Parallel measurement of transcriptomes and proteomes from same single cells using nanodroplet splitting

Abstract

Single-cell multiomics provides comprehensive insights into gene regulatory networks, cellular diversity, and temporal dynamics. Here, we introduce nanoSPLITS (nanodroplet SPlitting for Linked-multimodal Investigations of Trace Samples), an integrated platform that enables global profiling of the transcriptome and proteome from same single cells via RNA sequencing and mass spectrometry-based proteomics, respectively. Benchmarking of nanoSPLITS demonstrates high measurement precision with deep proteomic and transcriptomic profiling of single-cells. We apply nanoSPLITS to cyclin-dependent kinase 1 inhibited cells and found phospho-signaling events could be quantified alongside global protein and mRNA measurements, providing insights into cell cycle regulation. We extend nanoSPLITS to primary cells isolated from human pancreatic islets, introducing an efficient approach for facile identification of unknown cell types and their protein markers by mapping transcriptomic data to existing large-scale single-cell RNA sequencing reference databases. Accordingly, we establish nanoSPLITS as a multiomic technology incorporating global proteomics and anticipate the approach will be critical to furthering our understanding of biological systems.

Fig. 1. Overview of the nanoSPLITS-based single-cell multiomics platform.

Schematic illustration showing the workflow including cell sorting, lysis, droplet merging/mixing, and droplet splitting for downstream scRNAseq and scProteomics measurement.

Fig. 2. Qualitative and quantitative assessment of nanoSPLITS for transcriptome and proteome measurements.

a Mean number of detected genes and proteins for each modality. Error bars indicate standard deviations ( ± s.d.). b CCO distributions of genes (scRNAseq) and proteins (scProteomics) identified in the single-cell data. c Distributions of the coefficients of variation (CV) for all genes and proteins with at least 2 observations across replicates. Indicated values represent median CVs, which are also shown at the center point within each distribution. Whiskers extending from the center point represent standard deviation ( ± s.d.). d The ratios of average protein abundances were calculated for comparisons between the different pooled cell samples. Experimental medians are indicated at the black crossbar while the theoretical ratio for each comparison is shown at the red dotted line within each boxplot. Boxplot boundaries extend from the first (25%) and third (75%) quartile, representing the interquartile range (IQR). Boxplot whiskers extend to 0.75*IQR while outliers are not shown. e Pearson correlation heatmap with clustering of transcriptomics and proteomics results. Only the proteomic data showed complete clustering of 11,3 and 1 cell samples, indicated by the inscribed white squares. Self-comparisons along the diagonal are excluded (white). For (a–d), and (e) sample size for all analyses were n = 6 for 11, 3, and 1 C10 cell, with the exception the 1 C10 cell group analyzed by scRNAseq which had an n = 7.

Fig. 3. Underlying cell phenotype signatures are maintained after nanoSPLITS.

a Pearson correlation heatmap with clustering of scRNAseq and scProteomic results for both single C10 (n = 26, paired) and SVEC (n = 23, paired) cells. b Histogram of mRNA-protein correlations for each gene quantified in both modalities with at least 4 observations (performed separately for C10 and SVEC cell types). Statistical testing was performed by a two-sided Wilcoxon rank-sum test. Solid and dashed line indicates median of randomized and experimental correlations, respectively. c Top 5 gene markers from scRNAseq data and protein markers from scProteomics data for each cell type. Candidate marker features were determined using a two-sided Wilcoxon rank-sum test (adjusted p-values < 0.001). d Weighted-nearest neighbor (WNN) Uniform Manifold Approximation and Projection (UMAP) generated using Seurat to integrate scRNAseq and scProteomic data. Middle and right panels are colored based on H2-K1 gene (purple) and protein (red) abundance, respectively. e Feature-based UMAP generated for C10 cells using cell-cycle markers derived from scRNAseq data. Middle and right panels are colored based on Cdk1 gene (purple) and protein (red) abundances, respectively. All abundance values shown in (c–e) are Z-scores after scaling and centering of data.

Fig. 4. Protein, mRNA, and phosphopeptide cell cycle features from same single cells.

a Clustergram of log2 centered intensities for differentially abundant proteins from scProteomic data with FDR < 0.01 and log2FC of +/−0.5 (327 proteins). Columns (cells) are clustered by K-means (k = 2), while rows cluster proteins (k = 6). Colored areas along the y-axis indicate protein clusters. b Volcano plot of G2/M arrested cells/untreated cells for scProteomic data (3182 proteins quantified), scRNAseq data (4186 genes with < 50% missing values), and peptide level (16,938 peptides). Differential abundance statistical tests were performed using a two-sided, limma/Empirical Bayes t-test with p-values adjusted for multiple hypothesis tests. c, d Integration of scRNAseq, scProteomic, and phosphoproteomic data by comparing relative abundances from each modality for VIM and HNRNPU, respectively. Statistical differences were determined using a two-sided Wilcoxon rank-sum test. Adjusted p-values are presented above boxplots and “ns” indicates not significant. The center line in boxplots represents the median while the box boundaries extend from the first (25%) and third (75%) quartile, representing the interquartile range (IQR). Boxplot whiskers extend to 1.5*IQR while outliers are not shown. Sample sizes for each cell type were n = 36 for G2M arrested cells and n = 32 for untreated cells.

Fig. 5. Multiomics analysis of dissociated pancreatic islet cells.

a Left panel: UMAP of Azimuth human pancreatic reference map consisting of 35,289 cells. Cell types are indicated by color and inset text labels. Right panel: UMAP of 106 dissociated pancreatic islet cells from two human donors based on label transfer from Azimuth reference map to nanoSPLITS scRNAseq data. Cells with annotation scores of > 0.8 are shown. b Abundance heatmap of selected protein markers identified from nanoSPLITS scProteomics data for alpha cells (n = 64), beta cells (n = 30), delta cells (n = 12), immune cells (n = 4), and unclassified cells (n = 8), compared across all annotated cell types. Statistically significant markers were determined using a two-sided, limma/Empirical Bayes t-test (adjusted p-values < 0.01) or a hypergeometric test to identify markers based on missing values across cell types (adjusted p-values < 0.01). c Clustergram of Pearson correlations for secretory granule proteins hierarchically clustered for alpha (left panel) and beta cells (right panel). “Overlap” indicates if proteins were observed in both alpha and beta cell secretory granule clusters or only one cluster (alpha or beta).