Direct reprogramming of human smooth muscle and vascular endothelial cells reveals defects associated with aging and Hutchinson-Gilford progeria syndrome.

Vascular dysfunctions are a common feature of multiple age-related diseases. However, modeling healthy and pathological aging of the human vasculature represents an unresolved experimental challenge. Here, we generated induced vascular endothelial cells (iVECs) and smooth muscle cells (iSMCs) by direct reprogramming of healthy human fibroblasts from donors of different ages and Hutchinson-Gilford Progeria Syndrome (HGPS) patients. iVECs induced from old donors revealed upregulation of GSTM1 and PALD1, genes linked to oxidative stress, inflammation and endothelial junction stability, as vascular aging markers. A functional assay performed on PALD1 KD VECs demonstrated a recovery in vascular permeability. We found that iSMCs from HGPS donors overexpressed bone morphogenetic protein (BMP)-4, which plays a key role in both vascular calcification and endothelial barrier damage observed in HGPS. Strikingly, BMP4 concentrations are higher in serum from HGPS vs. age-matched mice. Furthermore, targeting BMP4 with blocking antibody recovered the functionality of the vascular barrier in vitro, hence representing a potential future therapeutic strategy to limit cardiovascular dysfunction in HGPS. These results show that iVECs and iSMCs retain disease-related signatures, allowing modeling of vascular aging and HGPS in vitro.

Introduction

Physiological and pathological aging represent a major risk factor for the onset of cardiovascular diseases, the leading cause of death worldwide (Benjamin et al., 2019). The in vitro generation of human vascular cells (e.g. smooth muscle cells (SMCs), vascular endothelial cells (VECs)) represents a promising approach to regenerate compromised vascular beds (Leeper et al., 2010), as well as to model specific features of vascular dysfunction (e.g. blood brain barrier damage) (Vatine et al., 2019). Among the available strategies, direct reprogramming of skin fibroblasts into functional cells was previously reported for blood progenitor cells (Szabo et al., 2010), SMCs (van Tuyn et al., 2005) and VECs (Han et al., 2014; Morita et al., 2015). Direct reprogramming of cells from skin biopsies offers a critical advantage compared to other approaches such as the differentiation of induced pluripotent stem (iPS) cells. Indeed, although iPS cells have significantly contributed to the discovery of novel disease mechanisms (Park et al., 2019) as well as to the pre-clinical screening of promising drug candidates (Zhang et al., 2019), the differentiation process of iPS cells requires a mandatory transit through an embryonic-like state which was reported to reset the aging profile of the cells of origin (Mertens et al., 2015). Conversely, direct reprogramming from skin fibroblasts was shown to maintain the cellular signature of aging and to highlight critical factors of human aging in neurons (Mertens et al., 2015).

Previous studies reported that single or combined expression of specific transcription factors were able to directly reprogram skin fibroblasts into SMCs (Long et al., 2008) or VECs (Chen et al., 2019), which then expressed cell identity genes and were able to integrate within the mouse vasculature following subcutaneous or intramuscular implantation (Han et al., 2014; Morita et al., 2015). Strikingly, it was demonstrated that ETV2 alone, which is a master regulator of VEC development and early vasculogenesis (Qiu and Hirschi, 2019), was sufficient to in vitro reprogram human skin fibroblasts into functional VECs (Morita et al., 2015). Similarly, the induced expression of myocardin (MYOCD) was able to strongly induce both smooth and cardiac muscle genes in human foreskin fibroblasts and mesenchymal stem cells (van Tuyn et al., 2005).

Since directly reprogrammed cells, but not iPS cells, preserve the aging signature of the donor (Mertens et al., 2015) and can be largely expanded from an easily obtained skin biopsy compared to other rare cell sources (e.g. blood-derived endothelial progenitor cells) (Mathur et al., 2019), we hypothesized that these cells could represent an effective tool to study cardiovascular aging and disease. In particular, Hutchinson-Gilford Progeria Syndrome (HGPS) is a rare age-related disease induced by mutation in the Lamin A/C gene (Prokocimer et al., 2013). This mutation not only alters the architecture of the nuclear envelope but also impacts the genome organization and gene expression leading to severe cardiovascular damage (e.g. atherosclerosis, vascular calcification), which is responsible for the premature death of patients at an average age of 14 years. In this scenario, the generation of directly reprogrammed vascular cells from HGPS patients would represent an invaluable tool to develop potential new therapies counteracting HGPS vascular degeneration, with the broad possibility to translate them for the treatment of vascular dysfunctions during physiological aging. Hence, the aim of this study was to generate, characterize, validate and use directly reprogrammed vascular cells, both induced VECs (i.e. iVECs) and induced SMCs (i.e. iSMC), to identify gene expression and functional differences between young and old individuals, as well as between healthy and HGPS donors. Our group has a strong expertise in the molecular and transcriptional characterization of human skin fibroblasts, which we recently combined with a machine learning approach to predict the biological aging of human donors (Fleischer et al., 2018). Based on these premises, here we combined RNAseq and functional studies of vascular permeability in 3D microscale models which allowed us to identify potential new biomarkers of vascular aging and to explore the pathological cross-talk between SMCs and VECs in HGPS patients.

Results

iSMCs and iVECs express cell-identity markers and form 3D vascular structures in vitro

Direct reprogramming of skin fibroblasts into tissue-specific cells (e.g. neurons, vascular cells, blood cells) is a relatively new research field which can potentially affect both tissue engineering applications and in vitro disease modeling. Recently, several attempts have been made to generate reprogrammed vascular cells (Han et al., 2014; van Tuyn et al., 2005), but a complete characterization of their gene expression profile compared to the original fibroblast population, as well as to primary vascular cells, is still missing.

Since SMCs and VECs play an important role in human aging and HGPS (Hamczyk et al., 2019; Li et al., 2018), we directly induced iSMCs and iVECs by isolating skin fibroblasts from human donors and overexpressing either the master regulators MYOCD or ETV2, respectively. iSMCs expressed cell-identity markers including alpha smooth muscle actin (αSMA) (Figure 1A, expressed by about 70% cells) and calponin (Figure 1B, expressed by about 80% cells), both in monoculture and when co-cultured with VECs (Figure 1C). Importantly, αSMA-expressing iSMCs contributed to the formation of microvascular networks in 3D fibrin matrices (Figure 1D), thus mimicking the behavior of fibroblasts and differentiated mesenchymal stem cells observed in previous vascular models developed by our (Bersini et al., 2016; Jeon et al., 2014) and other groups (Zheng et al., 2012). In particular, no differences were found in terms of vascular area fraction when VECs were co-cultured with fibroblasts or iSMCs (Figure 1—figure supplement 1). In parallel, we found that CD31+ iVECs (Figure 1—figure supplement 2) were able to self-assemble into vessel-like structures when co-cultured with skin fibroblasts (Figure 1E) or iSMCs (Video 1) in 3D fibrin matrices. These vessel-like structures were patent, as demonstrated by cross-sectional views of the lumens (Figure 1F).

Figure 1.

Characterization of directly reprogrammed vascular cells.

iSMCs express the identity markers αSMA (A) and calponin (B). iSMCs contribute to the formation of endothelial monolayers (C) and 3D microvascular networks (D). iVECs form 3D vessel-like structures in fibrin matrix (E) and show open lumens (F). Both iSMCs and iVECs express cell identity genes showing a progressive transition from fibroblasts to mature vascular cells (G–J). iSMC gene expression profile compared to primary SMCs and fibroblasts (analysis based on 1,852 DE genes between primary SMCs and fibroblasts). N = 6 donors for iSMCs and fibroblasts, N = 3 donors for SMCs. Z score = ± 3. DE genes with log2FC > 1 and FDR < 0.05 (G). Gene ontology analysis reveals that iSMCs upregulate genes associated with muscle development, contraction and blood pressure regulation compared to fibroblasts (H). iVEC gene expression profile compared to primary VECs and fibroblasts (analysis based on 1,780 DE genes between primary VECs and fibroblasts). N = 6 donors for iVECs, primary VECs and fibroblasts. Z score = ± 4. DE genes with log2FC > 1 and FDR < 0.05 (I). Gene ontology showing that iVECs upregulate genes associated with chemotaxis, blood vessel morphogenesis and VE cadherin–VEGFR complex compared to fibroblasts, indication of ongoing differentiation towards and endothelial phenotype (J). Scale bars: 25 µm.

Region P4 refers to CD31 positive cells labeled with AlexaFluor488 secondary antibody.

Figure 1—figure supplement 1.

Quantification of the vascular area fraction in 3D matrices embedding a co-culture of VECs and fibroblasts or iSMCs (Student’s t-test with p=0.61).

Figure 1—figure supplement 2.

Representative FACS plot showing iVEC sorting based on CD31 expression.

Region P4 refers to CD31 positive cells labeled with AlexaFluor488 secondary antibody.

Video 1.

Representative movie showing a 3D reconstruction of in vitro co-cultured iVECs (white) and iSMCs (red) in a fibrin matrix.

To further define the reprogramming status of iSMCs and iVECs, we performed RNAseq analysis on these induced cells and compared them with the donor fibroblast population as well as with human primary smooth muscle and endothelial cell lines. Specifically, we initially compared the expression profile of skin fibroblasts (N = 6 donors) vs. primary SMCs (N = 3 donors, data obtained from GEO) or skin fibroblasts (N = 6 donors) vs. primary VECs (N = 6 donors, data generated for this study using commercially available skin microvascular endothelial cells) (Supplementary file 1). This analysis allowed us to identify about 1800 differentially expressed (DE) genes in both comparisons, which defined the identity of fibroblasts, SMCs and VECs (Supplementary file 2, 3 and 4). Then, we used these DE genes as a benchmark to analyze the gene expression of iSMCs (Figure 1G and H) and iVECs (Figure 1I and J). The comparison among skin fibroblasts, primary vascular cells and reprogrammed cells highlighted multiple gene clusters representative of specific gene expression signatures. Strikingly, we found that the SMC identity gene ACTA2 (Figure 1G, cluster 2) and multiple endothelial-specific genes (e.g. FLT1, KDR, CDH5, TIE2, ROBO4, NECTIN2) (Figure 1I, cluster 2) were enriched in both reprogrammed cells and primary vascular cells when compared to the original skin fibroblasts. As typical for induced cells, they did not express all cell identity markers of primary vascular cells. For example, iVECs expressed less CD31 and TJP1 compared to primary endothelial cells (Figure 1I, cluster 3). However, the expression level of KDR (an important gene defining endothelial cell identity) was 6.62 log2FoldChange (FC) higher in primary VECs vs. fibroblasts and 6.38 log2FC higher in iVECs vs. fibroblasts. At the same time, another endothelial-specific gene (i.e. CDH5) was upregulated in both primary cells (log2FC = 6.7) and iVECs (log2FC = 4) compared to fibroblasts, although at different levels (Supplementary file 5). Similarly, both primary and reprogrammed SMCs upregulated key identity genes including ACTA2 and MYH10 compared to fibroblasts (Supplementary file 5). Notably, the expression of MYH10 was lower in reprogrammed vs. primary SMCs. Furthermore, key skin fibroblast cell identity genes (e.g. FSP1, SLUG, TWIST1, TWIST2) (Figure 1G, cluster 4; Figure 1I, cluster 1) were downregulated in both reprogrammed and primary vascular cells (Supplementary file 5).Finally, gene ontology analysis revealed that iVECs upregulated genes linked to blood vessel morphogenesis and VE-cadherin-VEGF receptor complex when compared to fibroblasts (Figure 1J), while iSMCs upregulated genes associated with muscle contraction and muscle development (Figure 1H).

Taken together, the majority of the markers expressed by reprogrammed cells remained fibroblast-like. After carefully considering this critical point, we would like to highlight that reprogrammed SMCs and VECs expressed typical cell identity genes, self-organized into 3D vascular structures mimicking simple vascular networks and showed a global gene expression signature representing a gradual transition towards fully mature cells.

Induced vascular cells show gene-expression and functional differences between young vs. old donors

Previous studies demonstrated that directly reprogrammed, but not iPS cell-derived neurons, were able to retain the transcriptional aging signature of the donor (Mertens et al., 2015). We therefore asked if iSMCs and iVECs could be employed to identify genes involved in the vascular degeneration occurring during physiological aging (Ungvari et al., 2018). Aging is characterized by a progressive decrease in the functionality of the endothelial barrier (Palmer and Ousman, 2018), formation of atherosclerotic lesions, reduced contractility and calcifications of SMCs (Yao et al., 2013). Hence, we specifically aimed at identifying genes potentially involved in these processes. To answer this question, we established fibroblast cell lines from skin biopsies of N = 3 young (19–30 y.o.) and N = 3 old (62–87 y.o.) healthy donors (Supplementary file 1) and directly reprogrammed them into iSMCs or iVECs using the master regulators MYOCD and ETV2, respectively. It is important to highlight that the expression of MYOCD and ETV2 was comparable in reprogrammed cells derived from young and old donors (log2FC = 0.03 for MYOCD and log2FC = 0.19 for ETV2 comparing young and old cells; no statistical difference comparing young and old cells). DE analysis highlighted a set of genes (21 DE genes for iSMCs; 9 DE genes for iVECs) that changed according to the age of the donor for either iSMCs (Figure 2A) or iVECs (Figure 2B). We then analyzed the expression of a few cell identity genes in iSMCs or iVECs derived from young and old donors without finding obvious age-related differences (Supplementary file 6 and Figure 2—figure supplement 1). Together with the similar expression of MYOCD and ETV2 in cells derived from young and old donors, these data suggest that the reprogramming efficiency should not be affected by the age of the donor and that potential differences between young and old cells do not seem to be due to different levels of reprogramming.

Figure 2.

Vascular cells reprogrammed from young vs. old donors show gene expression and functional differences.

Heat-map representing DE genes between reprogrammed vascular cells from young (N = 3) vs. old (N = 3) donors (iSMCs (A), iVECs (B)). Z score = ± 1.5. DE genes with log2FC > 1 and FDR < 0.05. Quantification of BMP2 (C) and PALD1 (D) expression in human skin biopsies from young vs. old donors (N = 2 donors per condition; N = 10 tissue sections per condition; Student’s t-test with p<0.001 (***) and p<0.05 (*)). Representative images of BMP2 (E) and PALD1 (F) from skin biopsies obtained from young vs. old donors. Scale bars: 50 µm. Quantification of vascular permeability using young vs. old reprogrammed cells (G–J). Schematic showing the generation of an endothelial monolayer covering the interface with a 3D matrix embedding SMCs. Vascular permeability was quantified by measuring the variation of 70 kDa dextran fluorescent intensity across the interface (G). Representative images of dextran flow (red) through the endothelial monolayer in presence of young vs. old reprogrammed cells. Scale bar: 100 µm (H). Quantification of vascular permeability in presence of young vs. old iVECs (I, at least N = 40 independent measurements per condition in three biological replicates; ANOVA with Holm-Sidak test; comparison with primary VECs (** is p<0.01 and *** is p<0.001); comparison with young iVECs (♦♦♦is p<0.001); comparison with old iVECs (••• is p<0.001)) or young vs. old iSMCs (J, at least N = 40 independent measurements per condition in three biological replicates; ANOVA with Holm-Sidak test; * is p<0.05, ** is p<0.01 and *** is p<0.001). Quantification of vascular permeability in presence of PALD1 KD (K) or BMP2 KD (L) in primary VECs and SMCs, respectively (at least N = 60 independent measurements per condition in three biological replicates for each KD experiment; Student's t-test; *** is p<0.001).

Fibroblasts were included as control. Z score = ± 1. All these genes were not DE comparing young and old donors.

Representative image showing young (mCherry MYOCD) and old (EGFP MYOCD) iSMCs. Experiments were also performed with the reciprocal tagging. At least N = 25 independent measurements in three biological replicates. Student’s t-test, p<0.01 (**). Scale bar: 50 µm.

Scale bars: 100 µm.

Biological processes upregulated in old vs. young mouse VECs (pool of bone marrow, brain and skin VECs). It is interesting to observe the upregulation in old mouse VECs of biological processes associated with immune system activation and inflammation, consistent with the general concept that aging is associated with chronic inflammation. Cells were derived from N = 3 animals per age group and then pooled for RNAseq.

The higher amount of Y658 staining in old iVECs (Student’s t-test with p<0.01 (**) from N = 3 biological replicates) suggests that cell-cell junctions might be compromised in reprogrammed cells derived from old donors. Nuclei: blue. Scale bars: 25 µm.

VECs stained with VE-cadherin (white) cover the lateral channel and the interface with the 3D matrix embedding iSMCs (red, αSMA). iSMCs from young donors migrate towards the endothelial layer. Nuclei: blue. Scale bar: 50 µm.

Note that BMP2 KD was validated using qPCR due to technical problems with any of the tested anti-BMP2 antibodies (Western Blot). The use of these antibodies resulted in multiple aspecific bands.

Figure 2—figure supplement 1.

Heat-maps representing key cell identity genes for both iSM cells and iVECs derived from young and old donors.

Fibroblasts were included as control. Z score = ± 1. All these genes were not DE comparing young and old donors.

Intriguingly, among the 21 DE genes, iSMCs from older donors showed upregulation of bone morphogenetic protein 2 (BMP2), which is involved in the vascular calcification of blood vessels during aging and disease (Petsophonsakul et al., 2019). Significantly, we found that αSMA+ cells in human skin biopsies from old (N = 2) vs. young (N = 2) donors expressed higher levels of BMP2 (217.4 ± 10.9% vs. 100.0±10.4% after normalization, Figure 2C and E; Videos 2 and 3), in agreement with the RNAseq findings. Noteworthy, the pattern of αSMA+ cells suggests that these cells are actually surrounding blood vessels rather than being isolated in the skin extracellular matrix (Figure 2E). Finally, we co-cultured iSMCs generated from a young (19 y.o.) and an old (67 y.o.) donor finding higher expression of BMP2 in the iSMCs derived from the old donor (Figure 2—figure supplement 2), confirming observations seen with RNAseq and immunofluorescence of skin biopsies.

Video 2.

Representative movie showing a 3D reconstruction of human skin biopsy from a young (21 y.o.) donor stained for BMP2.

From the left: panel one is BMP2, panel two is αSMA, panel three is DAPI and panel four is merge.

Video 3.

Representative movie showing a 3D reconstruction of human skin biopsy from an old (72 y.o.) donor stained for BMP2.

From the left: panel one is BMP2, panel two is αSMA, panel three is DAPI and panel four is merge.

Figure 2—figure supplement 2.

BMP2 expression by iSMCs reprogrammed from young (19 y.o.) vs. old (67 y.o.) donor.

Representative image showing young (mCherry MYOCD) and old (EGFP MYOCD) iSMCs. Experiments were also performed with the reciprocal tagging. At least N = 25 independent measurements in three biological replicates. Student’s t-test, p<0.01 (**). Scale bar: 50 µm.

In parallel, DE analysis performed on iVECs highlighted the upregulation of GSTM1 and PALD1 in cells derived from old vs. young donors. GSTM1 is generally activated in presence of oxidative stress and inflammatory conditions (Lopes-Aguiar et al., 2017), while PALD1 is an endothelial-specific gene inhibiting endothelial junction stability (Nitzsche, 2016). To test if these genes were actually dysregulated in human donors, we analyzed human skin biopsies and found that CD31+ VECs from old vs. young healthy donors indeed expressed higher levels of PALD1 (194.9 ± 30.0% vs. 100.0±24.3% after normalization, Figure 2D and F; Videos 4 and 5) and GSTM1 (Figure 2—figure supplement 3). This result shows that the gene expression differences observed in directly reprogrammed cells partially recapitulate signatures of aging in humans. Strikingly, DE analysis on primary VECs from mouse skin, brain and bone marrow also highlighted a clear upregulation of GSTM1 and PALD1 in cells obtained from old (18 months) vs. young (3 months) animals (N = 3 animals per condition, Figure 2—figure supplement 4). Together, these data suggest that both PALD1 and GSTM1 might represent evolutionary conserved vascular aging markers.

Video 4.

Representative movie showing a 3D reconstruction of human skin biopsy from a young (21 y.o.) donor stained for Paladin 1.

From the left: panel one is Paladin, panel two is CD31, panel three is DAPI and panel four is merge.

Video 5.

Representative movie showing a 3D reconstruction of human skin biopsy from an old (72 y.o.) donor stained for Paladin 1.

From the left: panel one is Paladin, panel two is CD31, panel three is DAPI and panel four is merge.

Figure 2—figure supplement 3.

Representative images of GSTM1 expression by endothelial cells within human skin biopsies from young vs.old donors (N = 2 donors per condition, 10 tissue sections per condition; Student’s t-test with p<0.01 (**)).

Scale bars: 100 µm.

Figure 2—figure supplement 4.

FPKM values for PALD1and GSTM1 from RNAseq performed on primary mouse endothelial cells from bone marrow, brain and skin of young (3 months) vs. old (18 months) mice.

Biological processes upregulated in old vs. young mouse VECs (pool of bone marrow, brain and skin VECs). It is interesting to observe the upregulation in old mouse VECs of biological processes associated with immune system activation and inflammation, consistent with the general concept that aging is associated with chronic inflammation. Cells were derived from N = 3 animals per age group and then pooled for RNAseq.

Next we wanted to test if these aging markers were functionally and mechanistically linked to vascular aging. Since abnormal expression of BMPs has been linked to vascular dysfunction (Cai et al., 2012; Csiszar et al., 2009) and upregulation of PALD1 to compromised VEC junctions (Nitzsche, 2016), we hypothesized that the identified DE genes might reflect a functional change in vascular permeability. To test this directly, we employed an in vitro 3D model where VECs lining a microfluidic channel form a continuous barrier separating the channel from a 3D fibrin matrix embedding SMCs (Figure 2G and H). The presence of old vs. young iVECs (N = 3 donors per condition) lining the interface with the matrix channel increased the vascular permeability (146.1 ± 5.4% vs. 100.0±4.6% after normalization for old vs. young iVECs). The difference in permeability can be explained by partial damage to cell-cell junctions, as indicated by the different pattern of Y658 VE-Cadherin in old vs. young iVECs (Figure 2—figure supplement 5). Comparison with positive control (primary VECs lining the channel) and negative control (fibroblasts lining the channel) revealed that permeability of young (131.9 ± 6.1%) and old (192.7 ± 7.1%) iVECs was significantly lower compared to fibroblasts (250.0 ± 12.1%), reaching values close to primary VECs (100.0 ± 4.7%) (Figure 2I, data normalized to the permeability of primary VECs). Similarly, the presence of old vs. young iSMCs (N = 3 donors per condition) in the fibrin matrix channel induced a significant increase in the leakiness of the endothelial barrier (149.8 ± 11.3% vs. 100.0±4.7%, data normalized to young iSMCs). iSMCs from young donors induced even lower levels of vascular permeability compared to primary SMCs (78.2 ± 3.7% vs. 100.0 ± 4.3%) (Figure 2J, data normalized to primary SMCs), demonstrating that direct reprogramming generates functional cells which contribute to the maintenance of the vascular barrier. Surprisingly, we also found that iSMCs from young vs. old donors were more migratory, being able to reach the endothelial barrier at the interface with the 3D matrix (Figure 2—figure supplement 6). These findings are not only in agreement with in vivo observations but directly link iSMCs to the appearance of functional and morphological changes during aging.

Figure 2—figure supplement 5.

Representative images and quantification of Y658 VE-Cadherin (white) in iVECs derived from young and old donors.

The higher amount of Y658 staining in old iVECs (Student’s t-test with p<0.01 (**) from N = 3 biological replicates) suggests that cell-cell junctions might be compromised in reprogrammed cells derived from old donors. Nuclei: blue. Scale bars: 25 µm.

Figure 2—figure supplement 6.

Representative images of microfluidic devices seeded with iSMCs reprogrammed from young or old fibroblasts.

VECs stained with VE-cadherin (white) cover the lateral channel and the interface with the 3D matrix embedding iSMCs (red, αSMA). iSMCs from young donors migrate towards the endothelial layer. Nuclei: blue. Scale bar: 50 µm.

To test the functional significance of two of the identified DE genes that might play a role in vascular permeability (i.e. PALD1 and BMP2), we performed knockdown (KD) experiments on VECs and SMCs (Figure 2—figure supplement 7). Collected data showed that PALD1 KD in VECs decreased vascular permeability compared to luciferase control (51.5 ± 3.4% vs. 100.0 ± 4.3%) (Figure 2K, data normalized to the permeability with luciferase control). On the other side, KD of BMP2 in SMCs induced a weak, but reproducible, decrease in permeability compared to luciferase control (86.3 ± 3.8% vs. 100.0 ± 4.3%) (Figure 2L, data normalized to luciferase control).

Figure 2—figure supplement 7.

Validation of KD of PALD1 and BMP2 in VECs and SMCs, respectively (N = 2 biological replicates).

Note that BMP2 KD was validated using qPCR due to technical problems with any of the tested anti-BMP2 antibodies (Western Blot). The use of these antibodies resulted in multiple aspecific bands.

Overall, while we do not claim that reprogrammed cells are transcriptionally and functionally identical to primary cells, these data clearly demonstrate that reprogramming induces a minimum set of transcriptional and functional features which allow to capture a fraction of the vascular changes occurring during aging and disease. This aging signature could be in principle used to longitudinally identify features of vascular degeneration with minimally invasive skin biopsies.

Induced vascular cells recapitulate hallmarks of vascular dysfunction of HGPS patients

HGPS is a rare form of accelerated aging due to mutation in the Lamin A/C gene. Patients exhibit severe cardiovascular damage which leads to premature death (Stehbens et al., 2001). In particular, HGPS is characterized by dysfunctional SMCs. Phenotypic changes in SMCs are often associated with vascular calcification (Wei et al., 2018) and hypertension (Miriyala et al., 2006) in a broad spectrum of diseases. For these reasons we sought to determine if iSMCs directly reprogrammed from HGPS vs. healthy fibroblasts could retain a signature of vascular dysfunction, with the ultimate goal to identify new potential therapies to slow down the process of vascular degeneration.

Following the same strategy used to identify signatures of vascular aging, we performed RNAseq on iSMCs derived from N = 3 healthy donors vs. N = 8 HGPS donors (Supplementary file 1) revealing 93 DE genes (Figure 3A). Among the identified DE genes, we found upregulation of BMP4, which is generally overexpressed during the early onset of vascular calcification (Wei et al., 2018). Moreover, secreted BMP4 enhances the activation of the endothelium and induces production of reactive oxygen species by VECs (Lowery and de Caestecker, 2010), compromising the vascular barrier. To test the clinical significance of BMP4, we obtained human blood serum from HGPS and age-matched healthy controls (N = 7 donors per group) from the Progeria Research Foundation. Interestingly, we found that BMP4 was higher in healthy vs. HGPS donors below 11 years old, while a reverse trend was observed for children above 12 years old (Figure 3B). Analyzing these data from an aging perspective, we found a significant decrease in the level of BMP4 in older vs. younger healthy individuals (p=0.04) which was lost in the HGPS samples. This abnormal trend suggests that a potential correlation might exist between the vascular damage in HGPS patients and circulating BMP4. Indeed, it is reported that the level of BMP4 is high during early development, being a key regulator of musculo-skeletal system formation and vascularization. BMP4 decreases during late childhood/young adulthood and then increases again during physiological aging (Meyers et al., 2016), as well as in pathological conditions including metabolic syndrome (Son et al., 2011), leukemia (Voeltzel et al., 2018) or bone fractures (van Baardewijk et al., 2013). Given the extremely limited number of available samples from HGPS patients, we then collected blood from HGPS mouse models carrying the progerin mutation G609G (N = 3, 13–14 weeks old) and we compared the level of circulating BMP4 with age-matched healthy mice (N = 3, 15 weeks old) as well as with extremely old mice (N = 5, about 2 years old). Collected data revealed a higher level of BMP4 in the blood of HGPS mice (Figure 3C), hence further corroborating the analyses performed on human samples.

Figure 3.

iSMCs reprogrammed from HGPS fibroblasts show signatures of vascular dysfunction.

Heat-map representing DE genes between iSMCs reprogrammed from healthy (N = 3) vs. HGPS (N = 8) donors. Z score = ± 2. DE genes with log2FC > 1 and FDR < 0.05 (A). ELISA measurement of BMP4 in the serum of HGPS and age-matched healthy human donors (B, Student’s t-test with p=0.4 for normal vs. HGPS 5–11 y.o.; p=0.575 for normal vs. HGPS 12–18 y.o.; Student’s t-test with p=0.04 for 5–11 y.o. vs. 12–18 y.o. normal donors; Student’s t-test with p=0.4 for 5–11 y.o. vs. 12–18 y.o. HGPS donors), as well as in the serum of HGPS, young (age-matched) and old mice (C, ANOVA with Holm-Sidak test with p=0.79 for young vs. old; p=0.001 for young vs. HGPS; p<0.001 for old vs. HGPS). Quantification of vascular permeability in presence of iSMCs reprogrammed from young (N = 6), old (N = 6) and HGPS (N = 6) donors. At least N = 95 independent measurements per condition in three biological replicates. ANOVA with Holm-Sidak test with p<0.001 (***) (D). Schematic showing the vascular permeability assay in presence of antibodies specifically blocking secreted BMP4 and Jagged 1 (E). Quantification of vascular permeability comparing iSMCs from HGPS donors (ctrl) or the same cells treated with antibodies blocking BMP4 or Jagged 1. At least N = 60 independent measurements per condition in three biological replicates. ANOVA with Holm-Sidak test with p<0.001 (***) (F). Quantification of the effect of BMP4 and Jagged one blocking antibody on Y658 VE-cadherin area fraction. N = 6 independent measurements per condition in three biological replicates. ANOVA with Holm-Sidak test with p<0.05 (*) or p<0.01 (**); p=0.897 for BMP4 Ab vs. Jagged1 Ab (G). Representative images of Y658 VE-cadherin in presence of BMP4 and Jagged one blocking antibodies (H–J). Scale bars: 25 µm.

These supplementary data refer to N = 1 biological replicate.

Multiple dysregulated pathways contribute to the age-related and pathological vascular remodeling observed in human patients. In this context, the Jagged 1-Notch pathway plays a pivotal role. Indeed, recent studies demonstrated that Jagged 1-mediated Notch activation induces hyper-permeability of blood vessels in a model of diabetic retinopathy and that blocking Jagged 1 can prevent retinal edema (Miloudi et al., 2019). Intriguingly, we found that iSMCs from HGPS vs. healthy donors upregulated JAG1, suggesting its potential contribution in driving the pathological remodeling of blood vessels in HGPS patients.

Based on these findings, we hypothesized that the identified DE genes in reprogrammed iSMCs from HGPS vs. healthy donors could partially explain potential changes in the vascular barrier functionality. To test this hypothesis, we employed microfluidic devices embedding iSMCs reprogrammed from fibroblasts of young, old or HGPS donors (N = 6 donors per group) in co-culture with old VECs. Strikingly, we found that the presence of HGPS iSMCs determined a significant increase in the vascular leakiness compared to young donors (197.4 ± 7.5% vs. 100.0±4.6% after normalization, Figure 3D).

To verify that BMP4 and Jagged 1 were actually playing a key role in the regulation of vascular permeability, we treated iSMCs from HGPS donors with antibodies blocking secreted BMP4 or Jagged 1 (Figure 3E). Strikingly, we found a significant decrease in the vascular leakiness upon antibody treatment in both conditions compared to control (27.8 ± 1.6% vs. 100.0 ± 4.4% for BMP4 antibody after normalization and 61.5 ± 3.8% vs. 100.0 ± 4.4% for Jagged 1 antibody after normalization; Figure 3F and Videos 6, 7 and 8). Noteworthy, BMP4 antibody was able to reduce the vascular permeability to a value even lower than the level detected in presence of iSMCs from young donors (indeed we quantified 72.2% decrease in vascular permeability with BMP4 antibody-treated HGPS cells and only 49.3% decrease comparing iSMCs from young vs. HGPS donors). Immunofluorescence staining of the endothelial cell-cell junction marker VE-cadherin did not show any macroscopic difference between HGPS iSMCs with and without BMP4/Jagged 1 blocking antibody. However, we did observe an increase in the amount of Y658 VE-cadherin (indication of cell-cell junction destabilization) in presence of untreated iSMCs, in agreement with the higher vascular leakiness (Figure 3G–J).

Video 6.

Representative movie showing 70 kDa dextran diffusion through the endothelial barrier in presence of iSMCs reprogrammed from HGPS donors.

Video 7.

Representative movie showing 70 kDa dextran diffusion through the endothelial barrier in presence of iSMCs reprogrammed from HGPS donors and treated with BMP4 blocking antibody.

Video 8.

Representative movie showing 70 kDa dextran diffusion through the endothelial barrier in presence of iSMCs reprogrammed from HGPS donors and treated with Jagged 1 blocking antibody.

Surprisingly, we also found that primary, healthy VECs upregulated BMP4 and other genes associated with calcification, skeletal system development and extracellular matrix organization when co-cultured with HGPS fibroblasts (Figure 3—figure supplement 1, N = 1 biological replicate).

Figure 3—figure supplement 1.

Biological processes upregulated in primary, healthy VECs when 3D co-cultured with HGPS vs. healthy fibroblasts for 25 days.

These supplementary data refer to N = 1 biological replicate.

Overall, using our validated approach we identified two previously unexplored target proteins which are secreted by HGPS SMCs and directly affect the functionality of the vascular barrier. These secreted proteins can be targeted through blocking antibodies to revert the phenotype of VECs, suggesting a potential application to restore the barrier function in HGPS patients.

Discussion

We have described a powerful strategy to study vascular aging and human pathologies affecting the cardiovascular system through directly reprogrammed vascular cells. To validate our strategy, we have provided to our knowledge the most complete characterization of reprogrammed vascular cells by comparing them with the skin fibroblasts from which they are derived and with primary vascular cells. While both iSMCs and iVECs express key cell identity markers and contribute to the formation of in vitro 3D vascular structures, their degree of maturation does not allow ta claim of a complete replica of their primary counterparts. Indeed, most of the markers expressed by reprogrammed cells remained fibroblast-like and all the results here presented should be viewed cwith this point in mind. Nevertheless, reprogrammed cells showed indications of progressive transition from the fibroblast to the vascular lineage considering a small set of cell identity genes. These cells could then be successfully employed to identify genes involved in the vascular degeneration observed during physiological aging. Despite the limited number of DE genes comparing reprogrammed cells from young vs. old donors, the identified targets reflect typical changes expected during vascular aging, as demonstrated by their validation within human skin biopsies. While we were surprised for the low number of identified DE genes between young and old donors, this observation is consistent with other studies comparing cells from donors of different biological age (Fleischer et al., 2018; Mertens et al., 2015).

Strikingly, we found that reprogrammed cells from young vs. old donors were also characterized by functional differences, as demonstrated by the compromised barrier function in presence of iSMCs or iVECs from old individuals. We noticed that the levels of permeability were about 50% higher in presence of reprogrammed cells from old vs. young donors. The observed difference in permeability seems to reflect intrinsic differences between young and old cells, considering that the expression level of a small set of cell identity genes was not affected by the age of the donor. It is important to keep in mind that our microscale models do not fully recapitulate the vascular niche, including tissue-specific cells and immune cells, which might contribute to changes in vascular functionality during aging and disease. The absence of these components might partially explain why we observed limited (~50%) differences in vascular functionality. Further, it is important to mention that the contribution of SMCs to the vascular barrier could be in terms of both paracrine secretion and direct support to the endothelial monolayer. While confocal images showed that iSMCs from old donors did not physically interact with iVECs, differential expression analysis highlighted that iSMCs from old donors upregulated genes which might interfere with the integrity of the barrier (e.g. BMP2 which was identified through RNAseq and then through immunofluorescence of skin biopsies). Although we cannot separate the contribution of these two aspects, it is likely that the combination of physical support and secreted factors contributes to the observed differences in the barrier functionality in presence of young vs. old iSMCs.

Once validated in the context of physiological aging, we applied our strategy to investigate potential mechanisms involved in the vascular degeneration of HGPS, which is the leading cause of death for patients affected by this disease. HGPS is characterized by a mutation in the Lamin A/C gene causing structural defects in the nuclear envelope and leading to the appearance of signs of accelerated aging (Prokocimer et al., 2013). One of the main cardiovascular complications observed in HGPS is the loss of SMC coverage of blood vessels (Stehbens et al., 2001; Varga et al., 2006). Dysfunctional SMCs have been associated with several vascular damages, including vascular calcification (Wei et al., 2018) and hypertension (Miriyala et al., 2006). However, the molecular mechanisms leading to this vascular degeneration in HGPS and other cardiovascular diseases have not yet been completely identified.

In a recent tissue engineered blood vessel model, iPS cell-derived SMCs from a HGPS donor showed increased calcification and apoptosis when compared to cells obtained from a healthy donor (Atchison et al., 2017). However, no genetic analyses were performed to investigate potential targets or biological mechanisms involved. Herein we combined RNAseq and functional assays of vascular permeability to compare iSMCs from healthy vs. HGPS donors and we identified novel targets associated with HGPS vascular dysfunction. We did not find overlapping genes comparing old vs. young and HGPS vs. healthy cells. However, we were surprised to find that BMP2 was upregulated in old donors while BMP4 was highly expressed by HGPS cells. BMP2 and BMP4 share a 62% sequence similarity and are functionally equivalent. Hence, BMP2/4 dysregulation might underlie a common mechanism leading to vascular degeneration during physiological aging and disease. In this context, BMP4 is not only linked to vascular calcification. Indeed, it directly affects the stability of the endothelial barrier by increasing the vascular permeability both in vitro and in vivo (Helbing et al., 2017). Hence, BMP4 might represent a pleiotropic factor secreted by SMCs which compromises different components of blood vessels. Surprisingly, activation of BMP signaling was reported in presence of mutations of Lamin A/C and other nuclear proteins in skeletal muscle cells (Dubinska-Magiera et al., 2013). Moreover, depletion of Lamin A/C in mouse embryonic fibroblasts determined upregulation of several genes including BMP4 (Singh et al., 2013). Together, these data suggest that Lamin A/C could be directly or indirectly involved in the expression of BMP4, hence establishing a potential link between HGPS and BMP4 dysregulation. Our measurement of BMP4 in blood serum from HGPS and age-matched healthy donors revealed opposite trends in the transition between early/late childhood. The lower level of BMP4 detected during early childhood in HGPS patients could be correlated with the unphysiological development of their musculo-skeletal system. Indeed, the largest growth rate of long bones occurs between 5 and 11 years old (Abrahamyan et al., 2008) and relies on multiple signaling molecules including BMP4. Conversely, abnormal levels of BMP4 in adulthood are correlated with vascular dysfunctions including calcification, atherosclerosis and vascular leakage. It is possible then that the higher level of BMP4 in HGPS vs. control donors during late childhood could be associated, together with other dysregulated pathways, with their cardiovascular decline. Despite these tantalizing observations, we acknowledge that additional analyses are required to give clinical significance to our findings, considering the limited number of samples that were employed (N = 7 donors per group). However, we would like to acknowledge that it is extremely difficult to obtain serum form HGPS patients due to the extremely low number of HGPS samples which are available for basic research.

Surprisingly, we also found that primary, healthy VECs upregulated BMP4 and other genes associated with calcification, skeletal system development and extracellular matrix organization when co-cultured with HGPS fibroblasts. Although very preliminary, these data suggest that different HGPS vascular cells could autonomously upregulate or induce the expression of bone-related genes.

In addition to BMP4, we also found that iSMCs from HGPS donors showed upregulation of the Notch ligand JAG1. Notch signaling plays a critical role in vascular homeostasis and previous reports demonstrated that Jagged 1 increases the vascular permeability of the blood brain barrier (Yao et al., 2011). Furthermore, Jagged 1 was recently identified as one of the contributors of the aging signature of healthy human donors in a large study comparing protein expression and chronological age (Tanaka et al., 2018). To our knowledge, no previous studies demonstrated the dysregulation of two key mediators of vascular permeability in HGPS SMCs. Hence, our approach has the potential to identify novel therapeutic targets in a wide set of diseases characterized by cardiovascular dysfunction, including Alzheimer Disease and Diabetes. In this scenario, the use of reprogrammed vascular cells could be extended to monitor the personalized response of a patient to a given therapy. However, in order to reach these goals it will be essential to optimize current protocols promoting the full maturation of reprogrammed vascular cells and to develop high-throughput systems compatible with the functional screening of hundreds or even thousands of experimental conditions in parallel.

Materials and methods

Cell culture and human serum

Cell lines used in the study are reported in Supplementary file 1. Details are available in Supplementary Materials and methods. Briefly, iVECs and primary endothelial cells were cultured in EGM-2MV (Lonza) on collagen coated dishes, iSMCs were cultured in EGM-2 (Lonza) and fibroblasts were cultured in DMEM+10%FBS+1% P/S. Human primary dermal fibroblast cell lines and serum samples were obtained from The Progeria Research Foundation (PRF) Cell and Tissue Bank. The HGPS cell lines were HGADFN178, HGADFN188, HGADFN164, HGADFN143, HGADFN169, HGADFN122, HGADFN127, HGADFN367. The control cell lines were HGADFN368, GM03349, GM01652. The HGPS serum samples were 205S1, 009, 231, 224, 82, 113, 171. The control samples were 237, 148, 207, 205, 203, 197, 232. Details are reported in Supplementary file 1 and 7.

RNAseq: library preparation and data analysis

The quality of cDNA libraries was checked with Tape Station (Agilent) and the amount of cDNA quantified through Qubit (ThermoFisher). HiSeq2500 was used to sequence the samples (single-end reading). Reads were mapped by STAR [v2.5.3a, ref: 10.1093/bioinformatics/bts635. pmid:23104886] to the hg19 reference genome with default parameters. Homer [v4.9.1, Heinz et al., 2010; http://homer.ucsd.edu/homer/] was used to quantify gene expression by counting uniquely mapped reads across all exons of RefSeq genes. Differential expression analysis was performed using edgeR for all iSMC related comparisons and DESeq2 for all iVECs related comparisons. Genes with adjusted p-value<0.05 and absolute log fold-change >1 were identified as significantly differentially expressed genes between two conditions. Gene ontology maps were generated with Metascape [http://metascape.org]. Heatmaps were generated using package gplots (https://cran.r-project.org/web/packages/gplots/index.html) in R (R Core Team (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/) on the row-wise z-scaled log-transformed count per million values (CPM). A pseudo-count of 5 was added to each CPM value before log-transformation to reduce noise. Dendrograms on the heatmap were calculated using Ward. D2 clustering algorithm and 1-Pearson’s correlation as distance measurement.

Statistics

Details on number of biological replicates, independent measurements/technical replicates, data normalization and statistical tests are reported in each figure legend. Error bars are standard error of the mean (s.e.m.).

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for choosing to send your work, "Direct reprogramming human vascular cells reveals defects associated with aging and Hutchinson-Gilford Progeria Syndrome", for consideration at eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Jessica Tyler as the Senior Editor.

Bersini et al. provide a conceptual advance in the use of direct reprogramming to model age-associated dysfunction of vascular cells. This is a novel and exciting proposed approach to study mechanisms of age-related vascular dysfunction and may provide actionable in vitro models of age-related vascular disease. The data presented, however, do not support efficacy of the reprogramming approach at the transcriptional or functional level. Critical validation of the approach is lacking. In the functional comparisons between reprogrammed cells from young and old donors, it is unclear that reprogramming efficiencies are comparable. Substantial weaknesses in the statistical analyses of the data were identified. Some of the supplementary materials provided may be superfluous and could be replaced by informative panels included in the main figures. BMP2 is identified in one set of experiments, but subsequent experiments focus on a different gene, BMP4. From an RNA-seq experiment that was performed data from only 2 genes are shown. Overall, while the focus of the studies reported is of high interest, these and other weaknesses decrease the priority for publication of these studies in their current form. A revised manuscript adequately addressing the main concerns raised in the peer-review process, that will require additional experimental work, might be suitable for publication. A revised Tools and Resources article also needs to provide additional insights into how the authors' reprogramming techniques and biological resources will advance the field.

Reviewer #1:

Bersini et al. report the generation of induced vascular endothelial cells and smooth muscle cells by direct reprogramming of skin fibroblasts through enforced expression of single master regulator genes for the endothelial and smooth muscle fates respectively. The stated goal of the study is to characterize and validate directly reprogrammed vascular cells, then use to ascertain age-specific differences in their biology. The goal is important since the authors state that transcriptomic studies on direct reprogrammed cells are lacking. The studies in directly reprogrammed skin fibroblasts hinge on the effectiveness of direct reprogramming. This is not supported by the transcriptomic data of Figure 1G and I. Even though (as the author state) induced cells do not express all cell identity markers of the primary cell types that they model, the paucity of cell type-specific genes that are expressed in the induced smooth muscle and vascular endothelial cells that are presented in Figure 1G strongly suggest that these cells may still be largely of fibroblast nature. It could be argued that a handful of vascular-specific genes are sufficient to direct the functionally relevant reprogramming of fibroblasts. However, the lack of direct comparisons in functional barrier assays (i.e. lack of data from “negative” and “positive” control samples using primary fibroblasts (negative control) and primary vascular endothelial or smooth muscle cells (positive controls)) in all subsequent studies that use cells directly reprogrammed from young and old donors make interpretation of those data difficult. Further, the differences in permeability between barriers established with induced cells from old and young donors are very small, and may be related to small differences in integrity or health of induced cells that retain (as suggested by the transcriptomic studies) most of their fibroblast-like features. These are important concerns that need to be addressed.

Reviewer #2:

The current manuscript by Bersini et al. provides a conceptual advance in the utility of direct reprogramming and modeling of aging-associated dysfunction in SMCs and VECs. Interestingly, the authors identified BMP2 and PALD1 as genes differentially regulated in aged iSMC and iVEC, respectively, which remarkably was reflected in human biopsy samples. By taking a similar experimental approach, the authors define the upregulated expression of BMP4 in reprogrammed cells from Hutchinson-Gilford Progeria Syndrome (HGPS) patients, which underlies dysfunction of iSMCs from HGPS patients. The idea underlying the study itself and the perspective of utilizing iSMCs/iVECs for studying the effect of aging is exciting. The manuscript in its current form, however, will benefit from clarifications for some of the data presented, which would further strengthen the author's conclusions.

Based on the heatmap data shown in Figure 1, it is not clear whether the transcriptome of reprogrammed cells represent fully reprogrammed and functional iVECs/iSMCs. The expression data indicate that both iSMCs and iVECs express more genes in common with fibroblasts than with primary VECs or SMCs. The authors state "transition towards full mature cells." The readers will be better informed if the authors clarify what is the measurement for transcriptomics and function for partial versus fully reprogrammed cells, or present additional data that either reject or support fully mature versus partial reprogramming. It is understandable that the goal here is to induce SMCs or VECs to the point that can capture some of the age-associated changes, not necessarily generating cells that will be identical to primary cells. Still, this will be useful information to the readers.

The choice of some of the SMC markers was not clear. For example, Calponin (shown in Figure 1B) appears to be expressed in cells that do not express the reprogramming factor, MYOCD, suggesting that Calponin could be also expressed in human fibroblasts. As such, it will be informative to the readers if the authors show the same immunostaining data for Calponin in fibroblasts.

The main focus in Figure 2 is to define age-associated changes modeled through direct reprogramming of fibroblasts of young and old age groups. It is then essential to provide experimental evidence that the comparative analyses were performed in the realm of similar reprogramming efficiencies between young and old groups. This point is particularly crucial as fibroblasts from different ages may not reprogram with the same efficiency.

It will be good to clarify whether the difference in dextran permeability between young and old cells shown in Figure 2H may be influenced by differences in reprogramming efficiency. Looking at Figure 2H, there appears to be a difference in dextran permeability, but the raw data in Figure S6 show there are very few SMA+ cells (just one red cell) in the old sample and none at the interface with iVECs, in contrast to the young sample. To further dissect the difference in permeability due to aging, not the lack of reprogrammed cells in the old samples, it will be informative to provide clear pictures showing the reprogrammed cells at the interface of the permeability assay.

As the end of Figure 2 was currently left with the correlation studies, additional mechanistic insights into the identified DE genes will greatly signify the authors' findings, for example by either i) reducing the expression of BMP2 or PALD1 in old cells, or ii) overexpressing those aging-associated genes in young cells to pinpoint whether the identified DE genes are underlying the observed phenotypes.

BMP2 was discovered to be upregulated in aging in Figure 2, but all the HGPS experiments in Figure 3 were associated with a different gene, BMP4. Based on the existing datasets, is BMP4 also upregulated in aged induced cells? Also, JAG1 is introduced in Figure 3, while there is no mention of PALD1 from Figure 2. The text will benefit from a more in-depth discussion of the link between findings of Figure 2 to Figure 3.

Reviewer #3:

Direct reprogramming of fibroblasts from young, aged, and HGPS donors into vascular endothelial cells (VECs) and smooth muscle cells (SMCs), combined with transcriptomics analyses and co-culture assays, is a novel and interesting approach to study age-related alterations in EC-SMC crosstalk and discover new potential therapeutic targets to fight vascular dysfunction in normal and premature aging. However, RNAseq results from the physiological aging-related part lacks functional validation, and RNAseq results from the HGPS-related part presents weak in vivo/ex vivo validation.

1) Regarding cells derived from normally aged donors, the authors show validation of three hits obtained from the RNAseq assays by immunofluorescence analysis of skin biopsies. However, no studies were performed to analyze the functional significance of such expression changes on endothelial permeability, as done for HGPS-related studies (e.g., using blocking antibodies or shRNA).

2) Although the HGPS-related part of the paper includes in vitro functional studies, ex vivo and in vivo validation of the RNAseq results appears rather limited. BMP4 ELISA data in Figure 2B comparing age-matched controls and HGPS patients is very far from reaching statistical significance (p=0.4), thus making very difficult to draw conclusion. I concur with the authors that these experiments are very challenging considering the limited number of available HGPS samples, and mouse studies may help overcome this limitation. Regarding mouse studies in Figure 3C, the authors used the t-test to compare serum BMP4 levels in young, aged and HGPS groups. However, t-test is not appropriate for studies including more than 2 experimental groups. These results should be re-analyzed using ANOVA and post-hoc analysis. Given the limitation of the human studies, a more thorough validation using the mouse model would significantly contribute to strengthen the biological significance of the results. For example, it would be very informative to perform qPCR and/or immunofluorescence assays to assess BMP4 and Jagged 1 levels in mouse arteries. I also miss experiments to examine in vivo endothelial permeability and the effects of neutralizing anti-BMP4 and anti-Jagged 1 antibodies in the mouse models of normal and premature aging.

3) Progerin expression should confirmed in HGPS-derived cells.

4) Several references appear to be incorrect in the introductory paragraph in the section of Results focused on HGPS. The authors state: "In particular, HGPS is characterized by dysfunctional SMCs, which promote vascular calcification (Wei et al., 2018), hypertension (Miriyala et al., 2006) and loss of vascular barrier functionality (Helbing et al., 2017)". The cited papers address the role of BMPs in vascular calcification, hypertension and loss of vascular barrier functionality, but I believe none of them include studies with either HGPS patients or HGPS experimental models. The same comment regarding the studies by Wei et al., 2018 and Miriyala et al., 2006 applies to the Discussion: "The main cardiovascular complications observed in HGPS patients include loss of SMC coverage of blood vessels (Varga, et al., 2006), abnormal vascular calcification (Wei, et al., 2018) and hypertension (Miriyala, et al., 2006)." Also note that all results in Varga et al., 2006 were derived from a mouse model of HGPS, not from HGPS patients, therefore references for SMC loss in patients should be cited instead of Varga et al., 2006 (or 1-2 review articles).

5) Figure 1A, B: Quantitative data should be presented (eg. % aSMA/Myocardin-immunoreactive and % Calponin/Myocardin-immunoreactive cells).

6) Figure 1D: The authors state that iSMCs contribute to the formation of microvascular networks in 3D fibrin matrices (Figure 1D), but no control without iSMCs is shown, which is needed to ascertain if iSMC promote this process.

7) The images in Figure 2E and 2F do not seem to be very representative according to the quantitative data shown in Figure 2C and 2D (massive upregulation in images, but only 2-fold induction in the graphs).

8) Figure 2G-J: To confirm that age-related increased permeability is due to compromised VEC junctions, confocal microscopy experiments should be performed in young and old VECs to visualize junctions within neighbouring cells, ideally in monoculture and in co-culture with young and old SMCs.

9) Figure 3D, F and G: The experiments include 3 groups, therefore t-test is not adequate. Results should be re-analyzed using ANOVA and post-hoc analysis.

10) Figure 3H-J: Quantitative results should be shown.

11) Figure 2—figure supplement 3: Quantitative results of GSTM1 expression in young versus old ECs should be shown (as done for BMNP2 and PALD1 in Figure 2C and 2D, respectively.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Direct reprogramming human vascular cells reveals defects associated with aging and Hutchinson-Gilford Progeria Syndrome" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Jessica Tyler as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

Direct reprogramming of skin fibroblasts is a valuable strategy to obtain large amounts of specific cell types that are centrally involved in human disease processes. This manuscript reports direct reprogramming of skin fibroblasts to address age-specific and disease-specific changes in vascular cells during normative aging and in accelerated aging in Hutchinson-Gilford Progeria Syndrome. The authors report the transcriptional and functional characterization of vascular cell types generated by direct reprogramming. This strategy may have substantial impact by providing a tool to study changes in vascular function and potentially that of other tissues during physiological and pathological aging.

Essential revision requirements:

The authors addressed most comments appropriately with new experimental results or new statistical analyses. However, the question of similar reprogramming efficiency between young and old samples (an essential point to the manuscript), was still not fully answered. In the revised manuscript, the authors compared the expression of ETV2 and MYOCD and found no statistical difference. However, these are comparisons of the expression level of reprogramming factors, not the reflection of similar reprogramming efficiency between young and old samples – that is, whether the reprogrammed state into iVECs or iSMC as an output, is similar between young and old samples. This is a central point that needs to be addressed.

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Bersini et al. provide a conceptual advance in the use of direct reprogramming to model age-associated dysfunction of vascular cells. This is a novel and exciting proposed approach to study mechanisms of age-related vascular dysfunction and may provide actionable in vitro models of age-related vascular disease. The data presented, however, does not support efficacy of the reprogramming approach at the transcriptional or functional level. Critical validation of the approach is lacking. In the functional comparisons between reprogrammed cells from young and old donors, it is unclear that reprogramming efficiencies are comparable. Substantial weaknesses in the statistical analyses of the data were identified. Some of the supplementary materials provided may be superfluous and could be replaced by informative panels included in the main figures. BMP2 is identified in one set of experiments, but subsequent experiments focus on a different gene, BMP4. From an RNA-seq experiment that was performed data from only 2 genes are shown. Overall, while the focus of the studies reported is of high interest, these and other weaknesses decrease the priority for publication of these studies in their current form. A revised manuscript adequately addressing the main concerns raised in the peer-review process, that will require additional experimental work, might be suitable for publication. A revised Tools and Resources article also needs to provide additional insights into how the authors' reprogramming techniques and biological resources will advance the field.

We have extensively revised the previous version of the manuscript by performing additional analyses and experiments. Specifically, we have addressed the concerns regarding the reprogramming efficiency of cells derived from young and old donors and the differentiation level of directly reprogrammed cells. We have performed new statistical analyses suggested by reviewers #2 and #3 and we have provided additional quantifications such as pathway analysis for the RNAseq data on organ-specific mouse VECs, expression of GSTM1 in skin biopsies, expression of Calponin in reprogrammed SMCs and analysis of endothelial cell-cell junction quality.

We have also addressed comments related to the validation of the approach: i) inclusion of negative (fibroblasts) and positive (primary VECs) controls for the vascular barrier assay; ii) permeability measurements with PALD1 and BMP2 knock-down on primary cells (both VECs and SMCs) to show the functional relevance of these genes.

We also provide a clear rationale for the transition from physiological aging to HGPS (including a discussion on the shift from BMP2 to BMP4) and a discussion on the contribution of the reprogramming approach for the advancement of the field.

Finally, in a specific comment by reviewer #3 regarding the need for additional statistical analyses for the level of BMP4 in mouse serum, the reviewer suggested that we perform in vivo experiments to test vascular permeability after treatment with blocking antibody. We agree with the reviewer that an in vivo validation could pave the way for the identification of a robust biomarker for vascular degeneration in HGPS patients (with potential consequences for the development of new therapies). Unfortunately, we only have access to the serum of this mouse model, not the mice themselves. We also feel that these experiments might go beyond the scope of this manuscript as a Tool and Resource article, which is “assessed in terms of the potential to facilitate experiments that address problems that to date have been challenging or even intractable”, as reported in the guide for authors.

Reviewer #1:

Bersini et al. report the generation of induced vascular endothelial cells and smooth muscle cells by direct reprogramming of skin fibroblasts through enforced expression of single master regulator genes for the endothelial and smooth muscle fates respectively. The stated goal of the study is to characterize and validate directly reprogrammed vascular cells, then use to ascertain age-specific differences in their biology. The goal is important since the authors state that transcriptomic studies on direct reprogrammed cells are lacking. The studies in directly reprogrammed skin fibroblasts hinge on the effectiveness of direct reprogramming. This is not supported by the transcriptomic data of Figure 1G and I. Even though (as the author state) induced cells do not express all cell identity markers of the primary cell types that they model, the paucity of cell type-specific genes that are expressed in the induced smooth muscle and vascular endothelial cells that are presented in Figure 1G strongly suggest that these cells may still be largely of fibroblast nature. It could be argued that a handful of vascular-specific genes are sufficient to direct the functionally relevant reprogramming of fibroblasts. However, the lack of direct comparisons in functional barrier assays (i.e. lack of data from “negative” and “positive” control samples using primary fibroblasts (negative control) and primary vascular endothelial or smooth muscle cells (positive controls)) in all subsequent studies that use cells directly reprogrammed from young and old donors make interpretation of those data difficult. Further, the differences in permeability between barriers established with induced cells from old and young donors are very small, and may be related to small differences in integrity or health of induced cells that retain (as suggested by the transcriptomic studies) most of their fibroblast-like features. These are important concerns that need to be addressed.

We thank the reviewer for the detailed comments on our manuscript. We agree with the reviewer that an effective reprogramming protocol is fundamental to obtain meaningful data and we apologize for the absence of key data supporting this issue in the original manuscript. We have now included additional results showing that:

1) The expression of the master regulators ETV2 and MYOCD is comparable in reprogrammed cells derived from young vs. old donors (log2FC=0.19 for ETV2 and log2FC=0.03 for MYOCD comparing young and old cells; no statistical difference comparing young and old cells). This point demonstrates that the reprogramming efficiency is not affected by the age of the donor and that the observed differences between young and old cells are not due to different levels of reprogramming. Details have been included in the manuscript. We would like to highlight that permeability values were about 50% higher in presence of reprogrammed cells from old vs. young donors. This difference was consistent across multiple replicates and donors. However, we need to highlight that our microscale models do not include several components of the vascular niche, including tissue-specific cells and immune cells, which might contribute to changes in vascular functionality during aging and disease. The absence of these components might partially explain why we observed limited (~50%) differences in vascular functionality. We have included a paragraph in the manuscript to discuss this specific point.

2) The expression of most key cell identity genes for both iVECs and iSMCs is comparable in reprogrammed and primary cells. In addition, the expression of key fibroblast genes is significantly lower in both reprogrammed and primary cells compared to the fibroblast population. Details have been included in Supplementary Information (Supplementary file 2). Together, these data show that reprogrammed cells are not only expressing cell identity genes, but also downregulating well established fibroblast markers.

As pointed out by the reviewer, a “handful of vascular-specific genes might be sufficient to direct the functionality of reprogrammed cells”. However, in the original manuscript there was no direct comparison of vascular permeability among reprogrammed cells, primary VECs (positive control) and fibroblasts (negative control). We have performed additional experiments showing that the permeability of reprogrammed cells is significantly lower compared to fibroblasts, reaching values close to the permeability of primary VECs. See Figure 2I.

Although we cannot claim that reprogrammed cells are transcriptionally and functionally identical to primary VECs, these data clearly demonstrate that reprogrammed cells have acquired a “minimum set of transcriptional and functional features”, which allows to employ them to capture a fraction of the vascular changes occurring during aging and disease. We have included these new data and their discussion in the manuscript. Please note that we performed experiments to measure the barrier functionality of iVECs compared to primary VECs and fibroblasts. Indeed, fibroblasts represent an appropriate negative control for the study of vascular functionality, since they do not contribute to the formation of the inner wall of blood vessels (which is formed by VECs). However, regarding SMCs we have only performed experiments comparing iSMCs and primary SMCs. The main reason is that fibroblasts normally populate the extracellular matrix surrounding blood vessels and directly contribute to the homeostasis of VECs. Hence, a comparison including fibroblasts would have been confusing. Collected data show that the vascular permeability in presence of young iSMCs is even lower compared to primary SMCs, demonstrating that reprogramming generates functional cells directly contributing to the maintenance of the vascular barrier (potentially through paracrine secretions). See Figure 2J.

Reviewer #2:

[…] Based on the heatmap data shown in Figure 1, it is not clear whether the transcriptome of reprogrammed cells represent fully reprogrammed and functional iVECs/iSMCs. The expression data indicate that both iSMCs and iVECs express more genes in common with fibroblasts than with primary VECs or SMCs. The authors state "transition towards full mature cells." The readers will be better informed if the authors clarify what is the measurement for transcriptomics and function for partial versus fully reprogrammed cells, or present additional data that either reject or support fully mature versus partial reprogramming. It is understandable that the goal here is to induce SMCs or VECs to the point that can capture some of the age-associated changes, not necessarily generating cells that will be identical to primary cells. Still, this will be useful information to the readers.

We thank the reviewer for the useful comments on our manuscript. We have included additional data to better clarify the concept of “partial maturation” of the reprogrammed cells. First, we have analyzed the expression level of key cell identity genes for both iVECs and iSMCs and we have compared it with primary cells and fibroblasts. We have included in Supplementary Information a detailed table with these genes to provide a quantification of our concept of “partial maturation” (Supplementary file 2). For example, the expression level of KDR (an important gene defining endothelial cell identity) was 6.62 log2FC higher in primary VECs vs. fibroblasts and 6.38 log2FC higher in iVECs vs. fibroblasts. At the same time, CDH5 was upregulated in both primary cells (log2FC=6.7) and iVECs (log2FC=4) compared to fibroblasts, although at different levels. Similarly, both primary and reprogrammed SMCs upregulated key identity genes including ACTA2 and MYH10 compared to fibroblasts. However, the expression of MYH10 was lower in reprogrammed vs. primary SMCs. Finally, we would like to highlight that reprogrammed cells downregulated the expression of fibroblast-specific genes (e.g. FSP1, TWIST, SLUG), again indicating a transition towards fully mature vascular cells.

As a second step, we have performed additional experiments to compare the functionality of iVECs with primary VECs (positive control) and original fibroblasts (negative control). Our data show that the permeability of iVECs is significantly lower compared to fibroblasts, reaching values close to the permeability of primary VECs. Regarding SMCs we have only performed experiments comparing iSMCs and primary SMCs. The main reason is that fibroblasts normally populate the extracellular matrix surrounding blood vessels and directly contribute to the homeostasis of VECs. Hence, a comparison including fibroblasts would have been confusing. Collected data show that the vascular permeability in presence of young iSMCs is even lower compared to primary SMCs, demonstrating that reprogramming generates functional cells directly contributing to the maintenance of the vascular barrier (potentially through paracrine secretions).

We have included these data in the manuscript. See Figure 2I and J.

Overall, we think that these additional quantifications provide a better picture of vascular reprogramming and highlight that iVECs and iSMCs show transcriptional and functional features of VECs and SMCs, respectively, although they do not fully replicate primary cells. A discussion paragraph has been included in the manuscript.

The choice of some of the SMC markers was not clear. For example, Calponin (shown in Figure 1B) appears to be expressed in cells that do not express the reprogramming factor, MYOCD, suggesting that Calponin could be also expressed in human fibroblasts. As such, it will be informative to the readers if the authors show the same immunostaining data for Calponin in fibroblasts.

The image reported in Figure 1B showed MYOCD positive cells expressing Calponin. We noticed that a few cells can express Calponin despite the weaker MYOCD signal. We apologize if this aspect was confusing. To further clarify the expression of smooth muscle cell markers we have quantified the percentage of MYOCD positive cells co-expressing αSMA and Calponin. We have found that 70% MYOCD positive cells expressed αSMA, while 80% expressed Calponin. We have introduced these data in the manuscript.

The main focus in Figure 2 is to define age-associated changes modeled through direct reprogramming of fibroblasts of young and old age groups. It is then essential to provide experimental evidence that the comparative analyses were performed in the realm of similar reprogramming efficiencies between young and old groups. This point is particularly crucial as fibroblasts from different ages may not reprogram with the same efficiency.

We agree with the reviewer. We have provided additional data (Supplementary Information) showing that the reprogramming efficiency is comparable in cells obtained from young and old donors (log2FC=0.19 for ETV2 and log2FC=0.03 for MYOCD comparing young and old cells; no statistical difference comparing young and old cells).

It will be good to clarify whether the difference in dextran permeability between young and old cells shown in Figure 2H may be influenced by differences in reprogramming efficiency. Looking at Figure 2H, there appears to be a difference in dextran permeability, but the raw data in Figure S6 show there are very few SMA+ cells (just one red cell) in the old sample and none at the interface with iVECs, in contrast to the young sample. To further dissect the difference in permeability due to aging, not the lack of reprogrammed cells in the old samples, it will be informative to provide clear pictures showing the reprogrammed cells at the interface of the permeability assay.

We thank the reviewer for the possibility to clarify this aspect. As discussed in the previous comment, we have provided additional data showing that the reprogramming efficiency was similar in cells obtained from young and old donors. Hence, differences in permeability are not due to different reprogramming efficiency.

Regarding the distribution of iSMCs in the microscale assay, we realized that our description in the manuscript was not sufficiently clear. Indeed, iVECs and iSMCs are seeded in two separate channels, with iVECs forming a monolayer at the interface with the channel containing iSMCs (embedded in a 3D matrix). iSMCs homogeneously distribute inside the matrix channel. iSMCs from young donors tend to migrate towards the endothelial monolayer. However, the same migratory behavior is not observed with cells from old donors. Here we report a new Figure 2—figure supplement 6 showing a detail of the endothelial monolayer (VE-cadherin, white) between the pillars of the microscale model; smooth muscle actin (red) positive cells are close to the monolayer only in young donors (nuclei are shown in blue; single slice from a confocal stack).

It is important to highlight that the contribution of SMCs to the vascular barrier could be in terms of both paracrine secretion and direct support to the endothelial monolayer. Overall, IF images showed that iSMCs from old donors did not physically interact with iVECs; at the same time, differential expression analysis highlighted that iSMCs from old donors upregulated genes which might interfere with the integrity of the barrier (e.g. BMP2 which was identified through RNAseq and, at the protein level, through IF of human skin biopsies). Although we cannot separate the contribution of these two aspects, it is likely that the combination of physical support and secreted factors contributes to the observed differences in the barrier functionality in young vs. old conditions. We have included a paragraph in the Discussion section to better explain this point.

As the end of Figure 2 was currently left with the correlation studies, additional mechanistic insights into the identified DE genes will greatly signify the authors' findings, for example by either i) reducing the expression of BMP2 or PALD1 in old cells, or ii) overexpressing those aging-associated genes in young cells to pinpoint whether the identified DE genes are underlying the observed phenotypes.

We have performed additional experiments to demonstrate the functional significance of the identified genes. We thought that KD of PALD1 and BMP2 in primary VECs and SMCs, respectively, would have been even more relevant to clarify the contribution of these genes in the vascular functionality. Collected data showed that KD of PALD1 in VECs decreased vascular permeability compared to luciferase control (51.5±3.4% vs. 100.0±4.3%, at least N=60 independent measurements per condition in 3 biological replicates, Student’s t-test with p<0.001). On the other side, KD of BMP2 in SMCs induced a weak, but reproducible, decrease in permeability compared to luciferase control (86.3±3.8% vs. 100.0±4.3%, at least N=60 independent measurements per condition in 3 biological replicates, Student’s t-test with p=0.07).

Together, these results clearly show that PALD1 expression by VECs is a key determinant of vascular permeability. On the other side, the contribution of BMP2, which is secreted by SMCs, is more limited. Based on this finding, it is possible that the effect of SMCs on the regulation of vascular permeability is the result of the action of multiple genes, each giving an incremental or synergistic contribution (Figure 2A includes other potential candidates). Overall, these experiments show that reprogrammed cells are a promising tool to identify potential contributors to vascular aging. Data have been included in the manuscript. See Figure 2K and L.

BMP2 was discovered to be upregulated in aging in Figure 2, but all the HGPS experiments in Figure 3 were associated with a different gene, BMP4. Based on the existing datasets, is BMP4 also upregulated in aged induced cells? Also, JAG1 is introduced in Figure 3, while there is no mention of PALD1 from Figure 2. The text will benefit from a more in-depth discussion of the link between findings of Figure 2 to Figure 3.

We thank the reviewer for the possibility to clarify this point. We did not find overlapping genes comparing young vs. old and healthy vs. HGPS cells. This is the reason why JAG1 was not mentioned in Figure 2. However, it is interesting to highlight that BMP2 (identified in Figure 2) and BMP4 (identified in Figure 3) share a 62% sequence similarity and are functionally equivalent. Hence, BMP2/4 dysregulation might underlie a common mechanism leading to vascular degeneration during physiological aging and disease. We have clarified these aspects in the Discussion section.

Reviewer #3:

Direct reprogramming of fibroblasts from young, aged, and HGPS donors into vascular endothelial cells (VECs) and smooth muscle cells (SMCs), combined with transcriptomics analyses and co-culture assays, is a novel and interesting approach to study age-related alterations in EC-SMC crosstalk and discover new potential therapeutic targets to fight vascular dysfunction in normal and premature aging. However, RNAseq results from the physiological aging-related part lacks functional validation, and RNAseq results from the HGPS-related part presents weak in vivo/ex vivo validation.

1) Regarding cells derived from normally aged donors, the authors show validation of three hits obtained from the RNAseq assays by immunofluorescence analysis of skin biopsies. However, no studies were performed to analyze the functional significance of such expression changes on endothelial permeability, as done for HGPS-related studies (e.g., using blocking antibodies or shRNA).

We agree with the reviewer that RNAseq data from the physiological aging part were not coupled with a functional validation of the identified genes. Hence, we have provided additional data showing that KD of PALD1 and BMP2 in VECs and SMCs, respectively, is coupled with a decrease in vascular permeability. More in detail, KD of PALD1 in VECs decreased vascular permeability compared to luciferase control (51.5±3.4% vs. 100.0±4.3%, at least N=60 independent measurements per condition in 3 biological replicates, Student’s t-test with p<0.001). On the other side, KD of BMP2 in SMCs induced a weak, but reproducible, decrease in permeability compared to luciferase control (86.3±3.8% vs. 100.0±4.3%, at least N=60 independent measurements per condition in 3 biological replicates, Student’s t-test with p=0.07).

Together, these results clearly show that PALD1 expression by VECs is a key determinant of vascular permeability. On the other side, the contribution of BMP2, which is secreted by SMCs, is more limited. Based on this finding, it is possible that the effect of SMCs on the regulation of vascular permeability is the result of the action of multiple genes, each giving an incremental or synergistic contribution (Figure 2A includes other potential candidates). Overall, these experiments show that reprogrammed cells are a promising tool to identify potential contributors to vascular aging. Data have been included in the manuscript. See Figure 2K and L.

2) Although the HGPS-related part of the paper includes in vitro functional studies, ex vivo and in vivo validation of the RNAseq results appears rather limited. BMP4 ELISA data in Figure 2B comparing age-matched controls and HGPS patients is very far from reaching statistical significance (p=0.4), thus making very difficult to draw conclusion. I concur with the authors that these experiments are very challenging considering the limited number of available HGPS samples, and mouse studies may help overcome this limitation. Regarding mouse studies in Figure 3C, the authors used the t-test to compare serum BMP4 levels in young, aged and HGPS groups. However, t-test is not appropriate for studies including more than 2 experimental groups. These results should be re-analyzed using ANOVA and post-hoc analysis. Given the limitation of the human studies, a more thorough validation using the mouse model would significantly contribute to strengthen the biological significance of the results. For example, it would be very informative to perform qPCR and/or immunofluorescence assays to assess BMP4 and Jagged 1 levels in mouse arteries. I also miss experiments to examine in vivo endothelial permeability and the effects of neutralizing anti-BMP4 and anti-Jagged 1 antibodies in the mouse models of normal and premature aging.

We appreciate that the reviewer concurs with us that experiments on HGPS are challenging considering the limited availability of samples and models. Based on the reviewer comments and on comments by another reviewer, we have re-analyzed data on serum from human donors by considering the aging perspective. While healthy donors show a significant decrease in BMP4 over age (t-test, p=0.04), this trend is lost in HGPS samples. This finding offers a novel way to analyze the potential contribution of BMP4 in vascular degeneration and has been included in the manuscript.

We also agree with the reviewer that a t-test was not ideal to analyze the level of BMP4 in mouse samples given the presence of 3 experimental groups. For this reason, we have performed an ANOVA test followed by Holm-Sidak test finding that HGPS samples were statistically different compared to old (p<0.001) and young (p=0.001) samples. However, there was no difference comparing old and young samples (p=0.79). We have included these data in the manuscript.

We agree with the reviewer that in vivo experiments (e.g. measurements of vascular permeability in presence of anti-BMP4 antibody) would significantly strengthen the conclusions of the manuscript. BMP4 might represent a robust biomarker of vascular damage in HGPS, with potential consequences for the development of new therapies. However, we think that these experiments might go slightly beyond the scope of this manuscript as a Tool and Resource article, which is “assessed in terms of the potential to facilitate experiments that address problems that to date have been challenging or even intractable”.

3) Progerin expression should confirmed in HGPS-derived cells.

We have provided additional information about the source of HGPS cells. All HGPS cell lines were obtained from the Progeria Research Foundation and screened for mutation in the LMNA gene, as certified by the Progeria Research Foundation (https://www.progeriaresearch.org/wp-content/uploads/2019/11/PRF-AVAILABLE-CELL-LINES-1.pdf).

4) Several references appear to be incorrect in the introductory paragraph in the section of Results focused on HGPS. The authors state: "In particular, HGPS is characterized by dysfunctional SMCs, which promote vascular calcification (Wei et al., 2018), hypertension (Miriyala et al., 2006) and loss of vascular barrier functionality (Helbing et al., 2017)". The cited papers address the role of BMPs in vascular calcification, hypertension and loss of vascular barrier functionality, but I believe none of them include studies with either HGPS patients or HGPS experimental models. The same comment regarding the studies by Wei et al., 2018 and Miriyala et al., 2006 applies to the Discussion: "The main cardiovascular complications observed in HGPS patients include loss of SMC coverage of blood vessels (Varga, et al., 2006), abnormal vascular calcification (Wei, et al., 2018) and hypertension (Miriyala, et al., 2006)." Also note that all results in Varga et al., 2006 were derived from a mouse model of HGPS, not from HGPS patients, therefore references for SMC loss in patients should be cited instead of Varga et al., 2006 (or 1-2 review articles).

We apologize for the incorrect references. We have modified the text accordingly.

5) Figure 1A, B: Quantitative data should be presented (eg. % aSMA/Myocardin-immunoreactive and % Calponin/Myocardin-immunoreactive cells).

We have quantified the percentage of MYOCD positive cells co-expressing αSMA and Calponin. We have found that 70% MYOCD positive cells expressed αSMA, while 80% expressed Calponin. We have introduced these data in the manuscript.

6) Figure 1D: The authors state that iSMCs contribute to the formation of microvascular networks in 3D fibrin matrices (Figure 1D), but no control without iSMCs is shown, which is needed to ascertain if iSMC promote this process.

We apologize for the misunderstanding. We wanted to highlight that iSMCs physically interact with VECs and partially wrap segments of capillary structures. Based on the reviewer comment, we have re-analyzed data generated in our lab regarding the area fraction covered by VECs (vascular structures) in presence of fibroblasts (co-culture model; fibroblasts are considered an optimal source of growth factors to promote the formation of in vitro vascular networks). The same culture conditions (cell culture medium, 3D matrix, cell densities, time) were applied in the co-culture model with iSMCs. We have then quantified the area fraction covered by VECs in presence of fibroblasts or iSMCs considering regions of interests of the same dimensions. Our data show that there is not a statistical difference (t-test with p=0.61) when VECs are co-cultured with fibroblasts or iSMCs, supporting the idea that iSMCs contribute to the formation of vascular networks in vitro (Author response image 1).

Author response image 1.

Quantification of the vascular area fraction in 3D matrices embedding a co-culture of VECs with fibroblasts or iSMCs (Student’s t-test with p=0.61).

7) The images in Figure 2E and 2F do not seem to be very representative according to the quantitative data shown in Figure 2C and 2D (massive upregulation in images, but only 2-fold induction in the graphs).

We apologize, the images were selected to be representative for the differential expression of selected proteins. Image quantification is based on thresholded images (binarized) where low intensity pixels (which can be barely detected by eyes in the images) might be amplified resulting in the lower difference reported in the graphs.

8) Figure 2G-J: To confirm that age-related increased permeability is due to compromised VEC junctions, confocal microscopy experiments should be performed in young and old VECs to visualize junctions within neighbouring cells, ideally in monoculture and in co-culture with young and old SMCs.

We have included additional data showing the expression of Y658 VE-Cadherin (marker of cell-cell junction de-stabilization) in iVECs derived from young and old donors. Quantification of Y658 area fraction showed a significant over-expression of this marker in iVECs from old donors, complementing data on vascular permeability reported in Figure 2—figure supplement 5.

9) Figure 3D, F and G: The experiments include 3 groups, therefore t-test is not adequate. Results should be re-analyzed using ANOVA and post-hoc analysis.

We have re-analyzed the data using ANOVA test. Details are reported below.

Figure 3D: ANOVA with Holm-Sidak test gives p<0.001 for all comparisons

Figure 3F: ANOVA with Holm-Sidak test gives p<0.001 for all comparisons

Figure 3G: ANOVA with Holm-Sidak test gives p=0.01 for control vs. Jagged1; p=0.043 for control vs. BMP4; p=0.897 for BMP4 vs. Jagged1

Results have been included in the manuscript.

10) Figure 3H-J: Quantitative results should be shown.

Quantification for Figure 3H-J was already present in Figure 3G.

11) Figure 2—figure supplement 3: Quantitative results of GSTM1 expression in young versus old ECs should be shown (as done for BMNP2 and PALD1 in Figure 2C and 2D, respectively.

Quantification for GSTM1 has been included in Supplementary Information.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Essential revision requirements:

The authors addressed most comments appropriately with new experimental results or new statistical analyses. However, the question of similar reprogramming efficiency between young and old samples (an essential point to the manuscript), was still not fully answered. In the revised manuscript, the authors compared the expression of ETV2 and MYOCD and found no statistical difference. However, these are comparisons of the expression level of reprogramming factors, not the reflection of similar reprogramming efficiency between young and old samples – that is, whether the reprogrammed state into iVECs or iSMC as an output, is similar between young and old samples. This is a central point that needs to be addressed.

We realize that our answer regarding the question about similar reprogramming efficiency between young and old samples by showing expression of the reprogramming factors ETV2 and MYOCD did not fully satisfied the reviewers. We now provide evidence showing that the reprogramming efficiencies are similar. We have performed additional analyses to show that the expression of key cell identity genes did not change comparing cells derived from young and old donors, considering both induced endothelial cells (iVECs) and induced smooth muscle cells (iSMCs). We have included a Supplementary File 3 reporting the log2FoldChange (FC) and the False Discovery Rate (FDR) for these genes. As you can see, the expression of these genes is not statistically different comparing young and old induced cells.

In addition, we have generated heat-maps to better visualize how the expression of all these genes is not affected by the age group. As a reference, we have also included fibroblasts derived from young and old donors. These heat-maps have been included as Supplementary Figure (Z-score: ±1) (Figure 2—figure supplement 1). If you think that these additional images might better fit in the main text, we are open to include them in Figure 1.

The main text has now been modified considering these details. We think that these data together with the expression of ETV2 and MYOCD will clarify the concept that differences between young and old reprogrammed cells are not dependent on their identity.

We have now changed the way supplementary figures, videos and tables are cited. Moreover, supplementary methods and legends are now included in the main text. The Key Resources Table has now been included at the beginning of the Materials and methods Section. Regarding cell lines, relevant information have been added to the Supplementary Materials and methods section. In addition, we have provided references in the Key Resources Table to clarify that all cell lines were characterized for cell identity and tested negative for mycoplasma. We have provided a more detailed ethics statement, which is reviewed and approved yearly by the IRB. Regarding plasma from mice, we would like to highlight that we did not perform any kind of animal experiment. Plasma was obtained from Prof. Belmonte at Salk Institute. Relevant sequencing data are uploaded and publicly available in GEO (accession number GSE140898) and analyzed data are also included in Source Data 1, Source Data 2 and Source Code 1.

Key resources table

Reagent type (species) or resource	Designation	Source or reference	Identifiers	Additional information
Cell line (Homo-Sapiens)	Dermal fibroblasts (normal)	Coriell	Details in Supplementary file 1	https://www.coriell.org/0/Sections/Support/Global/QCcells.aspx?PgId=409
Cell line (Homo-Sapiens)	Dermal fibroblasts (normal and HGPS)	Progeria Research Foundation	Details inSupplementary file 1	https://www.progeriaresearch.org/wp-content/uploads/2020/04/PRF-AVAILABLE-CELL-LINES.pdf
Cell line (Homo-Sapiens)	Microvascular endothelial cells (normal)	Angioproteomie	Details in Supplementary file 1	https://www.angioproteomie.com/commerce/ccp1073-human-dermal-microvascular-endothelial-cells-cap-0005.htm
Cell line (M. musculus)	Microvascular endothelial cells (normal)	Cell Biologics	Details in Supplementary file 1	https://cellbiologics.com/index.php?route=product/product&path=2_47_89_91&product_id=2240
Transfected construct (human)	pQCXIB EGFP-MYOCD	This paper		Retroviral construct for stable expression
Transfected construct (human)	pQCXIB ETV2	This paper		Retroviral construct for stable expression
Transfected construct (human)	On Target Plus Human PALD1 siRNA (Smart Pool)	Horizon	L-026434-00-0005
Transfected construct (human)	On Target Plus Human BMP2 siRNA (Smart Pool)	Horizon	L-011219-00-0005
Biological sample (Homo-sapiens)	Human skin samples	This paper		Isolated from human donors by the Cooperative Human Tissue Network or by the Clinical and Translational Research Institute of UCSD
Antibody	Anti-human CD31 (Mouse monoclonal)	ThermoFisher	Cat# MA3100 RRID:AB_223516	IF(1:100)
Antibody	Anti-human collagen type I (Mouse monoclonal)	ThermoFisher	Cat# MA126771 RRID:AB_2081889	IF(1:250)
Antibody	Anti-human VE-cadherin (Rabbit monoclonal)	Cell Signaling	Cat#2500S RRID:AB_10839118	IF(1:100)
Antibody	Anti-human Y658 VE-cadherin (Rabbit polyclonal)	ThermoFisher	Cat#441144G RRID:AB_2533583	IF(1:100)
Antibody	Anti-human alpha smooth muscle actin (Mouse monoclonal)	eBioscience	Cat#14-9760-82 RRID:AB_2572996	IF(1:2,500)
Antibody	Anti-human calponin (Mouse monoclonal)	Santa Cruz	Cat#sc-58707 RRID:AB_781770	IF(1:100)
Antibody	Anti-human BMP2 (Rabbit polyclonal)	Novus	Cat#nBP1-19751 RRID:AB_2227877	IF(1:100)
Antibody	Anti-human Paladin (Rabbit polyclonal)	ThermoFisher	Cat#PA5-53591 RRID:AB_2645183	IF(1:100)
Antibody	Anti-human GSTM1 (Mouse Monoclonal)	ThermoFisher	Cat#MA5-17085 RRID:AB_2538556	IF(1:100)
Antibody	Anti-human BMP4 (Mouse Monoclonal)	R and D Systems	Cat#MAB3552 RRID:AB_2065677	Blocking (2 ug/ml)
Antibody	Anti-human Jagged 1 (Goat polyclonal)	R and D Systems	Cat#AF1277 RRID:AB_354713	Blocking (1 ug/ml)
Sequence-based reagent	CD31 forward	This paper	PCR primers	TGGTCAAGAAAAGCAACACAG
Sequence-based reagent	CD31 reverse	This paper	PCR primers	GATTCGCAACGGACTTCG
Sequence-based reagent	CD31 forward	This paper	PCR primers	ACAACGAGGGCATCATCAAG
Sequence-based reagent	CD31 reverse	This paper	PCR primers	GAAGTGGTAGAAAGGCTGCTG
Sequence-based reagent	BMP2 forward	This paper	PCR primers	CATGCTAGACCTGTATCGCA
Sequence-based reagent	BMP2 reverse	This paper	PCR primers	TGTTTTCCCACTCGTTTCTGG
Sequence-based reagent	BMP2 forward	This paper	PCR primers	GCCCTTCGAGCACCACGCA
Sequence-based reagent	BMP2 reverse	This paper	PCR primers	TGGCTTGTAGTGCCGCTGCTG
Commercial assay or kit	BMP4 ELISA Kit	Sigma-Aldrich	RAB0029
Software, algorithm	Fiji			https://fiji.sc/