Multiomic analyses of healthy versus disease samples

To analyze the data generated from the above described workflow, t-SNE analysis was performed on four CLL samples and three healthy donor samples using mRNA expression, surface protein expression or a combination of both. The t-SNE plot generated based on AbSeq data showed higher level of separation among different samples compared to using RNA-Seq data alone. A combination of both mRNA and AbSeq data greatly improved the resolution of different samples. Four CLL samples formed distinct clusters individually whereas the healthy samples overlapped to a greater extent in the t-SNE plot, suggesting the clinical course of CLL patients can be heterogeneous. These results also highlight the power of multiomics in analyzing single-cell profiles.

Analysis of differentially expressed markers across healthy and CLL samples

To understand the markers underlying the distinct clustering between healthy and CLL samples, a differential expression analysis was performed using single-cell mRNA and AbSeq data. This analysis identified markers that were decreased in CLL samples compared to healthy samples (Group 1). Genes or proteins that were differentially increased in all CLL samples are listed in Group 2 (all CLL), while Group 3 lists those which are up in some CLL samples. CD5, the marker aberrantly expressed in CLL patients, was higher in CLL03 and CLL04 compared to other samples. Also, a subset of markers that were specifically associated with CLL04 (Group 4) was also discovered, such as CD11c. Interestingly, not all cells produce CD11c at the same level, suggesting intra-tumoral heterogeneity of CLL.

T cell analyses 

Distinct clusters are resolved for each sample and the resolution is equivalent for clustering based on protein only, mRNA only or a combination of both. The transiently stimulated cell clusters (Day 14) appear separately from both fresh T cells and chronically stimulated cells, suggesting a distinct signature, in contrast to previous observation, suggesting that they are phenotypically and functionally similar to fresh cells. Chronically stimulated cells (Day 3, 7 and 14) form clusters that are separate from fresh cells. The three samples form distinct clusters, suggesting they each have unique signatures.

Concordance of BD^® AbSeq Marker data with flow cytometry

The kinetics of protein expression as demonstrated by the percent of CD8 T cell expression markers and the respective level of these cell surface markers detected by both flow cytometry and AbSeq was consistent. These data strongly demonstrate the concordance and utility of both these technologies to study protein expression.

Characterization of Treg differentiation by Monocle analysis

Unsupervised high-dimensional data analysis was used to refine the identification of Treg subsets. The Monocle algorithm was used to perform single-cell trajectory analysis and define transitional states of differentiation processes. Based on the expression of 22 protein and 399 mRNA targets, eight distinct states were identified, likely representing distinct differentiation states of Tregs. The single-cell heatmap in the figure below (B) displays selected proteins and genes differentially expressed across the eight states. Cells belonging to each state were arranged, from left to right, based on progressive loss of CD45RA and upregulation of HLA-DR, to represent a progression from naïve to activated cells. In addition to the differential levels of CD45RA expression, these three states could be further discriminated because CD31 is exclusively expressed in state 7 cells, consistent with the signatures of RTEs. On the other hand, the exclusive expression of CD95 and concurrent downregulation of CD45RA in state 5 may suggest a transition from naïve to apoptosis-sensitive and activated cells.

Differences in major ILC subgroups revealed by multiomic analyses

Data analysis was initially performed by manually gating the three major groups of circulating ILCs conventionally defined as CD117− CD294− ILC1, CD117+/−CD294+ ILC2 and CD117+CD294− ILC progenitors/ILC3 (ILCP/ILC3). A differential expression analysis was performed between the three ILC groups using single-cell AbSeq and mRNA data and identified unique protein and mRNA signatures for each group. This analysis also revealed expected expression patterns for key transcription factors defining ILC groups. For example, the gene encoding T-bet (TBX21) was exclusively expressed in a subset of ILC1 cells, whereas GATA3 was expressed at highest levels in ILC2 cells. As recently reported, RORC, encoding RORγδ, was partially expressed by circulating ILC progenitors displaying an ILC3 phenotype (CD117+CD294− ILCP/ ILC 3 cells).

Unbiased cell clustering identifies multiple cell phenotypes across different samples 

t-SNE was generated using the multiomic data generated from the above experiment for data visualization and an unbiased clustering algorithm PhenoGraph, SeqGeq™ Software plug-in, for detection of cell populations. PhenoGraph identified 12 cell clusters across the four different tissue types. Of the 12 clusters, four showed differences in cell density and distribution in the adipose tissue of HFD mouse compared to control clusters 1, 5, 7 and 12. 

Monocle analysis of CD8+ T cells from HFD adipose tissue suggests that these cells underwent cell exhaustion

Specific analysis of CD8+ T cells in the adipose tissue using Monocle discriminated five developmental states. The CD8+ T cells from the control mouse belonged to states 1 and 2 whereas cells from the HFD mouse further followed a trajectory toward states 3, 4 and 5. States 2, 3, 4 and 5 from the HFD mouse presented an elevated expression of exhaustion markers such as CD279 (PD-1) protein and mRNA (Pdcd1) as well Tigit.