Multilineage contribution of CD34⁺ cells in cardiac remodeling after ischemia/reperfusion injury

The human infarcted heart samples were collected from patients diagnosed with MI prior to heart transplantation. The control heart was collected from a generous healthy donor for another organ transplantation who gave consent for heart sample collection for research purpose. All samples were taken from left ventricles during the surgery. Written informed consent was obtained from the patients. All samples were collected from the Nanjing Drum Tower Hospital (China) or the First Affiliated Hospital of Zhejiang University (China) between September 2019 and December 2020. The baseline characteristics of patients are summarized in Table S6. All experiments were approved by the research ethics committees of the two aforementioned hospitals. All procedures had local ethical approval, and all experiments were conducted according to the 2013 Declaration of Helsinki [36].

All animal procedures conformed to the Guide for Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH; 8th edition, 2011) and were approved by the Institutional Ethics Committee and performed according to the guidelines set by the Nanjing Drum Tower Hospital, China.

The mouse lines used in this study were as follows: Cd34-CreER^T2 mice was a self-generated transgenic mouse line generously provided by Prof. Qingbo Xu’s Laboratory (Zhejiang University and Queen Mary University of London). The C57BL6/J, R26-tdTomato (JAX: 007909), and R26-eGFP-DTA (JAX: 006331) mouse lines were purchased from the Shanghai Biomodel Organism Co., Ltd. The Cd34-CreER^T2;R26-tdTomato mice were obtained by crossing Cd34-CreER^T2 with Rosa26-tdTomato mice. The Cd34-CreER^T2;R26-DTA/tdTomato mice were obtained by crossing Cd34-CreER^T2 mice with Rosa26-tdTomato mice and R26-eGFP-DTA mice. The tissue was lysed with proteinase K, precipitated using isopropanol, washed with 70% ethanol, and prepared for genotyping.

MI or myocardial I/R model was established by treating wild-type mice or Cd34-CreER^T2;R26-tdTomato mice, aged 8–12 weeks, with tamoxifen. The procedures were performed according to established previous studies [18].

All data were determined from multiple individual biological samples and presented as mean values ± standard error of the mean. The “n” in the study represented the number of biological replicates, which is indicated in the figure legends. The GraphPad Prism 9 was used to perform the statistical analysis. The data were first analyzed for normality test, followed by the unpaired-sample Student’s t test or one-way analysis of variance test, which are indicated in the figure legends. A P value < 0.05 indicated a statistically significant difference. All mice were randomly assigned to different experimental groups.

A Cd34-CreER^T2; R26-tdTomato lineage-tracing mouse model (Cre/TDT, the Cd34-lineage cells were permanently labeled with tdTomato upon tamoxifen administration) was developed and subjected to I/R surgery and single-cell RNA sequencing (scRNA-seq, Fig. 1A) to fully explore the roles of CD34⁺ cells in vivo. We performed echocardiography imaging as a noninvasive method to evaluate cardiac physiology in knock-in animals [19]. Infarction was established by echocardiogram-measured reduced cardiac function and Masson’s trichrome–stained fibrosis (Fig. S1). All the indexes, including left ventricular ejection fraction, left ventricular end-systolic diameter, and left ventricular end-diastolic diameter, exhibited similar results between Cre/TDT and wild-type C57BL6/J animals (Fig. 1B, C). Successful labeling of Cd34-lineage tdTomato (tdT)⁺ cells was observed in the heart and major organs, with partially co-expressed endothelial cell (EC) marker CD31 but seldom co-stained with leukocyte marker CD45 (Fig. 1D–H). The accumulation of tdT⁺ cells was observed in the border zones of the infarcted area, which indicated the involvement of CD34⁺ cells in tissue repair after I/R injury (Fig. 1I). Meanwhile, tdT⁺ cells co-stained with Ki67 and EdU were also observed, indicating their proliferating state (Fig. 1J).

The infarcted hearts were harvested at sham, 1 and 2 weeks after the I/R surgery, fully digested, and applied to scRNA-seq to illustrate the cellular changes after the I/R injury at the single-cell level. A median detection of 2000–2500 genes per cell was established by scRNA-seq (Fig. S2). We first combined the total-cell datasets with published murine MI datasets [6] to compare the differences in the two disease models. Cd34 was mainly expressed in mesenchymal cell and EC populations in both models (Fig. 2A). Interestingly, the I/R model possessed unique cell components and gene profile so that a large number of ECs were preserved compared with those in either the sham or MI group (Fig. 2B). The mesenchymal cells and ECs, as the dominant and Cd34-expressing cell types in heart, showed decreased expression of progenitor genes Cd34 and Ly6a (stem cells antigen-1, Sca-1) but higher expression of EC markers (Pecam1 encoding CD31 and Cdh5 encoding vascular endothelial–cadherin) in the I/R injury (Fig. 2C). Referring to the upregulated genes or comparative DEGs in mesenchymal cells and ECs from I/R or MI datasets (Fig. 2D, E), profibrogenic genes were either seldom shared [Periostin (Postn), Fn1, and Sfrp2] between models or largely enriched in MI (Ctgf, Cthrc1, Cilp, Lum, and Col8a1). However, the angiogenic genes (Cdh5, Gja1, Rgcc, Icam1, Cav2, and Klf4) were principally enriched in the I/R dataset, leading to a unique biological process (BP) of vasculature development predicted in the I/R injury (Fig. 2F), in which Cd34-lineage cells might actively take part. The following immunostaining on tissue sections from both diseases also confirmed the participation of CD34⁺ cells in angiogenesis after cardiac injury (Fig. 2G). As Cd34 was mainly expressed on mesenchymal cells, we further compared our dataset with canonical Pdgfra-lineage cardiac mesenchymal cells using a public dataset from a Pdgfra^GFP/+ mouse line to extensively address the biological properties of Cd34⁺ cell [5]. Cd34 was exclusively expressed in ECs and a minor population of immune cells compared with the Pdgfra⁺ cells, while Pdgfra⁺ cells were almost only mesenchymal cells (Fig. S3A and S3B). The Cd34-lineage cells also exhibited exclusive capacities of vasculature development and cell differentiation (Fig. S3C and Table S1).

The infarcted ventricles were harvested and digested at indicated time points, sorted based on the tdT expression, and applied to scRNA-seq to illustrate the dynamics of Cd34-lineage cells after the I/R injury (Fig. S4A). A total of 26,320 single cells from 6 datasets (total/tdT⁻ cells and sorted tdT⁺ cells from 3 time points) were integrated, and 7 major cell types were identified (Fig. S4B). The Cd34-lineage tdT⁺ cells were wildly distributed in mesenchymal cells, ECs, and monocyte/macrophage populations. Previous studies confirmed the role of bone marrow-originated Cd34-derived myeloid cells in cardiovascular diseases, which is a distinct developmental origin from their tissue-resident mesenchymal cells and EC counterparts [13].

The I/R injury activated the mesenchymal cells to produce extracellular matrix (ECM), which formed a long-term persistent fibrotic scar in the wounded area [16]. As a major component of Cd34-lineage cells, a total of 14,125 mesenchymal cells were divided into 7 finer subclusters by focused clustering analysis (Fig. 3A and Table S2). The number of Postn-expressing activated mesenchymal cell subclusters (M_Act/Act2) significantly increased after the injury (Fig. 3B and Fig. S5A and S5B), which were believed to be major contributors to cardiac fibrosis [5]. Although tdT⁺ mesenchymal cells shared similar dynamic changes to tdT⁻ ones (Fig. S5B), the tdT⁺ cells mostly consisted of disease-responding mesenchymal cell subclusters, including M_Act/Act2, interferon-stimulated mesenchymal cell (M_IFN), and especially Sca-1 highly expressing ones (M_SH, 74.11%), rather than the Wnt-expressing subcluster (M_Wnt), which was anti-fibrotic (Fig. 3C) [5]. Pseudotime trajectory analyses using SCORPIUS and Monocle methods exhibited an initiative cell state of M_SH and a potential differentiation process toward M_Act/Act2 (Fig. 3D, E and Fig. S5C). The upregulated genes at the branch point where M_SH differentiated into M_Act/Act2 (Fig. 3F) revealed distinct cellular functions. The M_SH was characterized to be involved in BPs, including cell migration, cell adhesion, active response to stimuli or cytokines/chemokines, and regulation of vasculature development. However, the M_Acts clusters showed a pro-fibrotic phenomenon such as ECM organization (Fig. 3G and Table S1). We also evaluated our findings in vivo through immunostaining on tissue sections. Noticeable tdT⁺ cell-involved cardiac fibrosis was observed after I/R injury according to vimentin and collagen I staining (Fig. S5D), and tdT⁺ M_SH/M_Acts/M_IFN were also identified (Fig. 3H) with evident co-expression of tdT and Sca-1, Postn encoded, and Ifitm1/2. Of note, according to the aforementioned data, Cd34-lineage mesenchymal cells dominated the M_SH subtype (Fig. 3C). However, the initial active player M_SH, co-expressing Ackr3, Cd34, Cd248, and Pi16 (Fig. 3B, F), was reported as cardiac mesenchymal-like stem cells with multilineage differentiation and self-renewal capacities, and promoted heart failure [2, 5]. Unlike most classic myofibroblasts, which normally faded within 2 weeks [8], the M_Acts subclusters (Postn⁺, Acta2^low), which were also tdT⁺ cells largely constituted subtypes, kept increasing until 2 weeks after I/R injury and expressed Cilp (encoding CILP1), indicating a pro-fibrotic phenotype crucial in cardiac remodeling and fibrosis [21]. The other tdT⁺ cell highly contributed to the M-IFN subcluster, which, despite comprising a minor population and remaining stable over time, might slow down the progression of cardiac fibrosis [15]. Collectively, Cd34-lineage mesenchymal cells played dominant roles in post-I/R cardiac fibrosis.

We also performed high-resolution analyses on the other main components of tissue-resident Cd34-lineage cells, the cardiac ECs. The ECs were further characterized into eight subclusters according to their gene profile and putative biological functions (Fig. 4A and Fig. S6A–S6D and Tables S3 and S4). The angiogenic EC subcluster, highly expressing tip/stalk cell markers (Nrp2, Apln), and the ECM gene-expressing endothelial-to-mesenchymal transition (EndoMT) subcluster were two major tdT⁺ cell-contributing cell types (over 60%; Fig. 4B). Comparative DEG analyses between tdT⁺ and tdT⁻ ECs revealed specifically enriched BPs, including blood vessel development and ECM organization in tdT⁺ ECs (Fig. 4C and Fig. S6E and Table S1). As the subtype that tdT⁺ cells contributed the most, the angiogenic EC possessed higher abilities of angiogenesis, cell junction organization, and focal adhesion, with several canonical angiogenesis-related pathways (Rap1 signaling and Apelin signaling) enriched (Fig. 4D and Fig. S6F). Besides, tdT⁺ angiogenic ECs exhibited amplified DEGs compared with their tdT⁻ counterparts, indicating a more active cellular response to I/R injury (Fig. 4E). Pseudotime trajectory analyses also showed an initial cell state and differentiating potential from angiogenic ECs toward other cell subtypes (Fig. 4F and Fig. S6G). Although we found that the angiogenic ECs might be the source of postinjury angiogenesis, the coronary ECs were also distributed at the primitive cell state on the trajectory. DEGs were acquired according to different gene profiles at the branch point 1 to delineate the true source of angiogenesis, and two distinct gene modules were identified in the prebranch primitive state (Fig. S6H). Combining BP and pathway enrichment analyses, gene module 2 was found to be related to angiogenesis by its enriched vasculature/endothelium development BPs and pathways, while gene module 4 was more like a pattern relevant to disease response (Fig. 4G and Fig. S6I and Table S1). Additional examination of DEGs in the two modules further revealed that module 2 (Aplnr) represented angiogenic ECs while module 4 (Nfkbiz) was a representative to coronary ECs, which strongly indicated that Cd34-lineage angiogenic ECs were the actual source of post-I/R angiogenesis (Fig. 4G and Fig. S6J).

After cardiac I/R injury, both resident and circulating monocytes/macrophages infiltrated into the injured area to exert heterogenous effects such as modulating inflammation response, engulfing cell debris, regenerating cardiac tissues, and initiating fibrotic scar formation [4]. Nine finer subclusters of monocytes/macrophages were identified in our scRNA-seq datasets, with distinct biological functions predicted according to their gene profiles (Fig. 5A and Fig. S7A, and Table S5). The monocytes and Spp1⁺ subclusters were classified as recruited circulating myeloid cells via the expression of canonical Spp1, Ccr2, and Cxcr2 markers. Another two Folr2/Lyve1/Ace-expressing subclusters were identified as resident-like macrophages. Other minor populations included interferon-stimulated gene (ISG) macrophages (Isg15 and Ifit3), antigen-presenting macrophages (H2-Aa and H2-Eb1), and dividing macrophages (Mki67 and Top2a). Comparing the percentage of tdT⁺ cells in each subcluster, the recruited monocytes/Spp1⁺ macrophages dominated the contribution of Cd34-lineage myeloid cells, with enriched cellular function of classical response to external stimulus, which was believed to be a trigger to postinjury inflammation (Fig. 5B). We next sought to specify the cellular characteristics of the macrophage system by assessing a previously established classification of classically activated (M1) and alternatively activated (M2) macrophages [24]. Major recruited macrophages were more like M1 macrophages, while resident macrophage subclusters acted as M2 macrophages, supporting their pro-inflammatory or anti-inflammatory properties, respectively. The antigen-presenting properties were seen mostly in the antigen-presenting subcluster and slightly in ISG macrophages, Ace⁺ resident-like macrophages, and dividing macrophages (Fig. S7B). Further enrichment analyses based on DEGs in each subcluster revealed a pro-inflammatory response of recruited macrophages and anti-inflammatory property of resident-like ones (Fig. 5B and Fig. S7C and S7D). The accumulation of abundant CD68⁺ tdT⁺ macrophages in the heart chamber in response to I/R injury was also validated in vivo (Fig. S7E).

As mentioned earlier, we previously confirmed a bone marrow origin of Cd34-lineage myeloid cells in a pathological vessel disease [13]. A bone marrow transplantation mode was performed and chimeric C57BL6 mice with Cre/TDT bone marrow cells were generated to address whether bone marrow-originated Cd34-lineage cells also participated in the I/R injury. Through tamoxifen administration, only bone marrow-originated Cd34-lineage cells could be labeled tdT expression in such chimeric mice (Fig. 5C). After 3 days post-I/R injury, a large number of tdT⁺ CD68⁺ macrophages were found in the chimeric heart sections, while the tissue-resident components (vimentin⁺ mesenchymal cells and CD31⁺ ECs) were rarely seen, which validated a bone marrow source of Cd34-lineage myeloid cells in vivo (Fig. 5D).

The existence of intercellular interaction among cardiac cells was addressed. The mesenchymal cells were able to promote angiogenesis [33], and ECs had an impact on the phenotypic switch of mesenchymal cells [17]. We first assessed the proliferating property of Cd34-lineage cells and found a high proportion of EdU⁺tdT⁺ cells in all mesenchymal cells, ECs, and macrophages, which indicated their active participation after the I/R injury (Fig. S8). We next aimed to decipher intercellular communicating patterns between these two cell populations, especially in Cd34-lineage tdT⁺ cells. A thorough evaluation of cellular communicating networks revealed that the vascular endothelial growth factor (VEGF) and SEMA3 signaling, with the Vegfa-Vegfr1/2 and Sema3c-Nrp1/2 ligand–receptor pairs, were found to be the major influencers in M_SH-angiogenic EC interaction. The EndoMT cells mainly received the Gas6 signal via their Axl receptors from M_Acts. Angiogenic and coronary ECs, in turn, communicated with M_SH cells via CXCL (Cxcl12) and MIF (Mif) signaling, while Ackr3 expressed on M_SH played a crucial role in receiving signals from ECs (Fig. S9A). Especially, tdT⁺ M_SH was the most active pro-angiogenesis VEGF signal sender to take effect on ECs, while tdT⁺ angiogenic ECs obtained unique Apelin signaling to interact with other EC subtypes (Fig. 6A). Further analyses also revealed I/R injury-related signaling (Fig. S9B), such as fibrosis-related pleiotrophin, periostin signaling, and inflammation-related SPP1 signaling in the I/R 1w (1 week after I/R injury) dataset, in which M_Acts vigorously took part. In the I/R 2w (2 weeks after I/R injury) group, the noncanonical WNT signaling regulated the cross talk between M_SH and M_Acts, and a late-phase angiogenic response was found to be enhanced by tdT⁺ angiogenic ECs via Apelin signaling (Fig. S9C and S9D).

The M_SH cells, especially the Cd34-lineage tdT⁺ ones, exhibited the highest and strongest cell–cell communicating activity across all cell types (Fig. S9E). Of note, unlike traditional concepts that inflammation induces postinjury angiogenesis, our findings of intercellular communication revealed that tdT⁺ mesenchymal cells (especially tdT⁺ M_SH) dominated a pro-angiogenesis effect on ECs compared with immune cell types. Interestingly, the capacity of Vegfa expression in tdT⁺ M_SH, despite displaying a decrease 1 week after I/R, restored in the second week (Fig. 6B and Fig. S9F). Compared with the MI dataset, M_SH exhibited unique Vegfa expression, the highest in the second-week I/R group (Fig. 6C, D), which might explain the higher percentage of EC reserved in I/R than MI. In summary, Cd34-lineage tdT⁺ M_SH and angiogenic ECs led the cross talk among cardiac cells.

We next sought to find out how Cd34-lineage cells were involved in I/R injury in vivo. A Cd34-CreER^T2; R26-eGFP-DTA (Cre/DTA) mouse line was generated to conduct a tamoxifen-induced diphtheria toxin (DT)-mediated specific cell depletion of Cd34-lineage cells (Fig. S10A). No difference in cardiac function was found in Cre/DTA mice compared with the wild-type mice under normal conditions. The cell ablation was evaluated by immunostaining and flow cytometric analyses on both cardiac and bone marrow cells. Fewer CD34⁺ mesenchymal or EC were also found on heart tissue sections (Fig. S10B–S10E). The Cre/DTA mice were then subjected to I/R injury (Fig. 7A), and we surprisingly observed a recovered cardiac function and reduced fibrosis in Cre/DTA mice (Fig. 7B, C, and Fig. S10F). The cellular changes of cardiac cell components showed a significant decrease in mesenchymal cells, while no differences were found in macrophages or ECs between control and Cre/DTA mice (Fig. 7D). According to the aforementioned findings, this improvement in the cardiac function could be attributed to the annihilation of Cd34-lineage cell-conducted cardiac fibrosis. Unlike most clinical trials on accumulating circulating CD34⁺ cells, the intervention using ablating tissue-resident CD34⁺ cells exhibited a beneficial therapeutic effect in treating cardiac ischemic disease.

Finally, we collected healthy and ischemic human cardiac samples for scRNA-seq from patients undergoing heart transplantation (Fig. S11A and Table S6) to investigate the evolutionary similarities and variations in gene expression profiles between human and mouse hearts. We compared human and mouse total-cell datasets using the LIGER algorithm (Fig. 8A). CD34 was restrictively expressed in a mesenchymal cell and EC populations despite some cell-cluster variations between species (Fig. 8B, C and Fig. S11B), which was validated and also found to be enhanced in ischemia through immunostaining on cardiac tissue sections (Fig. 8F). Significant shared and dataset-specific gene markers were further recognized by the LIGER joint cluster analyses. A gene set in factor 27 showed high expression in cells with classical CD34 expression pattern. Species-shared and species-specific genes of factor 27 were illustrated. The shared expression of CD248, SEMA3C, and ACKR3 indicated a conserved existence of Cd34-lineage M_SH cells (Fig. 8D). Enriched analysis using genes from factor 27 showed shared BPs of wound healing and cell development, while vessel development was more active in mouse dataset and cell adhesion and found to be more enriched in human dataset (Fig. 8E).