Unveiling impaired vascular function and cellular heterogeneity in diabetic donor-derived vascular organoids

Abstract

Vascular organoids (VOs), derived from induced pluripotent stem cells (iPSCs), hold promise as in vitro disease models and drug screening platforms. However, their ability to faithfully recapitulate human vascular disease and cellular composition remains unclear. In this study, we demonstrate that VOs derived from iPSCs of donors with diabetes (DB-VOs) exhibit impaired vascular function compared to non-diabetic VOs (ND-VOs). DB-VOs display elevated levels of reactive oxygen species (ROS), heightened mitochondrial content and activity, increased proinflammatory cytokines, and reduced blood perfusion recovery in vivo. Through comprehensive single-cell RNA sequencing, we uncover molecular and functional differences, as well as signaling networks, between vascular cell types and clusters within DB-VOs. Our analysis identifies major vascular cell types (endothelial cells [ECs], pericytes, and vascular smooth muscle cells) within VOs, highlighting the dichotomy between ECs and mural cells. We also demonstrate the potential need for additional inductions using organ-specific differentiation factors to promote organ-specific identity in VOs. Furthermore, we observe basal heterogeneity within VOs and significant differences between DB-VOs and ND-VOs. Notably, we identify a subpopulation of ECs specific to DB-VOs, showing overrepresentation in the ROS pathway and underrepresentation in the angiogenesis hallmark, indicating signs of aberrant angiogenesis in diabetes. Our findings underscore the potential of VOs for modeling diabetic vasculopathy, emphasize the importance of investigating cellular heterogeneity within VOs for disease modeling and drug discovery, and provide evidence of GAP43 (neuromodulin) expression in ECs, particularly in DB-VOs, with implications for vascular development and disease.

Keywords: blood vessel organoids; cardiovascular diseases; diabetic vasculopathy; induced pluripotent stem cells; regenerative medicine; vascular disease modeling.

Graphical Abstract

Figure 1.

Blood vessel organoid generation and culture from iPSCs. (A) Colony of human iPS cells in 80%-90% confluence were dissociated for cell aggregate generation. (B) Aggregate formation efficiency. The size of aggregates is on average 200 µm. (C) Cell aggregates become bigger to form mesodermal floating spheroids. (D) The floating spheroids start to bud after induction by VEGF-A and forskolin. (E) Phase-contrast image of the sprouting vessels that appeared 2 days after embedding in the 3D Collagen-Matrigel matrix. Every 2-4 vascular spheroids were cut to allow for making full, rounded, mature organoids till day 20 (F). Live staining of the mature organoids by Calcein-AM in green. Scale bars of A-F = 250 µm. The corresponding day, the aim of each stage of differentiation, and the respective materials are indicated below each image (A-F). Successful and reproducible generation of vascular organoids (VOs) from iPS cells that contain both small capillary networks (G) and bigger arteriole (H). Scale bar = 100 µm. Video of the whole confocal series is represented in Supplementary Materials. The confocal imaging showed the presence of both endothelial tubes (CD31⁺, red), mural cells (PDGFRb⁺, gray), and basement membrane (CollagenIV⁺, green) within the VOs (ND5 VO as a representative image). The inset (Scale bar = 50 µm) shows the interaction and alignment of ECs with mural cells. (I) 3D projection of 69 confocal images of a mature VO showing a nice alignment of mural cells with endothelial tubes. Scale bar = 200 µm. Size range of organoids 0.5-1.5 mm. (J) Flow cytometry results indicate the percentage of endothelial cells (CD144+) compared to pericytes (PDGFRB+) in the ND-VO (ND19, upper quadrant) in contrast to the DB-VO (DB14, quadrant).

Figure 2.

Functional assays of diabetic vascular organoids (DB-VOs) versus non-diabetic (ND-VOs). Enhanced ROS production in DB-VOs is represented by confocal imaging live (A), and a boxplot of flow cytometry analysis of 12 independent experiment of 6 DB-VOs against 9 independent experiment of 5 ND-VOs donors, all listed in Table 1 (B). Ac-LDL uptake showing by confocal imaging (C), and quantified by flow cytometry from 2 independent experiments of 6 DB-VOs (DB07, DB09, DB11, DB12, DB13, and DB14) and 4 ND-VOs (ND5, ND19, ND20, and ND22) represented by boxplot (D), which were significantly higher in DB-VOs as compared to ND-VOs. Scale bar in (A) and (C) is 100 and 200 µm, respectively. (E) Quantification of MitoTracker Red Live Staining of VOs in the bar chart showing significantly higher accumulation of mitochondria in DB-VO as compared to ND-VOs. These data are based on 6 independent experiment of 3 DB-VOs and ND-VOs, P-values are shown: ****P < .0001 (unpaired, 2-tailed t-test). (F) Confocal imaging confirms the result of E. (G) Transmission electron microscopy shows larger mitochondria in DB-VOs as compared to ND-VOs. Error bars represent mean ± SEM (n = 3). The scale bars for ND are 200 nm and 1 µm, for DB 500 nm and 2 µm, left to right, respectively. P-values are shown: **P < .01, ****P < .0001 (unpaired, 2-tailed t-test). The human protein array of both conditioned media (minus FBS), 1.5 mL per well for 24 hours, the mixture of 3 DB-VOs (DB07, DB13, and DB14) versus 3 ND-VOs (ND5, ND19, and ND22) 6 weeks post-treatment (H), and 3 independent DB-VOs and ND-VOs individually 3 weeks post-treatment (I). Forty-three angiogenesis-related proteins were compared. Nineteen proteins were up/downregulated in CM’s of DB at day 42 of diabetogenic media treatment, from which 12 of them were antiangiogenic or proinflammatory cytokines (shown in blue text on the arrays’ membrane and quantified in the table). DB-VOs were treated with diabetogenic media (glocuse + TNF-a + IL-6) and control only with mannitol. Left panel: array 1, and right panel: array 2. Values in the table represent the fold change ratio of each protein in DBs versus NDs. Dashed rectangles on the array membranes show the positive internal control for the normalization of each array. Bar plot in (I) showing the comparison of protein extracts from 3 independent DB-VOs versus 3 independent ND-VOs. Only bFGF and TIMP-2 were significantly different between DBs and NDs at 3 weeks post-treatment. A table showing the layout of all 43 antibodies against 20 proteins on array 1, and 23 proteins on array 2, along with the array membrane have shown in Supplementary Figure S6. Unpaired t-test nonparametric (Mann-Whitney U test) was used to statistically examine differences between DBs and NDs. Mito or M denotes mitochondria. Abbreviation: CM, conditioned media.

Figure 3.

Comparison of diabetic (DB-VOs) and non-diabetic vascular organoids (ND-VOs) in hind limb ischemia recovery and host vasculature integration. (A) Laser Doppler imaging was used to assess blood reperfusion in the lower limb of SCID mice at days 0, 1, and 14 post-surgery following transplantation of DB-VOs (DB07, DB9, and DB14) and ND-VOs (ND5, ND19, and ND22). The yellow rectangle highlights the ischemic leg, with the blood perfusion recovery quantified in the line chart (n = 6 per condition, 2-way ANOVA, P-value < .01). The average blood perfusion recovery was 41.5% for ND-VOs and 21.8% for DB-VOs. (B) Injected labeled cells were tracked at 3 locations around the hind limb ischemia area by Bruker fluorescent imaging on day 14 after sacrificing mice. The muscles were sampled precisely from the injected area and embedded into Tissue-Tek O.C.T. compound within cylinders. The cylinders were then tracked for labeled human cells by Bruker imaging on dry ice to avoid heat damage to snap-frozen tissue samples. (C) Cryosectioning of the cylinders containing injected cells of labeled VOs was performed, followed by IHC and confocal imaging to assess the integration of injected human cells (red, labeled with LuminiCell Tracker 670) of ND-VOs and DB-VOs into host vasculature. The results indicated that injected cells of ND-VOs were better integrated with host vasculature than those of DB-VOs.

Figure 4.

Characterization of the iPSC-derived vascular organoids (VOs) in single-cell resolution. (A) 3D UMAP projection of the cells’ pool of VOs derived from diabetic (DB07, DB09, and DB11) and non-DBs (ND05, ND19, and ND22) donors, coloured by the sample name. (B) Mural cells’ subpopulation showing the expression of PDGFRb and ACTA2 (a-SMA). (C) Endothelial cells’ subpopulations showing the expression of PECAM1 (CD31) and CDH5 (CD144). (D) Differentially expressed analysis of total single cells of DB-VOs versus ND-VOs showed no significant differences in the expression of typical markers of endothelial cells (CDH5 and PECAM1), vSMCs (TAGLN), and pericytes (PDGFRB and CSPG4), represented by violin plots, except for vSMCs’ marker ACTA2 with FDR < 0.009 and fold change = −2.2 (See Supplementary Excel Sheet). The expression level is normalized log2-UMI and each dot represents a cell. (E, F) Pseudo time analysis shows a common ancestor cell expressing mesenchymal stem cell markers (CD44, CD73/NT5E, CD90/THY1, and CD105/ENG), further developed into the more differentiated cell during the time (darker blue). (G) VOs represent 3 cell states with a branch point, reflecting 2 differentiated cell populations with a common progenitor cell. (H) Overlying cell identities found in (B) and (C) on the trajectory revealed that the dichotomy is mural cells in the left wing and ECs in the right wing. (I) 2D UMAP projection of the VOs’ single cells coloured by the sample name. (J) Seventeen graph-based unsupervised clusters with their respective cell number and percentage in DB-VOs and ND-VOs in the table. (K) Endothelial cells (ECs) were distinguished into 4 clusters based on the expression of PECAM1 and CDH5. (L) Pericytes were identified based on the expression of CSPG4 (NG2) and PDGFRB. (M) vSMCs were characterized by the expression of ACTA2 (a-SMA) and TAGLN (SM22). (N) Classification of the whole 16 706 cells into a total of 9 clusters based on the distinct clusters of ECs, VSMCs, and pericytes. A small population of macrophages was identified based on the expression of CD14 and CD86 (data not shown). (O) The percentage of each vascular cell subtype within DB-VOs and ND-VOs. Endothelial-4 and pericytes-2 are specific for DB-VOs. (P) The relative expression of some selected markers in each of the 9 main subpopulations.

Figure 5.

Lack of full similarity of EC populations in vascular organoids (VOs) to organ-specific EC subtypes. The heatmap shows the degree of similarity between ECs in VOs and organ-specific subtypes of ECs found in the heart, brain, lung, kidney, and liver. Biomarkers for each organ were obtained from Table 3 of a recent publication (Trimm et al, 2023, Nature). The heatmap reveals that the endothelial 1-4 in VOs only partially express a few markers from each organ, indicating that they do not match any specific organ perfectly. These results highlight the need for developing organ-specific differentiation protocols for iPSCs to effectively produce VOs that closely resemble a desired organ. Abbreviations: aCap, aerocyte capillary; EC, endothelial cell; E, embryonic; gCap, general capillary; LSEC, liver sinusoidal endothelial cell.

Figure 6.

Cell-cell signaling network in diabetic vascular organoids (DB-VOs). (A) Analysis of the signaling sent from each cell group using the CellChat algorithm (R package). Nodes represent cell groups, and edges indicate signaling interactions, with thickness reflecting signaling strength. (B) The heatmap represents 33 significantly secreted signaling molecules in the cell-cell communication network. This highlights which signals contribute the most to the outgoing or incoming signaling of certain cell groups, providing a comprehensive view of the signaling roles played by various molecules.

Figure 7.

Differentially expressed hallmarks and genes in diabetic vascular organoids (DB-VOs). (A) Bubble plot showing the 35 hallmarks (y-axis) that were either overrepresented or underrepresented in at least one of the conditions (x-axis). Differentially expressed genes (DEGs) were identified by comparing the 9 main subpopulations/clusters of DBs against NDs, and significant upregulated and downregulated genes were used for preranked gene set enrichment analysis (GSEA) performed by fgsea (R package) and using MSigDB Hallmark dataset. Diabetic-specific clusters, endothelial-4 (EC4) and pericytes-2, were compared versus the rest of ECs (rECs) and rest of the mural cells, respectively. Three additional conditions were included in the analysis: total (t) mural cells of DB vs ND, total ECs of DB vs ND, and total DB cells vs ND cells. However, no significant hallmarks were identified in the latter 2 comparisons, and therefore they are not shown in the bubble plot. Circle size indicates the number of genes in each hallmark dataset. The GSEA plot on the left shows the normalized enrichment score and adjusted P-value (padj) for the ROS pathway and angiogenesis hallmarks, which were significantly overrepresented and underrepresented, respectively, in the diabetic-specific EC4 cluster. Bar plots show the leading edge genes’ expression in the DEGs dataset. Pericytes-2 showed no positively enriched hallmarks. (B) GAP43 was identified as the top DEG in the comparison of total ECs of DB versus total ECs of ND, with a fold change of 13.2 and FDR of 2.88 × 10⁻⁵. (C) GAP43 expression was highest in the diabetic-specific EC4 cluster. (D) GAP43 expression overlay on the uniform manifold approximation and projection (UMAP) plot, with the arrow indicating the DB-specific cell clusters. (E) Violin plot showing that GAP43 has 7.8 times significantly higher expression in total single cells of DB-VOs compared to ND-VOs in scRNA-seq data. (F) Box plot showing the log2 normalized expression value of GAP43 in an independent bulk RNA-seq dataset of pure DB iPSC-derived ECs (DB iPS-EC, n = 6, DB03, DB06, DB07, DB09, DB11, and DB23) versus ND iPS-EC (n = 2, ND19 and ND22), further confirming the scRNA-seq results. (G) Flow cytometry demonstrating the purity of iPS-EC using CD144 antibody before and after selection of endothelial cells by autoMACS for subculture and analysis of GAP43 expression by immunostaining and Western blotting. (H) Immunostaining of DB iPS-EC with typical EC markers CD144 and CD31, showing higher expression of GAP43. (I) Western blotting to quantify the relative expression of GAP43 in DB iPS-ECs versus ND iPS-ECs, with the box plot representing the normalized signal intensity values of GAP43 in Western blot bands of DB iPS-ECs (n = 5) versus ND iPS-ECs (n = 8).