scRNA-seq identifies three molecularly distinct lung perivascular populations
Pericytes, a mesenchymal cell type closely linked to vascular smooth muscle cells (VSMCs) (2), remain poorly characterized. We investigated the ontogeny of lung pericytes using whole-lung scRNA-seq at multiple stages of mouse lung development, ranging from E10.5 to E18.5. Through unbiased clustering and cell type annotation, we identified endothelial (Pecam1+), epithelial (Epcam+), mesothelial (Wt1+), immune (Ptprc+), and mesenchymal (Pecam1-, Epcam-, Ptprc-) cell compartments (Fig. S1A-B). Mesenchymal cells were subsequently extracted and analyzed separately (Fig S1C). This revealed numerous transitional cells as well as more mature cell types (Fig. S1D). To characterize pericyte-specific developmental changes, we performed RNA velocity analysis of the mesenchymal populations (Fig. S1E). Based on these predictions, we isolated and reclustered closely related smooth muscle and non-smooth muscle cell types. This revealed eight transcriptionally distinct populations, including pericytes/perivascular type III cells (Cspg4+, Pdgfrb+), airway smooth muscle (Foxf1+, Acta2+, Myh11+), myofibroblasts (Tgfbi+, Pdgfra+), lipofibroblasts (Tcf21+, Gyg+), and chondrocytes (Sox9+, Col2a1+) (Fig. 1A-C). The fraction of cells and mean expression of signature genes for each population were visualized (Fig. 1D).
We identified an Ebf1 + population as early as E10.5 (Fig. 1B-C). At E12.5, two molecularly distinct Ebf1 + populations emerged, one of which we identified as true pericytes/type III perivascular cells (perivascular cell III) due to the expression of general pericyte markers Pdgfrb, Cspg4, and Mcam (Fig. 1D). The second Ebf1 + cluster, which we defined as type II perivascular cells (perivascular cell II), was found to be molecularly distinct from pericytes, although it did co-express many pericyte markers, albeit to a lower extent. Both Ebf1 + populations expressed Notch3, a key regulator of VSMC maturation and vessel stabilization (11, 26). The top differentially expressed markers of the perivascular II cells included Cxcl12 and Sparc, both of which are part of the pericyte secretome and are involved in pulmonary artery muscularization and extracellular matrix (ECM) formation and angiogenesis, respectively (Fig. 1C) (27, 28).
Recently, Ebf1 positive fibroblasts were identified as a novel pulmonary mesenchymal subpopulation in the embryonic mouse lung from gestational age 14.5 (E14.5) onwards (29). Although their transcriptomic profile closely resembled that of pericytes, suggesting that they may share a common lineage, a distinct pericyte population could not be identified. Ebf1 has also been shown to be pericyte specific and to contribute to pericyte and VSMC cell commitment. By contrast, Foxf1, an airway smooth muscle cell (ASMC)–specific transcription factor, has been found to drive airway smooth muscle cell (ASMC) development (30, 31). Here, we identified two different smooth muscle cell lineages, one of which was identified as the VSMC compartment due to the expression of Ebf1 and Heyl (perivascular cell II and III), the other which was determined to represent the ASMC compartment due to the expression of Foxf1 (airway smooth muscle and myofibroblast, Fig. 1D).
Both perivascular type II and III cells were found to specifically express Heyl, another transcription factor tied to VSMC cell fate determination (31). Interestingly, only the perivascular II cells expressed smooth muscle cell (SMC) markers Acta2 and Tagln, which implies that they might be pulmonary VSMC progenitors.
We uncovered a third distinct mesenchymal population which was only detected between E10.5 and E14.5 These cells, which have previously been described as adventitial fibroblasts (32), were termed perivascular type I cells (perivascular cell I), were defined by the expression of Col1a1, which was recently shown to regulate pericyte proliferation, migration, and differentiation, and Dlk1, a more recently discovered pericyte marker (Fig. 1C) (27, 33). Based on these findings, we hypothesized that these perivascular type I cells might give rise to type II and III perivascular cells. To define the temporal trajectories of the perivascular clusters, we performed RNA velocity analysis on our scRNA-seq data. Both the dynamical and stochastic models implied a developmental relationship between the three perivascular cell types, although a single progenitor could not be clearly established (Fig. 1E, S1F).
DCM-time machine tracks pericyte cell state changes throughout lung development
To determine the development of pulmonary perivascular cells, we applied DCM-time machine (DCM-TM) technology (25). This recently developed method offers superior flexibility as it is not constrained by temporal resolution or gene detection limits, which strongly contrasts with RNA velocity. Consequently, it allows for comprehensive whole-transcriptome lineage tracing of target cell populations.
To explore these trajectories, we conducted four distinct pulse-chase experiments, each involving a 48-hour regimen of doxycycline administration to induce the DCM tagging system, followed by the isolation of Cspg4+/Pdgfrb + perivascular cells (perivascular cells I and II) and Pecam1 + endothelial cells (ECs) at E18.5 using FACS (Fig. 2). Sequencing of methylated DNA (MeD-seq) was performed on genomic DNA isolated from these cells, followed by the identification of methylation specific DNA regions (see genome browser plots in supplementary Fig. 2) and normalization for DCM induction efficiency (Fig. S3A, table S2) (34). The time points, ranging from E8.5 to E16.5, were explicitly chosen to reflect the early and late embryonic and pseudoglandular stages of lung development, during which most of the pulmonary vascular growth takes place.
DCM methylation levels for each time point were analyzed to determine to what extent perivascular cell development could be traced retrospectively. Gene meta-analysis indicated that DCM methylation levels of gene bodies in both perivascular cells and ECs were found to be similar to those observed in controls at E8.5-10.5, indicating that the methylation of active genes at his time point was insufficiently maintained after 8 days (Fig. S3B). Consequently, we set the cutoff to E10.5-12.5.
By leveraging the DCM labeling of active genes across the early (E10.5-12.5), middle (E12.5-14.5), and late (E14.5-16.5) time intervals, we meticulously tracked changes in gene expression within pericyte precursors over time. Notably, the DCM labeling of Col1a1 and Dlk1 confirmed their enrichment at the early and middle time points, respectively (Fig. 3A). Furthermore, we observed that genes specific to pericytes attained their peak expression levels at the late time point, with Mcam, Notch3, and Ebf1 demonstrating higher expression levels in comparison to Cpsg4 and Pdgfrb (Fig. 3A). Additionally, examination of smooth muscle cell markers revealed an increasing expression of VSMC-specific genes in the perivascular cells, including Heyl, Egfl6, and Mef2c, thus providing further confirmation that pericytes follow the developmental trajectory of VSMCs (Fig. 3B).
To map temporal gene activity throughout pericyte differentiation, we clustered genes with a DCM signal significantly higher than background levels based on their peak day. We visualized the average expression of these gene clusters on the uniform manifold approximation and projection (UMAP) of our mesenchymal scRNA-seq data, which shows enrichment at specific clusters (Fig. 3C).
This analysis revealed that genes with a temporal DCM methylation profile peaking at the early time point were enriched in perivascular type I cells, while genes peaking at the late time point were enriched in perivascular type III cells (Fig. 3C). Genes with maximal temporal signals at the middle time point were broadly expressed across all three perivascular clusters, indicating that this peak pattern mostly represents ubiquitously expressed genes. Gene ontology (GO) analysis of the DCM labeled genes highlighted gene set enrichment for stem cell pluripotency at the early time point and for VSMC contraction at the late time point (Fig. 3D-F).
In summary, we conclude that perivascular type I cells serve as pericyte and VSMC progenitors, giving rise to both perivascular type II cells, termed intermediate pericytes, and pericytes (Fig. 3G).
Pathways determining pericyte differentiation and pulmonary vascular development
To gain deeper insights into cell state changes during pericyte differentiation, we conducted a KEGG pathway analysis on DCM-labeled genes. Mesenchymal Wnt, Shh, and Vegf signaling pathways play crucial roles in the development of various mesenchymal lineages, including airway and vascular smooth muscle cells and pericytes, and the establishment and expansion of the pulmonary vascular plexus (35–38). Consistent with this, we observed the early enrichment of these signaling pathways during lung development (Fig. 4A-B).
At the middle time point, we observed enrichment of the Notch signaling pathway, aligning with prior reports highlighting its significance in distal angiogenesis, arterial specification, and vessel maturation (39). By contrast, the late time point revealed enrichment of Tgfb and Hippo signaling pathways, both of which are well-known regulators of mesenchymal ECM production, a process that directly follows blood vessel formation (40).
To unravel the cell-cell signaling mechanisms driving pericyte differentiation, we examined the expression of ligands and receptors by performing Cellphone DB analyses on our scRNA-seq data (41). These analyses highlighted the pericyte progenitors and intermediate as sources of ligands that not only signal to each other but also communicate with mature pericytes. This finding implies a regulatory role for these pericyte progenitors on mature pericytes (Fig. 4C). The number of predicted ligand-receptor interactions also revealed that mature pericytes exhibit a lower number of interactions compared to their progenitors, showing that progenitor cells can process more signaling cues due to their ability to differentiate into different cell types. Among the ligands expressed by the pericyte progenitors were Wnt2, Wnt5b, while intermediate pericytes emerged as the primary source of Tgfb signaling effectors, including Tgfb2 and Tgfb3 (Fig. 4D). Interestingly, retinoic acid signaling was highly enriched between the three perivascular populations, with pericyte progenitors emerging as the main source of retinoic acid (Fig. 4E).
We further aimed to elucidate the interplay between ECs and pericyte progenitors in the context of pulmonary vascular development through receptor-ligand analysis of our whole-lung scRNA-seq data. The interaction between ECs and pericytes is essential for the development and expansion of the pulmonary vasculature (6). In line with this, our finding indicated consistent interactions between ECs and both intermediate pericytes and mature pericytes, primarily involving well-established angiogenic pathways including Pdgfb, Angiopoetin, and Notch (Fig. 4F) (6). Notably, we observed that Vegf signaling, a key regulator of early vasculogenesis, mostly occurred between ECs and intermediate pericytes, corroborating the notion that intermediate pericytes represent an early and less differentiated type of pericyte (Fig. 4G).
Temporal expression of known and novel pulmonary pericyte markers
While Cspg4 and Pdgfrb are commonly used molecular markers for pericytes, they have generally failed to serve as unique markers (2, 11, 42, 43). The absence of a single, definitive marker hinders our understanding of pulmonary pericyte ontogeny and differentiation capacity. Based on the differentially expressed genes found in our scRNA-seq and DCM-TM experiments, we evaluated the specificity of novel potential pulmonary pericyte-specific markers using immunocytochemistry. Notch3, Ebf1, and Mcam were analyzed for their temporal and spatial expression in Cspg4/Pdgfrb double positive cells within mouse lungs at E12.5, E15.5, and E18.5.
Consistent with prior research by our group, Cspg4 maintained perivascular expression throughout lung development, while Pdgfrb expression became increasingly restricted to the vasculature by E18.5 (13) (Fig. 5A, S4A, S5). Notably, very few Cspg4/Pdgfrb double positive cells were detected.
At all time points, we found Notch3 expression in all Cspg4 positive cells, with the highest expression occurring in Csp4/Pdgfrb double positive cells (Fig. 5B, S4A). A similar expression pattern was observed for Ebf1 (Fig. S5). This observation confirmed that Ebf1 and Notch3 serve as more selective markers for VSMCs and pericytes in the lung when compared to Cspg4 and Pdgfrb.
To determine whether pericytes and perivascular type II cells could be distinguished from one another, we co-stained lungs for Notch3, Ebf1, and Mcam. Mcam was identified as highly specific for lung perivascular cells at all time points (Fig. 5B, S4B). Furthermore, not all Ebf1 + cells co-expressed Mcam, demonstrating that Mcam can effectively differentiate between intermediate and mature pericytes in the lung.