Mitochondrial variant enrichment from high-throughput single-cell RNA sequencing resolves clonal populations

Abstract

The combination of single-cell transcriptomics with mitochondrial DNA variant detection can be used to establish lineage relationships in primary human cells, but current methods are not scalable to interrogate complex tissues. Here, we combine common 3' single-cell RNA-sequencing protocols with mitochondrial transcriptome enrichment to increase coverage by more than 50-fold, enabling high-confidence mutation detection. The method successfully identifies skewed immune-cell expansions in primary human clonal hematopoiesis.

Figure 1.. Targeted enrichment of mitochondrial transcripts enables discrimination between genetic clones.

A. Schematic shows the procedures for lineage inference from single-cell transcriptomes using MAESTER. Following mRNA capture and whole transcriptome amplification, part of the cDNA is used for standard scRNA-seq, and another part is used for PCR-based enrichment of mitochondrial transcripts. 300 bp sequencing reads maximize mitochondrial genome coverage to call variants. B. Diagram depicts the circular mitochondrial genome with annotated features. The green triangles indicate where MAESTER primers bind. C. Barplot shows the number of mtDNA variants that were detected by ATAC-seq (blue bars) and their recovery by MAESTER (red line). DNA (ATAC) and RNA (MAESTER) were acquired from the same K562 cells using the 10x multiome workflow. D. Barplot shows coverage of the mitochondrial genome with and without amplification from Seq-Well libraries using MAESTER. Mean coverage of 2,482 K562 and BT142 cells is shown. Each of the transcripts (UMIs) was sequenced ≥3 times. E. UMAPs show detection of cell type-specific (top) gene expression from scRNA-seq and (bottom) homoplasmic mtDNA variants from MAESTER. F. Heatmap depicts separation of 1,523 K562 and BT142 cells (columns) based on six mtDNA variants (rows). Cell type annotation from scRNA-seq is shown on top. G. Correlation matrix shows cell similarity based on the allele frequencies of six homoplasmic variants (rows and columns depict 1,523 cells). Unsupervised clustering identified two clusters that correlate with cell annotations from scRNA-seq. H. Heatmap shows VAF of 21 mtDNA variants detected by MAESTER (rows) for 588 K562 cells (columns, 44.4% of all K562 cells) with informative variants. Homoplasmic K562 variant 2141T>C is shown for comparison. Heatmap is organized by clonal structure (Methods). For E-H, only cells with >3-fold coverage of the indicated variants are shown. SMART and NXT are specific primer binding sequences, SA: streptavidin, CB: cell barcode, UMI: unique molecular identifier.

Figure 2.. Genetic clones exhibit lineage bias in clonal hematopoiesis.

A. UMAP of all 9,346 cells profiled by 10x scRNA-seq from a bone marrow aspirate from an individual with clonal hematopoiesis. B. Heatmap shows VAF of 26 informative mtDNA variants detected by MAESTER and maegatk (rows) for 1,397 cells (columns) with at least 1% VAF for one of the 26 variants. Heatmap is organized by clones and sorted by clone size. Cell type is listed on the bottom by color according to legend in A. C. UMAPs display VAF in each cell for mtDNA variants 683G>A (left) and 2593G>A (right). D. Stacked bar graph of the number of cells in each clone, with cell type denoted by color according to legend in A. E. Schematic depicts multimodal analysis we performed on the same single cells. F. Plot shows cells (rows) in which both mtDNA and TRB clonal markers were detected. Clones are indicated by colors and defined by mtDNA variants and the TRB variable region, respectively. G. Confusion matrix shows concordance between mtDNA clones and TRB clonotypes. H. UMAPs of 8,382 T-cells in the clonal hematopoiesis sample, with state annotated by transcriptional signatures (left), and selected mtDNA clones (right). I. Heatmap of T-cells (columns) in the selected mtDNA clones with mtDNA VAF (top), TRB sequence (middle), and state-defining transcript expression (bottom). ARI: Adjusted Rand Index.