Enhancing single-cell western blotting sensitivity using diffusive analyte blotting and antibody conjugate amplification

Abstract

While there are many techniques to achieve highly sensitive, multiplex detection of RNA and DNA from single cells, detecting protein contents often suffers from low limits of detection and throughput. Miniaturized, high-sensitivity western blots on single cells (scWesterns) are attractive since they do not require advanced instrumentation. By physically separating analytes, scWesterns also uniquely mitigate limitations to target protein multiplexing posed by affinity reagent performance. However, a fundamental limitation of scWesterns is their limited sensitivity for detecting low-abundance proteins, which arises from transport barriers posed by the separation gel against detection species. Here we address sensitivity by decoupling the electrophoretic separation medium from the detection medium. We transfer scWestern separations to a nitrocellulose blotting medium with distinct mass transfer advantages over traditional in-gel probing, yielding a 5.9-fold improvement in limit of detection. We next amplify probing of blotted proteins with enzyme-antibody conjugates which are incompatible with traditional in-gel probing to achieve further improvement in the limit of detection to 10³ molecules, a 520-fold improvement. This enables us to detect 85% and 100% of cells in an EGFP-expressing population using fluorescently tagged and enzyme-conjugated antibodies respectively, compared to 47% of cells using in-gel detection. These results suggest compatibility of nitrocellulose-immobilized scWesterns with a variety of affinity reagents — not previously accessible for in-gel use — for further signal amplification and detection of low abundance targets.

Introduction

Variability among single cells (heterogeneity) is a fundamental property of cell populations. Cell functional states change rapidly, especially during development and differentiation ^1,2. Changes in cell function affect nearby cells, leading to development of heterogeneous tissues from initially identical cells ^2–4. Disease is also characterized by changes in cell state heterogeneity that affects its progression and complicates development of disease models and treatments ^3,5–7. Proteins and their post-translational modifications are known to regulate cell functional states more proximally than mRNA and thus provide more information about cell fate decisions than single-cell transcriptomics ^8,9. However, quantifying single cell proteome states remains a significant technical challenge in large part because proteins and PTMs — unlike mRNA — cannot be amplified to aid detection. Bridging this capability gap between single-cell proteomics and transcriptomics would enable higher-resolution functional mapping of single cells.

Two main groups of methods are currently employed in single-cell proteomics, including highly sensitive antibody-based methods and highly-multiplexed mass spectrometry ^10–17. Among antibody-based methods, microfluidic assays that run on open-faced hydrogels have significant advantages in simplicity and modularity, enabling several analytical advances^18–20. Hydrogel matrices can sieve, concentrate, or immobilize proteins, enabling quantitation of species in downstream assay steps. However, hydrogel properties favorable for one step are often not favorable for others and can create conflicts of performance between them. For example, in single-cell western blotting (scWesterns), cells are loaded into microwells stippled into a thin polyacrylamide (PA) gel, lysed, and their protein contents separated by electrophoresis¹⁸. Proteins are then covalently photocaptured in the gel using photoactive benzophenone-methacrylamide (BPMAC) co-monomers and probed using fluorescently-labeled antibodies^21,22. The protein capture step fixes analytes in place, avoiding a rapid erosion of spatial resolution due to diffusion. However, this integration of separation and in-gel detection creates mass-transfer conflicts between analyte sieving during separation, which requires gel pore sizes on the order of protein sizes, and in-gel probing, which favors larger pore sizes to increase the penetration of antibodies^18,23. Small pore-size polyacrylamide gels tend to prevent antibody penetration due to partitioning, which limits their in-gel concentration at equilibrium^23,24. This problem is exacerbated when using bulkier detection reagents such as antibody-enzyme conjugates that could otherwise add sensitivity or analyte multiplexing benefits. Even for fluorescently-labeled antibodies, ~10-fold higher free-solution concentrations are required compared to traditional membrane immunoblotting to achieve sufficient in-gel concentrations¹⁸. This in turn can result in relatively high background staining of gels due to physical trapping of unbound antibody, limiting assay performance and increasing cost²⁵. Diffusional resistance can be circumvented using out-of-plane electrotransfer of antibodies²⁶ or chemical modification of gel pore size (including between assay steps)^27,28. However, each of these require significant increases in assay complexity and likely still limit compatibility of scWesterns with alternative detection schemes. Therefore, significant potential remains to circumvent in-gel probing in scWesterns while retaining the ability to ‘lock in’ spatial resolution of separated proteins on the microscale.

Here we replace in-gel probing with an alternative diffusive transfer step of analytes to a large pore-size nitrocellulose membrane better suited for antibody probing (Figure 1). This not only significantly increases antibody probing sensitivity but also allows the use of enzyme-antibody conjugates for amplified target detection, which has not been possible in polyacrylamide gels. We find that nitrocellulose blotting of scWestern separations operates at >90% transfer efficiency for a wide range of protein molecular weights (30-150 kDa) and in an immobilization-dominated regime relative to diffusion. This better limits band spreading during blotting relative to traditional in-gel scWesterns. We find a 5.9-fold improvement in the limit of detection for EGFP quantitation on paper relative to in-gel quantitation using primary and fluorescently-labeled secondary antibodies. Additionally, we used a 10-fold lower antibody concentration, thus reducing non-specific background and assay cost. Leveraging the relatively accessible pore sizes of nitrocellulose that allows for bulkier detection species, we assayed substrate turnover from enzyme-linked antibodies to achieve 520-fold improvements in limit of detection, down to 10³ molecules. These improvements were sufficient to detect a > two-fold wider range of expression heterogeneity in EGFP from HEK293 cells than achievable in polyacrylamide gels.

Figure 1: Diffusive blotting to porous nitrocellulose paper for single-cell western blotting with enhanced sensitivity.

Stage 1 - scWesterns separate proteins from polyacrylamide microwell-localized single cells or purified proteins by non-reducing SDS-PAGE (here EGFP, 27 kDa; tdTomato, 55 kDa from a microwell containing a single lysed HEK293 cell). Stage 2 - In traditional scWesterns, analytes are photocaptured inside the sieving gel. Here we instead perform blotting to nitrocellulose paper by diffusion.

Experimental Section

Protein immobilization on nitrocellulose membranes and protein transfer efficiency.

1.5 μL of 5 μM GFP (27 kDa), Ovalbumin-AlexaFluor 555 (46 kDa), or IgG-AlexaFluor 555 (150 kDa) were spotted on a 7%T (total monomer) 3.3%C (weight % bisacrylamide crosslinker relative to total monomer) PA gel ^18,22,29, incubated for 5 min at room temperature and transferred upon contact with 0.2 μm pore-size nitrocellulose for 1-4 min. Fluorescence of protein spots on gel pre- and post-transfer was compared to determine transfer efficiency. See Supporting Information for detail.

Diffusive spreading of immobilized protein.

Diffusive spreading of single-cell proteins spatially separated by single-cell SDS-PAGE and immobilized on 1) BPMAC-functionalized polyacrylamide or 2) nitrocellulose via transfer from plain polyacrylamide gel for 1-4 min was quantified by increase in spot size. See Supplementary Information for detail.

Microwell concentration calibration.

PDMS microchannels were fabricated with channel height 30 μm (identical to polyacrylamide gel thickness) and adhered to glass slides (see Supplementary Materials). 10 μL of GFP-Alexa 555 at concentrations of 0.001, 0.01, 0.1, 1, 5, and 10 μM were added to microchannels and imaged on a confocal microscope to generate a calibration curve. Fluorescence was plotted against EGFP molecule number for calibration and quantification of limit of detection.

Protein partition coefficient measurement.

40 μL of 0.1 μM or 1 μM of EGFP-Alexa Fluor-555 in RIPA-like buffer (0.5% SDS, 0.1% v/v Triton X-100, 0.25% sodium deoxycholate in 12.5 mM Tris, 96 mM glycine, pH 8.3, 0.5× from a 10× stock) were added to a BPMAC-functionalized polyacrylamide gel with 20 μm-diameter wells, sandwiched in a hybridization cassette. The gel was then briefly rinsed, sandwiched with a glass slide, and EGFP fluorescence was measured in gel and in microwells on a confocal microscope and correlated. The partition coefficient was calculated from the slope of the best fit line and used to determine in-gel and in-well concentrations relative to EGFP solution concentrations.

Fluorescently-labeled antibody probing calibration.

On 1) scWestern gel; 0.5 μL of 0.01, 0.0193, 0.0373, 0.072, 0.1389, 0.2683, 0.5179, and 1 μM solutions of EGFP-AlexaFluor 555 in RIPA-like buffer were spotted and fixed by irradiating with 254 nm UV light (7 mW cm⁻²) for 60 s. Antibody probing with either 1:200 goat anti-GFP or 1:400 rabbit polyclonal anti-GFP antibody both at 0.05 mg ml⁻¹ for 2 hr was then performed at room temperature and appropriate secondary antibody-AlexaFluor 647 at 1:500 dilution at 0.004 mg ml⁻¹ for 1 hr at room temperature. On 2) nitrocellulose, 0.5 μL of the identical protein concentrations were spotted on scWestern gel and transferred to nitrocellulose for 3 min. The membrane was fixed, blocked and probed with antibodies as described above. See Supplementary Materials for details.

Enzyme-linked antibody probing calibration.

0.5 μL of 1 μM, 100 nM, 10 nM, 1 nM, 100 pM, 10 pM, 1 pM, 100 fM, and 0 μM EGFP in RIPA-like buffer were spotted on a 7% PA gel and transferred to nitrocellulose for 3 min. The membrane was fixed, blocked and probed with antibody as described above except secondary probing was with 1:10,000 dilution of horseradish peroxidase (HRP)-conjugated rabbit anti-goat secondary antibody. We then cut nitrocellulose with protein blots and placed them face down in 48-well microplate wells. Detection was performed via QuantaRed Chemifluorescence kit (ThermoFisher Scientific 15159) to convert ADHP substrate to resorufin in the presence of HRP and H₂O₂ applied as a 100 μL reaction cocktail per well. Resorufin fluorescence was detected by widefield microscopy. Slopes of fluorescence vs. time plots were used to create a calibration curve. LOD was determined as the lowest analyte concentration at which the fluorescence increase was greater than the increase in standard deviation of the blank in a reaction time equal to the time t_D that resorufin diffuses a characteristic distance of 100 μM (the approximate size of single-cell protein plumes in scWestern separations). See Supplementary Materials for details.

Single-cell primary and secondary antibody-based and enzyme-linked antibody protein detection.

We tested and compared 1) standard BPMAC in-gel immobilization and probing with primary and secondary fluorescently tagged antibody vs 2) the same for nitrocellulose blotting, 3) enzyme-linked antibody-based EGFP detected in scWestern blots bearing single-cell separations. For 1) BPMAC-immobilized single-cell proteins, gel was incubated with 1:20 polyclonal goat anti-GFP antibody for 2 hr at room temperature, washed and incubated with 1:50 dilution of donkey anti-goat Alexa-555 antibody in 1% BSA in TBST for 60 min at room temperature. For 2) nitrocellulose-based detection, nitrocellulose was probed with 1:200 primary antibody for 2 hr at room temperature, washed and incubated with 1:500 donkey anti-goat Alexa-555 antibody for 60 min at room temperature. For 3) Enzyme-linked detection on nitrocellulose in scWesterns, primary antibody probing was performed with 1:200 goat anti-GFP for 2 hr at room temperature, washed and probed with rabbit anti-goat secondary IgG antibody conjugated to HRP for 60 min at room temperature. 100 μL of QuantaRed kit was added to nitrocellulose bearing single-cell Western protein separations, sandwiched with a glass slide and signal imaged on a widefield microscope. We determined background-adjusted average fluorescence values of single cell EGFP spots immediately after immobilization. Next, we found background adjusted average fluorescence of corresponding ROI readouts for 1) BPMAC-immobilized protein secondary antibody signal, 2) nitrocellulose-immobilized secondary antibody signal and 3) enzyme-linked antibody resorufin signal increase. We defined LOD as 3.3 times standard deviation of background for 1) and 2) and 3.3 times standard deviation of blank sample at t_D for 3) enzyme-linked detection. We then plotted initial EGFP fluorescence values to corresponding readout values and determined % of cells below LOD. See Supplementary Materials for details.

Results and Discussion

Transfer of proteins from polyacrylamide gel to nitrocellulose is efficient and retains spatial resolution.

Previous scWestern implementations noted significant sensitivity barriers for in-gel antibody probing, since IgG species suffer from a 5.9x lower equilibrium concentration in 8%T polyacrylamide relative to free solution due to partitioning ¹⁸. Additionally, in-gel antibody diffusion required equilibration time of 4τ ~ 3 min during each probing and washing step ¹⁸. Diffusive transfer of protein species from polyacrylamide gels to blotting media has been reported for traditional separation assays^30–32. However, its performance has not been validated at spatial scales relevant to scWesterns. We reasoned that transferring protein species from single-cell separations to a nitrocellulose paper interface^33–36 would increase sensitivity by mitigating subsequent mass transfer limitations during probing. It would also enable alternative assay readout strategies such as enzyme-linked antibody detection that is not compatible with the restrictive pore sizes of polyacrylamide gels. Transfer to nitrocellulose paper would also simplify the scWestern workflow by allowing the use of traditional polyacrylamide gel formulations rather than those with photoactive capture functionality. Finally, analyte detection could achieve higher signal-to-noise ratios on-paper due to relatively low non-specific antibody binding ^37,38.

We began by performing scWestern separations of fluorescent EGFP and tdTomato proteins expressed in HEK 293T cells, finding that proteins appeared to blot successfully after immediately interfacing the separation gel with nitrocellulose (Figure 2A,B). Two performance parameters are crucial to high-fidelity blotting - transfer efficiency and spatial resolution. We first sought to evaluate transfer efficiency for purified proteins embedded in scWestern gels and diffusively transferred to a 0.2 μm pore size nitrocellulose membrane for 1-4 min (see Experimental Section). This time range bracketed a theoretical transfer time ⊤ ~ L²/D for a characteristic diffusion length equal to gel thickness (30 μm) and EGFP diffusivity of ~38 μm² s⁻¹ in 7%T 3.3%C PA gel, such that ⊤ ~ 24 s and 5⊤ ~ 2 min ^{22,23,39–41}. Next, we imaged the gel for residual protein and quantified transfer efficiency as:

$$\%transfer=100*\frac{AU{C}_{pretransfer,gel}-AU{C}_{posttransfer,gel}}{AU{C}_{pretransfer,gel}}$$ (1)

where AUC is background-subtracted fluorescence area under the curve for the ROI. Transfer efficiency exceeded 90% for ~150 kDa and smaller proteins after 4 min, with even the heavier IgG marker exceeding 80% efficiency after 2 min (Figure 2C). These data indicate efficient analyte efflux from the gel during nitrocellulose blotting for the studied range of protein molecular mass. This range spans the median protein molecular weight of ~50 kDa ⁴², predicting highly efficient transfer for many single-cell protein targets using the nitrocellulose blotting approach.

Fig. 2: Nitrocellulose blotting achieves high transfer efficiency and improved band dispersion compared to traditional scWesterns.

(A) Schematic and fluorescence micrographs of single-cell protein separation and blotting to nitrocellulose. (B) Fluorescence micrographs showing single-cell protein separations post-blotting. (C) Plot of protein spot transfer efficiency from gel to nitrocellulose paper for varying blotting times. (D) Loss in band resolution determined by increase in full-width at half-maximum (ΔFWHM) for pre- vs. post-blotted EGFP bands after single-cell separation. Experiment data are overlaid on simulation trajectories for different Damköhler numbers showing expected band dispersion and % of EGFP molecules immobilized over time.

One concern for microscale scWestern blotting relative to traditional blotting is that the rate of analyte diffusion may become significant relative to the rate of immobilization to the blotting medium at very small characteristic lengths. This would decrease separation resolution, reduce concentrations of target protein available for analysis due to diffusive band spreading, and negatively affect assay limit of detection. We therefore sought to quantify band spreading during transfer of EGFP bands from single-cell separations to nitrocellulose over a range of blotting durations (Figure 2D). The increase in full-width-at-half-maximum (FWHM) values was modest over the blotting time — generally less than 50 μm for initial EGFP band widths of ~100 μm. Spreading appeared to saturate after 3 min, implying that analyte immobilization was complete in that time. The kinetics of band spreading are a function of a Damkohler number^22,43,44 (Da = kL²/D), which represents the ratio of the characteristic timescales of diffusion (L²/D) and immobilization (1/k) where k is a first-order immobilization rate constant (s⁻¹), L is a characteristic diffusion length that we take to be the initial band width (FWHM), and D is diffusivity of EGFP in nitrocellulose (we assume similar to that in free solution at ~90 μm² s⁻¹, ref. ^22,39–41, see Supplementary Methods). To more quantitatively understand if band spreading is diffusion- or immobilization-dominated during blotting, we simulated 1D diffusion and immobilization of a Gaussian band of EGFP in paper (initial FWHM of 100 μm), ignoring mass transfer resistance from gel to paper:

For the free solution fraction:

$$\frac{\partial [{\mathrm{GFP}}_f]}{\partial t}=D\frac{{\partial }^2[{\mathrm{GFP}}_f]}{\partial x^2}-k\left[{\mathrm{GFP}}_f\right]$$ (2)

For the immobilized fraction:

$$\frac{\partial [{\mathrm{GFP}}_i]}{\partial t}=k\left[{\mathrm{GFP}}_f\right]$$ (3)

Where the subscripts f and i denote ‘free-solution’ and ‘immobilized’, respectively. Since the immobilization rate constant k is unknown, we simulated a range of values creating a corresponding range of Damkohler number values, where Da < 1 indicates kinetics are diffusion-dominated, and Da > 1 indicates kinetics are immobilization-dominated. We found that the EGFP single cell transfer data fell within a Damkohler number range of ~5-10 from the model, predicting favorable immobilization-dominated kinetics. In practice, we also fixed transferred proteins to the blotting medium using paraformaldehyde to preserve high transfer efficiency and protein spot bandwidth. We then performed a similar analysis comparing nitrocellulose blotting against BPMAC-mediated in-gel immobilization in traditional scWesterns (Figure 2D). In contrast to nitrocellulose blotting, transport appeared to be far more diffusion-dominated for BPMAC-mediated immobilization as indicated by higher band spreading. Sufficient sensitivity to analyze the spatial distribution of the bound EGFP fraction was only possible at 1 and 2 minutes due to band spreading and, potentially, UV-induced bleaching. To evaluate diffusion and immobilization kinetics in BPMAC-mediated protein capture, we created a second version of the model, where D was the diffusivity of EGFP in a 7%T scWestern gel (~38 μm² s⁻¹, ref. ^22,39–41). Despite its lower diffusivity in-gel, EGFP immobilization fell within a Damkohler number range of ~0.1-3, predicting greater diffusive band spreading relative to immobilization compared to nitrocellulose blotting. These data validate rapid and efficient immobilization of microscale protein bands to the nitrocellulose blotting medium, retaining spatial resolution in scWestern separations.

Nitrocellulose blotting increases detection sensitivity and reduces antibody use compared to in-gel immobilization.

Having validated favorable blotting performance, we sought to compare analyte detection sensitivity for in-gel probing vs. probing on nitrocellulose blots (Figure 3A). Direct comparison between the two methods for single cells is not possible since fluorescent fusion protein expression is heterogeneous in cell cultures. Instead, we mimicked immobilization from single cells in BPMAC gels vs. after transfer to nitrocellulose using 0.5 μl spots of a serial dilution of purified AlexaFluor 555- labeled EGFP. We labeled EGFP to increase its capture efficiency by BPMAC²² to make it as competitive as possible with nitrocellulose blotting. For BPMAC-based immobilization, equilibrated spots were photo-immobilized in BPMAC polyacrylamide scWestern gels through a photomask, antibody probed, and imaged. We then determined spot signal-to-noise ratio (SNR, see Supplementary Information) and plotted it against number of EGFP molecules determined by estimating in-gel [EGFP] using its partition coefficient (Figure S1) and taking a gel volume equivalent to that of a ~100 μm diameter single-cell spot. For nitrocellulose-based immobilization, spots equilibrated in plain polyacrylamide scWestern gels were transferred from the gel to nitrocellulose for 3 minutes before performing a similar analysis. Primary and fluorescently-labeled secondary antibody concentrations were optimized for the lower concentration range typical of traditional nitrocellulose blotting rather than the substantially higher concentrations (~10-100-fold) used in in-gel scWesterns¹⁸. We found an optical limit of detection (LOD) of ~10^6.5 molecules of EGFP for traditional in-gel detection, and ~10^5.7 molecules for detection on nitrocellulose; a 5.9-fold improvement in detection sensitivity using nitrocellulose blotting (see Supplementary Methods for calculation details). With a second primary anti-GFP antibody we obtained a similar LOD of 10^6.9 EGFP molecules for traditional in-gel detection and 10⁶ molecules for nitrocellulose detection (7-fold improvement) (Figure S2A). These data reveal significant improvement in detection sensitivity for EGFP blotted to nitrocellulose relative to photocapture and probing in-gel.

Figure 3: Nitrocellulose blotting increases assay sensitivity and enables enzyme-based amplification of scWestern readouts.

(A) Left, immunofluorescence micrographs of calibration spots of EGFP transferred to nitrocellulose or immobilized in-gel prior to probing with goat anti-EGFP primary and anti-goat Alexa-647 secondary antibody. Pixel intensity was log-transformed. Note that contrast settings are equivalent within each image row, but different between rows. Right, corresponding calibration curves. (B) Left, immunofluorescence kymographs of calibration spots of EGFP transferred to nitrocellulose and detected with HRP-conjugated secondary antibodies for enzymatic amplification. Dotted line shows initial protein spot boundary before resorufin diffusion. Inset shows kymographs for the first 10 s. Right, corresponding calibration curve compared to fluorescently-labeled secondary antibody detection. (C) Left, intensity of secondary antibody signal detecting EGFP bands vs initial immobilized EGFP intensity, log transformed, from single-cell separations on BPMAC-immobilized polyacrylamide gel. Top right, EGFP fluorescence micrograph immediately post immobilization in BPMAC gel. Bottom right, secondary antibody-Alexa 555 fluorescence micrograph for the same ROI. (D) Left, intensity of secondary antibody signal detecting EGFP bands vs initial immobilized EGFP intensity, log transformed, from single-cell separations on nitrocellulose. Top right, EGFP fluorescence micrograph immediately post immobilization in nitrocellulose. Bottom right, secondary antibody-Alexa 555 fluorescence micrograph for the same ROI. (E) Left, resorufin signal increase over t<t_D detecting EGFP bands vs initial immobilized EGFP intensity, log transformed, from single-cell separations on nitrocellulose. Top right, EGFP fluorescence micrograph immediately post immobilization in nitrocellulose. Bottom right, resorufin fluorescence micrograph for the same ROI at t = t_D.

Nitrocellulose blotting enables 520-fold improvement in sensitivity via enzyme-linked antibody detection.

Free from the limitations of in-gel probing, we next attempted to improve detection sensitivity. We employed HRP-conjugated secondary antibody and assayed for ADHP substrate (10-acetyl-3,7-dihydroxyphenoxazine) conversion to fluorescent resorufin in the presence of H₂O₂ to read out the assay directly on nitrocellulose paper bearing individual EGFP protein spots. The rate of resorufin fluorescence increase here is expected to be proportional to the local HRP concentration, given excess ADHP ^45–49. Calibration showed a dynamic range of ~10³-10⁸ molecules of EGFP (Figure 3B). Since the resorufin product is soluble, analyte detection is a race between resorufin production and diffusion, with characteristic times t_R and t_D respectively. In other words, successful analyte detection while retaining acceptable spatial resolution would only occur when t_R < t_D. Across the assay dynamic range, t_R (time for resorufin signal to exceed 3.3*S.D. of background) was smaller than t_D = 10.4 s estimated for resorufin to diffuse an equivalent distance to a ~100 μm band in scWestern applications (see Supplementary Information). These data show that enzyme-linked detection strategies yield favorable performance in scWestern assays down to a detection limit of ~10³ molecules of EGFP without significant loss in protein band resolution. This yields ~520 fold improvement in LOD over fluorescently-tagged antibody detection, enabling future application to detection of low abundance proteins. Additionally, we verified that enzyme-linked detection failed in scWestern gels (Figure S2B). Nitrocellulose blotting of single-cell protein separations may therefore enable a variety of amplification strategies that are incompatible with in-gel detection.

Nitrocellulose blotting improves measurement of single-cell protein expression heterogeneity.

To test the performance of our diffusive analyte blotting approach for single-cell protein detection, we assayed for EGFP in scWestern blots using fluorescently-tagged antibody detection and the enzyme-antibody amplification scheme (Figure 3C–E). We detected only 46.6% of EGFP+ cells previously FACS sorted for high EGFP expression in BPMAC-immobilized blots (70 of 150 cells). On nitrocellulose blots, assaying the same cell culture using fluorescently-labeled secondary antibody successfully detected EGFP in 86.5% of cells (138 of 160 cells). Finally, EGFP was detected in 100% of the cells assayed using enzyme-linked antibody amplification on nitrocellulose (160 of 160 cells). We note that the observed positive linear correlation between resorufin fluorescence increase and initial fluorescence is weaker than that detected by fluorescently-tagged antibodies on nitrocellulose (R² of 0.36 vs 0.77 respectively). We hypothesize that ADHP mass transfer limitations may contribute to this additional variability given that ADHP may become limiting in areas of scWestern blots with locally denser separations. While enzyme-antibody conjugate-based detection of single-cell proteins is highly sensitive here, further innovation may be needed to remove mass transfer limitations on local substrate concentration to increase assay linearity and quantitative capability. Overall, our data demonstrate successful application of alternative antibody conjugate-based detection in scWesterns. This innovation improved detection sensitivity and may offer advantages in analyte multiplexing over in-gel probing in future work.

Conclusions

We introduce diffusive scWestern blotting to nitrocellulose for microscale protein bands, improving assay sensitivity and adaptability to different detection schemes relative to traditional in-gel antibody probing. We found that nitrocellulose blotting has favorable analyte retention and band spreading characteristics, validated by reaction-diffusion modeling. Head-to-head comparison of detection limits for fluorescently-labeled antibody detection of EGFP revealed a 5.9-fold improvement using on-paper analyte blotting relative to in-gel photocapture. Nitrocellulose blotting allowed us to use 10-fold lower antibody concentrations due to the lack of antibody probe partitioning that limits in-gel immobilization. This significantly reduces background, increases signal-to-noise ratio and improves characterization of single-cell expression heterogeneity. Additionally, nitrocellulose-based immobilization significantly reduces assay complexity.

Limits of detection improved to as few as 10³ molecules using enzyme-antibody conjugates for detection, a 520-fold improvement in sensitivity. This amplification strategy had previously not been compatible with in-gel scWesterns due to the prohibitive pore-size of the combined separation/blotting medium. Transferring separations to a large pore-size medium may enable future application of a larger diversity of amplification strategies for in situ detection beyond those applied here, including chemiluminescence, rolling circle amplification, hybridization chain reaction, nanoparticle/quantum dot-based techniques etc. These may also ease mating of scWesterns to other detection approaches such as mass spectrometry¹⁶. Future work will aim to complement these advances to improve scWestern multiplexing, which would have transformative impacts in single-cell proteomics.