Organization of the lung endothelium from single-cell data
To improve endothelial cell viability and enrichment for high-resolution analysis, we isolated CD31-positive cells from rat lungs using anti-CD31 antibody and magnetic beads (Fig. 1a), followed by processing with 10x genomics single-cell RNA sequencing (scRNAseq) (Supplementary data Fig. 1a,b). After mapping and annotating isolated ECs on the UMAP plot (Fig. 1b, Supplementary data Fig. 1c-g), we identified five novel populations, termed gCaps A-E. The lineage of these newly discovered ECs was determined based on the expression of the Apelin receptor, a previously reported gCaps marker1. Notably, novel gCaps demonstrated the formation of two unique zonations between pulmonary artery and vein ECs (AV-zonation) involving gCaps A, B, D, and E phenotypes (Fig. 1b). This finding suggests a gradual phenotypic transition in capillary cells between arterial and vein ECs. Each cell type expressed a set of markers, as shown in the heatmap (Fig. 1c), highlighting the uniqueness of the identified cell types. To annotate cells and AV zonation, we utilized various markers based on the most distinguishing features. By employing gene expression cut-offs to emphasize regions of high expression, we effectively illustrated the phenotype transition along AV-zonation in Fig. 1d. The macrovascular ECs from the pulmonary artery and vein express the Von-Willibrant factor (Vwf). The endothelial cell cluster block began at the converging right corner of the two AV zonations, with Elastin (Eln) expression required in a high-pressure environment in mature arterial ECs from large pulmonary arteries6. Pulmonary artery ECs continued from this corner, marked by the previously established gap junction gene Gja5, and formed two zonation arms transitioning into general capillary ECs (Kit, AplnR positive). Ackr1 and N2rf2 gene expressions visualized pulmonary vein ECs at the left corner of the endothelial clusters in zonation. The lymphatic (Mmm1+) and aerocyte (Ednrb+, Apln+) ECs showed separate clusters outside of AV zonation, implying the distinct and highly specialized phenotypes for these endothelial cells not associated with the endothelial transition from artery to vein (Supplementary Fig. h).
Novel populations of general capillary cells
Our data indicated five novel phenotypes of what was previously considered a single general capillary cell phenotype (Fig. 1d). Two large clusters of gCaps form the lower arm with gCapA with the highest expression of the chloride channel (Clic4+), and the upper arm, gCapD, highly expressing the sodium channel (Scn7a+). The zonations converge with the venous-capillary gCapE (Lingo2+), which connects to the pulmonary vein ECs. Consequently, gCapsA, D, and E create a continuous capillary zonation from the artery to the vein. A small yet persistent subset of gCapB (Flot1+) is situated at the interface between pulmonary artery cells and gCapsA&D, also on (Supplementary File 1, 3D UMAP). RNA velocity analysis identifies this cell type as the "root" or origin for gCapsA&D and arterial ECs (Supplementary data Fig. 1i,i’). Due to the highly mitotic nature of these cells and activation of transcriptional factors involved in EC proliferations such as Jun and Fos7 (Supplementary data Fig. 1f,j), we hypothesize that gCapsB may play a role in repairing capillary ECs and contribute to capillary rejuvenation and neo-capillary genesis. Finally, gCapsC displays distinct separation of the cluster outside of the AV-zonation, highly expressing Fabp4. Comparing the new gCaps with arterial and venous endothelial cells revealed similarities between gCapsA(Clic4+), B(Root), and D(Scn7a+) with arterial ECs, and gCapE(Lingo2+) with venous ECs (Fig. 1e). In contrast, gCap C exhibited a completely different expression pattern. We used SCENIC to calculate regulon specificity scores, which show (Fig. 1f, Supplementary data Fig. 1j) that the novel phenotypes of gCaps have different transcriptional programs, representing distinct cell populations. We also build UMAP using the activity of transcription factors (Supplementary data Fig. 1j,k) that resembles the structure of a gene-based UMAP on principal components (Fig. 1b). Finally, we found the top 5 most specific transcription factors for each phenotype using regulon specificity scores (RSS) (Supplementary data Fig. 1l). We didn't observe a distinct separation of capillary cells in publicly available human or mouse datasets. Thus, we also added our mouse dataset into cross-species integration, in which cells were isolated by FACS using endothelial expression of GFP (Tg(TIE2GFP)287Sato/J mouse). We achieved a three-fold higher resolution for rat endothelial cells (30k cells) compared to mouse sets (8k)8 or human sets (2k) data9,10 (Fig. 1g). We analyzed and visualized the integrated data (Fig. 1h), including cell type markers. Notably, the publicly accessible human or mouse datasets possess no clear distinction of pulmonary capillary cells. In particular, human datasets lacked capillary markers Aplnr and Kit but had high Vwf, Eln, Gja5, Nr2f2 expressions related to large vessels ensothelial cells, indicating a marked loss of capillary cells9,10 (Supplementary data Fig. 2a). This effect can be attributed to the harvesting of cells from post-mortem lung tissues and the preparation process for whole lung sequencing, which is associated with a significant loss of viable capillary cells. In contrast, the publicly available mouse dataset was enriched with capillary cells, but only showed a partial separation of cellular phenotypes8. In addition, the expression of cell type markers in publicly available mouse data was weak compared to our mouse dataset. Incorporating our mouse dataset collected from mice expressing endothelial GFP, which allowed FACS-based endothelial enrichment, enhanced a cross-species integration. We showed that our isolated mouse and rat endothelial cells had a better quality in terms of the number Unique molecular identifier counts (UMI counts) and/or genes detected in each cell (Supplementary data Fig. 2b). This allowed us to successfully identify the newly described phenotypes of gCaps in mouse datasets in addtion to rats. We also found a significant intersection of cell markers among our mouse and rat datasets (Supplementary Table 2).
Characteristics of novel general capillary endothelial cells
gCapC (Fabp4+)
Fabp4 + gCaps form a distinct cluster on the UMAP, resembling lymphatic ECs or aerocytes clusters rather than being present at the AV zonation. This observation implies their increased specialization and perhaps indicates a reduced structural participation in the capillary network. The Fabp4 + cells classify as gCaps because they express Aplnr, Cdh5, Myo10, and Ccdc85a, along with other phenotypes of gCaps (Fig. 1d). Further analysis shows that Fabp4 + cells exhibit a high enrichment of genes responsible for ameboid-type migration and cell motility (Fig. 2a, Supplementary Table 3). These cells also express genes involved in lipid metabolism (Pparg, CD36, Fabp4/5) and oxidative phosphorylation (Fig. 2b,c), indicating their high energy demand compared to other endothelial cells. This finding suggests that these cells rely on lipid-centered metabolism for their angiogenic capabilities. Previous studies have demonstrated that angiogenesis depends on lipid metabolism and is primarily impaired in Fabp-deficient mice 11. A comparison of the gene list for tip-cell markers revealed that most tip-cell markers are highly expressed in Fabp4 + capillary cells (Fig. 2d). Tip ECs play a crucial role in angiogenesis, guiding the sprouting of new capillaries/vessels and facilitating branching angiogenesis12. Our study, for the first time, identifies a specific cell population within gCaps in the lung responsible for capillary angiogenesis. These cells are also rendered in published mouse data and humans (Supplementary data Fig. 2a). However, Fabp4 + cells are sometimes mistaken for macrophages due to their high motility and expression of various cytokines. Nevertheless, Fabp4 + cells do not express macrophage markers such as CD68 (Supplementary data Fig. 3a. Supplementary Table 4, 5). This study contributes to the annotation of this particular phenotype of gCaps. To examine the angiogenic properties of Fabp4 + cells, we employed cell sorting to isolate CD31 + cells and CD31 + cells lacking Fabp4 + cells (Supplementary data Fig. 3b). The sprouting experiment conducted on matrigel revealed a decreased angiogenic capacity in Fabp4 + population deficient CD31 + cells (Fig. 2e-g). Fluorescent images displayed distinct features of these cells, including lamellipodia-like structures (Fig. 2h, Supplementary data Fig. 3c).
gCapB (Root/Flot1+)
A minor population of gCapB (Flot1+) root cells is situated within the AV-zonation at the interface of three cell types: small artery endothelial cells, gCapA (Clic4+), and gCapD (Scn7a+) (Fig. 1b, d, Supplementary data Fig. 1i, i'). We deduced root and end cells from the velocity graph using a Markov chain implemented in CellRank, and identified several root states (Supplementary Fig. 1i Root and End cells), with gCapB among the root cells. The partition-based graph abstraction (PAGA) analysis, enhanced by velocity-inferred directionality, revealed a high transcriptional similarity between gCapB cells and Scn7a+, Clic4+, Lingo2+, and pulmonary artery cells, which exhibited less similarity to each other than to gCapB (Supplementary Fig. 1.i’). Moreover, gCapB cells appear to be the principal root cells, as all other cells, except for pulmonary veins, are descendants of them based on directed transition probabilities. Consequently, we hypothesize that gCapB cells serve as the primary root cells responsible for regeneration. Our data indicate several important transcriptional factors (TFs) are highy expressed in Root cells, such as Fos and Jun, members of the AP-1 family. These essential transcription factors promote endothelial cell proliferation and angiogenesis (Fig. 2i)13. Conversely, the Foxo transcription factor and Foxo pathways (Fig. 2j) were upregulated in Root cells. The Foxo pathway is a crucial regulator of cellular homeostasis, playing a role in cell cycle regulation, apoptosis, and oxidative stress responses14. In endothelial cells, Foxo1 regulates angiogenesis by controlling the expression of VEGF and other angiogenic factors. Additionally, Foxo1 has been implicated in regulating endothelial cell migration, a critical step in angiogenesis15. Overall, Fos, Jun, and Foxo transcription factors are essential regulators of endothelial cell proliferation and angiogenesis, and their dysregulation can contribute to the development of various angiogenic diseases. Thus, the Root cells may play a central role in regulating capillary and arteriole formation and repair by being the source of endothelial cells. Indeed, the Root cells showed a high G2M and S score; the G2M score is a critical marker for indicating mitotically active cells (Fig. 2k). This suggests the role of Root cells as the origin of capillary and pulmonary artery endothelium. Our data also showed high expression of CDKn1a,1c (p21, and p57) isoforms (Fig. 2j) is necessary for stem cell maintenance and differentiation to harness high proliferation rate.
gCapA (Clic4+) and gCapD (Scn7a+)
Clic4 + and Scn7a + cells constitute the two primary populations of gCaps, which play a significant role in constructing the capillary network around the alveoli to facilitate gas transport throughout the organism. These endothelial types are present in two distinct AV-zonations, connecting a small artery to vein endothelium via gCapE (Lingo2+) (Fig. 1d). This raises two critical questions: First, do they form distinct capillaries? Second, what unique roles does each type play? Our confocal fluorescent imaging indicates that Scn7a + and Clic4 + cells can form extensive capillary structures with both phenotypes, particularly around aerocytes (Fig. 3a, Supplementary data Fig. 4a,b).
Numerous genes in Clic4 + cells within the top 50 differentially expressed regulate barrier maintenance, function inflammatory cell and ion/fluid trafficking, inflammation, and coagulation control (Supplementary Table 1). Analysis of differentially expressed genes indicates that Clic4 + cells actively regulate cell growth control through upregulation of Bgt2 and Socs3. Elevated expression of Atf3 controls the metabolism of Clic4 + cells and has been implicated in endothelial cell activation.
Conversely, Scn7a + cells exhibit a greater representation of hormonal/soluble factors-based regulation through various receptors. They display highly expressed Vipr1 and Npr3 receptors, which regulate vasodilation16. The Adgr family of G-protein coupled receptors, including CD97, Latrophillin, and Gpr116, are involved in cell-cell interactions, cell adhesion, and migration (Fig. 3b)17. Ephrin-B2 and Calcrl receptors modulate angiogenesis and cell adhesion18,19. Scn7a + cells also express receptors responsible for proliferation and survival, such as c-Kit, Tie2, Bmpr2, Lifr, and co-receptor Eng (Fig. 3b). Lastly, two tyrosine phosphatase receptors, Ptprm and Ptprb, and integrin Itga1, play roles in cell adhesion and maintaining barrier function 20. Therefore, we may conclude that transcriptional regulators predominantly program Clic4 + cells, while Scn7a + cells are primarily controlled by paracrine/autocrine signaling via numerous receptors. Further clarification regarding the differences between these cell types can be found in Fig. 2k, which illustrates the highly mitotically active Clic4 + cells alongside the much less active Scn7a + cells. This suggests that these cell types represent different stages of capillary cell maturation. Intriguing is that one capillary population (Scn7a+) mainly expresses a sodium transporter, whereas another population (Clic4+) highly expresses the chloride channel. Both channels are essential for controlling the exchange of molecules (ions, solutes, and water) between the blood and the surrounding lung tissue, which is the primary physiological role of capillaries. Moreover, these channels can maintain the electrochemical gradient leading to membrane polarization and, perhaps, promote a tighter barrier.
gCapE (Lingo2+)
Lingo2 + cells are situated in the AV zonation between two gCaps (Clic4 + and Scn7a+) types and venous ECs, which characterizes them as venous-capillary cells. These cells express unique genes, primarily indicative of capillary endothelial characteristics. Firstly, Lingo2 + cells express the Lingo2 gene, which is involved in the Nogo signaling pathway21. Although Nogo signaling mostly pertains to neuronal system development and maintenance, NogoA/B has been reported to contribute to angiogenesis and the inflammatory response in endothelial cells22,23. Consequently, unusually high Lingo2 expression may suggest these cells participate in angiogenesis and chemokine response. Furthermore, Lingo2 + cells express the Ackr3 (Cxcr7) receptor, unlike venous ECs, which predominantly represent Ackr1. Ackr3 specifically interacts with Cxcl12 24, modulating angiogenesis, while venous Ackr1 scavenges all CC and CxC chemokines, aiding inflammation regulation25.
Endothelial cells perform various functions, such as maintaining barrier integrity, vasodilation, inflammation, and angiogenesis26. Our study identified a high degree of heterogeneity among capillary cells, prompting us to investigate whether proteins with essential roles are differentially expressed across endothelial cell types, reflecting a higher level of specialization in contrast to their previously assumed shared functionality among endothelial cells. We employed Western blot analysis to assess protein levels in the EC populations (Fig. 3c). Our findings revealed considerable functional heterogeneity among these endothelial cell types, supporting the notion of specialized endothelial functions within distinct populations27.
Interactions between populations of endothelial cells
Pulmonary circulation is a highly specialized vascular network, and its proper functioning depends on the coordinated interplay among various cellular components. These components include endothelial cells, smooth muscle cells, pericytes, fibroblasts, and inflammatory cells, which interact via cell-cell communication and local oaracrine signaling. Endothelial cell interactions are crucial in forming and stabilizing new blood vessels during angiogenesis and establishing the blood-tissue barrier28. Throughout this process, endothelial cells communicate using various signaling molecules to regulate cell adhesion, migration, and proliferation29. Understanding the heterogeneity in endothelial cells can help us analyze cell-cell interactions in more detail.
The most significant communication signals emerge from Scn7a + cells in paracrine and autocrine manners (Fig. 4a). Clic4 + cells, aerocytes, and pulmonary artery cells also substantially contribute to cell-cell interactions in the lungs. Additionally, several intriguing cellular communications, suggesting novel phenotypes, are discovered. For instance, Scn7a + cells communicate with aerocytes using Sema3 and pulmonary arteries using Sema6 for vessel guidance and cellular signaling. The angiogenic chemokine Cxcl12 exhibits outbound signaling from Fabp4 + and arterial ECs. Proliferating c-kit signaling initiates from aerocytes and targets two primary gCaps: Scn7a + and Clic4+. Interestingly, Root cells play a vital role in endothelin signaling to aerocytes. All outgoing and incoming signaling patterns are illustrated in Fig. 4a, while the complete connectome analysis can be found in supplementary Fig. 5.
Sex difference
The overall connectome analysis revealed that Esam and Reelin (Reln) signaling are preferentially upregulated in males, while non-canonical WNT, Visfatin (Nampt), Progranulin (Grn), EphA/B, and Semaphorin 3 are upregulated in females (Fig. 4b). As a result, the Reelin and Visfatin systems appear to regulate the inflammatory response in a sex-dependent manner. Furthermore, Esam and Ephrins (EphA/B) modulate sex-dependent adhesion to endothelial cells18. Females also possess two pro-survival systems, non-canonical WNT and GRN, which regulate vascular cell proliferation and repair. Sex-specific information about connectome for the different cell type and differential expression genes are available in supplementary Fig. 6–8 and supplementary table 6.