Single-cell transcriptome analyses reveal novel targets modulating cardiac neovascularization by resident endothelial cells following myocardial infarction

1Chronic heart failure following acute myocardial infarction (MI) has reached global epidemic proportions, and new therapeutic approaches are vital. Functional neovascularization in the infarct border may delay progression to heart failure and improve outcome, although the underpinning mechanisms remain unclear. We show that resident cardiac endothelial cells (ECs), and not bone marrow cells, mediate therapeutic angiogenesis post-MI, and provide a unique single-cell atlas of the transcriptional signature of these regenerative cardiac EC. We predict that these data will inform future design of clinical strategies aimed at promoting vascular perfusion and long-term functional viability of the ischaemic heart.

2Chronic heart failure following acute myocardial infarction (MI) has reached global epidemic proportions, and new therapeutic approaches are vital. Functional neovascularization in the infarct border may delay progression to heart failure and improve outcome, although the underpinning mechanisms remain unclear. We show that resident cardiac endothelial cells (ECs), and not bone marrow cells, mediate therapeutic angiogenesis post-MI, and provide a unique single-cell atlas of the transcriptional signature of these regenerative cardiac EC. We predict that these data will inform future design of clinical strategies aimed at promoting vascular perfusion and long-term functional viability of the ischaemic heart.

3Chronic heart failure as a consequence of left ventricular impairment following acute myocardial infarction (MI) has reached epidemic proportions affecting more than 23 million patients worldwide. Patients have a poor prognosis with a mortality of 45–60% by 5 years. Early reperfusion following MI is important to limit infarct size, however, patients who present late with extensive myocardial injury benefit little from these treatments. In this high-risk group, new therapeutic approaches are required to enhance myocardial perfusion, limit infarct expansion, and promote cardiac repair and regeneration. As such, extensive research has focused on harnessing the regenerative value of exogenous cell transplantation, and on elucidating the endogenous mechanisms that promote structural repair in the injured heart.

4Experimental studies have produced disparate results regarding the regenerative potential of transplanted bone marrow cells in the setting of MI, including their contribution to vascular structures in vivo (reviewed in Refs,). Moreover, clinical trials using autologous bone marrow-derived cell transplantation aimed at promoting recovery following acute MI have largely failed to show sustained clinical benefit., Recently, endothelial cell (EC) clonal expansion in the post-ischaemic myocardium was shown using a Cdh5-CreERT lineage-tracing mouse. However, Cre-recombinase can be detected in bone marrow cells as well as endothelium in Cdh5-Cre lines and therefore the origin of these proliferative EC remains somewhat inconclusive. In the same study, clonally expanded EC co-expressed endothelial and mesenchymal markers, leading to the conjecture that ‘partial’ endothelial-to-mesenchymal transition (EndMT) may be associated with new vessel growth post-MI. However, a direct role for EndMT in therapeutic angiogenesis has not been conclusively shown, to date. Therefore, the cellular origin and mechanisms of neovascularization following MI remain unresolved, including whether reparative EC derive from a bone marrow niche or reside locally within the cardiac microvasculature.

5Here, we have used an EC-specific multispectral lineage-tracing mouse (Pdgfb-iCreERT2-R26R-Brainbow2.1) coupled with single-cell RNA sequencing to collectively investigate the origin, proliferative dynamics and transcriptional profile of EC that mediate neovasculogenesis in the adult mouse heart. We show that a subpopulation of resident EC with progenitor-like functional properties directly contribute to new blood vessel formation at 7 days following MI, with evidence that bone marrow-derived cells and EndMT are unlikely to contribute. We have identified multiple discrete heterogeneous EC states in the healthy and post-ischaemic heart, and characterized gene expression profiles of each EC state at the single-cell level. These data identified plasmalemma vesicle-associated protein (Plvap) as a novel endothelial-specific marker of cardiac neovasculogenesis, which was further validated in cardiac samples from patients with ischaemic heart disease and following gene silencing in vitro. Together, these data provide a comprehensive single-cell index of gene expression by resident cardiac EC that contribute to endogenous neovascularization in the infarcted myocardium and will assist identification of new therapeutic targets for heart disease.

6See Supplementary material online for extended experimental procedures.

7We first evaluated Pdgfb expression in the adult mouse coronary vasculature, which has not been widely characterized. We used the presence of a sequence coding for enhanced green fluorescent protein (EGFP) located downstream of iCreERT2 in Pdgfb-iCreERT2 mice (Figure 1A, Supplementary material online, Figure S1A–C) to show that Pdgfb-EGFP expression was widespread and endothelial-specific, confirmed by co-localization of EGFP with isolectin B4 (Figure 1B). We further showed using flow cytometry that Pdgfb was expressed by 94.5 ± 5.6% of CD31+ vascular EC in the adult Pdgfb-iCreERT2-R26R-Brainbow2.1 mouse heart (Supplementary material online, Figure S1C). Expression of the four fluorophores encoded by the Brainbow2.1 transgene (YFP, RFP, nGFP, mCFP) was specific to Pdgfb-EGFP+ cells (Figures 1C; Supplementary material online, Figure S1D and E). Cre-recombination efficiency was 46.6 (±9.3) % of total cardiac Pdgfb-EGFP+ cells, and no reporter expression was observed in mice administered peanut oil as a vehicle control for tamoxifen or in Cre-negative mice administered tamoxifen (Supplementary material online, Figure S1D–G).

8Clonal proliferation by EC was observed in the healthy heart at 21 days post-tamoxifen, highlighting that Pdgfb-lineage cardiac EC are not quiescent in physiological conditions (Figure 1C). Clone size was significantly increased in the infarct border region at 7 days post-MI, with clones of >50 cells frequently present (cells per clone = 4.0 ± 2.1 vs. 10.3 ± 10.6, P < 0.0001; Figure 1C–E; Supplementary material online, Movie S1). Representative histological images of injured and healthy hearts are shown in Supplementary material online, Figure S2. Perfusion of vessels formed by clonal proliferation of Pdgfb-EC was confirmed by fluorophore co-expression with isolectin-B4, infused intravenously just prior to cull (Figure 1B). Consistent with other studies using the R26R-Brainbow2.1 mouse, we observed a bias in fluorophore distribution in both groups with YFP, RFP, nGFP, and mCFP expressed by 51.9 ± 13.5, 25.7 ± 8.7, 11.3 ± 6.2, and 11.8 ± 5.8% of total reporter expressing cells in the healthy heart, and by 52.9 ± 26.5, 18.8 ± 15.0, 26.5 ± 29.8, and 8.8 ± 7.8% of total reporter expressing cells in the injured heart (Supplementary material online, Figure S1H and I). However, clone size did not differ between each fluorophore in either group (Supplementary material online, Figure S1J and K). Therefore, patches were likely to be formed through clonal expansion, and not by the merger of multiple smaller clones of the same colour (see Supplementary material online file for detailed methods used for quantitative analysis of clonal proliferation).

9Vessels with a polychromatic endothelium were frequently observed in healthy and infarcted hearts and most likely represented pre-existing vessels composed of labelled EC that had not undergone proliferation within the study timeframe, or new vessels formed from the merger of non-proliferative labelled EC (Figure 1F). Polychromatic vessels were significantly less abundant than monoclonal vessels in both groups (healthy hearts = 25.5 ± 4.1% vs. 74.5 ± 4.1%, P < 0.001; MI hearts = 9.3 ± 15.3% vs. 90.7 ± 15.3%, P < 0.001), implying that clonal proliferation is a dominant mechanism of cardiac EC turnover in homeostasis and in response to injury. The proportion of Confetti+ EC present as single cells, i.e. that had not undergone division during the preceding 21 days, was similar in healthy and MI groups (54.6 ± 6.6% vs. 63.4 ± 4.7%, P = 0.28). Therefore, vascular turnover and neovascularization by clonal proliferation appears be restricted to a subpopulation of Pdgfb-lineage EC, presumably with progenitor-like properties. The thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU) was administered 1 h prior to cull to label cells in S phase. EdU+ total cardiac cells were significantly increased in the infarct border at 7 days post-MI compared with the healthy left ventricle, as expected (0.09 ± 0.04% vs. 1.9 ± 0.6%, P < 0.0001). The percentage of Confetti+ EC that co-expressed EdU was also increased in the infarct border compared with the healthy left ventricle (28.5 ± 4.8% vs. 58.5 ± 7.6%, P = 0.0005), underlining our evidence that the infarct border is a region of dense neovasculogenesis with maximal clonal proliferation by Pdgfb-EC (Figure 2A–F).

10We observed minimal Brainbow2.1 reporter expression in femoral bone marrow cells in healthy and MI groups (0.04 ± 0.02% vs. 0.03 ± 0.009%, P = 0.79) (Figure 2F and G). Therefore, we postulated that Pdgfb-lineage EC reside locally within the cardiac vasculature. This was corroborated by our observation that the founding number of recombined events (i.e. a Confetti+ cell or clone) did not differ significantly between groups (P = 0.46) (Figure 2H) and thus migration of Pdgfb-lineage Brainbow2.1 fluorophore expressing EC from outside the heart was likely to be rare.

11Pdgfb-lineage vascular EC (Confetti+ CD31+ podoplanin−) were isolated as single cells from dissociated ventricles of healthy and ischaemic hearts at 7 days post-MI for single-cell RNA sequencing (Supplementary material online, Figure S3A). Mean 3195 (range 2637–3535) and 3955 (range 3326–4744) cells were retained for analysis from the healthy and MI groups, respectively (P = 0.11) after filtering of cells with a low gene count (<400 genes per cell) and a high mitochondrial transcript ratio (threshold of 20%) (Supplementary material online, Figure S3B). Mean number of unique molecular identifiers (UMIs) was 2232 (range 2035–2398) and 5041 (range 4165–5604) (P = 0.03), and median gene number per cell was 1180 (range 1046–1217) and 2133 (range 1807–2251) (P = 0.03) for the healthy and MI groups (Supplementary material online, Figure S3B). Dimension reduction by principal component analysis of the whole transcriptome gene signature followed by 3D t-distributed stochastic neighbour embedding (t-SNE) projection assigned cells into 10 transcriptionally distinct clusters, i.e. discrete heterogeneous EC states (Figure 3A). Cluster distributions were conserved in each biological replicate animal in each group (Supplementary material online, Figure S3C). The proportion of cells in each cluster, the contribution of cells from both groups to individual clusters, as well as a heat map of top differentially expressed genes in each cluster are shown in Figure 3B, C, and E. An isolated cluster of cells that expressed the haematopoietic marker Ptprc (CD45) but did not express endothelial markers Pecam-1 (CD31) or Kdr (VEGFR2) was removed from the analysis, as it was likely due to cellular contamination during FACS. Following this, we confirmed enrichment of Pecam-1 and Kdr in all clusters and a minimal presence of lymphatic markers in our analysis (Figure 3D). A very small proportion of Ptprc+ cells from the MI group that co-expressed Pecam-1 (0.29% total cells) or KDR (0.19% total cells) was dispersed throughout several clusters (Figure 4A). Correspondingly, we observed rare in situ expression of CD45 by Pdgfb-lineage EC post-MI (Figure 4B). These CD45+ cells may be EC that have transitioned towards an inflammatory phenotype in response to injury or, more likely, scarce infiltrating bone marrow cells, described above.

12We used Gene Ontology (GO) term enrichment analysis (PANTHER, GOrilla and GeneMANIA) to reveal significant terms associated with the top differentially expressed genes in each cluster, which allowed us to extrapolate information regarding the functional identity of each EC state (Table 1). For example, cluster 10 predominantly comprised cells from the MI group with a high and restricted expression of cell cycle and proliferation markers (e.g. Mki67, Top2a, Cenpf, Cks2, Birc5, Cenpa, Ube2c, Cdc20,). Our aforementioned observations of heightened clonal proliferation post-MI by Pdgfb-lineage EC compared with the healthy heart, substantiated this gene expression profile (Figure 1D). Cluster 3 showed selective expression of genes encoding ventricular cardiac muscle morphogenesis (Myl2, Mb, Myl3, Tnnt2, Tnni3, Actc1), suggesting that resident adult cardiac EC may switch on cardiomyocyte lineage genes in response to MI. This was supported by a study showing cardiac muscle cells derived from embryonic EC following their direct injection into the ischaemic adult mouse heart. We confirmed rare in situ existence of cells that co-expressed a Brainbow2.1 fluorophore and the cardiomyocyte marker, troponin-T, in the infarct border region, but not in the healthy heart (Figure 4C and D). Therefore, an endothelial to cardiomyocyte fate shift may occur post-MI, although did not contribute significantly to myocardial regeneration at the studied timepoint.

13Endothelial markers were not reduced after MI in our data (Figure 3D), and therefore, we investigated the role of ‘partial EndMT’, recently hypothesized to play a role in neovascularization in the ischaemic mouse heart. However, the overall EndMT signature was unchanged in both groups (Figure 5A and B) despite an increase in some EndMT markers after MI (Icam1, Vcam1, Vim, Fn1 and Smtn) (Figure 5C). This indicates that EndMT and partial EndMT are unlikely to participate significantly in our model at the studied timepoint.

14Clusters 6, 7, and 8 predominantly comprised cells from the MI group (Figure 3B), indicating that their gene expression profiles may be relevant to neovasculogenic pathways. Gene expression of Plvap was highest in clusters 6, 7, and 8 (Figure 6A and B). Plasmalemma vesicle-associated protein encodes an EC-specific marker, although its role in neovasculogenesis in the adult heart is unknown. We quantified Plvap expression in the healthy and infarcted mouse heart, as well as in healthy and ischaemic human cardiac tissue sections. Plvap expression was specific to CD31+ EC and was increased in the infarct border compared with the healthy left ventricle, thus validating our RNA sequencing data at the protein level (% Plvap+ EC = 70.5 ± 19.9% vs. 38.7 ± 28.2%, P = 0.002, Figure 6C, E–G). Plvap expression was also increased in EC adjacent to regions of fibrosis and scarring in the ischaemic human heart, compared with healthy human hearts (% Plvap+ EC = 36.9 ± 10.1% vs. 11.1 ± 8.8%, P = 0.002, Figure 6D, H–O). In order to better understand whether Plvap can directly modulate neovascularization we carried out in vitro siRNA gene silencing on human umbilical venous EC (HUVEC) where Plvap expression was significantly reduced at the mRNA level compared with control siRNA (RQ = 3.0 ± 0.9 vs. 0.2 ± 0.2, P = 0.003) and confirmed at the protein level by western blot (Figure 6P, S). Proliferation was assessed using an EdU incorporation assay and was significantly inhibited following Plvap gene silencing compared with control siRNA treatment (% EdU+ HUVEC = 60.7 ± 3.9% vs. 21.1 ± 11.0%, P = 0.0038; Figure 6Q and R), providing strong evidence for a direct functional role of PLVAP in endothelial proliferation.

15In this study, we demonstrate that structural integrity of the endothelium of the adult heart during physiological conditions and following MI is maintained through clonal proliferation of a subset of resident EC with progenitor-like functional properties with minimal, if any, contribution by bone marrow cells. We further present the first comprehensive single-cell atlas of gene expression by 10 individual EC subpopulations in the adult mouse heart after MI. Finally, we show that Plvap is likely to represent a novel target to augment new vessel formation in the ischaemic human heart.

16Our observation that clonal expansion of resident cardiac EC is the predominant mechanism driving neovasculogenesis expand on recent findings of He et al. who used genetic lineage tracing to show that new cardiac blood vessels are formed from pre-existing EC. This may, in part, explain the disappointing outcomes of clinical studies using autologous bone marrow-derived cells for therapeutic neovascularization in patients with ischaemic heart disease., It is important to note that a contribution of Pdgfb-lineage EC to cardiac neovasculogenesis post-MI cannot be discounted, i.e. due to an inactive Cre-driver or the migration of non-recombined cells. However, despite a 46.6% Cre-recombination efficiency in cardiac ECs we observed Brainbow2.1 reporter fluorophore expression in fewer than 0.05% of bone marrow cells. Furthermore, the founding number of recombined events in the cardiac vasculature did not change before or after ischaemic injury. This led us to postulate that Pdgfb-lineage EC contributing to neovasculogenesis via clonal proliferation reside locally within the cardiac vasculature with minimal, if any, contribution from bone marrow cells. Therefore, future research and clinical focus should aim to enhance the regenerative capacity of endogenous resident cardiac EC to promote therapeutic neovasculogenesis in the injured heart.

17Using single-cell RNA sequencing, we have defined endothelial heterogeneity in the healthy and injured mouse heart through characterization of 10 EC states with distinct gene expression signatures, and by mapping the transcriptional changes arising in each subset following injury, at the level of a single cell. Transcriptional profiling of the non-myocyte cardiac cellulome of healthy mice was recently reported, although EC were strategically depleted prior to sequencing to enable focused analysis of fibroblasts and rarer cardiac lineages such as glia and mural cells. Single-cell RNA sequencing was also recently applied to map cell populations in the healthy and injured mouse heart 3 days after ischaemia–reperfusion surgery. This study identified two EC clusters, although detailed EC gene expression analysis was not provided as this study focused on changing cardiomyocyte subpopulations. Therefore, we believe our data is the first in-depth characterization of the molecular profiles that demarcate cardiac endothelial heterogeneity and plasticity in response to ischaemic injury.

18We defined changes in EndMT gene signatures between healthy and injured cardiac EC using a broad panel of markers., However, as we found no inhibition of endothelial genes nor a significant change in the EndMT gene signature, we concluded that EndMT does not play an obvious role in our model at the timepoint studied. This contrasts with a recent report suggesting that partial induction of EndMT may support cardiac neovascularization after ischaemia. This discrepancy may be due to a transient nature of EndMT, as the authors did not clarify whether partial EndMT genes were enriched at 7 or 14-days post-MI. Therefore, further temporal analysis of EndMT gene signatures following injury, as well as investigation into whether EndMT directly facilitates cardiac neovasculogenesis are warranted.

19Plasmalemma vesicle-associated protein was selected for further study, due to its cluster-restricted increase in MI, and the lack of understanding of its role in endothelial regulation in the adult heart. Plasmalemma vesicle-associated protein is specifically expressed by EC and increased in some inflammatory states as well as in pathogenic angiogenesis. Here, we showed increased Plvap expression specifically within EC in clonal neovessels in the ischaemic border region of the mouse heart, and in regions of fibrosis and scarring in the ischaemic human heart, compared with non-ischaemic hearts. We further demonstrated that targeted silencing of Plvap in vitro directly inhibits EC proliferation. Therefore, Plvap shows early promise as a gene of interest for therapeutic strategies to promote cardiac regeneration. Future studies must now systematically evaluate the molecular pathways and regenerative significance of enhanced Plvap expression in EC in the infarct border region at the time of injury. However, at present, methods for cardiac endothelial-specific gene delivery to increase expression are challenging, and therefore, genetic mouse models and/or a pharmacological approach would be required.

20The current study provides a high-resolution single cell gene expression atlas of resident cardiac EC in both physiological conditions and following permanent ligation of the left anterior descending coronary artery at 7 days post-MI. This model results in massive cardiomyocyte death, coagulative necrosis and scarring, with evidence of significant neovasculogenesis in the infarct border region. However, this model is unlikely to stimulate EC-mediated myocardial regeneration, and therefore, molecular targets identified from the analyses presented herein should be rigorously interrogated to investigate whether they promote myocardial regeneration and prevent cardiomyocyte death through enhanced tissue perfusion, rather than merely supporting scar formation. One could venture that a potential limitation of this study lies within the vast wealth of data generated from a single snap-shot in time, and care must be taken to extract and validate future targets that most effectively tackle the clinical problem. Therefore, consulting publicly available EC transcriptomics data, such as the endothelial database, EndoDB, will assist the construction of novel hypotheses for future studies.

21In summary, this study provides a rich resource of targets and pathways that can be interrogated to reveal their role in promoting neovascularization by endogenous cardiac EC and thus may guide novel therapeutic strategies aimed at enhancing myocardial repair and regeneration.