The Macrophage Landscape Across the Lifespan of a Human Cardiac Allograft

Abstract

Background:

Much of our knowledge of organ rejection after transplantation is derived from rodent models.

Methods:

We used single-nucleus RNA sequencing (snRNA-seq) to investigate the inflammatory myocardial microenvironment in human pediatric cardiac allografts at different stages post-transplantation. We distinguished donor from recipient-derived cells using naturally occurring genetic variants embedded in snRNA-seq data.

Results:

Donor-derived tissue resident macrophages, which accompany the allograft into the recipient, are lost over time post-transplantation. In contrast, monocyte-derived macrophages from the recipient, populate the heart within days post-transplantation and form two macrophage populations, recipient MP1 and recipient MP2. Recipient MP2s have cell signatures similar to donor-derived resident macrophages, however, they lack signatures of pro-reparative phagocytic activity typical of donor-derived resident macrophages and instead express profibrotic genes. In contrast, recipient MP1s express genes consistent with hallmarks of cellular rejection. Our data suggest that recipient MP1s activate a subset of natural killer (NK) cells, turning them into a cytotoxic cell population via feed forward signaling between recipient MP1s and NKs.

Conclusions:

Our findings reveal an imbalance of donor-derived and recipient-derived macrophages in the pediatric cardiac allograft which contributes to allograft failure.

INTRODUCTION

After an organ transplant, innate immunity mediates allograft rejection by activating adaptive immunity, including alloreactive T cells. Macrophages are the main innate immune cells in the heart, displaying remarkable functional diversity that is closely related to their developmental origin. Lineage-tracing studies in mice established the existence of a dual origin of cardiac macrophages ^1–3. Monocyte-derived macrophages originate from adult hematopoietic progenitors and are replenished by ongoing monocyte infiltration into the heart. In contrast, tissue-resident macrophages are derived from embryonic hematopoietic progenitors and are thought to be maintained independently of monocyte input. Monocyte-derived macrophages orchestrate the inflammatory response by activating neutrophils and T cells, whereas tissue-resident macrophages suppress inflammation and promote tissue repair.

It is unclear whether the origin-coupled functional divergence of macrophages is conserved in the human heart; evidence is currently lacking. In addition, the precise roles of the macrophage subpopulations in pediatric cardiac allografts have not been described. The early post operative period remains an unknown period for changes in macrophage populations, with donor macrophages and recipient macrophages co-existing. Our understanding of the changes in various macrophage populations in the immediate post-transplant period is especially poor due to lack of human heart tissue during that period. Furthermore, to understand the immune response to an allograft at the molecular level, we must be able to accurately classify donor- and recipient-derived macrophages in the transplanted heart. Despite previous efforts to differentiate donor-derived cells from recipient-derived cells in human heart transplants ⁴, no reliable method has previously been used to achieve this in an unbiased and high-throughput manner. Single-nucleus RNA sequencing (snRNA-seq) not only allows quantification of gene expression, but also for identifying of naturally occurring small nucleotide variances (SNVs) in each cell. In practice, SNVs have been successfully used to assign identity to individual cells in a multiplexed human samples ^5,6. A transplanted heart is a naturally occurring tissue source where cells from two different individuals co-exist, making a SNV based identification of cell identity potentially feasible.

In this study we performed snRNA-seq to profile the transcriptional landscape of transplanted pediatric hearts across a wide time span. We were able to differentiate donor-derived cells and recipient-derived cells within the transplanted human heart. We were then able to delineate changes in the donor and recipient macrophages within the heart. These findings establish the landscape of macrophage diversity in pediatric cardiac allografts.

METHODS

Research ethics for donated tissues

Cardiac tissues samples used in this study were collected during cardiothoracic surgeries performed at Texas Children’s Hospital (Houston, Texas). The protocols for the procurement and use of these patient samples were approved by the Institutional Review Board for Baylor College of Medicine and Affiliated Hospitals (Protocol Number H-26502). With the help of the Heart Center Biorepository at Texas Children’s Hospital, consent was obtained from patients undergoing repeat heart transplantation. The anatomic location of tissue collected is left ventricle apex. Detailed patient information can be found in Supplementary Materials.

Sample Collection and Nuclear Isolation

Cardiac tissue was collected in the operating room during repeat heart transplantation. Cardiac tissue samples were kept in cold saline on ice during transfer to the laboratory for preservation. Cardiac tissue samples were carefully dissected into multiple aliquots, which were flash-frozen and stored at –80 °C Nuclear isolation was performed as described previously ⁷. Briefly, frozen cardiac tissue was dissociated by using a Dounce homogenizer. Single nuclei were isolated via fluorescence-activated cell sorting (FACS).

Single-nucleus RNA Sequencing

SnRNA-seq was performed by using the 10X Genomics platform. Isolated nuclei and cells were loaded into the 10X Genomics Chromium Controller to obtain the gel beads in emulsion. The sequencing libraries were then prepared according to the manufacturer’s protocols for the Single-cell 3’ Reagents Kits v2 or v3. Sequencing was performed by using the NextSeq 500 and NovaSeq 6000 systems.

RESULTS

Immune Cells in Pediatric Cardiac Allografts

To examine the evolving landscape of immune cells in the cardiac allograft over time, we collected left ventricular tissue from cardiac allografts at the time of repeat heart transplantation from 3 patients: five days post-transplantation (5dPT), 15 months PT (15mPT), and 12 years PT (12yPT) (Table S1). Pathological analyses indicated that all three transplants exhibit mild acute cellular rejection (ACR 1R), and none show evidence of antibody-mediated rejection (AMR). Notably, no evident cardiac allograft vasculopathy (CAV) is observed in the 5dPT heart, while the 15mPT and 12yPT hearts present CAV with moderate to marked loose fibroinflammatory intimal hyperplasia, accompanied by biventricular chronic ischemic changes and fibrous expansion (detailed in Supplementary Information). Histological examination revealed high tissue quality and structural integrity (Figure S1A). We performed snRNA-seq profiling (Figure 1A and Table S1) on patient tissues and used available snRNA-seq datasets from 4 age-matched donor hearts as controls ^7,8. Principal-component analysis (PCA) of pseudobulk data indicated that when compared with age or sex, post-transplantation status had the greatest influence on gene expression variance (Figure 1B).

Figure 1. Pediatric Cardiac Allografts Harboring Donor- and Recipient-derived Immune Cells.

(A) Schematic illustrating the design of the snRNA-seq experiments. Graft biopsies were collected from three pediatric patients at 5 days (5dPT), 15 months (15mPT), and 12 years (12yPT) after undergoing transplantation. Tissues from the apical anterior wall of the left ventricles were obtained during repeat heart transplantation. (B) Principal component analysis (PCA) bubble plots of pseudobulk snRNA-seq data colored by individual patient (top), age (middle), and sex (bottom). The bubble size in the top plot represents the amount of quality-controlled single nuclei. (C) Global uniform manifold approximation and projection (UMAP) of an integrated dataset showing 75,941 nuclei combined from control hearts and 20,690 nuclei combined from transplanted hearts. (D) Global UMAP showing 9 distinct cardiac cell types based on unbiased clustering and reference-assisted annotation. (E) Cell composition of nonimmune, lymphoid, and myeloid cells at each time point. Each dot in a box represents the fold-change over one of the four controls. (F) Nuclei in the same global UMAP colored by their inferred donor or recipient cell origins. Inference of only a few nuclei was ambiguous. (G) Percentages of donor-derived lymphoid and myeloid cells captured in the allografts at different timepoints. PT: post-transplantation.

Using a comprehensive computational pipeline (Figure S1B), we processed the snRNA-seq data and performed rigorous quality control steps to filter for high-quality single nuclei (Figure S1C and S1D) and potential doublets (Figure S1E). Then, 96,631 high quality nuclei from control hearts and cardiac allografts we integrated using deep probabilistic modeling methods to correct for technical variations (Figure S1F and S1G). Finally, we visualized the integrated nuclei in a 2-dimensional uniform manifold approximation and projection (UMAP) (Figure 1C). Unbiased clustering and annotation revealed nine cardiac cell types, including cardiomyocytes (CMs), fibroblasts (FBs), endothelial cells (ECs), pericytes (PCs), smooth muscle cells (SMCs), neuronal cells (NCs), and immune cells such as myeloid and lymphoid cells (Figure 1D). Cell type annotation was assisted by unsupervised mapping to a reference atlas ⁹ (Figure S1H) and confirmed by manual examination of marker gene expression (Figure S1I and Table S2). Compared with controls, allografts contained large populations of myeloid and lymphoid cells. Myeloid cells were increased by 3-fold and lymphoid cells by 10-fold over controls (Figure 1E).

Expressed variants in snRNA-seq distinguishes donor from recipient cells

To classify donor and recipient-derived cardiac cells using snRNA-seq data, we used Demuxlet, an algorithm for demultiplexing pooled scRNA-seq data to cluster and call cell origins based on known common genetic variations within sequencing reads ^5,6. To demonstrate the robustness of this technique, we have shown two variant alleles on NEAT1 and MALAT1 transcripts (Figure S1J). Classification of donor versus recipient derived cardiac cells was based on a consensus of ~3 million such expressed variants. Nearly all 20,690 cell nuclei from the allografts were clearly classified, except for only 27 nuclei that were labeled as ambiguous (Figure 1F, Table S3). As an internal control, cardiac cells (e.g., CMs, FBs, ECs, SMCs, PCs, and NCs) had the same, “donor” genotype. In contrast, most immune cells had an alternative, “recipient” genotype. As further proof of principle, using the 15mPT heart, a case of female donor-to-male recipient transplantation, we scored the cells according to male-specific Y chromosome gene expression. The score agrees with the inferred origin, providing further validation of our methods (Figure S1K). We also performed a parallel inference using Souporcell, a demultiplexing method capable of de novo variant detection ¹⁰. The independently inferred cell origins were highly concordant, with an overall agreement of 97% (Figure S1L) revealing that snRNA-seq effectively distinguishes donor-derived cells from recipient-derived cells. Distinguishing donor and recipient myeloid cells at single-cell resolution enabled us to further decipher the immune response of the cardiac allograft microenvironment.

At the 5dPT time point, our analysis of cell origins revealed that ~14% of myeloid cells and ~2% of lymphoid cells were donor-derived (Figure 1G). Around 86% myeloid cells in the 5dPT allograft were recipient-derived, which had rapidly infiltrated the transplanted allograft. In contrast, at 15mPT and 12yPT, essentially all cardiac immune cells were recipient-derived (Figure 1G).

Heterogeneous Macrophage Populations in Cardiac Allograft

To investigate myeloid cell heterogeneity within the allograft, we first focused on myeloid cells in the integrated data irrespective of cell origin. We subclustered myeloid cells into four populations based on their transcriptional signatures: CD16⁺ monocytes (Mono), IL1B⁺ macrophage cluster 1 (MP1), MRC1⁺ macrophage cluster 2 (MP2) and Mast cells (Figure 2A, Figure S2A, and Table S4). Next using inferred donor versus recipient cell origin, we found that donor-derived myeloid cells belong primarily to the MP2 cluster (Figure 2B), constituting 24% of total MP2s in the 5dPT allograft (Figure 2C). In all allografts, Mono and MP1 are almost exclusively recipient-derived. Combining both transcriptional signatures and cell origins, we conclude that the recipient Mono, recipient MP1, recipient MP2 and donor MP2 are the 4 major myeloid populations in a cardiac allograft (Figure 2D).

Figure 2. Recipient MP2s Incompletely Acquire the Tissue Resident Macrophage Cell State.

(A) A representative uniform manifold approximation and projection (UMAP) of the myeloid cell subset. Reclustered nuclei are colored by myeloid subtypes. MP: macrophage. (B) Myeloid cells at three timepoints after transplantation (5dPT, 15mPT and 12yPT) colored by donor or recipient origin. Dashed-line encircles the MP2 cluster. (C) Bar graph of Mono and MP cell counts, highlighting the substantial abundance of donor MP2s at the 5dPT timepoint and the lack of donor-derived cells in late-stage post-transplantation. ^3,4(D) Schematic illustrating the major classes of monocytes and macrophages in cardiac allografts, based on transcriptomic clustering and cell origin inference. (E) Box plot showing the expression z-scores of previously reported resident macrophage signature genes in monocytes and donor- and recipient-derived MP2s. (F) Violin plots showing expression levels of general MP2 genes MRC1, MERTK and CD163, donor MP2-specific genes LYVE1 and CYBB, and recipient MP2-specific gene CD163L1.

(G) Expression z-scores of MHC-I and MHC-II genes in the main classes of monocytes and macrophages from control and allografts. *: two-sided unpaired Student’s t test, p<0.001. n.s.: non-significant, adjusted p>0.05.

The Mono population most closely resembles circulating monocytes identified in human peripheral blood mononuclear cells (Figure S2B). In control pediatric hearts, we found myeloid cells were split between ~30% Mono and ~70% MP2 (Figure S2C), which is similar to the composition of adult hearts ¹¹. Monocytes significantly increased in the 5dPT allograft (Figure S2C), indicating that monocytes from the recipient infiltrate the allograft within the first 5 days post-transplantation.

Recipient MP2s Incompletely Acquire the Tissue Resident Macrophage Cell State

We scored all myeloid cells, regardless of site of origin, using defined markers ^3,4 and found that MP1s expressed monocyte-derived macrophages markers, whereas MP2s expressed tissue resident macrophage markers (Figure S2D). The combined MP2s expressed genes encoding the “TLF factors” (TIMD4, LYVE1 and FOLR2), which are established markers of tissue resident macrophages ³ (Figure S2E). In addition, MP2s had low or undetected expression of CCR2 and CD16, two markers of infiltrated monocyte-derived macrophages. These gene expression signatures support the conclusion that the MP2 cell state is closely related to the alternatively activated “M2” macrophages, which function in damage repair, immune surveillance, and tissue homeostasis ¹.

Direct comparison of donor MP2s and recipient MP2s revealed that donor MP2s expressed the highest levels of tissue-resident macrophage markers and lowest level of monocyte-derived macrophage markers (Figure 2E). In contrast, recipient MP2s expressed both tissue-resident macrophages markers and monocyte-derived macrophage markers revealing an intermediate cell state. Nonetheless, recipient MP2 robustly expressed MP2 genes (e.g., MRC1, MERTK, CD163), which are critical for tissue-resident macrophages function, at levels similar to donor MP2s (Figure 2F)^12,13. Interestingly, previously used tissue-resident macrophages marker genes showed differentially expression. For example, LYVE1 and CYBB are only highly expressed in donor MP2s but not recipient MP2s, and CD163L1 in recipient MP2s but not donor MP2s, indicating that these genes are likely linked with the developmental ontogeny of MPs (Figure 2F). Furthermore, recipient MP2s express higher levels of MHC-I and MHC-II than control MP2s or donor MP2s, suggesting that they present antigens similarly to monocytes (Figure 2E). Thus, although recipient MP2s are transcriptionally similar to tissue-resident macrophages, they lie within an intermediate cell state between tissue-resident macrophages and monocytes. We concluded that the donor MP2s identified in our dataset are bona fide tissue-resident macrophages, whereas recipient MP2s are a monocyte-derived reparative macrophage population.

Donor-derived Macrophages Fail to Self-Renew and are Replaced by Recipient-Derived MP2s in the Transplanted Tissue

In the 5dPT sample, the donor MP2 population constituted 14% of all myeloid cells in the allograft. However, in the 15mPT sample, donor MP2s constituted less than 2% of myeloid cells and remained low in the 12yPT sample (Figure 3A). A similar loss of donor MP2s over time was observed in heterotopic heart transplant tissues in mice ¹⁴. Conversely, the percentage of recipient MP2s increased over time after transplantation, eventually dominating the myeloid population by approximately 80% in the 12yPT sample (Figure 3A). To confirm the decreasing abundance of donor MP2 in the transplant tissue, we probed donor MP2-specific CYBB transcript with co-staining of general MP2 surface marker CD163. We detected abundant donor MP2 cells in the 5dPT but few in the 15mPT and 12yPT samples (Figure S3A and 3B). Consistently, we detected robust signals of recipient MP2-specific CD163L1 transcripts in CD163⁺ MP2s in the 15mPT and 12yPT tissues but not in the 5dPT tissue (Figure S3A and 3C). The estimated relative abundance of donor MP2s and recipient MP2s by staining showed highly agrees with that from the snRNA-seq data (Figure 3D).

Figure 3. Diminished donor MP2 replaced by recipient MP2s.

(A) The relative abundance of donor MP2s and recipient MP2s across the post-transplantation time points. (B and C) Representative images of RNA fluorescence in situ hybridization (FISH) for CYBB (B) and CD163L1 (C), co-stained with immunofluorescent antibody against CD163, DAPI and wheat germ agglutinin (WGA). Yellow arrowheads point to donor MP2s (CD163⁺CYBB⁺) (B) and recipient MP2 (CD163⁺CD163L1⁺) (C). Exemplary cells encircled by dashed boxes are shown in zoomed-in view. (D) Percentage of donor MP2s (CD163⁺CYBB⁺) and recipient MP2 (CD163⁺CD163L1⁺) quantified by FISH in allografts at each time point. Each data point represents a different tissue area. *: p<0.05, **: p<0.01, ***: p<0.001. P values determined by two-sided unpaired Student’s t test. (E) Distribution of proliferative MP2s from aggregated controls and each transplanted heart at each time points, quantified by expression of cell cycle-associated genes. (F) Cell-cell transition potential determined by RNA velocity overlayed on the UMAP of myeloid cells. The arrows illustrate the bifurcated transitions from monocytes to either recipient MP1s or recipient MP2s. (G) Aggregated expression pattern of a co-expressed gene set correlated with the monocyte-to-recipient MP2 differentiation trajectory. Each line represents the mean expression of a gene across monocytes and recipient MP2s, ranked by diffusion pseudotime. Six genes were highlighted as examples. (H) A dendrogram illustrating the similarity among donor MP2s and early and late recipient MP2s, measured by the distance between their transcriptomes. (I) Expression z-scores of reported tissue-resident macrophages markers (top) and the Mono-to-recipient MP2 differentiation trajectory genes shown in G in the donor MP2s and recipient MP2s across post-transplantation time points. *: two-sided unpaired Student’s t test, p<0.001.

Because local proliferation has been considered the major source of reparative macrophage persistence ^1,15, we scored all MP2s based on cell cycle gene expression. MP2 cell proliferation was nearly abolished in all allografts, which is in contrast to their substantial proliferation (~14%) under homeostasis (Figure 3E). The loss of self-renewal in recipient MP2s may have resulted from continuous immune suppression treatment, such as calcineurin inhibitors and steroids. The loss of donor MP2s in transplant tissues may have deleterious effects on allograft outcomes, as shown in a mouse model of heterotopic heart transplantation ¹⁴.

We speculated that the increase in recipient MP2s results from the differentiation of continuously infiltrating monocytes. Cell-specific RNA velocity estimates showed a clear continuum of cell state transitioning from monocytes to both recipient MP1s and recipient MP2s (Figure 3F). The starting point of the trajectory coincided with cardiac monocytes that were similar to circulating human monocytes, substantiating the accuracy of the trajectory inference (Figure S2B). To independently validate these findings, we inferred a pseudotemporal trajectory based on the low-dimensional projection of diffusion components (DC) ¹⁶. The resulting diffusion map recapitulated the bifurcating trajectories depicted by RNA velocity (Figure S3B). To describe the temporal transcriptional shift that occurred during the Mono to recipient MP2 cell polarization, we deconvoluted the upregulated gene expression as a function of progression through the inferred pseudotemporal axis. A module of 239 genes was identified that closely correlated with recipient MP2 differentiation (Figure 3G and Table S5). Many of these genes (e.g., RBPJ, F13A1, and MAF) have a critical influence on macrophage differentiation, maturation, and survival ^17–19. Genes that have yet to be functionally linked with cardiac macrophage physiology, such as FLI1, were also identified in this module.

We next examined whether recipient MP2s were transcriptionally similar to donor MP2s at different time points after transplantation. PCA showed that in contrast to early-stage recipient MP2s from the 5dPT sample, recipient MP2s from the 15mPT and 12yPT samples (referred to as “late-stage”) have transcriptional profiles more similar to that of donor MP2s (Figure 3H and S3C). Like donor MP2s, late-stage recipient MP2s also express high levels of cardiac resident MP marker genes and genes correlated with the Mono to recipient MP2 cell polarization trajectory (Figure 3I). We concluded that newly infiltrated monocytes polarize into recipient MP2s shortly after transplantation and within the first year progressively replaced donor MP2s.

Late-Stage Recipient MP2s Lack a Phagocytotic Signature and are Fibrogenic

Although the transcriptomes of late-stage recipient MP2s and donor MP2s have similarities, recipient MP2s expressed lower levels of MP2 polarization signatures and tissue resident macrophage markers compared to donor MP2s (Figure 3I). This suggests that late-stage recipient MP2s incompletely acquire the functions of donor MP2s. To address this further, we performed a focused analysis comparing recipient MP2s and donor MP2s (Figure 4A). Compared with donor MP2s, significantly upregulated genes in recipient MP2s include LYVE1, FOXO3, and ZBTB16, which are central to the reparative phenotypes and dampening of the inflammatory response (Figure 4B and Table S6) ^20–22. Upregulated genes in donor MP2s were indicative of active phagocytosis, vesicle organization, and dampened immune leukocyte activation (Figure 4C).

Figure 4. Late-Stage Recipient MP2s Lack a Phagocytotic Signature and are Fibrogenic.

(A) Split view of myeloid UMAP of two MP2 populations from all allografts: Donor MP2s (left) and Recipient MP2s (right). (B) Volcano plot highlighting the differentially expressed genes between donor MP2s and recipient MP2s. The Y-axis was capped at 20 as the maximum. (C) Gene ontology analysis showing enriched biological processes in donor MP2s. (D) Expression z-scores of phagocytosis pathway genes among donor MP2s and recipient MP2s. (E) Chord diagrams of predicted significant ligand-receptor interactions between macrophages and cardiac cells (CM, cardiomyocytes; EC, endothelial cells; FB, fibroblasts; PC, pericytes; and SMC, smooth muscle cells). Potential ligands sent from donor MP2s (left), recipient MP2s at 15mPT (middle), and recipient MP2s at 12yPT (right) and received by multiple cardiac cell types in the respective samples. The thickness of the chord within each signal group represents relative signaling strength. Signaling pathways known to induce cardiac fibrosis are highlighted in blue. (F) Box plot of fibrosis scores in cardiac fibroblast cells from controls and allografts. Score were calculated on the basis of the expression of previously reported cardiac fibrosis signature genes in the fibroblast populations. (G) Percentage of fibrotic tissue area quantified from the trichrome staining images shown in Figure S1A. *: two-sided unpaired Student’s t test p<0.01. (H) Representative images of RNA fluorescence in situ hybridization (FISH) for PDGFB, co-stained with immunofluorescent antibody against CD163, DAPI and wheat germ agglutinin (WGA). Pink arrows point to non-MP cells that express PDGFB (CD163^–) and yellow arrowheads point to PDGFB-expressing MPs (CD163⁺PDGFB⁺). Exemplary cells encircled by dashed boxes are shown in zoomed-in view. (I) Percentage of recipient MP2 (CD163⁺PDGFB⁺) quantified by FISH in controls and allografts at each time point. Each data point represents a different tissue area. *: p<0.05, **: p<0.01, ***: p<0.001. P values determined by two-sided unpaired Student’s t test.

Upregulated genes in recipient MP2s suggested the induced immune tolerance and the activation of a specific pathway of chondroitin sulfate (CS) proteoglycan metabolism (Figure S4A). CS has been implicated in the anti-inflammatory activity of bone–marrow–derived macrophages and has been shown to attenuate inflammatory processes by limiting NF-κB signaling ^23,24. Consistently, the expression scores of the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway genes corroborated that donor MP2s express significantly higher levels of genes involved in phagocytosis than do recipient MP2s (Figure 4D), whereas recipient MP2s express higher levels of genes involved in the CS glycosaminoglycan biosynthesis pathway (Figure S4B). These findings support the conclusion that recipient MP2s have an anti-inflammatory gene expression profile. Furthermore, because recipient MP2s constitute nearly the entire reparative MP2 population in the late-stage graft, they may functionally compensate for loss of donor MP2s. However, the elevated tolerance induction and decreased phagocytic capabilities of recipient MP2s suggest that these cells do not completely mirror the function of donor MP2s.

Donor MPs and Recipient MPs signal differently to the allograft myocardium

Because donor MP2s and recipient MP2s have signatures indicating they function differently in suppressing the inflammatory response, we reasoned that they also signal to the myocardium differently to modulate the myocardial microenvironment. CellChat, a cell-cell communication inference suite ²⁵, was used to model the significantly enriched ligand-receptor pairs between subsets of MP2s and other cardiac cell types. We found that both donor MP2s and recipient MP2s exhibit a wide spectrum of ligand-receptor communications with all major cardiac cell types, indicating a profound influence on the myocardial microenvironment (Figure 4E). In the 5dPT heart, outgoing signals from donor MP2s included FGF13, HBEGF, and VEGFB. Cells that received these signals included CMs, FBs, ECs, SMCs, and PCs (Figure 4E and Figure S4C). These signaling events were specific to donor MP2s and were not predicted as significant interactions initiated by recipient MP2s. Conversely, in both the 15mPT and 12yPT samples, we identified overlapping outward signals from recipient MP2s to the myocardium (Figure 4E). IGF1–IGF1R were predicted as a strong signal received by CMs (Figure S4D), which is consistent with previous reports showing that reparative macrophages secrete IGF1 to promote murine cardiomyocyte regeneration after injury ². Interestingly, many of the signaling interactions predicted in recipient MP2s are traditionally considered to be fibrogenic. Here, we highlighted several examples, including the PDGF-B ligand received by receptors PFGFRA and PFGFRB on fibroblasts and mural cells, and TGF-β received by fibroblasts expressing TGFBR1 and TGFBR2. WNT5B has been less studied in context of cardiac cells but is closely related to homologous WNT5A, which is a well-known pro-fibrogenic signal in the stressed and injured heart ^26,27. In late-stage transplants, the increase in fibrotic signaling contributed by the recipient MP2s coincided with a significant increase in the fibrotic signature genes expressed by cardiac fibroblasts (Figure 4F), as well as an increase of tissue fibrosis detected by trichrome staining (Figure S1A and 4G). To further corroborate the pro-fibrosis signals from recipient MP2s, we probed PDGFB transcript with co-staining of MP2 marker CD163. Because fibroblasts and endothelial cells also express PDGFB, we found specific PDGFB signal in both MP (CD163⁺) and non-MP cells (CD163^–) (Figure 4H). Quantification revealed that CD163⁺PDGFB⁺ MPs are significantly more abundant in the 15mPT and 12yPT tissues compared to the 5dPT time point or the control (Figure 4I).

Next, we asked whether the observed functional shift of MP2s is a result of post-transplantation remodeling or of heart failure. In a publicly available snRNA-seq data set from the failing hearts of 17 adult patients with dilated cardiomyopathy (DCM) and adult donor controls ²⁸, we labeled a MP2 cell population based on the expression score of previously identified MP2 signature genes (Figure S4E). The average expression of phagocytosis pathway genes in MP2 cells remained at similar levels in DCM failing hearts compared to the controls (Figure S4F). While the fibrogenic signaling genes identified in late-stage transplants are expressed at a lower level in MP2 cells from the DCM failing hearts, their expression did not significantly alter compared to the controls (Figure S4G). Taken together, we concluded that recipient MP2s have important similarities to donor MP2s and become dominant in the transplant at later stages post transplantation. However, recipient MP2s shifted their functional paradigm from surveillance and clearance, which is typical of donor MP2s, to one of immune tolerance and fibrosis induction. These phenotypical changes of transplant MPs are likely part of graft remodeling and are independent of heart failure.

Recipient MP1s Associate with the Inflammatory Allograft Response

After heart transplantation, recipient monocytes infiltrate the graft to promote alloimmune rejection ²⁹. However, the detailed cellular events contributing to this process in humans have not been described. We found that recipient MP1s were abundant in the 5dPT heart but substantially decreased at later timepoints and almost undetectable in steady state controls (Figure 5A). In contrast to monocytes, which were maintained at relatively stable levels after transplantation, MP1s exhibit a highly specific cell signature representing the early-phase immune response after transplantation (Figure 5B). Using multiple modeling methods, we visualized a clear trajectory of cell-state transitioning from monocytes to MP1s, indicating that a group of recipient monocytes polarized into recipient MP1s after transplantation (Figure 3F and S3B).

Figure 5. Recipient MP1s Associate with the Inflammatory Allograft Response.

(A) UMAP of the myeloid population, colored by cell subtypes and split by healthy controls and post-transplant timepoints. Each panel displays 500 randomly selected cells. Dashed lines highlight the Recipient MP1s in 5dPT samples. (B) Cell composition of monocytes and Recipient MP1s in allografts at different timepoints after transplantation. Fold-change of cell composition in allografts was calculated by comparing with controls. Note red dashed line at 0 represents baseline in controls. (C) Aggregated expression trend of a co-expressed gene set correlated with the Mono-to-Recipient MP1 differentiation trajectory. Each line represents the mean expression of a gene in monocytes and Recipient MP1s, ranked by diffusion pseudotime. Six of the 312 genes were highlighted as examples. (D) Gene set enrichment analysis showing enriched biologic processes (top) and molecular signature database (MSigDB) terms (bottom) in Recipient MP1s. Representative genes enriched in the terms are listed on the right. (E) Significant CCL3/4–CCR5 signaling predicted between Recipient MP1s and T cells (TC) and natural killer (NK) cells. Vertex diagram illustrates the direction (arrow) and strength (thickness) of interactions, and violin plots show specific expression of ligand and receptor genes. (F) Similar to E, vertex diagram and violin plots demonstrate the predicted IL1–IL1R signaling from Recipient MP1s to monocytes, Recipient MP2s, and Recipient MP1s themselves. (G) Principal component analysis (PCA) plot showing the distinct transcription profiles of Recipient MP1s in allografts compared to macrophages in failing hearts with dilated cardiomyopathy. (H) Heatmap showing the aggregated expression of representative genes from the allograft rejection hallmark gene set (left) and previously reported signature genes in cardiac allograft rejection (right).

To further understand the regulatory events that occur during the Mono–MP1 transition in the 5dPT sample, we examined which genes were upregulated as a function of progression through the pseudotemporal trajectory and identified a module of 312 genes (Figure 5C and Table S7). Within this MP1 activation module, we found key components of TNF-α and interferon signaling transduction. In addition, we used a single-cell regulatory network inference and clustering (SCENIC) approach to identify the potential transcription regulons (transcription factors and all co-expressed target genes) that were activated during Mono to MP1 polarization. Several predicted regulons corroborated the genes identified in the MP1 activation gene module, such as TNF-α (RELB and NFKB1/2) and interferon pathways (IRF1/2/8/9 and STAT1/2/4) (Figure S5A). Collectively, these results suggest that TNF-α and interferon signaling comprises fast-acting cues that activate and polarize monocytes during the first few days after transplantation.

We next examined the functional significance of recipient MP1s in the microenvironment of the graft myocardium. The gene expression signature of MP1s suggests a pro-inflammatory profile, highlighted by the expression of macrophage activation molecules (CD80 and CD83), interleukins that mediate the inflammatory response (IL-1a/b, IL-15, and IL-18), and chemokines that recruit lymphocytes (CCL3/4 and CXCL2/3) (Figure S5B). The enrichment analyses of MP1 signature genes corroborated a pro-inflammatory phenotype and, importantly, predicted functions related to T cells (TCs) and natural killer (NK) cell activation, as well as allograft rejection in general (Figure 5D). Cell-to-cell communication inference confirmed that CCL3 and CCL4 ligands were specifically produced by MP1s that recruit TCs and NK cells expressing the CCR5 receptor into the graft myocardium (Figure 5E). Autocrine signals from IL-1a and IL-1b were received by MP1s and to an even greater extent by monocytes and recipient MP2s (Figure 5F). IL-1β is a major proinflammatory “alarm” cytokine that promotes the infiltration and activation of myeloid cells ³⁰. These two examples of signaling highlight the function of MP1s in mediating both the innate and adaptive immune response.

The expansion of pro-inflammatory immune cells in the myocardium has been observed in adult patients with heart failure caused by myocardial stress, excessive cell death, and related inflammation ²⁸. To examine whether MP1s expand as a result of cardiac stress from heart failure, we performed a comparative analysis of gene expression in MP1s and the available snRNA-seq data from DCM failing hearts ²⁸. PCA showed that the allograft MP1s are transcriptionally distinct from macrophages in failing hearts with DCM and are in a different cell state (Figure 5G). To further validate that the MP1 population has a cellular signature of allograft rejection rather than of heart failure, we tested genes that were previously reported as hallmark genes of pan-organ or cardiac allograft rejection ³¹. As a result, MP1s specifically expressed allograft rejection hallmark genes, such as CXCL9, IL15, CD5, and CDH4, at higher levels than did macrophages in failing hearts with DCM (Figure 5H). These results further support that MP1s may be specifically associated with the microenvironment that leads to allograft rejection. Thus, recipient MP1s and their signaling pathways provide attractive early targets for therapeutically ameliorating host pro-inflammatory response towards the graft.

Recipient MP1s Induce a Hypertoxic NK Cell State

Lymphoid cells, especially NK cells, have profound implications in alloimmune responses and are critical in mediating tissue rejection. We next sought to examine lymphoid cell transcriptional signatures of allograft rejection at different time points after transplantation. By subclustering the lymphoid cell population in the integrated data, we identified populations of CD4⁺ and CD8⁺ T cells, as well as a large population of NK cells (Figure 6A). NK heterogeneity in the pediatric heart has not previously been characterized in detail. Here, we showed that at least two transcriptionally distinct subsets of NK cells, NK1 (KLRC2⁺) and NK2 (KLRF1⁺), co-exist in pediatric hearts at homeostasis (Figure 6A and S6A). In addition, an allograft-specific subset of NK cells (CRTAM⁺), called NK3, was detected in the transplanted heart (Figure S6A and Table S8). Because most NK cells lack traditional T cell markers such as CD2, CD4, or CD8, transcriptomic profiling was used to accurately separate these subsets of NK cells and T cells ⁹ (Figure S6B). Among the NK and T cells, the vast majority (~98%) was inferred to be of recipient origin (Figure 6B).

Figure 6. Recipient MP1s Induce a Hypertoxic NK Cell State.

(A) Representative UMAP of T cells and NK cells in the integrated data. Nuclei are colored by T cell and NK subtype annotations. (B) UMAP of all T and NK cells from the allograft 5 days after transplantation (5dPT), colored by inferred donor and recipient origin. (C) Cell composition of T cell and NK cell subsets in the allograft at various timepoints after transplantation. The fold-change of cell composition in allografts was calculated by comparing with controls. Red dashed line at 0 represents baseline in controls. Note the significant enrichment and then gradual decrease of NK3 cells in the allografts. (D) Representative images of RNA fluorescence in situ hybridization (FISH) for CRTAM, co-stained with immunofluorescent antibody against CD163, DAPI and wheat germ agglutinin (WGA). Yellow arrowheads point to CRTAM⁺ NK3 cells. Exemplary NK3 cells encircled by dashed boxes are shown in zoomed-in view. (E) Expression z-scores of genes in KEGG pathways of NK cytotoxicity (top) and allograft rejection (bottom) in the 3 NK subtypes. *: two-sided unpaired Student’s t test, p<0.01. (F) Significant FASL–FAS signaling predicted between NK3 cells and almost all noncardiomyocytes. Vertex diagram illustrates the direction (arrow) and strength (thickness) of interactions. (G and H) Similar to F, vertex diagrams demonstrate the predicted IL-18 signaling from monocytes, Recipient MP1s, and Recipient MP2s to NK3 cells (G); and reciprocal CSF2 signaling from NK3 cells to monocytes and macrophages (H). (I) Principal component analysis (PCA) plot showing the distinct transcription profiles of NK3 cells in cardiac allografts compared with NK cells in failing hearts with dilated cardiomyopathy (DCM). (J), Box plot showing NK3 signature expression in NK cells at the individual level from the current and public snRNA-seq data sets. Each data point within a box represents the average expression value from NK cells in a single patient. (K) Expression z-scores of genes from the KEGG allograft rejection pathway in pediatric allograft NK3 cells, adult allograft endomyocardial biopsies, and DCM NK cells compared with their respective age-matched controls. *: two-sided unpaired Student’s t test, p<0.01. n.s.: non-significant, p > 0.05.

On average, we observed a ~10-fold increase in all T cell and NK subpopulations within allografts (Figure 6C). NK3 was absent in aggregated control samples that included more than 60,000 total nuclei. Interestingly, the composition of NK3 cells suggested that the abundance of these cells may decrease over time after transplantation (Figure 6C). The NK3 population expressed a subset of known functional NK genes, different from the NK1 and NK2 populations. For example, expression levels of GZMB, IFNG, and REL were high in NK3 cells and relatively low in NK1 and NK2 cells (Figure S6B). CRTAM, an activation marker for NK cells ³², is exclusively expressed in NK3 subsets and not detected in other cells (Figure S6C). We therefore utilized CTRAM probe to label NK3 in the transplants and showcased the abundant presence of NK3 at the 15mPT time point (Figure 6D).

To characterize the functional features of NK subsets, we first scored NK cells on the basis of the expression of genes related to NK cytotoxicity. NK2 and NK3 cells both expressed significantly higher levels of genes related to NK cytotoxicity than did NK1 cells (Figure 6E), suggesting that they are activated NK cells. However, when we examined detailed expression patterns, we found a strong polarization in the expression of genes that are critical for NK activation in the NK subsets. For example, NK3 cells specifically expressed genes encoding inflammatory cytokines such as interferon-gamma, TNF-α, and Fas-ligand, whereas NK2 cells expressed genes encoding CD244 and KIR2DL4, which are suggestive of NK inhibition ^33,34 (Figure S6D). NK3 cells also expressed significantly higher levels of allograft rejection pathway genes than did NK1 or NK2 cells (Figure 6E), suggesting a hyperactivated phenotype that elicits cell damage in the graft. Indeed, our analysis of cell-to-cell communication revealed a striking interaction pattern between Fas ligand from NK3 cells and nearly all non-CM cells that expressed Fas receptor (Figure 6F and S6E). These results suggest that although NK3 cells are a small population, they comprise a highly damaging cell population that mediates cellular rejection in cardiac allografts. In addition, NK3 cells persist after the 12yPT timepoint in the allograft, suggesting that they are incompletely responsive to the immunosuppression therapy.

IL-1β, IL-15, and IL-18 are potent cytokines that promote NK cell survival and enhance cytotoxicity ^35,36. Their strong expression in MP1s suggested that they facilitate the hyperactivation of NK3 cells in cardiac allografts. We showed that IL-18 was a significant ligand received by NK3 cells via IL-18R1 and IL-18RAP (Figure 6G and S6F). Importantly, we also identified a reciprocal interaction initiated by NK3 cells and sent to monocytes and macrophages via the CSF2 signaling axis (Figure 6H and S6G). CSF2, also known as granulocyte-macrophage colony-stimulating factor (GM-CSF), is a potent mitogen and pro-survival factor for myeloid cells. Therefore, MP1–NK3 cells may positively reinforce a sustained inflammatory microenvironment that has lasting adverse effects on the allograft.

To evaluate whether NK3 cells are related to end-stage heart failure, we performed integrative analysis with publicly available snRNA-seq data of heart failure induced by a variety of cardiac diseases. These include pediatric failing hearts with hypoplastic left heart syndrome (HLHS) ³⁷, and adult failing hearts with myocardial infarction (MI) sampled at the infarct and the remote zones ³⁸, as well as adult failing hearts with DCM ³⁹ and arrhythmogenic cardiomyopathy (ACM) ⁴⁰. As an example, integrative PCA with a DCM data set indicated that NK3 cells, but not NK1 or NK2 cells, were transcriptionally distinct from the NK cells profiled in failing hearts (Figure 6I). Systemic comparison of NK3 signature expression at individual patient level further revealed that none of the non-transplanted failing hearts harbor a substantial NK3-like population (Figure 6J).

Furthermore, we also analyzed an endomyocardial biopsy dataset from adult healthy cardiac allografts and failing allografts due to chronic rejection ⁴¹. We found that the rejecting adult transplants, but not healthy ones, contain NK3 cells at the similar level as in the pediatric data. Consistently, allograft NK3 cells, but not NK cells in non-transplanted failing hearts, showed significantly increased expression of allograft hallmark genes, such as NFAT5, IL4R, and CRTAM, when compared with age-matched controls (Figure 6K and S6H). These results substantiate that the NK3 cell state is indeed associated with allograft dysfunction and maybe independent of heart failure. We conclude that NK3 cells are a highly cytotoxic subset of NK cells that are activated by MP1s and likely play a vital host pro-inflammatory response to the allograft.

DISCUSSION

Pediatric heart transplantation is a life-saving treatment for children with end-stage heart failure. Because of its lifelong impact on children, achieving the best long-term outcomes for quantity and adequate quality of life are crucial goals for this patient population. Allograft rejection remains a major cause of mortality and morbidity in pediatric heart transplant patients. In addition, these patients have a high propensity for graft failure and poor long-term outcomes ⁴².

Using three rare heart samples after repeat heart transplantation, we show the utility of snRNA-seq for providing insight into the pathophysiology of human heart allograft dysfunction. Our group recently used snRNA-seq to capture the transcriptional profile of human pediatric hearts with various congenital cardiac conditions at the single-cell level and at unprecedented depth ⁷. To our knowledge, we are the first to describe at single cell resolution, the molecular cell states of a transplanted pediatric heart. A recent study reported results from adult transplant patients with cardiac allograft vasculopathy (CAV) ⁴³. However, it did not provide a detailed portrait of immune cells derived from donors versus recipients, nor did it temporally resolve cell state dynamics. Pathologic examination with hematoxylin and eosin and antibody staining fails to provide a complete picture of cardiac cell states in the context of chronic rejection and graft failure. We characterized cell states in transplanted hearts of pediatric patients at five days and up to 12 years after transplantation. We gained especially valuable insights from cells collected five days after transplantation, allowing us to better understand the immune response during an early stage after transplantation.

Our data reveal that recipient monocytes are capable of bipolarized differentiation in the allograft. Recipient MP1s are reminiscent of inflammatory mouse Ccr2⁺ macrophages induced by myocardial infarction, although recipient MP1s that we identified were CCR2^–. Similar to mouse Ccr2⁺ macrophages, we speculate that differentiation and activation of recipient MP1s are provoked by TNF-α and interferon signaling in the allograft, which are profoundly active in the graft only days after transplantation. Upon differentiation, recipient MP1s activate proinflammatory gene programs, resulting in the secretion of TNF-α, interleukins, and chemokines that activate cascades of innate and adaptive immune responses. Our data are consistent with the conclusion that recipient MP1s directly activate NK3 cells via IL-15 and IL-18 signals. NK3 cells, like MP1s, are a unique cell state found in the early stage of allograft failure. Transcription profiles of both recipient MP1 and NK3 cells were similar to hallmark gene signatures observed in allograft rejection.

NK3 cells exhibit substantial cytotoxicity toward the myocardium. Interestingly, NK3 cells express high levels of CSF2 (GM-CSF), and MP1s express high levels of the receptor genes CSF2RA/B. This suggests that there may be crosstalk among these cells in situ to synergize their expansion and ultimately amplify inflammation in the myocardium. Together, these innate immune mechanisms may play critical roles in causing excessive cell death and irreversible myocardial damage that impair allograft survival. Whether current immunosuppression therapies effectively target these innate immune components remains unclear. We hypothesize that the early differentiation of MP1s and activation of NK3 cells has cascading long-term functional effects independent of ongoing immunosuppression. Suppressing the amplified inflammatory response induced by MP1 cells and myocardial damage instigated by NK3 cells may be beneficial for extending the lifespan of a heart allograft. This strategy may be especially effective in preventing early graft failure and acute rejections.

While we demonstrated that NK3 cells appear specifically in dysfunctional allografts, additional functional insights into NK3 cells are needed to develop specific targeting strategies that can be used as additional preventive care for heart transplant patients. Although NK cells in the transplanted tissue exerted plasticity, the role of NK cells in the immune response to a solid-organ transplant remains highly debated. A subset of NK cells is well-known for promoting allograft injury by amplifying early graft inflammation, such as the NK3 cells we identified. However, recent studies point to alternative roles of NK cell subsets in the priming of allograft tolerance ^44,45. Because small numbers of NK1 and NK2 cells are present in steady state, they may have roles in immune surveillance under homeostasis. However, given the heterogeneity of allograft NK cells, it may be essential to selectively attenuate the adverse features of NK reactivity while preserving other regulatory functions.

Our study supports the feasibility of using demultiplexing methods to identify donor and recipient cell origin on the basis of millions of SNVs embedded in snRNA-seq reads instead of relying on SNVs derived from genome sequencing data generated in the same individual. This critical information enabled us to characterize the donor- and recipient-derived immune components and how they contribute to the tissue microenvironment within the allograft. In a recent study in which the origins of immune cell subsets in the human heart were determined by using Y-chromosome and cell surface marker gene expression (e.g., CCR2, CD68), directly pinpointing cardiac tissue-resident macrophages at single-cell resolution proved to be technologically challenging ⁴. We show the feasibility of identifying donor- versus recipient-derived cells with high fidelity in an unbiased manner, a methodology that is applicable to all transplanted organs.

We observed a transition of recipient monocytes that differentiated into recipient MP2s, which were phenotypically similar to donor MP2s. Recipient MP2s replenished the rapidly diminishing donor MP2s in the allograft and became the dominant immune cell population at late stages after transplantation. Unlike donor MP2s, recipient MP2s lack phagocytic cell signatures and instead functioned in the induction of immune tolerance. We also found that recipient MP2s induced a fibrotic microenvironment in late-stage transplants, which was achieved via the production of multiple fibrogenic factors, including PDGF, TGF-β, IGF, and non-canonical Wnt in the myocardium. Of note, CAV is a leading cause of heart failure and graft loss among patients who have undergone heart transplantation. A main characteristic of CAV is the concentric intimal thickening of the coronary arteries, which results from fibrosis and the proliferation of mural cells ⁴⁶. Our findings suggest that recipient MP2s are an unexpected source of fibrogenic signals that promote adverse remodeling of the allograft.

The paradigm unveiled by our study might be used clinically to target CAV and rejection after transplantation. Firstly, during the initial acute post-transplantation phase, there is merit in suppressing the amplified inflammatory response induced by MP1 cells and myocardial damage instigated by NK3 cells. This may be especially important to prevent early graft failure and acute rejections. Standard monoclonal antibodies or advanced targeting mechanisms such as nanoparticles could potentially be used to specifically target the inflammatory MP1 or NK3 populations. Secondly, we have shown that the replacement of diminished donor MP2 cells by recipient MP2 cells in a transplant may result in increased fibrosis due to decreased phagocytosis activity and increased fibrogenic signaling. Therefore, as an element of long-term management, patients with transplanted hearts could gain advantages from preserving the population of resident macrophages (donor MP2) within the transplanted heart. Targeted immunomodulation which is donor MP2 sparing will be beneficial to the long-term allograft health.

LIMITATIONS

Recipient-derived dendritic cells (DCs) are an important innate immune component and have an established role in heart transplant rejection ⁴⁷. Because of limitations in the current technology, we were unable to obtain transcriptional profiles of DCs; therefore, the existence and role of donor-derived DCs remain unexplored. Due to limited sample availability and that the 5dPT sample is an unusual case of graft failure, the depicted myocardial microenvironment at early post-transplantation stages will need further validation with more patients in the future. Due to the rarity of patient samples, our study was limited by the inclusion of three patients. It is important to note that the allografts we profiled had failed at the time of tissue collection, and our observations reflect the time-resolved microenvironment of allograft remodeling, instead of that in a normally functioning cardiac allograft. While more studies with the inclusion of larger patient cohorts and non-falling allografts are necessary, our work provides important hypotheses for functional dissection in future work.

Supplementary Material

Supplementary Table S1 - Patient demographics and data source

Table S1 – Patient demographics and data source.

Supplementary Table S2 - Main cell type markers

Table S2 – Main cell type marker genes.

Supplementary Table S3 - Cell Count Per Infered Origin and Cell Type

Table S3 – Cell counts stratified by inferred origins and cell types.

Supplementary Table S4 - Myeloid cell subtype markers

Table S4 – Myeloid cell subtype marker genes.

Supplementary Table S5 - Monocyte to Recipient MP2 temporal genes

Table S5 – Monocyte to Recipient MP2 trajectory correlated genes.

Supplementary Table S6 - Donor MP2 vs Recipeint MP2 DEG

Table S6 – Differentially expressed genes comparing Donor MP2 versus Recipient MP2.

Supplementary Table S7 - Monocyte to Recipient MP1 temporal genes

Table S7 – Monocyte to Recipient MP1 trajectory correlated genes.

Supplementary Table S8 - NK cell subtype markers

Table S8 – NK cell subtype marker genes.

CLINICAL PERSPECTIVE.

What is new?

• We used single-nucleus RNA sequencing to investigate the cardiac immune microenvironment in human pediatric cardiac allografts at different stages post-transplantation.

• We show that donor-derived resident macrophages are lost over time, and are replaced by recipient-derived macrophages – which are more fibrogenic and less phagocytic.

• We show that recipient-derived macrophages activate cytotoxic NK cells in failing post-transplant hearts.

What are the clinical implications?

• Current non-specific immunomodulators lead to decrease in donor-derived anti-inflammatory resident macrophages and leads to its replacement with ‘defective’ recipient-derived macrophages.

• Targeted immunomodulation to preserve donor-derived resident macrophages will be beneficial to the longevity of the allograft.

• During initial acute post-transplantation phase, there is merit in targeted immunomodulation to suppress inflammatory macrophages and cytotoxic NK cells.

NON-STANDARD ABBREVIATIONS AND ACRONYMS

ACM

arrhythmogenic cardiomyopathy

CAV

cardiac allograft vasculopathy

CM

cardiomyocyte

CS

chondroitin sulfate

DCM

dilated cardiomyopathy

Donor MP

donor-derived macrophages

EC

endothelial cells

EMB

endomyocardial biopsy

FACS

fluorescence-activated cell sorting

FB

fibroblast

HLHS

hypoplastic left heart syndrome

KEGG

Kyoto Encyclopedia of Genes and Genomes

Lym

lymphoid

Recipient MP

recipient-derived macrophages

MI

myocardial infarction

Mono

monocytes

Mye

myeloid

NC

neuronal cells

NK

natural killer

PC

pericyte

PCA

principal component analysis

SCENIC

single-cell regulatory network inference and clustering

SMC

smooth muscle cells

SNV

small nucleotide variances

TC

T cells

TLF

TIMD4, LYVE1 and FOLR2

UMAP

uniform manifold approximation and projection

DATA AND CODE AVAILABILITY

All deidentified snRNA-seq data are deposited in GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE225652). In-house computer code for reproducing all bioinformatics analyses is available at GitHub (URL: https://github.com/XL-Genomics/2022_macrophage_cardiac_allograft).