Development of Microfluidic Systems Enabling High-Throughput Single-Cell Protein Characterization

This article reviews recent developments in microfluidic systems enabling high-throughput characterization of single-cell proteins. Four key perspectives of microfluidic platforms are included in this review: (1) microfluidic fluorescent flow cytometry; (2) droplet based microfluidic flow cytometry; (3) large-array micro wells (microengraving); and (4) large-array micro chambers (barcode microchips). We examine the advantages and limitations of each technique and discuss future research opportunities by focusing on three key performance parameters (absolute quantification, sensitivity, and throughput).


Introduction
Quantification of single-cell proteomics (e.g., functional proteins such as cytokines, and structural proteins such as actin filaments) provides key insights in the field of cellular heterogeneity (e.g., immune response variations and tumor heterogeneity) [1][2][3][4][5][6]. Fluorescent flow cytometry is the dominant workhorse in the field of single-cell protein quantification, featured with high throughput and high levels of multiplexing [7][8][9][10]. However, this technique is a semi-quantitative approach, reporting the intensity of fluorescence rather than the copy number of target proteins for individual cells. Even within the incorporation of calibration beads, the capability of absolute quantification is still questionable [11][12][13][14][15].
Within the last ten years, we have witnessed huge developments in microfluidic platforms enabling single-cell protein expressions (see Table 1), which can be classified into four major types as follows: (1) microfluidic fluorescent flow cytometry; (2) droplet based microfluidic flow cytometry; Detection of the activities of enzyme alkaline phosphatase secreted by single E. coli encapsulated in microdroplets [49] Droplet Based Microfluidic Flow Cytometry Detection of cytokine (IL-10) secretion of single CD4+ CD25+ regulatory T cells in microdroplets over time [50] Droplet Based Microfluidic Flow Cytometry Detection of intracellular proteins of HRas-mCitrine, expressed within single HEK-293 cells and actin-EGFP expressed within single MCF-7 cells encapsulated in microdroplets [51] Droplet Based Microfluidic Flow Cytometry Detection of cytokine (IL-2, IFN-γ, TNF-α) secretion of single, activated T-cells in microdroplets over time [52] Large-Array Micro Wells (Microengraving) Detection of secreted cytokines (IL-6, IL-17, IFN-γ, IL-2, and TNF-α) of primary T cells at the secretion rate from 0.5 to 4 molecules/s [53] Large-Array Micro Wells (Microengraving) Detection of secreted cytokines (IFN-γ and IL-17) of individual CD4+ T cells with peptide-loaded MHC class II pre-coated on the surface of micro wells for on-chip activation [54] Large-Array Micro Wells (Microengraving)~2 00-fold improvement in the limits of detection of secreted cytokines using hybridization chain reactions [55] Large-Array Micro Wells (Microengraving) Detection of serial, time-dependent secreted cytokines (IFN-γ, IL-2, TNF-α) of primary human T cells, revealing that cells predominantly release one cytokine at a time rather than actively secret multiple cytokines simultaneously [56] Large-Array Micro Wells (Microengraving) Detection of secreted chemokines (ELR + CXC) from single colorectal tumor and stromal cells with polyfunctional heterogeneity located [57] Large-Array Micro Chambers (Barcoding Microchips)   Detection of 12 proteins including TNF-α, IFN-γ, IL-2, IL-1α, IL-1β,  IL-6, IL-10, IL-12, granulocyte-macrophage colony-stimulating factor, CCL-2, TGF-β and PSA of macrophages and cytotoxic T lymphocytes [58] Large-Array Micro Chambers (Barcoding Microchips) Detection of 11 proteins directly or potentially associated with PI3K signaling of three isogenic cell lines representing the cancer glioblastoma multiforme, at the basal level, under EGF stimulation, and under erlotinib inhibition plus EGF stimulation [59] Large-Array Micro Chambers (Barcoding Microchips) Detection of secreted proteins (IL-8 and VEGF) of circulating tumor cells [60]

Microfluidic Fluorescent Flow Cytometry
In microfluidics, the fluorescent micro flow cytometry is the first approach proposed to quantify single-cell protein expressions, which is a miniaturized version of conventional flow cytometry [61][62][63][64]. As shown in Figure 1, single cells with surface (fluorescence labeled antibodies targeting surface proteins) or intracellular staining (fluorescence labeled antibodies targeting cytosolic proteins) are flushed into microfabricated flow channels. By integrating laser sources and fluorescent detection units, the fluorescent intensities of labeled single cells are obtained and the copy numbers of targeted proteins are quantified using the calibration curves obtained by calibration beads [46,47].
Compared to conventional flow cytometry, microfluidic flow cytometry is featured with a reduction in cellular samples and reagents. As shown in Figure 1d, the proposed system was used to analyze small numbers of cells, with the quantified number of cells decreased from 20,000 to 625.
Thus, microfluidic flow cytometry is more suitable for protein quantification of small cell samples (e.g., primary cells or tumor cells obtained from biopsies). As to the capability of absolute quantification, both approaches can assay cell surface markers with absolute quantification by combining calibration beads. As to the measurement of intracellular proteins, due to the lack of effective calibration beads, flow cytometry-based approaches are still qualitative rather than quantitative, which needs further research efforts.

Droplet Based Microfluidic Flow Cytometry
Since microfluidic fluorescent flow cytometry cannot estimate the secreted proteins from single cells, droplet-based microfluidic flow cytometry [65][66][67] was modified to encapsulate single cells, enabling the quantification of specific proteins secreted by single cells [48][49][50][51][52][68][69][70][71][72][73][74][75]. As shown in Figure 2, single cells and functionalized capture beads are encapsulated in agarose-gel droplets, where beads function to bind cytokines secreted by single cells. Then, droplets are gelled and washed to break the emulsion, followed by incubation with fluorescently-labeled detection antibodies targeting cytokines bound on the beads. Subsequently, the signal intensities in the beads are quantified by flow cytometry. Based on this approach, 7415 single cells were analyzed, revealing that (1) there was a presence of eight different cellular subpopulations; and (2) 85% of all individual cells secreted one or more cytokines [52].  Although powerful, there are still several concerns for this technique, which need to be carefully addressed. (1) The cell loading efficiency in individual droplets needs to be further quantified and optimized. High concentrations of suspension cells before the emulsification process lead to loading of more than one cell in each droplet, while low concentrations of cell suspensions result in high percentages of empty droplets without single cells. Variations in fluorescent intensities resulted from uncontrolled cell numbers per droplet may compromise the quantitative results; (2) The local microenvironments of single cells trapped in individual droplets (e.g., nutrient levels and gas permeability) may be significantly different from in vivo situations and, thus, there is a doubt on the normal metabolism status of individual cells trapped in droplets. If these individual cells are confronted with challenging environments, the secreted protein numbers cannot indicate the normal status of these individual cells; and (3) the use of cytokine-capture beads also brings more issues. One consideration is the bead loading evenness among individual droplets, which may lead to variations in protein quantification. The second point is the calibration issue. Due to the use of cytokine-capture beads, there is an uneven distribution of fluorescence among individual droplets (intensity peaks around individual beads), which poses further obstacles in calibration.
In addition to the quantification of secreted proteins of single cells, droplet-based microfluidic systems have also been used to quantify intracellular proteins of single cells. As shown in Figure 3, cells are introduced into the device in suspension and are electrically lysed in situ. The cell lysate is subsequently encapsulated together with antibody-functionalized beads into droplets, which are then stored on-chip while the binding of intracellular proteins to the beads is monitored fluorescently. Based on this approach, the concentrations of specific intracellular proteins over five orders of magnitude (~50 pM to 1 μM) can be characterized [51]. Although powerful, there are still several concerns for this technique, which need to be carefully addressed. (1) The cell loading efficiency in individual droplets needs to be further quantified and optimized. High concentrations of suspension cells before the emulsification process lead to loading of more than one cell in each droplet, while low concentrations of cell suspensions result in high percentages of empty droplets without single cells. Variations in fluorescent intensities resulted from uncontrolled cell numbers per droplet may compromise the quantitative results; (2) The local microenvironments of single cells trapped in individual droplets (e.g., nutrient levels and gas permeability) may be significantly different from in vivo situations and, thus, there is a doubt on the normal metabolism status of individual cells trapped in droplets. If these individual cells are confronted with challenging environments, the secreted protein numbers cannot indicate the normal status of these individual cells; and (3) the use of cytokine-capture beads also brings more issues. One consideration is the bead loading evenness among individual droplets, which may lead to variations in protein quantification. The second point is the calibration issue. Due to the use of cytokine-capture beads, there is an uneven distribution of fluorescence among individual droplets (intensity peaks around individual beads), which poses further obstacles in calibration.
In addition to the quantification of secreted proteins of single cells, droplet-based microfluidic systems have also been used to quantify intracellular proteins of single cells. As shown in Figure 3, cells are introduced into the device in suspension and are electrically lysed in situ. The cell lysate is subsequently encapsulated together with antibody-functionalized beads into droplets, which are then stored on-chip while the binding of intracellular proteins to the beads is monitored fluorescently. Based on this approach, the concentrations of specific intracellular proteins over five orders of magnitude (~50 pM to 1 µM) can be characterized [51]. Compared to conventional fluorescent flow cytometry, the droplet-based microfluidic flow cytometry does not need the steps of intracellular staining when the numbers of cytosolic proteins have to be measured. This advantage can address the issue of non-specific binding in the process of intracellular staining, which is definitely beneficial in protein quantification. However, in droplet based microfluidic flow cytometry, the dilution of targeted proteins occurs since the volume of droplets is roughly 0.1-1 nL, two or three orders larger than individual cells (~pL). Thus, proteincapture beads are required to enrich the concentration of targeted proteins, which, again, leads to the lack of effective calibration.

Large-Array Microwells (Microengraving)
Both conventional and microfluidic flow cytometry cannot evaluate the same single cells at multiple time points, which is of importance in immunology. To address this issue, Love et al. proposed the design of large-array micro wells ("microengraving", 0.1-1 nL each) to isolate individual cells and quantify secreted proteins in a time sequence [53][54][55][56][57][76][77][78][79][80][81][82]. As shown in Figure 4, single cells, suspended in media, are deposited onto a large array (~20 × 50 μm 2 ) of PDMS microwells. With the removal of excess cells, the microwells are then inverted onto a glass slide coated with a specific capture reagent. After an incubation period, the microwells are removed and applied to a second glass slide coated with a different capture reagent. The resulting microarrays are interrogated with laser-based fluorescence scanners.
After microengraving was initially proposed to quantify cellular protein expressions [76], several technical improvements were further realized, which are (1) the coating of specific macromolecules on the surface of PDMS wells to stimulate immune cells [54]; (2) the usage of the hybridization chain reaction to amplify signals resulting from sandwich immunoassays [55]; and (3) the retrieval of targeted single cells in specific micro wells [77].
Although microengraving can enable absolute quantification of multiple proteins secreted by individual immune cells, compared to flow cytometry, it requires the careful manipulation of glass slides without stimulating or even dislodging single cells within individual wells. In addition, it is also questionable that the micro environments formed within each microwell may be different from in vivo situations and, thus, the characterized protein levels cannot reflect the in vivo metabolisms of immune cells. Compared to conventional fluorescent flow cytometry, the droplet-based microfluidic flow cytometry does not need the steps of intracellular staining when the numbers of cytosolic proteins have to be measured. This advantage can address the issue of non-specific binding in the process of intracellular staining, which is definitely beneficial in protein quantification. However, in droplet based microfluidic flow cytometry, the dilution of targeted proteins occurs since the volume of droplets is roughly 0.1-1 nL, two or three orders larger than individual cells (~pL). Thus, protein-capture beads are required to enrich the concentration of targeted proteins, which, again, leads to the lack of effective calibration.

Large-Array Microwells (Microengraving)
Both conventional and microfluidic flow cytometry cannot evaluate the same single cells at multiple time points, which is of importance in immunology. To address this issue, Love et al. proposed the design of large-array micro wells ("microengraving", 0.1-1 nL each) to isolate individual cells and quantify secreted proteins in a time sequence [53][54][55][56][57][76][77][78][79][80][81][82]. As shown in Figure 4, single cells, suspended in media, are deposited onto a large array (~20ˆ50 µm 2 ) of PDMS microwells. With the removal of excess cells, the microwells are then inverted onto a glass slide coated with a specific capture reagent. After an incubation period, the microwells are removed and applied to a second glass slide coated with a different capture reagent. The resulting microarrays are interrogated with laser-based fluorescence scanners.
After microengraving was initially proposed to quantify cellular protein expressions [76], several technical improvements were further realized, which are (1) the coating of specific macromolecules on the surface of PDMS wells to stimulate immune cells [54]; (2) the usage of the hybridization chain reaction to amplify signals resulting from sandwich immunoassays [55]; and (3) the retrieval of targeted single cells in specific micro wells [77].
Although microengraving can enable absolute quantification of multiple proteins secreted by individual immune cells, compared to flow cytometry, it requires the careful manipulation of glass slides without stimulating or even dislodging single cells within individual wells. In addition, it is also questionable that the micro environments formed within each microwell may be different from in vivo situations and, thus, the characterized protein levels cannot reflect the in vivo metabolisms of immune cells.

Large-Array Micro Chambers (Barcoding Microchips)
Although the microengraving setup can enable absolute quantification of secreted proteins of single cells, this approach cannot be effectively used to measure cytosolic proteins. To address this issue, Heath et al. proposed large-array micro chambers (single-cell barcoding microchips), which can be used to assay both cytosolic and membrane proteins of single cells. As shown in Figure 5, the single-cell barcoding microchips consist of thousands of individually addressed micro chambers to conduct single-cell trapping, lysis, capture of targeted proteins by pre-printed antibodies on the surface of the chambers, enabling quantitative measurement using fluorescent immunosandwich assays [58][59][60][83][84][85][86][87][88][89][90][91][92][93][94][95] (see Figure 5).
The key feature of this technique is the preparation of spatially-encoded antibody barcodes enabling the device multiplicity. Since antibodies suffer from long-term storage issues, the barcodes were initially patterned as ssDNA barcodes. A cocktail of antibodies labeled with complementary DNA oligomers were then used to transfer the DNA barcodes into antibody barcodes, just prior to running experiments. This approach has been used to monitor multiple key cytokines of single cells, locating significant cellular heterogeneities in both immunology [58] and tumor biology [59,88].
This approach is the most powerful microfluidic system enabling the absolute quantification of both surface and cytosolic proteins of single cells in large arrays. However, there are also several practical concerns. First, this microfluidic system is a leaky system and when running experiments, large portions of single cells may stick to the bottom surfaces of cellular inlets. This limitation renders the quantification of cell types with limited numbers (e.g., circulating tumor cells) full of challenges. The second concern is the potential denature of targeted antibodies in the step of cellular lysis. In addition, to form individual chambers for single-cell analysis, it cannot be easily further scaled up to increase the throughput. Since this system requests complicated operation steps to form individual chambers for single-cell analysis, it cannot be directly scaled up to further increase the throughput.

Large-Array Micro Chambers (Barcoding Microchips)
Although the microengraving setup can enable absolute quantification of secreted proteins of single cells, this approach cannot be effectively used to measure cytosolic proteins. To address this issue, Heath et al. proposed large-array micro chambers (single-cell barcoding microchips), which can be used to assay both cytosolic and membrane proteins of single cells. As shown in Figure 5, the single-cell barcoding microchips consist of thousands of individually addressed micro chambers to conduct single-cell trapping, lysis, capture of targeted proteins by pre-printed antibodies on the surface of the chambers, enabling quantitative measurement using fluorescent immunosandwich assays [58][59][60][83][84][85][86][87][88][89][90][91][92][93][94][95] (see Figure 5).
The key feature of this technique is the preparation of spatially-encoded antibody barcodes enabling the device multiplicity. Since antibodies suffer from long-term storage issues, the barcodes were initially patterned as ssDNA barcodes. A cocktail of antibodies labeled with complementary DNA oligomers were then used to transfer the DNA barcodes into antibody barcodes, just prior to running experiments. This approach has been used to monitor multiple key cytokines of single cells, locating significant cellular heterogeneities in both immunology [58] and tumor biology [59,88].
This approach is the most powerful microfluidic system enabling the absolute quantification of both surface and cytosolic proteins of single cells in large arrays. However, there are also several practical concerns. First, this microfluidic system is a leaky system and when running experiments, large portions of single cells may stick to the bottom surfaces of cellular inlets. This limitation renders the quantification of cell types with limited numbers (e.g., circulating tumor cells) full of challenges. The second concern is the potential denature of targeted antibodies in the step of cellular lysis. In addition, to form individual chambers for single-cell analysis, it cannot be easily further scaled up to increase the throughput. Since this system requests complicated operation steps to form individual chambers for single-cell analysis, it cannot be directly scaled up to further increase the throughput.

Discussions and Future Development
In this study, we summarize key developments of microfluidic platforms enabling the quantification of single-cell proteins. Although significant improvements have been made within the last decade, there is still significant room for research since an ideal tool of single-cell protein quantification featured with absolute quantification, ultra-high sensitivity, and throughput is still not available.
Absolute quantification is a key requirement for single-cell protein assays. Without the capabilities of absolute quantification, the protein levels measured by different approaches cannot be effectively compared. For instance, microfluidic fluorescent flow cytometry can only provide semiquantitative results by reporting levels of fluorescent intensities of flowing single cells. Although calibration beads have been widely used in the quantitative flow cytometry, they can only, to an extent, produce reliable results in quantifying membrane proteins, while we still lack effective calibration beads for absolute quantification of intracellular proteins in flow cytometry.
As to the issue of ultra-high sensitivity, large-array microfluidic devices suffer from the issue of sample dilutions since the reaction volumes for single cells are roughly 0.1-1 nL, two or three orders higher than the volumes of single cells (~1 pL). Further volume decrease in the trapping wells or chambers is not appropriate since it may lead to higher percentages of empty wells without cells and

Discussions and Future Development
In this study, we summarize key developments of microfluidic platforms enabling the quantification of single-cell proteins. Although significant improvements have been made within the last decade, there is still significant room for research since an ideal tool of single-cell protein quantification featured with absolute quantification, ultra-high sensitivity, and throughput is still not available.
Absolute quantification is a key requirement for single-cell protein assays. Without the capabilities of absolute quantification, the protein levels measured by different approaches cannot be effectively compared. For instance, microfluidic fluorescent flow cytometry can only provide semi-quantitative results by reporting levels of fluorescent intensities of flowing single cells. Although calibration beads have been widely used in the quantitative flow cytometry, they can only, to an extent, produce reliable results in quantifying membrane proteins, while we still lack effective calibration beads for absolute quantification of intracellular proteins in flow cytometry.
As to the issue of ultra-high sensitivity, large-array microfluidic devices suffer from the issue of sample dilutions since the reaction volumes for single cells are roughly 0.1-1 nL, two or three orders higher than the volumes of single cells (~1 pL). Further volume decrease in the trapping wells or chambers is not appropriate since it may lead to higher percentages of empty wells without cells and also affect the normal metabolisms of individual cells. To address this issue, fluorescent nanoparticles with higher fluorescent intensities may be used to replace conventional fluorescent probes. In addition, antibodies labeled with nuclear acids, which can be further amplified and quantified in digital PCR, can be further incorporated in the field of single-cell protein quantification.
The third issue is the throughput. Although the current microfluidic devices are treated as high-throughput approaches (e.g.,~1000 cells per second for flow cytometry), they are still not high enough. Taking the study of tumor heterogeneity as an example, individual cells within biopsy samples (~1 million cells) are preferred to be processed one by one. For the microfluidic flow cytometry, which inherently functions as a serial processor, further development should extend the capabilities to parallel processing. As for microfluidic large-array devices which are featured with parallel analysis of single cells, more work should focus on the device scale-up issue, definitely, requiring extensive optimization.