Robust differentiation of human pluripotent stem cells into mural progenitor cells via transient activation of NKX3.1 – Nature Communications

NKX3.1 activation efficiently differentiates mesodermal intermediates into mural cells

To evaluate the differentiation potential of NKX3.1, we first genetically engineered human iPSCs to express NKX3.1 in response to doxycycline (Dox) using a piggyBac transposon system. Puromycin-selected clones were screened for homogeneous expression of NKX3.1 upon administration of Dox (Supplementary Fig. 1). The engineered clones (termed iPSC-Dox-NKX3.1) remained pluripotent and maintained the expression of pluripotency markers OCT4, SOX2, and NANOG at comparable levels to the parental iPSC counterpart (Supplementary Fig. 1). Next, we developed a two-dimensional, feeder-free, and chemically defined protocol that relies on a timely transition of iPSCs through two distinct stages, each lasting 48 h (Fig. 1A). The first step entails the conversion of iPSCs into intermediate MePCs, which is mediated by the activation of Wnt signaling pathways using the glycogen synthase kinase 3 inhibitor CHIR99021 and is characterized by the transient activation of the TF TBXT (Supplementary Fig. 6B, D). The second step involves the activation of NKX3.1 for 48 h via the provision of Dox in the absence of any growth factors (Fig. 1B). Thereafter, the resulting cells, herein termed iMPCs, were grown in a serum-containing smooth muscle growth medium (SMGM) for additional passages.

Fig. 1: NKX3.1-induced differentiation of MePCs into iMPCs.
figure 1

A Differentiation Schema: Illustration of the stepwise differentiation from iPSCs to iMPCs, detailing mesodermal induction, mural cell specification, and expansion phases. B NKX3.1 Expression: Time-course qRT-PCR analysis demonstrating NKX3.1 upregulation during differentiation (n = 9; ***P < 0.001). C Morphological Progression: Phase-contrast microscopy revealing morphological evolution at various stages (Scale bars: 100 µm; insets 50 µm). D Flow Cytometry: Analysis of CD13 and CD140b (mural cell markers), and TRA1-81 (pluripotency marker) throughout differentiation stages. E Cytoskeletal Markers: Immunofluorescence of iMPCs for α-SMA, SM22, Vimentin, and Calponin with DAPI nuclear staining (Scale bars: 50 µm). F Smooth Muscle Markers: qRT-PCR quantification of SMC gene expression (n = 3, 4, 5, 7; *P < 0.05, ***P < 0.001). G Pericyte Markers: qRT-PCR analysis showing pericyte-specific gene expression (n = 3, 4, 5, 7; *P < 0.05, **P < 0.01, ***P < 0.001). All PCR data is normalized to GAPDH. All data are mean ± s.e.m. n are biological replicates (B, F, G). Statistics are one-way ANOVA with Bonferroni’s post-test analysis (B, F, G). A was partially created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

During the 4 days of differentiation, we observed significant morphological changes in the cells that progressively resembled those of mesenchymal cell types (Fig. 1C). Moreover, iMPCs morphologically resembled mural cells in culture, displaying a characteristic stellate shape (Fig. 1C). We traced the presence of mural populations during differentiation using flow cytometry by analyzing the expression of PDGFRβ (CD140b), a general mural cell marker, and aminopeptidase N (CD13), known to be expressed in mural cells in vivo16. Our two-step protocol rapidly and uniformly converted CD140b-/CD13- iPSCs into CD140b + /CD13+ iMPCs with exceedingly high efficiency. Indeed, 48 h after NKX3.1 activation, ~99% of the cells were CD140b + /CD13+ cells (Fig. 1D). Moreover, after 4 days of differentiation, less than 1% of the cells expressed the TRA1-81 antigen, indicating a negligible presence of undifferentiated iPSCs (Fig. 1D). This effective conversion of MePCs into PDGFRβ + iMPCs via activation of NKX3.1 was reproducible across iPSC lines from three distinct cellular origins (Supplementary Fig. 3).

After differentiation, indirect immunofluorescence confirmed the expression of mural-specific contractile cytoskeletal proteins in iMPCs, including alpha-smooth muscle actin (α-SMA), calponin, transgelin (SM22), and vimentin (Fig. 1E). The expression of these mural cell markers was highly uniform (>90%) and reproducible across iMPCs derived from three distinct iPSC lines (Supplementary Fig. 4). To further confirm mural cell specification, we compared mRNA expression of iMPCs with those of control human primary pulmonary vascular SMCs and adipose tissue-derived mesenchymal stem cells (MSCs) by qPCR (Fig. 1F, G). The expression of selected smooth muscle markers was either equal to (e.g., ACTA2, TPM1) or significantly higher (CNN1, TAGLN, MYOCD, MYH11) in iMPCs compared to SMCs and MSCs (Fig. 1F). Similarly, the expression of various pericyte markers (PDGFRB, CSPG4, DES, NDUFA4L2, PDE5A, and THY1) was either comparable or significantly upregulated in iMPCs, with the exception of PDGFRB expression, which was higher in MSCs (Fig. 1G). Importantly, in the absence of Doxycycline, expression levels of mural cell markers were significantly lower, supporting NKX3.1’s role in mural cell specification (Supplementary Fig. 2).

It is important to note that during the differentiation of MePCs to iMPCs, the expression of NKX3.1 was only transient (Figs. 1B and S6). This transitory expression enabled the possibility of using chemically modified mRNA (modRNA), thus developing a genomic footprint-free protocol. Indeed, transfection of unmodified iPSCs with modRNA encoding NKX3.1 enabled robust transient expression of NKX3.1 (Supplementary Fig. 5). Moreover, activation of NKX3.1 with modRNA in MePCs efficiently produced iMPCs that were indistinguishable from those generated by the Dox-inducible protocol, including a robust expression of mural cell markers (Supplementary Fig. 5).

Next, we compared iMPCs generated using our NKX3.1-induced protocol to SMCs derived from iPSCs through a chemically induced approach inspired by a previously reported study8. Specifically, these induced SMCs (herein termed iSMCs to be consistent with previous terminology) were generated by exposing MePCs to PDGF-BB (10 ng/mL) and Activin A (2 ng/mL) for 48 h (Supplementary Fig. 6). It is important to note that our protocol for generating iSMCs was inspired by, rather than directly following, the protocol described by Patsch et al.8. We adopted only the mesodermal to mural cell differentiation aspects of their protocol (i.e., the use of PDGF-BB and Activin A) to enable a comparable mesodermal stage and effectively compare the outcomes of our NKX3.1 induction protocol with a chemically induced protocol. Cells during both the NKX3.1-induced and chemically-induced protocols exhibited a similar sequential pattern of transient expression of TBXT and NKX3.1, coinciding with their transition through mesodermal and mural cell differentiation stages, respectively (Supplementary Fig. 6). Both iMPCs and iSMCs consistently expressed SMC markers at similar levels, except for ACTA2 and MYOCD, which were significantly higher in iSMCs (Supplementary Fig. 6). Meanwhile, the expression of pericyte markers was more prevalent in iMPCs than in iSMCs (Supplementary Fig. 6). Of note, both iMPCs and iSMCs exhibited minimal MYH11 expression (Supplementary Fig. 6), which is consistent with a subdued expression in SMCs expanded in culture17. Together, these data suggest that while both protocols are effective at generating mural cells, the chemically-induced method preferentially produces cells consistent with an SMC phenotype, whereas the NKX3.1-induced protocol generates iMPCs with characteristics of both SMCs and pericytes.

In summary, the transient activation of NKX3.1 expression in MePCs (via a Dox-inducible system or modRNA) effectively and efficiently converted human iPSCs into cells exhibiting a distinct mural cell phenotype.

Contractile and secretory competence of iMPCs

Next, we assessed the functional attributes of iMPCs, with a particular focus on characteristics commonly associated with mural cells. Key functions explored included calcium influx, contractile properties, extracellular matrix (ECM) synthesis, and the ability to interact with ECs and maintain vessel integrity in vivo.

We examined the contractility of iMPCs by stimulation with vasoconstrictive agents. Calcium imaging demonstrated that both endothelin-1 and carbachol significantly increased intracellular calcium levels in iMPCs, similar to control primary MSCs and SMCs (Fig. 2A, B). To further evaluate their contractile properties, we employed a three-dimensional collagen contractility assay. When exposed to U46619, a thromboxane A2 (TXA2) analog that acts as a potent vasoconstrictor, iMPCs contracted comparably to MSCs (Fig. 2C). These findings indicate that iMPCs can respond to vasoconstrictive stimuli, thus exhibiting a crucial functional characteristic of mural cells in vivo18.

Fig. 2: Functional assessment and secretory profile of iMPCs.
figure 2

A Calcium Imaging: Intracellular calcium flux in iMPCs at Day 4 visualized using a green fluorescent indicator following stimulation with endothelin-1, carbachol, or PBS. Pseudo-coloring indicates intensity, with blue and red representing lower and higher calcium levels, respectively (Scale bars: 40 µm). Quantitative analysis of peak calcium uptake is shown on the right. B Peak Calcium Response: Comparative uptake in MSCs, SMCs, and iMPCs upon endothelin-1 and carbachol stimulation, shown as delta fluorescence (n = 5; ***P < 0.001). C Collagen Gel Contraction: Image and analysis of gel contractility by iMPCs and MSCs in response to U46619 (TXA2 analog) (n = 3; ***P < 0.001). D, E Fibronectin Deposition: Immunofluorescence staining of iMPCs treated with TGFβ and TGFβ inhibitor SB431542 (Scale bars: 100 µm), with quantification of fibronectin intensity per cell shown in (E) (n = 6; **P < 0.01, ***P < 0.001). F FN1 Expression: RT-qPCR analysis of FN1 normalized to GAPDH (n = 4; ***P < 0.001). G Secretome Profiling: Analysis of 55 angiogenic proteins in conditioned media from iMPCs versus SMCs and MSCs, with quantification using Cellpose (Right) (n = 1). H Angiogenic Factor Quantification: Multiplex assay of angiogenic factors in conditioned media, including FGF-2, HB-EGF, HGF, PLGF, VEGF-A, and VEGF-C (n = 2). All data are mean ± s.e.m. n are biological replicates (B, C, E, F, H). Statistics are one-way ANOVA with Bonferroni’s post-test analysis (B, C, E, F, H).

We also assessed the deposition of extracellular fibronectin in iMPCs following treatment with increasing concentrations of TGF-β. Upon treatment with exogenous TGF-β, we observed a substantial increase in fibronectin production, evident at both the protein (Fig. 2D, E) and mRNA (Fig. 2F) levels. Additionally, the introduction of small molecules that inhibit TGF-β signaling (SB431542) effectively prevented fibronectin production, thereby confirming that fibronectin deposition in iMPCs is mediated by TGF-β. The capacity to deposit extracellular fibronectin represents a key functional property of mural cells19.

Central to mural cell function is their capacity to interact with ECs by producing angiogenic factors. We investigated the ability of iMPCs to modulate EC behavior through the secretion of paracrine pro-angiogenic factors and compared it to that of SMCs and MSCs by examining their respective conditioned media using an angiogenesis protein array (Fig. 2G) and quantitative Luminex assay (Fig. 2H). Notably, iMPCs secreted various pro-angiogenic factors, including VEGF-A, PLGF, HB-EGF, HGF, several members of the IGFBP family, as well as members of the serine protease inhibitor (serpin) superfamily of proteins (Serpin E1 and Serpin F1) and urokinase-type plasminogen activator (uPA), among others. The levels of these factors varied among groups, with some being significantly more abundant in iMPCs (e.g., PLGF, HGF), while others were less prominent in iMPCs compared to primary SMCs (FGF2, VEGF-A, VEGF-C). Nevertheless, the overall pro-angiogenic secretome of iMPCs was consistent with what is expected for mural cells20.

Modulation of EC function by iMPCs

To investigate whether proteins secreted by iMPCs could modulate EC activity, we exposed human umbilical cord blood-derived endothelial colony-forming cells (ECFCs, referred to herein as ECs) to medium conditioned by iMPCs (CM-(iMPCs)) in three different in vitro functional assays (Fig. 3). Firstly, using a standard proliferation assay, we observed that direct exposure to CM-(iMPCs) (Fig. 3A) promoted EC mitogenesis, as the number of ECs exposed to CM-(iMPCs) for 72 h was significantly higher than the number observed when cells were exposed to a basal control medium. Furthermore, EC proliferation was comparable to that observed upon exposure to CM-(SMCs) and CM-(MSCs). Similarly, CM-(iMPCs) significantly enhanced the ability of ECs to migrate and re-endothelialize scratched monolayers (Fig. 3B), as well as to assemble into capillary-like structures in three-dimensional cultures (Fig. 3C). Overall, our data indicate that iMPCs effectively modulate EC function in vitro through the secretion of paracrine factors, and their capacity to influence EC activity is comparable to that of control mural SMCs and MSCs.

Fig. 3: iMPCs regulate endothelial cell function.
figure 3

A Growth quantification in ECs exposed to 2-fold concentrated conditioned medium from SMCs, MSCs, and iMPCs (n = 6, 8; ***P < 0.001). B Scratch assay comparing EC migration in conditioned medium from iMPCs and basal medium at 24 h (Left; Scale bar: 200 µm), with a quantified migration percentage of gap closure normalized to the basal medium control (Right; n = 4; ***P < 0.001). C Tube formation assay on Matrigel using conditioned medium, with a representative image (Scale bars: 200 µm) and quantification of total tube length (Right; n = 9; **P < 0.01, ***P < 0.001). D Schematic representation and fluorescent images showing the coculture of ECs and iMPCs (P1) within a microfluidic on-a-chip model. GFP-labeled iMPCs and DsRed-labeled ECs were embedded in a fibrin gel, and the formation of vascular structures was observed after 2 days. (Scale bars: 500 µm). E Immunofluorescent staining of the vascular network formed within the microfluidic chip. ECs are marked by CD31 and VE-Cadherin (red), and iMPCs by α-SMA and SM22 (green). Nuclei are counterstained with DAPI (blue). The inset shows a magnified view of an endothelial lumen surrounded by mural cells (yellow arrowheads). (Scale bars: 100 µm). F Subcutaneous implantation of ECs with or without mural cells into nude mice, with explanted grafts visually assessed at day 7 (Scale bar: 4 mm). G H&E staining identifying perfused blood vessels in implants at day 7 (yellow arrowheads) (Scale bars: 50 µm). H Microvessel density analysis per mm² area (n = 7; *P < 0.05). I IHC showing human-specific ECs (h-CD31 + ) and human perivascular cells (h-Vimentin + ) (Scale bar: 50 µm; inset 10 µm). J iMPC Tracing: GFP and α-SMA staining to track GFP-labeled iMPCs within the perivascular niche in vivo. (Scale bars: 50 µm; insets 10 µm). K Quantification of the percentage of human vessels with human mural cell coverage, comparing ECs implanted with SMCs, MSCs, and iMPCs (n = 4). All experiments in this Figure used nascent iMPCs right after differentiation (96 h). All data are mean ± s.e.m. n are biological replicates (AC, H, K). Statistics are one-way ANOVA with Bonferroni’s post-test analysis (AC, H, K). D, F were partially created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

To evaluate a more physiologically relevant assay, we used an established in vitro model that cocultures iMPCs directly with ECs in a three-dimensional (3D) hydrogel. This microphysiological system—a microfluidic ‘on-a-chip’ model—facilitates the dynamic interaction of cells and the formation of a microvascular network through vasculogenesis. First, we combined GFP-labeled iMPCs and DsRed-labeled ECs within a fibrin gel and examined the ability of the iMPCs to enable vascular morphogenesis (Fig. 3D). This setup led to the formation of vascular structures lined by the DsRed+ ECs within 2 days (Fig. 3D). Furthermore, immunofluorescent staining confirmed the formation of a vascular network within the chip with a continuous endothelial lining marked by CD31 and VE-Cadherin and the presence of α-SMA+ and SM22+ iMPCs serving as perivascular cells adjacent to some of the EC-lined lumens (Fig. 3E). This ‘on-a-chip’ model confirmed the potential of iMPCs, when cocultured with ECs, to assemble complex vascular networks, hence supporting their functionality as mural cells.

To determine whether iMPCs can function as perivascular cells and support blood vessel formation in vivo, we implanted iMPCs together with ECs subcutaneously into immunodeficient mice (Fig. 3F–K). One week post-implantation, we removed the implants and analyzed them for the formation of human-specific vascular networks. Macroscopic examination of the explanted grafts revealed blood perfusion only in those containing ECs with mural cells (either SMCs, MSCs, or iMPCs) (Fig. 3F). Indeed, H&E staining confirmed that all implants seeded with mural cells had formed numerous perfused blood vessels containing murine erythrocytes (Fig. 3G), while grafts with ECs alone failed to form perfused vessels. Careful examination of the blood vessels showed no histological signs of hemorrhage or thrombosis (i.e., platelet aggregation and uniform fibrin deposition), indicating proper functionality. Quantification of the average microvessel densities at day 7 revealed no statistically significant differences between implants prepared with each of the different mural cell populations (Fig. 3H). The ability of iMPCs to support vascular networks in vivo was corroborated using another source of primary ECs. Grafts containing human umbilical vein endothelial cells (HUVECs) and iMPCs yielded mature and perfused vessels, as evidenced by human-specific CD31 staining and the presence of human perivascular cells (Supplementary Fig. 7).

The generation of EC-lined vascular structures depended on the presence of mural cells. Perfused vessels stained positively for human-specific CD31, indicating that the newly formed human vasculature had established functional anastomoses with murine host blood vessels (Fig. 3I). Perivascular involvement of α-SMA-expressing iMPCs was confirmed by human-specific vimentin staining observed in cells surrounding the human EC-lined microvessels (Fig. 3I). In designated experiments, we employed GFP-labeled iMPCs to track their in vivo location. Double staining of GFP and α-SMA revealed that, after 7 days in vivo, GFP-expressing iMPCs were primarily detected in proximity and immediately adjacent to lumenal structures (Fig. 3J), indicating their structural participation in the perivascular compartment of newly formed blood vessels. Quantification of mural cell investment revealed that a substantial majority (>90%) of the human vessels exhibited perivascular coverage, with a significant proportion of these vessels being invested by the transplanted iMPCs (Fig. 3K).

In summary, our data demonstrate that iMPCs were capable of modulating EC function in vitro and in vivo to the same extent as the control mural SMCs and MSCs.

Maturation of iMPCs upon interaction with ECs

Interactions between mural progenitor cells and ECs play a pivotal role in vascular blood vessel development, maturation, and stabilization21,22. Concurrently, these interactions drive the mural progenitors into mature terminally differentiated mural cell types23. To examine the maturation potential of our NKX3.1-induced iMPCs, we subjected them to a seven-day co-culture with ECs, resulting in co-iMPCs. We then compared iMPCs and co-iMPCs (sorted as CD31- cells from the co-culture) using bulk RNA-seq (Fig. 4). This comparison revealed thousands of differentially expressed genes between iMPCs and co-iMPCs (Fig. 4A). Moreover, hierarchical clustering analysis showed that co-iMPCs aligned transcriptionally more closely with primary SMCs and MSCs than iMPCs (Fig. 4D). Pairwise correlation (Fig. 4B) and principal component analyses (Fig. 4C) further confirmed this hierarchical association.

Fig. 4: Maturation of iMPCs upon interaction with ECs.
figure 4

A Co-Culture and Transcriptomics: Diagram showing iMPCs and ECs in co-culture for bulk RNA sequencing. The lower panel indicates up-regulated and down-regulated gene counts in co-cultured iMPCs (co-iMPCs) versus mono-cultured iMPCs. B Transcriptomic Correlation: Pearson’s correlation plot delineating transcriptional profiles among SMCs, co-iMPCs, MSCs, iMPCs, iSMCs, and iPSCs. C Principal Component Analysis: Transcriptional comparison of co-iMPCs with primary SMCs and MSCs relative to iMPCs and iPSCs (n = 3). D Gene Expression Heatmap: Differential gene expression patterns in co-iMPCs compared to iMPCs. E Gene Ontology Enrichment: Up-regulated genes in co-iMPCs associated with mature mural cell functions. F, G Marker Gene Expression: RT-qPCR analysis of (F) SMC and (G) pericyte markers, showing enhanced expression in co-iMPCs versus mono-cultured iMPCs (n = 3, 4, 6, 7, 9; *P < 0.05, **P < 0.01, ***P < 0.001). All PCR data is normalized to GAPDH. H, I Immunofluorescence Characterization of co-iMPCs: co-iMPCs were sorted as CD31- cells from the co-culture. H Sorted cells stained for α-SMA (green) with 3G5 (pericyte marker, red) and nuclei counterstained with DAPI (blue), demonstrating the presence of both SMCs (α-SMA + -/3G5-) and pericytes (α-SMA-/3G5 + ) (Scale bar: 100 µm). I Sorted cells stained for α-SMA (green) (green), MYH11 (red), and DAPI (blue). The co-localization of MYH11 and α-SMA is indicative of cells with a more mature SMC phenotype (yellow arrowheads). (Scale bar: 50 µm). All experiments in this Figure used nascent iMPCs right after differentiation (96 h). All data are mean ± s.e.m. n are biological replicates (F, G). Statistics are one-way ANOVA with Bonferroni’s post-test analysis (F, G). A was partially created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

We conducted gene ontology (GO) enrichment analyses to investigate these transcriptional differences in depth. We found a significant enrichment in co-iMPCs of genes related to mature mural cell functions, including ECM organization, regulation of vasculature, smooth muscle cell contraction, and connective tissue development (Fig. 4E). In addition, our qPCR analysis showed an evident upregulation of several mural cell genes in co-iMPCs relative to iMPCs. These included SMC-associated genes, such as ACTA2, CNN1, TAGLN, MYOCD, and TPM1 (Fig. 4F), as well as pericyte-related genes like CSPG4 and PDE5A (Fig. 4G). Of note, control iMPCs cultured in the same media for seven days without ECs did not exhibit the upregulation of mature mural markers observed when co-cultured with ECs (Supplementary Fig. 8). This general upregulation pattern in co-iMPCs mirrored that observed in primary MSCs after a seven-day co-culture with ECs, underscoring the widely recognized progenitor role of MSCs24.

Moreover, immunofluorescence staining of co-iMPCs confirmed the separate presence of both 3G5+ pericytes (the 3G5 ganglioside antigen is expressed on the cell surface of pericytes) and α-SMA + /3G5- SMCs (Fig. 4H). This 3G5 ganglioside antibody was previously validated to accurately label pericytes in both culture and clinical samples25,26. Moreover, studies have corroborated that 3G5 is not found in vascular SMCs and have utilized the 3G5 antibody for pericyte identification and isolation across various tissues, including human skin and mouse hearts27,28,29. Thus, the 3G5 ganglioside is accepted as a reliable marker for identifying pericytes.

Lastly, it is important to note that iMPCs exhibited only minimal MYH11 expression (a mature SMC marker) before co-culture with ECs at both the mRNA and protein levels (Fig. 4F, I). This is consistent with the well-documented observation that MYH11 expression is generally subdued in SMCs when cultured in isolation. Instead, robust expression of MYH11 is typically reported in vivo, in freshly isolated cells, or in coculture systems that facilitate interactions with ECs. Indeed, upon 7-day coculture of iMPCs with ECs, we observed some cells displayed high levels of both MYH11 and α-SMA, while others exhibited high MYH11 but low α-SMA (Fig. 4I), suggesting a heterogeneous mixture of mural cell phenotypes. This protein-level evidence supports the presence of MYH11+ SMCs among the generated mural cells and reinforces the contextual dependency of MYH11 expression in SMCs.

Taken collectively, these results indicate that NKX3.1-induced iMPCs are not passive EC function regulators. Instead, they display an active, dynamic response to EC interaction that leads to a significant maturation into mural cells. This ability to differentiate, characterized by a distinct upregulation of mature mural cell-associated genes, underscores the progenitor nature of iMPCs.

Recapitulation of mural cell heterogeneity by iMPCs

Our previous analyses using immunofluorescence and qPCR demonstrated that iMPCs express markers typical of multiple mural cell types, including SMCs and pericytes (Fig. 4). To further investigate the extent to which iMPCs can recapitulate mural cell heterogeneity, we employed single-cell RNA sequencing (scRNA-seq) to examine the differentiation of iPSCs into iMPCs as well as the interaction between iMPCs and ECs (Fig. 5). Specifically, we sampled four critical stages of our differentiation protocol corresponding to day 0 (iPSCs), day 2 (MePCs after mesodermal differentiation), day 4 (iMPCs after NKX3.1 activation), and day 11, following a seven-day co-culture of iMPCs with ECs (Fig. 5A). The goal was to determine whether co-culturing with ECs enhances iMPC maturation and diversifies the mural cell population into recognizable perivascular cell types. We used the 10X Genomics platform to obtain data from 10,000 cells at each differentiation stage. The Seurat software facilitated normalization across these time points and enabled integrated cell clustering, resulting in the identification of nine discrete cell populations based on the expression of characteristic cellular markers (Fig. 5B, D, and Supplementary Figs 912). The annotated populations included 1) iPSCs (Cluster #1, marked by OCT4, NANOG, and SOX2), 2) MePCs (Cluster #2, expressing TBX6, MSGN1, MIXL1 and TBXT), 3) iMPCs (Cluster #3) that exhibited high levels of NKX3.1, expressed CSPG4, PDGFRB, and DES, but had reduced expressions of genes encoding for CD73 (NT5E) and contractile proteins (ACTA2, CNN1, and TAGLN), and 4) a fibroblast-like population (Cluster #4) that was characterized by high NKX3.1, PDGFRB, ACTA2, and PDGFRA expression and lacked NT5E expression (Fig. 5D, and Supplementary Figs. 912). Of note, the expression of PDGFRA was prominent in cluster #4 compared to the other annotated mural cell populations, which aligns with common criteria used in the field for identifying fibroblasts30. The temporal evolution of these clusters was consistent with the progression from iPSCs (day 0) to MePCs (day 2) and then to iMPCs (day 4) (Fig. 5C).

Fig. 5: Delineation of mural cell heterogeneity and maturation in iMPCs via scRNA-seq.
figure 5

A Differentiation and Co-Culture Timeline: Schematic illustrating the progression from iPSCs through various stages to iMPCs and their subsequent co-culture with ECs, with an emphasis on the transition points sampled for scRNA-seq. c-SMCs and s-SMCs refer to contractile and synthetic SMCs, respectively. B Cellular Clustering: UMAP projection displaying identified clusters annotated into nine cell types, including iPSCs, MePCs, iMPCs, and various mural cells, based on gene expression markers. C Differentiation Trajectory: UMAP visualization tracking the differentiation from iPSCs to iMPCs and the emergence of mural cell clusters. D Marker Gene Expression: Dot plot summarizing the expression profiles of key markers across clusters, delineating cell identity. E Gene Expression Dynamics: Volcano plots displaying the up- and down-regulated genes in iMPCs after EC co-culture, with emphasis on genes related to contractility and ECM components. F Pseudotime Analysis: UMAP overlaid with pseudotime scores, indicating the developmental progression of cells. G Pseudotime Analysis: Differentiation trajectory from iMPCs to pericytes and SMCs, with annotations indicating distinct cell identities. H Schematic Summary: Illustration summarizing the differentiation of iMPCs into specialized mural cells, highlighting the impact of EC interaction on iMPC maturation and the establishment of mural cell heterogeneity. A, H were partially created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license (https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).

Additionally, we analyzed our scRNA-seq data for markers associated with paraxial mesoderm (TBX6, MSGN1), somites (FOXC2, MEOX2, TCF15), and sclerotome (PAX9, SOX9, NKX3.2)31,32, (Supplementary Fig. 10). At the mesodermal stage (MePCs at day 2, before NKX3.1 activation), TBX6 and MSGN1 were detectable, aligning with their expected expression in early mesodermal differentiation. However, post NKX3.1 activation at day 4 (iMPCs), these markers were not prominently expressed, suggesting a transition away from a general mesodermal identity towards a more defined lineage. Markers associated with somite and sclerotome differentiation showed negligible expression on day 2, and only FOXC2 and SOX9 showed some expression on day 4 in iMPCs (Supplementary Fig. 10). This pattern indicates a minimal influence of NKX3.1 activation on inducing somite or sclerotome identities directly from MePCs.

We also analyzed interactions between iMPCs and ECs. After 7 days of co-culture with ECs, iMPCs matured into three distinct mural cell subpopulations (clusters #5, #6, and #7 at day 11; Fig. 5B, C). These mural cell clusters no longer expressed NKX3.1, confirming its transient activation, but uniformly expressed general perivascular markers PDGFRB and NT5E (CD73). Of note, while iMPCs resembled nascent pericytes, mural cell clusters after co-culture with ECs resembled mature perivascular cells, including pericytes (cluster #5), contractile SMCs (c-SMCs; cluster #6), and synthetic SMCs (s-SMCs; cluster #7). These clusters, while sharing the expression of PDGFRB, exhibited important differences, particularly with respect to genes associated with contractile proteins and ECM production (Fig. 5E). Indeed, a direct comparison of differentially expressed genes revealed a significant upregulation in genes encoding for cell contractility (e.g., ACTA2, CNN1, TAGLN) and ECM proteins (e.g., FN1, COL1A1, COL1A2) in SMCs compared to pericytes (Cluster #5) (Fig. 5E), which is consistent with their perivascular roles in vivo1,33. Meanwhile, a direct comparison between the two clusters of SMCs revealed a clear distinction between the contractile (e.g., upregulation of ACTA2, MYL9, TAGLN) and the synthetic (e.g., FN1, COL5A1, COL4A1) phenotypes of c-SMCs and s-SMCs, respectively (Fig. 5E), consistent with the previous description of these two types of SMC manifestations33,34.

Additionally, we expanded our comparative analysis to evaluate the similarity between our iMPC-derived mural cells (i.e., after co-culture with ECs) and primary mural cells. First, we conducted comparative analyses with publicly available bulk RNA datasets to provide a more precise context. We specifically compared our scRNA-seq data from cells characterized as SMCs (clusters #6 and #7 in Fig. 5) to human aortic SMCs in public datasets. Similarly, our cells identified as pericytes (cluster #5 in Fig. 5) were compared to public datasets of primary human brain pericytes. Pearson correlation analysis of these comparisons demonstrated robust correlations (correlation coefficient ~0.6, p < 0.001) for both sets of comparisons, indicating a substantial transcriptional alignment of our derived mural cells with public datasets (Supplementary Fig. 13A, B).

Furthermore, to establish an additional unbiased benchmark, we used the comprehensive Tabula Sapiens Consortium’s vasculature dataset35. This dataset encompasses a diverse array of vascular endothelial and mural cell types. By overlaying our scRNA-seq data at day 11 (i.e., mural cells generated from iMPCs after 7 days of co-culture with ECs), we observed that our cells identified as SMC-like clusters (clusters #6 and #7 in Fig. 5) exhibited substantial overlap with the reference SMCs (4832 of our SMCs matched the reference Tabula Sapiens SMCs; Supplementary Fig. 13C). Similarly, most of our cells categorized as pericytes (cluster #5 in Fig. 5) closely matched with reference pericytes (808 of our pericytes matched the reference Tabula Sapiens pericytes; Supplementary Fig. 13D). This analysis indicates that most of our SMCs align strongly with established reference SMCs. It also shows that cells from our pericyte cluster show more similarity to the reference pericytes than to SMCs or fibroblasts.

These comparative evaluations demonstrate that the gene expression profiles of our iMPC-derived SMCs and pericytes exhibit significant congruence with established primary human mural cells and a detailed single-cell reference from The Tabula Sapiens Consortium, suggesting the relevance of our differentiation model to in vivo counterparts.

In order to gain further insights into the signals involved in mural cell differentiation and maturation, we examined the cell non-autonomous signals derived from ECs that promote mural cell maturation. Using CellChat to analyze our scRNA-seq data, we identified several key signaling pathways, notably NOTCH and TGF-β, which are known to be implicated in vascular and mural cell development36,37 (Supplementary Fig. 14A, B). To confirm the functional importance of these pathways, we conducted in vitro assays where iMPCs were co-cultured with ECs in the presence of specific pathway inhibitors (Supplementary Fig. 14C). The inhibition of NOTCH signaling with DAPT significantly impaired the maturation of iMPCs in both pericyte and SMC phenotypes, confirming the role of NOTCH in mural cell maturation (Supplementary Fig. 14D). Similarly, inhibition of TGF-β signaling with SB431542 selectively disrupted SMC maturation, indicating its pivotal role in this process while preserving a pericyte-like phenotype (Supplementary Fig. 14E).

Our analysis also included examining the differential gene expression profiles between nascent and more mature pericytes (clusters #5 at days 4 and 11, respectively). We identified distinct gene signatures differentiating early-stage pericytes from their mature counterparts. GO pathway analyses of these DEGs revealed that day 11 pericytes exhibited significant enrichment in genes associated with ECM organization, cellular adhesion, and TGF-β signaling pathways, indicative of a mature mural cell phenotype (Supplementary Fig. 15). In contrast, days 4 pericytes showed enriched expression in genes linked to cell proliferation, regulation of cell differentiation, and Wnt signaling pathways, reflecting their developmental stage closer to mesodermal progenitors (see DEGs in Supplementary Data 1). Moreover, GO and KEGG analyses of differential gene expression between pericytes (cluster #5) and c-SMCs (cluster #6) confirmed significant enrichment in functions associated with ECM organization, cell-matrix adhesion, cellular contractility, and various signaling pathways related to TGF-β signaling in c-SMCs (Supplementary Fig. 15), indicative of the contractile and structural roles typically associated with SMCs.

Next, we performed an in-depth analysis of our scRNA-seq data to elucidate the gene regulatory networks (GRNs) driving the differentiation of our mural cell populations from MePCs into iMPCs, pericytes, and SMCs (Supplementary Fig. 16). Using a combination of TF motif enrichment analysis and gene expression correlation mapping, we identified distinct GRNs that govern the progression from MePCs to mature mural cells, including pericytes and SMCs (Supplementary Fig. 16A). This approach leveraged regulatory elements predicted to be active in each cell state, providing a dynamic view of the transcriptional controls that shape cell fate decisions. For instance, in early-stage MePCs, these networks included regulators such as MIXL1 and MSX1, which are pivotal during mesodermal specification (GRN9, Supplementary Fig. 16B). Later, several GRNs were highly active in iMPCs compared to mature SMCs and pericytes (GRN4 and GRN12, Supplementary Fig. 16B), suggesting that genes included in these GRNs, such as TIMP1, TGFB1, JAK, and PIEZO1, could be influenced by NKX3.1. As the cells progressed toward a more defined mural cell fate, we observed a transition in the active GRN10, with an increased representation of motifs related to TGF-β signaling (TGFB2), ECM production, and contractile function—key aspects of mature mural cell phenotypes (Supplementary Fig. 16B).

Lastly, our trajectory analyses with pseudotime plots provided further insights into the temporal evolution from iPSCs to MePCs and subsequently to iMPCs (Fig. 5F). This analysis also confirmed that iMPC interaction with ECs promoted the development of the various mature mural cell subpopulations, starting with pericytes and progressing to c-SMCs and s-SMCs. Additional pseudotime analysis provided a more precise visualization of the developmental trajectory and maturation stages of different cell subsets derived from iMPCs (Fig. 5G). This pseudo-time trajectory analysis indicated that the pericyte cluster appears temporally closer to the iMPCs than the SMC clusters, suggesting an earlier stage in pericyte development. This is consistent with the notion that iMPCs resemble nascent pericytes. On day 11, the pericyte cluster (#5) manifests earlier than the synthetic SMC cluster (#7) (Fig. 5G), highlighting a developmental hierarchy. Furthermore, our analyses revealed insights into the origins of s-SMCs during co-culture maturation stages. Pseudotime trajectory analysis suggests that synthetic SMCs (s-SMCs) represent a later stage of mural cell differentiation, emerging from contractile SMCs (c-SMCs) under the influence of continuous endothelial interaction, highlighting the dynamic interplay of cell-autonomous and non-autonomous signals in mural cell diversification (Fig. 5G).

In summary, our scRNA-seq analysis substantiated the notion that iMPCs act as mural cell progenitors; after a week of co-culturing with ECs, the iMPCs diversified into distinct perivascular mural cell subpopulations, including pericytes and SMCs. This finding suggests that EC interaction is pivotal in the maturation of iMPCs, allowing for a robust recapitulation of mural cell heterogeneity (Fig. 5H).