Single cells are the fundamental components of organism, representing the complexity and factuality of life systems ^{[1][2][3][4][5]}[1,2,3,4,5]. Studying cells at individual level is intuitive and effective to investigate cell behavior and function. However, analysis of one true cell is difficult due to the ultra-small size of cells ^[6][7][6,7]. Indeed, limited by the development of instruments and equipment for a long time, the vast majority of cell studies have depended on the averages of cell ensembles based on the assumption of underlying normal distributions, which often mask the information available to identify the morphology and composition of individual cells [8]. With the in-depth study of cells, as well as many experimental examples, the heterogeneities between individual cells have been gradually unveiled ^[9][10][11][9,10,11]. Therefore, progress in cell analysis at the single-cell level can solve the problems arising from the heterogeneity of cells in physiological processes, such as metabolism, breeding mechanisms, signal transduction, etc. [12]. In the last couple of years, interest in exploiting platforms that can directly study single biological cells has increased dramatically. With the continuous innovation of analytical techniques and methods, researchers have put forward intelligent approaches to overcome the challenge of obtaining and measuring individual cells, which include cell isolation or sorting programs integrated with innovative analytical tools ^[13][14][15][13,14,15].

As microfabrication technologies have become more advanced and accessible in recent decades, microfluidic chips have become the most ideal devices for single-cell analysis. These miniaturized devices can precisely manipulate single cells of picoliter-scale volumes, enable the integration of necessary multistep single-cell handling processes, and shorten the operating time ^[16][17][18][16,17,18]. On microfluidic devices, single-cell manipulation processing (such as culturing, trapping, sorting, selecting, and isolating) and analysis (including chemical and biological content analysis, dynamic response, signal transduction pathways, etc.) can be achieved because the chip architecture can be designed flexibly, and accurate fluid control can be obtained through well-designed fluid characteristics ^[19][20][21][19,20,21]. Moreover, microfluidic devices can be easily implemented in combination with a variety of analytical devices for single-cell analysis with good sensitivity and reproducibility ^{[22][23][24][25]}[22,23,24,25]. As three of the most common detection methods, optical, electrochemical, and mass spectrometric techniques have been widely applied for cell analysis and integrated with microfluidic systems ^{[26][27][28][29]}[26,27,28,29]. Significantly, optical methods are hugely popular for single-cell analysis because of their ability to directly visualize the response process in cells using super-resolution microscopy, as well as to realize single-cell analysis in a label-free manner ^[30][31][30,31].

In wide-field fluorescence microscopy, a parallel beam produced by a single-wavelength mercury lamp or LED excites target fluorophores in the sample. Wide-field fluorescence imaging has been widely used for single-cell analysis due to its advantages including simple external connection equipment, convenient operation, and low price. Because single-cell signaling is a dynamic process that responds, in particular, to the complex extracellular microenvironment, it is challenging to simulate the complex extracellular microenvironment and capture single-cell dynamic signaling in real time. In recent years, efforts have been made by a group to probe dynamic signaling of single cells on a microfluidic chip system coupled with a CCD fluorescence imaging system. For example, a dynamic microfluidic cytometry method that combines the advantages of flow cytometry, microfluidic, and fluorescence microscopy was designed to realize cyclical cell trapping, stimulation, and automatic release ^[32][63]. On this integrated platform, G protein-coupled receptor signaling can be captured with high-throughput at single-cell resolution. The system consists of a microfluidic chip, two independent E/P transducers, a pressure control module, and a mercury lamp excited CCD fluorescence imaging module. The microfluidic chip includes cell culture channels, hydrodynamic gate channels, and concentration gradient channels, on which multiple parallel experiments can be conducted using different combinations of amplitudes and frequencies to probe adenosine triphosphate (ATP) and histamine (HA) response in single HeLa cells. Feng et al. ^[33][65] also investigated dynamic calcium signal transmission between a pair of single cells by designing an open microfluidic probe connected with a tumor microtube (TM) structure. The integrated system comprised an open microfluidic probe for in situ stimulation and a cell imaging platform for signal transmission observation; on this integrated platform, in situ regional stimulation on the single cell could be achieved without a complex cell capture process via the laminar flow effect in a stable flow region. As demonstrated, the signal transmission via TM structure, the heterogeneity of intercellular depolarization response, and the relationship between depolarization and opposite motion on tumor cells were explored in this assay. Because signaling pathways are related to cell growth, proliferation, and apoptosis, analysis of the cell signaling pathways is important in biochemical and cell biology. Lee et al. ^[34][66] studied the dynamics of yeast mitogen-activated protein kinase (MAPK) signaling pathways and their crosstalk using a microfluidic device coupled with quantitative microscopy. On this integrated platform, single yeast cells were trapped in a single focal plane, and the magnitudes of a given stress signal were modulated by microfluidic serial dilution while keeping other signaling inputs constant. With this device and single-cell analysis of microscopy images, concentration-dependent bimodal gene expression induced by osmotic stress was confirmed, and the crosstalk between the pheromone and cell wall integrity signaling pathways was quantified.

Compared with wide-field fluorescence microscopy, confocal laser scanning microscopy (CLSM) is inherently more complex and expensive. The greatest advantage of CLSM is the ability to obtain high-resolution fluorescence images, even revealing subcellular details, which is difficult using general optical microscopy. With confocal laser imaging, optical sectioning into the depth of a fluorescent sample or 3D tissue model can be achieved by collecting the light from a single focus plane. In recent years, CLSM coupled with microfluidic systems has been widely used for cell imaging. Pratt et al. ^[35][67] fabricated a drop-based microfluidic device coupled with CLSM to study the physiology of single microbial cells and to recognize the physiological differences of cells from a larger population with high throughput. Single cells were isolated into water-in-oil drops from a population with a 15 µm PDMS flow-focusing microfluidic drop-making device. The drops were arranged as a packed monolayer inside the PDMS microfluidic device, which enabled the maintenance of the drop position and volume during extended timelapse imaging. Using this strategy, the heterogeneity in growth rate and lag time of a large number of cells could be quantified. In addition, optofluidic scanning microscopy has been developed for live cell imaging with super-high resolution ^[36][68]. In this system, by taking advantage of the multi-focal excitation using the innate fluidic motion of the specimens, minimal instrumental complexity and full compatibility with various microfluidic configurations can be realized.

Fluorescence lifetime imaging (FLI), which is dependent of the intrinsic excited-state lifetime property of a fluorophore other than fluorescence intensity or local fluorophore concentration, has become an increasingly popular subject in the scientific research field in recent years. FLI microscopy can be implemented using wide-field microscopic or a laser scanning microscopic imaging modalities, and it has been actively explored for single-cell identification and kinetics analysis. Recently, Mathur et al. ^[37][69] used laser-excited FLI with open-space microfluidics to quantify spatially localized antibody binding kinetics on various biological substrates (such as cell/cells and tissues). The surface protein of the biological substrate has no associated lifetime, as it is not labeled. Accordingly, by monitoring the change in the component yield or amplitude in lifetime contribution, the proportion of the bound antibody can be tracked, and a kinetic curve can be rebuilt. In this way, p53 kinetics with differential biomarker expression in ovarian cancer tissue sections can be directly measured. Furthermore, using FLI microscopy, Lee et al. ^[38][70] developed a fast and label-free recognition platform for individual leukemia cells from blood in a high-density microfluidic trapping array. Two parts are contained in this single-cell leukemia recognition platform: a microfluidic chip that consists of hydrodynamic filters and a single-cell trap array and an FLI microscope coupled with a computer. Because the expression of the reduced form of nicotinamide adenine dinucleotide (NADH) in leukemia cells is different from that in normal white blood cells and different forms of NADH have unique lifetimes independent of their concentration, single leukemia cells can be indirectly recognized by quantification of variations between free/bound NADH through real-time monitoring of FLI microscopy images and a phasor algorithm approach.

Droplet technologies have long been popular for widespread applications on microfluidics platforms ^[39][71]. Droplet microfluidics can quickly and simply encapsulate single cells in monodisperse microcapsules by oil phase shear of the aqueous phase, offering a perfect microreactor for single-cell analysis. Therefore, this technique is especially favored for single-cell isolation and analysis in combination with various analysis techniques ^[40][72]. For example, the integration of droplet microfluidics and SERS provides a promising way to realize single-cell component analysis with high sensitivity and good reproducibility. In this integration, the aggregation time and the mixing degree of metallic nanoparticles can be controlled through a designable microfluidic channel structure and adjustable liquid flow rate. Using the SERS integrated microfluidic droplet method, Sun et al. ^[41][73] identified cell-to-cell heterogeneity by ultrasensitive analysis of single-cell alkaline phosphatase (ALP) activity. As a crucial disease biomarker, ALP is a zinc-doped glycoprotein that can dephosphorylate substrates containing phosphate groups. After collecting and incubating the droplets, BCI forms in the presence of ALP expression progress, and the droplets containing the BCI can be identified by detecting a strong SERS signal. In addition to ALP, telomerase is another important biomarker. Recently, a fluorescence dual-channel microfluidic droplet platform was integrated with SERS for in situ detection of telomerase activity with high sensitivity at single-cell resolution ^[42][74]. In this assay, a telomerase primer and Cy5-labeled complementary strand were modified onto AuNP surfaces. When the primer sequences were extended in the presence of the active telomerase, the Cy5-labeled strand was removed, resulting in turn-on fluorescence, as well as turn-off SERS, thus realizing the detection of telomerase activity in dual-channel mode. Using plasmonic imaging and a SERS dual-channel microfluidic droplet system, Cong et al. ^[43][75] detected a cancer biomarker, sialic acid (SA), in a single living cell level. In this system, a multi-functional plasmonic probe was performed by modification of 4-mercaptophenylboronic acid (MPBA, which can recognize SA and serves as a Raman reporter) on the surface of AgNPs. SA expression on single cells can be explored by SERS spectroscopy, as well as plasmonic energy loss imaging.

Because labeled SERS involves complex labeling chemistry, which may damage the probe structure, SERS that can be performed in a label-free manner is favored. Recently, Zhang et al. ^[44][76] used label-free and biocompatible plasmonic oxide nanoprobes to discriminate single cancer cells with high accuracy using a machine learning algorithm on a microfluidic platform. By combining the advantages of dopant-driven tunable plasmonics and enhanced electromagnetic-field-driven SERS, single human embryonic kidney 293 can be distinguished from human macrophage cell line U937, and THP-1 cells can be distinguished from peripheral blood mononuclear cells with a maximum accuracy of >90%. In addition, based on the SERS microfluidic droplet device, Sun et al. ^[45][77] implemented label-free detection of multiplexed metabolites at the individual cell level. AgNPs decorated with Fe₃O₄ magnetic nanoparticles by polyethylenimine (PEI) linker (Fe₃O₄@AgNPs) served as a magnetic SERS substrate. An Fe₃O₄@AgNP substrate and single cells were encapsulated into water-in-oil droplets on the microfluidic chip. During off-chip cell collection and incubation, various metabolites secreted by the cells were adsorbed directly on the Fe₃O₄@AgNP substrate and spontaneously entrapped in droplets due to the strong magnetism of Fe₃O₄. The SERS signal of metabolites on Fe₃O₄@AgNP substrates was explored when the dynamic adsorption of the metabolites reached an equilibrium state. In this way, label-free and simultaneous monitoring of lactate, pyruvate, and ATP was achieved via their intrinsic SERS fingerprints.

SPR based biosensors are intriguing because of their abilities in label-free detection and real-time monitoring of biological specimens ^[46][52]. Implementation of SPR sensing technology on microfluidics chip platforms provides a number of benefits, including precise control of each reaction stage, accelerating the reaction process and improving the sensing efficiency with high reproducibility. Inspired by the recent popular applications of SPR microfluidic chip detection units, Sugai et al. ^[47][78] designed a multichannel microfluidic chip sensing system for damage-free characterization of cells by pattern recognition. The PDMS chip was embedded with five Au films that could immobilize five cysteine derivative independently. When the cell culture media flowed into the chip, the cell-secreted target analytes interacted with the cysteine derivatives and generated five different SPR sensorgrams simultaneously. From the automatic statistical program of the SPR sensorgrams, optimal kinetic parameters could be selected as the “pattern information” for subsequent multivariate analysis. Borile et al. ^[48][79] also reported label-free, real-time, on-chip sensing of living cells via grating-coupled SPR. Grating-coupled SPR was established by combining a nanostructured gold grating with an SPR excitation and detection system, which could be further integrated in a PDMS microfluidic chamber for real-time monitoring of cell adhesion capability and cell-surface interaction with integral cell structures during the operational process. Nootchanat et al. ^[49][80] fabricated a miniature SPR sensor chip via the confined sessile drop method. By monitoring the deposition of layer-by-layer ultrathin films of poly(diallyldimethylammonium chloride)/poly(sodium 4-styrenesulfonate), the human immunoglobulin G could be detected with a higher sensitivity compared to common high-refractive-index glass SPR chips.

Because the LSPR peaks of metal nanoparticles can be tuned according to their size, shape, and composition, as well as the dielectric constant of the surrounding medium, attention on sensors based on LSPR continues to increase ^[50][81]. For example, a label-free LSPR-based biosensor was integrated with a biomimetic microfluidic chip platform for in situ monitoring of multiplexed cytokine secreted from obese adipose tissue with high-throughput ^[51][82]. The top PDMS layer of the integrated chip consisted of a cell culture chamber for adipocyte culturing and the formation of a crown-like structure induced by the macrophages. The bottom layer included an antibody-functionalized Au nanorod (AuNR) circular barcode, which was connected to the top layer for multiplexed LSPR detection. Multiplex determination of proinflammatory and anti-inflammatory cytokines secreted by the adipocytes and macrophages could be implemented simultaneously by recording different LSPR sensing signals on this “adipose-tissue-on-chip” nanoplasmonic sensing platform. Li et al. ^[52][83] also designed an LSPR biosensor on a PDMS chip for rapid erythrocyte counting and ABO blood group typing. In this biosensor, AuNPs functionalized with anti-A and anti-B antibodies were deposited on glass substrates of two independent sample cells for the identification of blood type (i.e., A, B, AB, and O); LSPR signals were obtained by a coupled ultraviolet-visible spectrometer.

The RI information of biological cells can indicate the size, constituent, and state of a living cell with high repetition and precision, making the interferometric method extremely appropriate for the selective identification of single cells ^[53][84]. As mentioned above, RI is very sensitive to environmental conditions, such as temperature, pressure, and liquid, etc.; it is highly susceptible to slight variations in the environment, such as moving or shaking of the sensing device. Moreover, it is challenging for operators to conduct living cell interference experiments without disturbing the surrounding buffer solution around the cells. These obstacles make the microfluidics integrated interferometry method widely favored for interferometric analysis of single cells ^[54][85]. In order to better understand the functions of single cells, an integrated label-free nanoplasmonic circular interferometric biosensor was developed on a PDMS chip to investigate molecular secretion from the immune cells ^[55][86]. The applied nanoplasmonic interferometer technology can address the remarkable drawback of conventional SPR or LSPR spectroscopy, such as poor spatial resolution and broad linewidth. In this way, the metalloproteinase 9 secreted from human monocytic cells could be effectively monitored at different time points after lipopolysaccharide simulation. More recently, single-cell capture, isolation, and long-time in situ label-free imaging were realized on a microfluidic chip using quantitative self-interference spectroscopy ^[56][87]. The cell incubator was integrated on the microfluidic chip for long-time single-cell culture after single-cell capture and separation in the microfluidic channels. The RI intensity changes induced by the multiple scattered light and transmitted light in the cell were reflected in the microscopic images. According to the quantitative analysis algorithm of the original RI data, the RI distribution of the single cell at the subcellular level could be obtained. In another work by the same group, monitoring of the dry mass of single cells with femtogram sensitivity was realized with a label-free and quantitative multichannel interferometric imaging technique ^[57][88]. Zhang et al. ^[58][89] proposed a microwave interferometric cytometry measurement system on a microfluidic chip for single yeast cell analysis. On this integrated platform, dielectric sensing can be applied for single S288C (a common Saccharomyces cerevisiae laboratory strain in) signal analysis at express speed by microwave interferometrics on a microchip, together with a software tool.

Moreover, an imaging technique without any exogenous labels has received wide attention. In particular, quantitative phase microscopy, which can precisely record the phase shift caused by heterogeneous samples, is extremely favored by scientific researchers. Interferometric reflectance imaging is a label-free imaging mode. It can realize the overall measurement of the height difference of biofilm layers induced by the accumulation of the biomass on the surface of a functionalized sensor. Additionally, it is capable of detecting single biological particles, such as cells, bacteria, or whole virus. Using interferometric reflectance imaging, Zaraee et al. ^[59][90] realized label-free and highly sensitive digital detection of whole E. coli cells. This detection even allows for the visualization of individual pathogens captured on the surface without requiring extra sample preparation or modification. By running the high-magnification imaging modality, the morphological characterization of single bacteria captured on the sensor surface can be visualized in this assay.