exRNA Atlas Analysis Reveals Distinct Extracellular RNA Cargo Types and Their Carriers Present across Human Biofluids

Abstract

To develop a map of cell-cell communication mediated by extracellular RNA (exRNA), the NIH Extracellular RNA Communication Consortium created the exRNA Atlas resource (https://exrna-atlas.org). The Atlas version 4P1 hosts 5,309 exRNA-seq and exRNA qPCR profiles from 19 studies and a suite of analysis and visualization tools. To analyze variation between profiles, we apply computational deconvolution. The analysis leads to a model with six exRNA cargo types (CT1, CT2, CT3A, CT3B, CT3C, CT4), each detectable in multiple biofluids (serum, plasma, CSF, saliva, urine). Five of the cargo types associate with known vesicular and non-vesicular (lipoprotein and ribonucleoprotein) exRNA carriers. To validate utility of this model, we re-analyze an exercise response study by deconvolution to identify physiologically relevant response pathways that were not detected previously. To enable wide application of this model, as part of the exRNA Atlas resource, we provide tools for deconvolution and analysis of user-provided case-control studies.

Keywords: ERCC; deconvolution; exRNA; exosomes; extracellular RNA; extracellular vesicles; lipoproteins; ribonucleoproteins.

Figure 1.. exRNA Atlas Resource

(A) Faceted charts for selecting exRNA profiles from the Atlas. The size of each slice (representing a profile count) has been log-transformed to aid usability. Any samples that have a health condition of “Unknown” are protected by the study design. RNA source categories follow the protocols established by the ERCC. (B) Bar charts describing the contents of the Atlas.

Figure 2.. Overview of exRNA Atlas Data Submission Process and exRNA Atlas Tools

(A) Workflow for submitting exRNA profiling data to the Atlas. Submissions consist of three different file types: data files (optional for qPCR), metadata files, and a manifest file. All files are processed through an FTP-based data submission pipeline, with exRNA-seq data being uniformly processed through exceRpt. After validation, processing, and deployment, all data and metadata are made available through the Atlas website. (B) DESeq2 analysis interface. The integrated DESeq2 tool allows users to discover differentially expressed miRNAs in Atlas data via pairwise differential expression analysis. Users can launch their own analyses via the results grid or view precomputed analyses via the Public Analysis Results page. (C) Pathway Enrichment via Pathway Finder on WikiPathways. Users can select miRNAs on the Atlas via their DESeq2 results or the Atlas census page and perform downstream pathway analyses via the Pathway Finder tool on WikiPathways. The Pathway Finder tool lists pathways that contain selected miRNAs and their targets. (D) PCA/tSNE tool interface. The integrated Dimensionality Reduction Plotting Tool allows users to visualize precomputed PCA/tSNE analyses on Atlas datasets. All analyses are available via the Public Analysis Results page. See also Figure S2. (E) BioGPS interface. Users can visualize individual miRNA read count expression via BioGPS for Atlas RNA-seq datasets. All Atlas BioGPS studies are available via the Datasets page. (F) Workflow for submitting data to a Genboree group for personal analysis. Submissions consist of exRNA-seq data files (FASTQ) and are processed via exceRpt on the Genboree Workbench. After processing is completed, results can be shared privately with collaborators. See also Figure S1.

Figure 3.. Description of Deconvolution Method

The exRNA expression profiles (transcript 1 to transcript m) of complex biofluid samples (S₁to S_n) are used as the input. The deconvolution algorithm estimates the exRNA expression profiles (transcript 1 to transcript m) of cargo profiles and the proportions (P₁ to P_k) of the cargo profiles in each sample (S₁to S_n) through an iterative algorithm for constrained matrix factorization using quadratic programming. The algorithm involves two steps, the first involving transformed transcript abundance values over an informative set of ncRNAs and the second step involving non-transformed abundance values over all ncRNAs (STAR Methods). See also Figure S3 and STAR Methods.

Figure 4.. Deconvolution of exRNA Atlas Datasets

Top self-heatmap represents the correlation scores of the 68 estimated cargo profiles, 2 HDV profiles, 2 LDV profiles, and 3 HDL profiles (75 total profiles) from the deconvolution of 21 individual analysis datasets in the exRNA Atlas. The top dendrogram represents the hierarchical clustering of the 75 profiles into 6 top-level clusters named Cargo Types: CT1, CT2, CT3A, CT3B, CT3C, and CT4. The table under Cargo Type names shows the biofluids where the cargo types are detected. The rows in the bottom of the figure show correlations of specific profiles (as indicated by the labels on the left) with the 75 profiles and Cargo Types. See also Figure S4, Figure S5, Figure S6, Figure S7, Table S1 and STAR Methods. *Fractions are deconvoluted profiles corresponding to that given fraction. Additional detail provided in Figure S5.

Figure 5.. Census Analysis of Abundant miRNAs

(A) Venn diagram representing the overlap between highly abundant miRNAs expressed greater than 10 mean RPMs in at least 50% samples for that biofluid within the Atlas. Black circle indicates the 44 miRNAs abundantly expressed within all 5 biofluids. Gray circle indicates the 50 miRNAs expressed within all biofluids except urine. (B) Heatmap representing the RNA expression (log10) level of the 94 highly abundant miRNAs across the predicted CT1–4. Color bar indicates if the miRNA was present in all 5 biofluid (black 44) or across 4 biofluids excluding urine (gray 50).

Figure 6.. Deconvolution Estimates Cargo Type Composition Among RNA Isolation Methods

(A) Top heatmap shows hierarchical clustering of per-sample proportions of CTs predicted through computational deconvolution of Plasma and Serum biofluid samples. The Biofluid and RNA Isolation Methods are color coded above the heatmap. Center heatmap shows hierarchical clustering results of sample gene expression profiles (clustering of miRNAs was performed; samples are ordered based on dendrogram of per-sample proportions). Groups 1–4, indicated to the right of the heatmap, are sets of miRNAs that are preferentially isolated by specific RNA isolation methods. Left heatmap shows expression profiles of the four CTs estimated through deconvolution across miRNA Groups 1–4. (B) Box-plot of per-sample proportions of four CTs estimated through computational deconvolution for all ten RNA isolation methods.

Figure 7.. Deconvolution of Exercise Case Study

(A) Heatmap representing the correlation between the 3 cargo profiles modeled for Atlas Dataset Accession ID: EXR-SADAS1EXER1-AN and the cargo profiles estimated from individual Atlas datasets that form the 6 CTs. (B) Difference in abundance of each cargo profile between baseline samples and post-exercise samples (*p-value = 0.001). See also Figure S7. (C) Number of differentially expressed miRNAs within each cargo profile. DESeq2 was used to identify differentially expressed miRNAs in the exceRpt-processed exercise dataset samples (Yellow circle: Shah et al., 2017). For methodological details see STAR Methods. (D) mirnaPath was used to identify pathway enrichment for miRNAs differentially expressed for each cargo profile. Yellow highlighted boxes indicate pathways related to energy metabolism. Pink highlighted boxes indicate pathways related to muscle contraction and cell motility. For methodological details see STAR Methods.