Single-nuclear RNA sequencing of endomyocardial biopsies identifies persistence of donor-recipient chimerism with distinct signatures in severe cardiac allograft vasculopathy

Cardiac allograft vasculopathy (CAV) is the leading cause of late allograft failure and mortality after heart transplantation (HT).¹ Histologically, CAV is chronic vascular rejection characterized by diffuse intimal thickening of macro- and microvasculature. While in vitro cellular models and in vivo histologic observations suggest coordinated responses of endothelial, fibroblast, and smooth muscle cells in CAV pathology, cell-specific transcriptional signatures among these in the transplanted human heart have not been studied. As current standards of diagnosis and treatment of CAV have significant limitations, understanding cell-specific responses may prove critical for developing improved detection strategies and novel therapeutics.

Recipients of HT undergo routine endomyocardial biopsies (EMB) for rejection surveillance, offering a unique opportunity to perform deep molecular phenotyping of human myocardium directly. Here, we demonstrate the feasibility of performing single nuclear RNA-sequencing (snRNA-seq) of human EMBs obtained during routine clinical practice. We compared tissue obtained at the time of re-transplantation from 4 individuals with severe CAV to EMB specimens from 3 individuals post-transplant without CAV (Figure 1A). Samples were obtained from the right ventricle (RV). In 3 out of 4 patients with severe CAV, left ventricular (LV) samples were also obtained (for a total 10 samples).

Figure 1.

A) Clinical characteristics of all seven patients studied. None of the patients had clinically-meaningful rejection at the time of sample acquisition. B) Uniform Manifold Approximation and Projection (UMAP) of 62,465 nuclei identified 17 major cell types using canonical marker genes. C) Cell compositional analyses were performed using scCODA v0.1.6. Labels correspond to patients described in Figure 1A. No significant difference in cell composition was noted using an automated reference cluster. D) Differential gene expression was performed using MAST. Volcano plot representing right ventricular CAV vs. control samples for the endothelial cell cluster. E) Biological pathway enrichment analysis of differentially upregulated genes in RV CAV endothelial cells using Metascape. F) Genotype-free inference of donor- versus recipient-derived nuclei was performed using souporcell v2.0. UMAP of donor- vs. recipient-derived nuclei. The clusters correspond to the same clusters annotated in Figure 1B. G) Donor-derived endothelial cells are enriched for markers of endothelial-to-mesenchymal transition. H) The monocyte/macrophage cluster is largely recipient-derived. Presence of distinct CCR2⁺ monocytes and CCR2⁻MCR1⁺ macrophages is highlighted using Nebulosa. LVEF = left ventricular ejection fraction; HTN = hypertension; DM2 = type 2 diabetes mellitus; CKD = chronic kidney disease; DSA = donor-specific antibodies; ACR = acute cellular rejection; AMR = antibody-mediated rejection; ER = extended release; MMF = mycophenolate mofetil; RV = samples from right ventricle; LV = samples from left ventricle; CAV = cardiac allograft vasculopathy; EndoMT = endothelial-to-mesenchymal transition.

Approximately 3–10 mg of EMB tissue and 25 mg of explanted tissue were mounted onto a Cryostat, followed by sectioning into 100 μm sections. The tissue was then disrupted by several rounds of dounce homogenization with loose and tight pestles before sequential filtration of debris (100um, 20um). Liberation of nuclei was monitored over the course of douncing to minimize nuclear damage. Nuclei were transferred to a cold resuspension buffer and counted on a hemocytometer before being loaded into a 10X Genomics single cell 3’ v3.1 platform for an estimated recovery of 8,000 nuclei per sample. Libraries were processed according to manufacturer’s instructions, with modifications as described previously.² Illumina sequencing targeted 25,000 reads per nucleus. Raw FASTQ files are deposited at the NIH NCBI GEO data repository (GSE203548). Detailed nuclear isolation methods and code used for analyses are deposited at https://github.com/learning-MD/CAV. This study was approved by the Vanderbilt University Medical Center’s Institutional Review Board.

We successfully isolated 62,465 nuclei and identified 17 major cell types with heterogenous distribution across the ten different samples (Figures 1B, 1C). When comparing RV samples, endothelial cells and fibroblasts in CAV exhibited increased expression of SERPINE1, which promotes neointimal hyperplasia and fibrosis. Endothelial cells were enriched for pathways involved in angiogenesis, cell migration, and extracellular matrix organization (Figures 1D, 1E). We identified similar cell-type specific gene expression changes across multiple clusters, along with a diverse array of immune cell types.

We inferred donor- and recipient-derived nuclei from each individual CAV sample using genetic variants detected in snRNA-seq reads. Using 5 of the 7 combined CAV samples (including both LV and RV tissue), 2,827 nuclei were confidently called as donor- or recipient-derived in the absence of genotyping (Figure 1F). Endothelial cells exhibited significant donor-recipient chimerism (21.8% recipient-derived). Interestingly, donor-derived endothelial cells were enriched for markers of endothelial-to-mesenchymal transition (EndoMT; SERPINE1, VIM, COL3A1; Figure 1G). In contrast, immune cells were largely replaced by those originating from the recipient (91.1% of macrophages/monocytes, 92.6% of NK cells, and 88% of T cells were of recipient origin). Recipient-derived macrophages included CCR2⁺ monocyte-derived macrophages and CCR2⁻ MRC1⁺ tissue resident macrophages, traditionally thought to be involved in cardiac repair³ (Figure 1H). Macrophages exhibited markers of activation, including HLA-DRA and CD74, and increased expression of TGFB1, a potential driver for the EndoMT observed in donor-derived endothelial cells.

This study is the first to successfully demonstrate feasibility of using human EMBs to perform snRNA-seq. While findings are those expected from an ischemic allograft, of relevance to the larger scientific community is the methodology for isolating nuclei from EMBs. We highlight several unique findings: 1) cell composition amongst EMB samples is highly heterogeneous and bulk RNA-seq approaches may exhibit high levels of variability due to sampling bias; 2) there are unique transcriptomic signatures of donor- versus recipient-derived cells highlighting putative novel avenues for investigation; and 3) the presence of recipient-derived CCR2⁻ macrophages warrants further study, as only a small percentage would be expected to be recipient-derived.³ It is possible that the CCR2 transcript may be undetectable in some cells. However, recent data suggest partial replacement of MHC-II^hiCCR2⁻ cardiac macrophages by monocytes, highlighting a still evolving understanding of macrophage subsets.⁴

Our study is limited by a small sample size and the use of samples derived from severe CAV. Importantly, however, these data demonstrate feasibility of performing snRNA-seq using frozen EMBs, presenting a unique opportunity that may have broad ramifications on the fields of heart transplantation and cardio-oncology/immunology. These data also lay the groundwork for ongoing experiments to study serial, routinely-collected EMB specimens after heart transplantation to identify novel biomarkers and pathways through which early CAV pathogenesis can be interrupted, thereby prolonging allograft survival.