Advanced Biosensing towards Real-Time Imaging of Protein Secretion from Single Cells

Protein secretion of cells plays a vital role in intercellular communication. The abnormality and dysfunction of cellular protein secretion are associated with various physiological disorders, such as malignant proliferation of cells, aberrant immune function, and bone marrow failure. The heterogeneity of protein secretion exists not only between varying populations of cells, but also in the same phenotype of cells. Therefore, characterization of protein secretion from single cell contributes not only to the understanding of intercellular communication in immune effector, carcinogenesis and metastasis, but also to the development and improvement of diagnosis and therapy of relative diseases. In spite of abundant highly sensitive methods that have been developed for the detection of secreted proteins, majority of them fall short in providing sufficient spatial and temporal resolution for comprehensive profiling of protein secretion from single cells. The real-time imaging techniques allow rapid acquisition and manipulation of analyte information on a 2D plane, providing high spatiotemporal resolution. Here, we summarize recent advances in real-time imaging of secretory proteins from single cell, including label-free and labelling techniques, shedding light on the development of simple yet powerful methodology for real-time imaging of single-cell protein secretion.


Introduction
Cells secrete proteins into the extracellular environment to achieve important physiological processes, such as transportation of nutrients, digestion of food, regulation of metabolic processes, etc [1][2][3][4][5]. Cell secretion can be classified into the constitutive and regulated secretory pathway [6][7][8]. The constitutive secretory pathway is associated with transportation of secretory vesicles to the cell surface and their release independently of stimulus [9][10][11]. The regulated secretory pathway produces secretory granules, stores them in the cytoplasm, and secretes them into the extracellular environment only upon receiving stimuli [12]. In human, more than 15% of the genome encoded proteins are associated with the secretion process [13]. These released proteins include neurotransmitters, protein hormones,

Label-free techniques
In previous works, the real-time monitoring of cell secretion was achieved by label-free techniques such as plasmonic sensors, or label-required techniques such as fluorescence microscopy. The principles for all these methods are demonstrated in Figure 1. The representative works are summarized as shown in Table 1.

Plasmonic sensors
Plasmonic sensors have received tremendous attention owing to the label-free process, high sensitivity and fast response [36][37][38]. Plasmonic sensors generate surface plasmons and evanescence field that is sensitive to the changes of refractive index on sensing interfaces [39,40]. By functionalization of the sensing interfaces, the change of refractive index upon specific binding of target molecules can be recognized. The plasmonic sensing could be implemented on varying materials, geometries and structures, such as flat metal films (surface plasmon resonance)

Surface plasmon resonance (SPR)
SPR is a sensitive label-free technique for characterization of molecular interaction and detection of molecules through the affinity recognition. Due to surface plasmons excited on the surface of a thin metal film and an evanescence field created on surface, the refractive index change in the vicinity of the sensor surface (within ~200 nm) can be detected. This relatively large sensing depth allows SPR to monitor the changes occurred not only on surfaces, but also in the bulk. SPR has been applied in various studies for molecules with molecular weight between 1000 Da~500 kDa [47]. Liu et al. [48] attached human ovarian carcinoma SKOV-3 cells on the ceiling of a flow cell chamber, and detected their vascular endothelial growth factor (VEGF) secretion by monitoring the SPR signal of the antibodiesimmobilized gold chip at the bottom of the flow cell. This SPR sensor has a linear dynamic range of 0.1-2.5 μg mL −1 . Of note, even though SPR features label-free measurement and high sensitivity, the lack of spatial resolution restricts its application in the mapping of protein secretion. In the meantime, SPR imaging is suitable to achieve this goal. Instead of measurement of the SPR angle shifts in SPR spectroscopy, plenty of SPR imaging systems use fixed angle and wavelength of incident light for the excitation of SPR so that the reflection intensities of multiple spots within imaging area are obtained simultaneous.
Milgram et al. [49] took advantage of SPR imaging to real-time monitor the secretion of immunoglobulins from B-cells hybridoma. They assembled an antigen microarray using electro-copolymerization of free pyrrole and pyrrole-modified antigen on a gold chip. The secreted immunoglobulins were captured by the antigen proteins, triggering the refractive index changes. The SPR intensity changes were consequently observed at a fixed angle using a 12-bit CCD. A sharp SPR kinetic curve was observed after several minutes of incubation, indicating fast and sensitive detection of immunoglobulins. Wu et al. [50] used anti-CD4 antibody to capture and immobilize human CD4 + T-cells on a sensing chip, and detected their in-situ secretion of Interferon gamma (IFN-γ) through functionalized anti-IFN-γ antibody located at the neighboring sites. The detection limit for IFN-γ was ~50 ng/mL. Stojanović et al. [51] applied SPR imaging to quantify the antibody production from single EpCAM hybridoma cells. Based on the measured SPR signal alteration, the overall secretion antibody amount from a single cell was calculated as 0.02 to 1.19 pg per cell per hour. However, this estimation is questionable because not all the antibodies secreted from cell were captured. To solve this problem, they performed simulation using COMSOL Multiphysics and found the captured antibodies by sensors accounts for 99% of the excreted antibodies, only 1% excreted antibodies are not detected [52].
A major limitation of intensity-based SPR imaging is its relatively low sensitivity. Compared to SPR of high sensitivity, intensity-based SPR imaging suffers from one order of magnitude lower of sensitivity [53]. This relatively low sensitivity is dominantly contributed by the fluctuations of the incident light intensity, photon statistics associated shot noise, and detector noise [54]. This issue could be improved by performing optical multilayer structured long-rang surface plasmons [55], angle-resolved or spectral SPR imaging [56], NIR light source [57], etc.

Nanoplasmonic biosensors
Different from SPR sensors utilizing propagating surface plasmons generated on flat metal film, nanoplasmonic sensors generate and manipulate localized surface plasmons on nanostructures [58]. Conventionally, nanoapertures or nanoparticles are fabricated to interact with light, leading to localized surface plasmon resonance (LSPR) or extraordinary optical transmission (EOT). As nanoplasmonic biosensors combine with cell imaging, a powerful tool is created for protein secretion imaging. Nanoplasmonic biosensors generate strong evanescence field in the vicinity of less than 30 nm, thus are highly sensitive to the local refractive index changes, allowing the detection of target molecules captured on surfaces. Since the excitation of these nanostructures did not require complex optical devices, the instrumentation can be simple and straightforward. By tracing the dynamic alteration of reflection or transmission spectra, these changes can be recorded in a simple and real-time manner. Therefore, this feature allows a simple-but-sensitive label-free detection.

Localized surface plasmon resonance (LSRP)
LSRP generated on nanostructures results in the collective oscillation of electrons at the interface of metallic structures. A localized electromagnetic field is sensitive to the nanostructure shape and changes of refractive index at the distance of 10-30 nm [59,60]. A molecular binding event on these nanostructures, causes a red shift in SPR peak, and gives rise to the scattering intensity of light at the same time. Due to a short electromagnetic field decay length, LSPR is insensitive to the changes of the refractive index in bulk, therefore the bulk effect can be minimized. This feature allows LSPR to measure temperature-dependent binding process, and investigate the effects of various environment factors on molecular interactions, such as solution pH and ionic strength [61].
LSPR measurement can be implemented by recording either SPR peak shift or scattering intensity changes. In the SPR peak shift based detection, Zhu et al. [62] combined microwell technique with LSPR to monitor Interleukin 6 (IL-6) secretion in the single cell level. The microwells were adapted to trap cells, a gold-capped nanopillar-structured cyclo-olefin-polymer film was covered on the top of microwells. The transmittance spectrum of the gold nanostructured surface provides real-time information on the absorption peak shift of nanogold during cell secretion. This nanostructured film further fabricated with Anti-IL-6 antibody realized a detection limitation of 10 ng/mL for IL-6.
In the scattering intensity based detection, Oh et al. [63] developed a multiplexed LSPR system for simultaneous measurements of pro-inflammatory cytokines (IL-2, IFN-γ, and TNF-α) and anti-inflammatory cytokines (IL-10) secreted by T cells. The cell culture and cytokine detection were conducted in independent steps. The sensitivity reached 20−30 pg/mL. Faridi et al. [64] applied similar system to characterize the secretion of IL-6 and tumor necrosis factor alpha (TNF), secreted from human monocytes and lymphocytes. The same principle has been applied to monitor neutrophil extracellular traps (NETs) and fibril released from single neutrophils [65]. Raphael et al. [66] achieved the real-time imaging of anti-c-myc antibodies secreted from single hybridoma cells with gold nanostructured arrays. The electron beam lithography was implemented to fabricate the square arrays of gold nanostructure on glass coverslips. The gold nanostructure arrays have a diameter of 70 nm, a height of 75 nm and a separation distance of 300 nm. The differing distances between the position of the settled single cell and the centers of arrays provide spatial resolution to observe the protein secretion. The c-myc peptides conjugated to plasmonic gold nanostructures captured the secreted anti-c-myc antibodies in a real-time manner. The caused binding displayed an increase in scattering intensity due to LSPR effect, which is measured through changes of light reflection. In parallel with reflected light based LSPR characterization, the transmitted light and fluorescence microscopy were integrated for live cell imaging. The transmitted light imaging enables observation of the position of single cell and its morphological change. The fluorescence microscopy allows the monitoring of membrane dynamics through a cell plasma membrane labeldye rhodamine DHPE, which distinguishes the signal due to occasional outward protrusions of lamellipodia from protein secretion signal. To model the singular cell secretion, the cell was assumed a spherical emitter producing a propagating pulse of antibodies with a Gaussian concentration profile, where the diffusion constant D = r 2 /6•t. This method provides a sensitivity of 100 pM (~4.88 ng/mL) for the detection of anti-c-myc antibodies.

Extraordinary optical transmission (EOT)
In EOT biosensor, regularly periodic nanohole structure of subwavelength in a metallic film results in significant enhancement of light transmission through the nanoholes. This phenomenon is associated with both localized and propagating surface plasmons on the nanohole structures. By collecting and analyzing the transmission spectra, the light frequency-dependent transmission enhancement could be easily recognized. The molecular binding event on surface is corresponding to the transmission spectral shift. Consequently, the binding event can be visualized in a real-time manner through tracking the changes of transmission spectra. Combining localized and propagating surface plasmons, the spectra of EOT provide a wealth of information with varying sensitivities at different regions of nanoholes. In addition, the nanohole structures allows the flow-through design, which changes the manner of the mass transfer of analytes, and increases delivering rate of analytes from bulk to the sensing surface [67].
To achieve real-time imaging of cell secretion, Li et al. [68] first developed a microfluidic device that separated a cell culture module and an EOT sensing module. The nanohole array sensor has a hole diameter of 200 nm and a periodicity of 600 nm, and was fabricated by deep-UV lithography and functionalized by biotinylated antibodies. This detection module is directly connected with an adjacent cell culture module made of a zigzag single-channel PDMS unit. This biosensor system achieved a detection limit of 145 pg/mL for VEGF. In this study, the VEGF secretion was detected from the media containing a group of cancer cells, and the mapping of secretion was disabled due to the separated configuration of detection and cell culture modules. Subsequently, they developed a microfluidic system suitable for the secretion imaging of a single cell [69]. Enclosed in a microchamber, a single cell was attached to nanohole arrays, functionalized by antibodies. To achieve the function of real-time imaging, spectrum profiles on a perpendicular 1D line was collected, and the selected positions of the region-of-interest were analyzed. This sensor achieved a detection limit of 39 pg/mL for interleukin-2 (IL-2) in complex media.

Photonic Crystal Resonator (PCR)
Similar with plasmonic sensors, Photonic Crystal Resonator (PCR) exploits an evanescence field to interact with and sense the surrounding medium, i.e. the changes of its refractive index. PCR is created by Photonic Crystals (PCs) that own periodically varied refractive indices, which forbid the light propagation of certain wavelength of light in certain directions inside the material. This causes constructive or destructive interference of the light, and therefore a minimum in the transmission spectrum and a maximum in the reflection spectrum could be observed. Dependent on material, geometry and the index contrast, the decay length of PCs ranges from dozens of nanometers to the order of a few microns [70]. PCs composed of silicon nitride (Si 3 N 4 ) or titanium dioxide (TiO 2 ) have been proved able to achieve cellular imaging with resolution 2-6 μm [71]. Juan-Colás et al. [72] monitored the secretion of thrombopoietin at single cell level using photonic crystal resonant imaging (Figure 2). The PCR surface microfabricated by electron beam lithography, consists of a grating with period of 555 nm on a 150-nm-thick Si 3 N 4 film. It displayed a penetration depth of ~200 nm. To realize hyperspectral resonance image, reflection spectra was taken in a sequence of illumination wavelengths with 0.25-nm wavelength step. Under each wavelength, the intensity value of each pixel was analyzed and fit into a Fano resonance curve to accurately obtain the resonance wavelength for each pixel. To create a sensitive surface for analyte capture, antibodies were functionalized on a 3D matrix consisting of branched glucan dextran, which increased the density of antibodies on the surface. This sensor was demonstrated to detect lower than 125 ng/mL of suspended recombinant Human thrombopoietin. The interaction of antibodyantigen was treated as a single adsorption process, and the protein secretion was modeled as a Langmuir adsorption distribution. By quantifying the total secretion area in a timely manner, the single-cell secretion rate was calculated as 22 μm2/h. In addition, the wide field of view allows parallel imaging of 30 cells on an area of 500 × 500 μm, so that their dynamics and the kinetics could be characterized simultaneously.

Interferometric Scattering Microscopy (iSCAT)
Interferometric Scattering Microscopy (iSCAT) is a single-molecule detection based approach. It relies on the light scattered by subwavelength objects. The signals come from the interference between the scattered light by the detected object and a reference light (Figure 3d). With the capability of single-molecule detection, iSCAT has shown its remarkably high sensitivity in cell imaging, singleparticle tracking, label-free imaging of nanoscopic (dis)assembly, and quantitative single-molecule characterization [73]. Meanwhile, iSCAT microscopy itself does not provide adequate chemical or biological specificity due to its nature in collection of scattered light from all small objects [74].
McDonald et al. [75] reported an iSCAT contrast method to distinguish proteins secreted from an Epstein−Barr virus (EBV)-transformed B cell line (Figure 3a). The observed contrast on an iSCAT image reflects the amplitude of the electromagnetic field scattered by proteins, which is directedly correlated with the scattering cross-sections of detected molecules (Figure 3c). Here, the iSCAT demonstrated its capability of monitoring secreted proteins with varying molecular  weight 100 kDa−1 MDa. To provide iSCAT the detection specificity for human Immunoglobulin G(IgG), the sensing surface was functionalized with anti-human IgG, which showed high specificity on the adsorption event of IgG (Figure 3b, d). The quantification of IgG was realized by counting over certain integration period. It was found that the secretion of IgG antibodies in single Laz388 cells proceeded at a rate of 100 molecules per second.

Secretome quantification and identification. (a) Cartoon of the detection region. (b) Histogram of detected proteins during a 125 s long measurement period in which detected contrasts are counted in bins of 1 × 10 -4 contrast (diagonal-hatched blue bars). 503 distinct proteins were counted in this period. The red arrow indicates the expected contrast corresponding to an IgG dimer. Superimposed over the cell secretion data is the contrast distribution of a purified IgG solution injected into the iSCAT FOV with a micropipette (diagonal-hatched rose bars, normalized for clarity). The detected secretion events from a single Laz388 cell on an anti-human IgGfunctionalized coverglass are also shown (125 s integration, hollow black bars). The anti-human IgG surface selectively binds all four IgG antibody subtypes and resists the adsorption of other proteins. (c) Comparison of 2.5 fps (400 ms) and 25 fps (40 ms) image acquisitions of a single
One of the challenges in iSCAT for cell secretion is that, the recognition of small molecules is difficult due to their limited iSCAT contrast. In above work, the minimum protein molecular weight from iSCAT was approximately 100 kDa. Most cytokines have molecular weight range from approximately 6 to 70 kD, which fall out of the iSCAT's detection range.

Fluorescence
Fluorescence based detection provides superb sensitivity of detection, and flexibility in labeling. Fluorescence microscopy features high spatial and temporal resolution, and capability of tracking multiple cells simultaneously [14]. Therefore, fluorescence based detection has been widely used for tracking protein secretion.

FLUOROSpot
A simple and widely used method of single cell secretion analysis is the enzymelinked immune absorbent spot (ELISpot) that utilizes the immunosandwich-based assay for the measurement of footprint of cells [76]. Cells are loaded into wells precoated with primary antibody. Secreted proteins are captured, and further bound with labeled secondary antibodies. The addition of substrate gives rise to the spots indicative of secretion footprint. Based on similar principle, FLUOROSpot was developed which uses fluorescent dyes to replace the enzyme labels. In this way, more than one protein could be analyzed at the same time [77]. In spite that ELISpot and FlUOROSpot have been used as common methods for the detection of cell secretion, several drawbacks hinder its applications in the real-time imaging of protein secretion, including spectral overlap, varied individual spots, limited temporal resolution caused by long incubation time (12-48 h), limited number of simultaneously tracked proteins, and cell lost during the process [78].

Microengraving
Inspired by the principle of FLUOROSpot, the microengraving method was developed to isolate and confine single cells in a planar array of microwells. The protein secretion of segregated cells can be tracked as protein detection microarrays are integrated with the microwells, which are conventionally made of elastomeric polymers such as polydimethylsiloxane (PDMS). These microfabricated wells have subnanoliter volumes. A glass side fabricated with antibodies is covered to the array of microwells, to capture the proteins secreted by confined cells. Subsequently, the slide is removed and incubated with fluorescently labelled secondary antibodies. The single-cell cytokine secretion analysis is implemented by fluorescence imaging. Bradshaw et al. [79] showed that the same cells in the microwell can be applied on two detection microarrays successively, allowing monitoring the secretion of four types of cytokines and antibodies. In the meantime, the segregated cells in the wells can also be directly stained by immunofluorescence to determine their lineages.
A drawback in this method is that, the tracked number of proteins is limited by the capability of differentiation of multiple fluorophores within a single spectrum.

Barcode chip
The number of secretion protein types from above methods is limited due to the spectral overlap as multiple fluorescence labels employed on the same detection chip. This spectral overlap was overcome by separating functionalized antibodies into different lines, using a barcode design. With a proper distance between, up to 15 parallel antibody lines can be contained within a nanoliter cell-trapping chamber. This improved spatial resolution results in an enhanced its capability of simultaneous multiplex detection [28,29,80]. Lu et al. [81] further extended the multiplexing capacity by further improving the spectral encoding, namely, in each isolated line, three antibodies with distinct fluorescent labels are contained. This strategy enables simultaneous analysis of 42 types of proteins secreted from a single cell.
Barcode chips provide valuable information for accumulated cell secretion over a period of time. However, the underlying limitations restrict its applications on the real-time monitoring of protein secretion. For instance, the lag between cell secretion and protein detection is inevitable since capture antibody and detection antibody have to play their part in distinct steps. Given their coexistence during cell secretion, strong background fluorescence signal is expected from the unbound labeled antibodies, thus secreted proteins are difficult to be distinguished from the background noise. In addition, to remove excess probes and avoid non-specific bindings, the sensor surface after incubation with cells requires intensive wash steps, which inhibits the shortening of analysis duration into a few hours.

Total internal reflection fluorescence microscopy (TIRFM)
Shiraskaki et al. [82] proposed a solution for the challenge mentioned above using total internal reflection fluorescence microscopy (TIRFM) combined with microengraving method. They deposited single cells on microwells, on the bottom of which anti-cytokine capture antibodies were fabricated. Instead of separate incubation of a sensor chip with a cell and the detection antibody, the fluorescent detection antibody is present in the cell culture medium, so the cytokine capture is in step with the binding of detection antibody. An objective lens of high numerical aperture was used to achieve high incidence angles, generating evanescence field. The near-field excitation by total internal refection enhances the fluorescence signal from the detection antibody in the sandwich immunocomplex, and reduces the background signal from the unbound detection antibodies in the culture medium (Figure 4a). With these features, imaging of cytokine secretion was achieved within a single step (Figure 4b). Meanwhile, the cell staining (calcein and SYTOX) was applied to investigate the membrane integrity during IL-1b secretion (Figure 4c). It was found the onset of IL-1b secretion was consistent with the onset of calcein disappearance and the second protein SYTOX influx (Figure 4d). This phenomenon indicated the loss of membrane during IL-1b secretion. The limit of detection of this approach reached down to 2000 cytokine molecules.

Nanoplasmonic fluorescence
The near-field excited fluorescence can also be realized by nanoplasmonic resonator. Wang et al. [83] developed a tunable nanoplasmonic resonator (TNPR) enhanced fluorescence immunoassay for imaging of IL-2 secretion in submicrometer resolution (Figure 5). In this study, the fluorescent secondary antibody was not present while the cytokine secretion was in progress. Instead, cytokine capture and detection step was separated by washing step. With a TNPR structure of 100 nm in diameter, and an optimized fluorescence enhancement at ∼10 nm from a gold surface, the fluorescence signal was enhanced 117-fold in the TNPR area. The limit of detection was lower than 100 ng/mL.

Bead biosensors
Another fluorescence based strategy is associated with bead-based biosensors. The cells are confined in microwells, along with antibody-labeled microbeads, and fluorescently-labeled secondary antibody. The capture of secreted proteins on beads was accompanied by the increasing density of secondary antibody, and increasing fluorescence intensity of microbeads. A key advantage of bead biosensors is that,  (SYTOX), that of a calcein-stained cell bodies was green (Calcein) and that of secreted IL-1β was yellow (CF660R anti IL-1β Ab). Merged images of these three fluorescence signals are also displayed (Merged). Each image was obtained at the described period. Scale bar, 20 μm. (c) Schematic of simultaneous monitoring of IL-1β secretion and cell membrane integrity using calcein and SYTOX staining. SYTOX influx and fluorescent calcein disappearance was observed due to compromised plasma membrane integrity. (d) Example of the signal time course during time-resolved monitoring. Grey bands represent the period when the monocytes were exposed to ATP. Arrows represent the transition time of the respective signals. Reprinted (adapted) with permission from Ref. [31]. Copyright 2014, Springer Nature.
it breaks the limit of target-molecular adsorption kinetics, which is controlled by diffusion rates, and exists in majority of immobilized or stationary biosensors [84]. In addition, the regeneration of biosensors can be as simple as removing used beads, and infusing new beads [85]. Son et al. [86] used microcompartments as small as 20 picoliter to confine single cells, suspended antibody-labeled microbeads, and fluorescently-labeled secondary antibody together. The quantitation of secreted cytokines was achieved by tracking the intensity changes of fluorescence on microbeads. They found that the number of microbeads confined within a single microcompartment did not significantly affect the fluorescence enhancement on a single microbead. An et al. [87] utilized a microwell device for cell confinement, and detected the fluorescence changes on functionalized detection beads which are co-incubated with single cells.
Other than the design of microwell, Cui et al. [88] developed an microfluidic immunoassay device that integrated a cell culture chamber, an array of cytokine detection units and an array of active peristaltic mixers for on-chip sample mixing. Cells were isolated, cultured and biochemically stimulated in the same chamber. The detection chambers were loaded with cytometric fluorescent beads. Upon sandwich structure formed after cytokine secretion, the fluorescence intensity changes were analyzed by flow cytometry. The continuous monitoring of cytokine secretion was achieved by the extraction of a small portion of the cytokine-containing culture media to the detection chamber. With this system, the secretion profilings of IL-6, IL-8, and TNF were observed with a detection limit of 20 pg/mL and a sample volume of 160 nl.

Cell-surface affinity sensors
Cell-surface affinity sensors functionalize cell surface to capture secreted molecules from cells. These targets can subsequently be detected by fluorescent labeling. An appealing advantage using this strategy is that, the effect of heterogeneous spatial distribution of secreted molecules is minimized [89,90]. As secreted molecules release from the cell surface, their diffusion and dilution pose a challenge to the sensitivity of sensors. As the cell surface is turned into sensor surface, these molecules could be captured prior to their diffusion. This immediate interaction enhances the sensitivity of secretion detection. Various cell secretions have been studied previously  utilizing cell-surface affinity sensors, including ATP [91,92], growth factors [93], cytokines [94], and antigens [95]. Liu et al. [96] developed a detection method that utilized fluorescent magnetic nanoparticles labeled secondary antibody, and achieved a detection limit of 0.1 pg/mL. The cell surface was biotinylated, and functionalized by neutravidin and a biotinylated IL-6 capture antibody. Upon the binding of cytokines secreted from cells, fluorescent magnetic nanoparticles labeled secondary antibodies are introduced for indication of the amount of cytokine secretion.
A challenge of cell-surface affinity sensors is the damage and cell viability caused by cell surface modification [97]. Due to the high background signal from free labeled secondary antibody, the cells after secretion have to undergo wash steps to remove free secondary antibodies. This limits its application on the real-time monitoring of cell secretion.

Droplet microfluidics
Droplet microfluidic techniques involve mixture of the aqueous phase and oil phase, and generation of water-in-oil microdroplets in the microfluidic chips [98]. Droplet shape offers encapsulation of a single cell in a picoliter droplet, which confines individual cell and entraps its secretion, restrains the interference from surrounding environment. To achieve real-time monitoring of cell secretion, droplets microfluidic techniques are conventionally integrated with cell-surface affinity sensors or bead-based biosensors. Qiu et al. [94] developed a membrane-anchored aptamer sensor and combined it with droplet microfluidics for IFN-γ detection. This aptamer sensor has a hairpin structure, resulting in quenched fluorescence. Upon binding with IFN-γ, the aptamer switches on the fluorescence signal by aptamer structure change. The limit of detection was approximately 10.0 nM. Due to the proximity of the aptamer sensor to the cell surface, this method achieved monitoring of cytokine secretion at a single-cell level. Chokkalingam et al. [99] took advantage of free capture beads in droplets to detect the secreted cytokines from cells. The cytokine capture step and fluorescent detection step were separated by the gelation of droplets, and de-emulsification. Wei et al. [100] combined LSPR with droplet technique, developing a sensor with a detection limit of 6-7 ng/ ml for VEGF and IL-8. The LSPR spectrum shift was obtained by dark-field spectroscope. Due to the advantages of droplet technique, the plasmonic droplets in a continuous flow allow high-throughput detection of 600-1800 droplets/min, 100-150 cells/min.

Challenges and future trends
Isolated cell culture in a confined environment for subsequent secretion detection allows good flexibility in the selection of sensing techniques, but such module-based strategy has not been achieved for real-time detection of individual cell secretion. Meanwhile, the integration of cell culture and secretion detection has shown promising results in real-time monitoring single cell secretion by physically presenting the target cells in close proximity to the sensing region. However, there are still significant challenges and difficulties ahead as indicated below.
1. Cell secretion dynamics are found associated with cell states, therefore, it is important for the sensor systems to precisely quantify the cell secretion. An obstacle for current methodology is that, the molecule secretion and molecule detection are both dynamic process occurring at the same time.
The concentrations of target molecules in solution are constantly changing, such that an equilibrium-state based calibration curves are difficult to obtain. Therefore, on one hand, novel or improved detection methods with more rapidly reached equilibrium state are required to minimize the effect of detection process. On the other hand, a comprehensive dynamic model involving secretion and detection dynamics is necessary to understand this complex process and precisely quantify the cell secretion [101,102].
2. The real-time imaging process requires highly specific and selective responses from sensors. Since the specificity and selectivity of biosensors might be compromised by non-specific binding in complex biological media used in experiments, significant efforts are needed to eliminate the matrix effect. Several blocking strategies that instantaneously passivate the sensor surface and do not affect the recognition chemistry are developed, including protein (like BSA), detergent (Tween, etc.), or different polymer blockers. In imaging experiments, a thoroughly optimized blocking strategy should be applied.
3. It is important to investigate the effects of important experiment conditions for the detection. Previous works mainly focused on the innovation of methodology, improvement of the sensitivity, enhancement of the capability of multiplex detection, etc. However, the influence of multiple important experimental conditions are lack of attention. For instance, in previous works, cells during their secretion were manipulated in various status, including suspended in media, immobilized on the ceiling of microwells, or directly absorbed on sensor surface. These distinct distances between sensing surface and cells are expected to influence the local concentrations of secretions on sensor surface, leading to considerate impact on detection sensitivity. 4. In current stage, multiplex real-time imaging of cell secretion is a great challenge. Though the fluorescence based barcode chip has demonstrated its capability in profiling up to 42 types of proteins secreted from a single cell, the real-time imaging of these proteins is incapable due to the lag between cell secretion and protein detection, and intensive wash steps which require several hours. Nanoplasmonic biosensors and SPR imaging have difficulty to broaden their multiplexity owing to their label-free characteristic. Therefore, to achieve the multiplex real-time imaging, the integration of these techniques might provide promising opportunities.
5. Current studies focus on the secretion of single type cells. The interactions of different cells, and their intercellular communication through proteins such as cytokines demand comprehension and further exploration. Therefore, a platform capable of monitoring of multiple cells' real-time secretion is demanded.

Conclusions
In the past two decades, cell secretion monitoring has gained increasing attention. To understand the heterogeneity of protein secretion in single cell level, various optic methods have been developed including labelling and label-free techniques. The label-free techniques such as plasmonic sensors, PCR, EOT, are able to recognize the secreted proteins by sensing the changes of refractive index on the sensor interfaces. The limit of detection differs from 20 pg/mL to 50 ng/ mL. The signal transduction during the analyte-ligand interaction does not require

Author details
Lang Zhou, Pengyu Chen and Aleksandr Simonian* Materials Research and Education Center, Department of Mechanical Engineering, Auburn University, Auburn, AL 36849, USA *Address all correspondence to: als@auburn.edu the involvement of any additional materials, therefore, the label-free techniques allow simple operation and fast response, suitable for real-time imaging. The labelling techniques such as fluorescence microscopy utilize fluorophore-labeled bioreceptors to recognize the proteins. Different from the label-free techniques, the fluorescence based methods conventionally use two recognition elements to form a sandwich complex, one of them as the capture element, the other as the signal transducer (the fluorophore-labeled bioreceptor). The limit of detection differs from 0.1 pg/mL to 20 ng/mL. Even though the two recognition element approaches involve additional processing and washing steps, which prolong the response time, the extraordinary sensitivity and low background improve the reliability and allow multiplex detection, making it very attractive to some applications.
For instance, TIRFM has demonstrated its possibilities on real-time imaging of protein secretion from single cells by utilizing both high sensitivity from fluorescence microscopy, and wash-free process due to the evanescent field created by total internal reflection. In spite of the achievements mentioned above, intensive work is demanded in future such as understanding of cell secretion dynamics, improving the detection specificity, developing multiplex real-time imaging platform, and investigating the interactions of different cells, and their intercellular communication.