska, Analyst, 2015, 140, 2321–2329.
371. L. Puppulin, S. Hosogi, H. Sun, K. Matsuo, T. Inui, Y. Kumamoto, T. Suzaki, H. Tanaka and Y. Marunaka, Nat. Commun., 2018, 9, 5278.
372. J. Zhang, Z. Liu, P. Lian, J. Qian, X. Li, L. Wang, W. Fu, L. Chen, X. Wei and C. Li, Chem. Sci., 2016, 7, 5995–6005.
373. B. Tang, F. Yu, P. Li, L. Tong, X. Duan, T. Xie and X. Wang, J. Am. Chem. Soc., 2009, 131, 3016–3023.
374. B. Dong, X. Song, C. Wang, X. Kong, Y. Tang and W. Lin, Anal. Chem., 2016, 88, 4085–4091.
375. J. Song, J. Zhou and H. Duan, J. Am. Chem. Soc., 2012, 134, 13458–13469.
376. J. Yin, X. Kong and W. Lin, Anal. Chem., 2021, 93, 2072–2081.
377. Y. Zhou, S. Yu, J. Shang, Y. Chen, Q. Wang, X. Liu and F. Wang, Anal. Chem., 2020, 92, 15069–15078.
378. D. W. Hwang, H. Y. Kim, F. Li, J. Y. Park, D. Kim, J. H. Park, H. S. Han, J. W. Byun, Y.-S. Lee, J. M. Jeong, K. Char and D. S. Lee, Biomaterials, 2017, 121, 144–154.
379. W. Fu, C. Yan, Z. Guo, J. Zhang, H. Zhang, H. Tian and W.-H. Zhu, J. Am. Chem. Soc., 2019, 141, 3171–3177.
380. R. Yan, Y. Hu, F. Liu, S. Wei, D. Fang, A. J. Shuhendler, H. Liu, H.-Y. Chen and D. Ye, J. Am. Chem. Soc., 2019, 141, 10331–10341.
381. X. Gao and N. Wu, Electrochem. Soc. Interface, 2016, 25, 79–81.
382. Point of Care/Rapid Diagnostics Market, https://www.marketsandmarkets.com/Market-Reports/point-of-care-diagnostic-market-106829185.html.

---