CD31 defines a subpopulation of human adipose-derived regenerative cells with potent angiogenic effects

Cellular heterogeneity represents a major challenge for regenerative treatment using freshly isolated Adipose Derived Regenerative Cells (ADRCs). Emerging data suggest superior efficacy of ADRCs as compared to the ex vivo expanded and more homogeneous ADRCs (= ASCs) for indications involving (micro)vascular deficiency, however, it remains unknown which ADRC cell subtypes account for the improvement. Surprisingly, we found regarding erectile dysfunction (ED) that the number of injected CD31+  ADRCs correlated positively with erectile function 12 months after one bolus of autologous ADRCs. Comprehensive in vitro and ex vivo analyses confirmed superior pro-angiogenic and paracrine effects of human CD31+ enriched ADRCs compared to the corresponding CD31− and parent ADRCs. When CD31+, CD31− and ADRCs were co-cultured in aortic ring- and corpus cavernous tube formation assays, the CD31+  ADRCs induced significantly higher tube development. This effect was corroborated using conditioned medium (CM), while quantitative mass spectrometric analysis suggested that this is likely explained by secretory pro-angiogenic proteins including DKK3, ANGPT2, ANAX2 and VIM, all enriched in CD31+  ADRC CM. Single-cell RNA sequencing showed that transcripts of the upregulated and secreted proteins were present in 9 endothelial ADRC subsets including endothelial progenitor cells in the heterogenous non-cultured ADRCs. Our data suggest that the vascular benefit of using ADRCs in regenerative medicine is dictated by CD31+  ADRCs.

CD31+ ADRCs exhibit enhanced ex vivo paracrine angiogenic potential.We next investigated the pro-angiogenic potential of CD31+ cell populations isolated from ADRCs from 5 donors using magnetic cell sorting (MACS).Their enrichment was validated by flow cytometry and qPCR analyses (Supplementary Fig. 1).
We next developed a co-culture ex vivo assay in which human ADRCs were cultured in inserts for 15 days with Matrigel-embedded mouse corpus cavernosum explants at the bottom of the wells (Fig. 2 and Supplementary Fig. 2).The ADRCs initiated explant sprouting at day four, while well-defined mesh-like structures had developed on day 15 (Supplementary Fig. 2b) enabling quantification of sprouting in the well-developed area as previously described for the aortic ring sprouting assay 37 .The CD31+ ADRCs alone elicited a robust and significantly greater sprouting response as compared to ADRCs (p < 0.01, n = 5) or CD31-ADRCs (p < 0.01, n = 5) (Fig. 2), suggesting a paracrine interaction with penile tissue explants.Furthermore, using a standard aortic ring co-culture assay, CD31+ ADRCs again elicited increased tubular formation as compared to ADRCs (p < 0.05; n = 4) or CD31-ADRC populations (p < 0.01; n = 4) (Fig. 3).Sprouting cells from corpora cavernosa stained positively with NG2 but not CD31, (Supplementary Fig. 4a) indicating the presence of pericytes in the capillary-like structures in agreement with previous findings that pericytes are among the first to invade newly vascularized tissue 38,39 .The CD31+ ADRCs superior ability to facilitate angiogenesis and vascular repair increases understanding of the clinical data because of the vascular element in the ED pathogenesis 38,39 .
CD31+ ADRCs secrete a unique set of proteins associated with angiogenesis.We next looked for differentially secreted proteins by quantitative proteome analysis using mass spectrometry (MS).Conditioned media (CM) from ADRCs, CD31+ and CD31− ADRCs, respectively (n = 3 patients), were analyzed based on the relative protein abundances measured by LC/MS-MS employed with tandem mass tag labeling for relative quantitation.Mass spectra were searched against the Swissport database restricted to humans.Proteins were deduced on basis of at least one unique peptide identified with high confidence (false discovery rate (FDR) < 0.01).
A total of 997 human proteins were detected in all samples, of which 680 were present at high enough concentrations to enable quantification (listed in Supplementary Table 3).Gene ontology (GO) enrichment analysis, using Database for Annotation, Visualization, and Integrated Discovery (DAVID version 6.8; https:// david.ncifc rf.gov/), confirmed enrichment of secreted proteins belonging to the extracellular exosome (GO:0070062, FDR = 8.57E−168).A list of GO terms with significantly overrepresentation of detected proteins are given in Supplementary Table 4.
The differential protein expression was further analyzed using the Perseus MaxQuant program.Following hierarchical clustering, we identified eighteen differentially regulated proteins between CD31+ and CD31− ADRC secretomes (Fig. 4a and Supplementary Fig. 5 and Supplementary Table 3).Of these, fourteen proteins appeared nominally upregulated, while four were downregulated in the CD31+ versus CD31− population (p < 0.05) (Fig. 4a and Supplementary Fig. 5).Five known secreted proteins were upregulated in CD31+ ADRC derived CM: Angiopoietin 2 (ANGPT2), Dickkopf 3 (DKK3), Annexin A2 (ANXA2), CMP-N_acetylneyraminate-beta-galactosamide-alpha 2,3-sialytrasferase 1 (ST3GAL1) and Vimentin (VIM).Interestingly, all five proteins have been associated with angiogenesis [40][41][42][43][44] .The expression profiles of all, except ST3GAL1 were confirmed at the mRNA level in ADRCs, CD31+ and CD31− ADRCs co-cultured with mouse aortic rings (Fig. 4b and Supplementary Fig. 3a).Higher mRNA expression levels of ANGPT2, ANXA2, VIM and DKK3 as well as PECAM1 and VWF, were also observed in single cultures of CD31+ ADRCs compared to CD31− ADRCs (Supplementary Fig. 3b).This pattern was sustained throughout the culture period that also revealed a notable decline in ANGPT2, PECAM1 and VWF levels between day 8 and day 15 whereas ANXA2 and VIM levels gradually declined as DKK3 expression increased (Supplementary Fig. 3b).Thus, CD31+ ADRCs secrete a unique set of proteins associated with angiogenesis, which may explain the observed superior vascular effect in vivo.(2) more distally, a well-developed area is characterized with higher structural organization of tubes and establishment of mesh-like structures.The border between these areas is indicated with a stippled line.Quantification of the sprouting was done in three fixed-sized squares, one of these is indicated by a solid-lined box and shown at higher magnification in the lower line of figure panels.(b) Quantification and statistical analyses of the sprouting from mouse corpus cavernosum explants, evaluated based on number of nodes/mm 2 , total length (mm)/mm 2 , number of meshes/mm 2 , and mesh coverage i.e., mesh area per area analyzed.Data is based on five experiments, using cells from three male donors and two female donors.Each experiment consisted of triplicates for each tested cell type.Statistical analysis: For each experiment and condition, outliers were identified by the Rout method before normal distribution was confirmed using D' Agostino-Pearson or Kolmogorov-Smirnov normality tests as appropriate.The means of each of the 4 conditions (Negative control, ADRC, CD31− and CD31+ ADRCs) for each experiment were calculated and subsequently compared using one-way ANOVA.Depiction of data: Each data point represents a mean from one experiment.The box represents the mean of 5 means ± standard deviation (SD).Statistically significant p-values are shown in the figure panels.
Single-cell RNA sequencing of CD31+ ADRCs reveals cell type heterogeneity and enriched angiogenic potential specifically in PECAM1 high expressing endothelial cells.To determine the cellular identity of the CD31+ ADRCs, we used high-resolution single-cell RNA sequencing (scRNA-seq) from samples obtained from 4 donors.The resulting quality-controlled, single-cell data included a total of 24,403 human CD31-selected ADRCs that were integrated and clustered using uniform manifold approximation and projection (UMAP) with 54 dimensions at resolution 1.0, resulting in identification of 31 cell clusters (Fig. 6a,b), all containing cells from each of the 4 samples (Fig. 6c).ADRCs expressing the CD31-encoding gene PECAM1 were present in all 31 clusters (Fig. 6b,e), with 14 clusters (Clusters C1-C13, and C29) showing high average     6d).We subsequently searched for cellular subsets with angiogenic potential.We identified the significantly upregulated genes in each of the 31 clusters, and secondly, clusters with over-representation of upregulated genes with the GO term: "GO:0001525 ~ angiogenesis".We identified 14 clusters (EC clusters C1-C9, IC cluster C24, PC clusters C25 and C26, and AC clusters C28 and C31), representing 67.09% of all CD31+ selected cells, with over-representation of genes associated with the GO term: "GO:0001525 ~ angiogenesis", suggesting that these clusters could play a role CD31+ ADRC angiogenic effect.As expected, the EC (and PC) clusters represent major contributors to the angiogenic signature, and noteworthy, a cluster of immature ECs (C7) contain the highest number of significantly upregulated angiogenesis-related genes (Supplementary Table 8).Similarly, the most stem cell-like ACs (C31), defined by expression of DPP4, CD55, and SFRP4 (Supplementary Fig. 6b), possess a strong angiogenic profile.Finally, we observed a contribution from immune cell cluster C24, a subpopulation of monocytes where 98.0% express CD14.The majority (90%) of C24 cells also express the mural cell markers RGS5, ACTA2, and TAGLN (Supplementary Fig. 6a), suggesting this cluster defines a transitional stage between circulating monocytes and pericytes 45 .
We then investigated expression of genes encoding the 14 proteins that were significantly upregulated in the CM from CD31+ ADRCs (Fig. 4a).Indeed, all 31 clusters of CD31+ selected ADRCs contain cells expressing at least one of the 14 genes.However, clusters C5 (postcapillary venule ECs) and C7 (adipose-resident endothelial progenitor cells and immature angiogenic ECs, described in more details in the next section) expressed the highest numbers, 9 and 11 genes, respectively, including the 4 genes encoding the secreted proteins ANGPT2, ANAX2, and VIM (Fig. 7a).However, these clusters are relatively small, thus clusters 5 and 7 represent 1.96% and 0.80% of all CD31+ selected cells, respectively (Supplementary Table 6).Finally, the EC clusters C1-C9 contained significantly higher number of the 14 genes compared to the remaining combined clusters C10-C31 (p < 0.0001) and the IC, PC and AC clusters (Fig. 7b,c).
Thus, despite having a common therapeutic vascular phenotype, CD31+ ADRCs originate along the entire arterio-venous axis of the microvasculature, and therefore remain heterogenic in origin and likely also in function.

Discussion
Herein, we identify and characterize a potent angiogenic subfraction within ADRCs, which may underlie their beneficial therapeutic effect in vascular repair.
Cell therapy for vascular regeneration include both autologous and allogenic cells obtained from various sources and with varying degrees of manipulation (i.e.selection, culturing) (reviewed in 23 ).Mesenchymal stem cells (MSCs) derived either from adipose tissue (termed ASCs) or bone marrow are the most frequently used cell therapeutic products in general with 1000 + clinical MSC-trials registered world-wide 53 .Due to its accessibility, the adipose tissue stromal vascular fraction (referred to as ADRC herein) is a preferred cell source and since ex vivo expanded ASCs lack MHC class II expression, allogenic administration is largely without safety issues 54 .Despite being suitable as an off-the-shelf product, ASCs do not per se represent an improved product in comparison to ADRCs.Conventional culturing conditions profoundly impact cell phenotype and compared to the minimally processed heterogeneous ADRCs, the angiogenic properties of the more homogenous ASCs are reduced resulting in poorer outcomes in comparative in vivo studies 55 .This clearly suggests the existence of a more potent cell type (or synergistic mechanisms between multiple cell types) within the ADRCs.Thus, identification and characterization of such a cell type(s) will aid at realizing and maximizing the full clinical potential of ADRCs and -descendants for vascular regeneration in particular.
The pathogenesis of post prostatectomy erectile dysfunction (post RP-ED) is known to include a vascular component involving penile arterial insufficiency and/or veno-occlusive dysfunction and is thought to involve apoptosis of corporeal stromal cells, vascular smooth muscle, and endothelial cells.Using bone marrow-mononuclear cells (BM-MNCs), Yiou et al. 56 observed significantly improved penile vascularization and normalized penile endothelial function in 8/11 post RP-ED patients in a dose-dependent manner.In agreement with preclinical ED studies 57,58 , Yiou et al. suggested the cellular therapy effect could be due to angiogenic repair.The BM-MNCs and ADRCs share many cell types, including mesenchymal stromal/stem cells and endothelial (progenitor) cells however, Yiou and co-workers did not address distributions of BM-MNC subpopulations in their study 56 .With this in mind, we re-visited our post RP-ED data from 15 urine continent men treated with ADRCs by a single intracavernous injection 2 .We found no significant relationship between total ADRC numbers and the patient-reported outcome measure IIEF-5, implying that the clinical effect could not be ascribed to the simple presence of a (critical) cell mass, but more likely to a specific subpopulation.Furthermore, the effect did not correlate with the number of CD34+ ADRCs (stem/progenitor cells among others), or CD73+/ CD90+ ADRCs (mesenchymal cell populations), which is intriguing since these subpopulations represent the in situ ASCs.In contrast, the number of injected CD31+ ADRCs significantly correlated with the IIEF-5 score, implying an important role for CD31+ endothelial-and/or endothelial progenitor cells, likely due to their angiogenic/pro-angiogenic properties.In support of this notion, our data clearly demonstrate that enriched human CD31+ ADRCs exhibit superior paracrine angiogenic potential compared to both the CD31-depleted ADRCs and parent ADRCs.In the ex vivo angiogenesis assays, freshly isolated ADRC populations were seeded at the same density in insert wells and co-cultured with explants of murine corpora cavernosum or aortic ring explants.Analysis of PECAM1 (encoding CD31) and VWF mRNA levels validated the CD31 enrichment strategy and showed sustained expression pattern in the insert cells until the endpoint (aortic ring assay day 8).Underscoring CD31+ ADRC potency, the total number of the actively transcribing insert CD31+ ADRCs were lower that of the corresponding ADRC and CD31− ADRCs due to higher growth rates of the two latter (data not shown).
The superior paracrine effect could be ascribed to components present in the CD31+ cells conditioned medium.Cell culture medium for quantitative proteomics was collected from single cultures of ADRC subsets to circumvent contribution from mouse explant tissue.As seen with the insert CD31+ ADRCs, PECAM1 and VWF mRNA levels were sustained in the single cultures of CD31+ ADRCs up to day 8, whereafter levels were markedly reduced by day 15.This is in agreement with studies showing that CD31-expressing cells become overgrown by ASCs after 10 days of culturing 17,29 .Regardless, CM from day 15 CD31+ ADRCs outmatched the corresponding CD31− ADRC CM and ADRC CM when tested in pericyte migration assays as well as in the aortic ring assay.Furthermore, the CD31+ ADRC secretome showed enrichment of the pro-angiogenic factors ANGPT2 40,59 , DKK3 42 , VIM 60 , and ANAX2 61 .These 4 factors may contribute individually and synergistically to the observed effect.DKK3 has been shown to restore erectile function in diabetic mice by enhancing angiogenesis 62 .Due to the presence of abundant bovine serum proteins and inherent limitations of the relative quantification method (iTRAQ), other differentially expressed secreted proteins are likely not to have been detected which is also the case for proteins solely present in one type of CM.All 14 upregulated proteins were most frequently expressed in EC subtypes as evidenced by single-cell RNA sequencing of ~ 24,000 CD31+ enriched ADRCs.Our data suggest that the upregulation is not a culturing artefact but that ECs at the time of seeding are actively transcribing the relevant mRNA species.This underscores the importance of in-depth characterization of ECs, which are also the more PECAM1 expressing cells in the enriched ADRC fraction.Notably, in the non-cultured CD31+ enriched ADRCs, DKK3 mRNA was only identified in a pericyte subpopulation (C26), in agreement with other data suggesting that DKK3 is rarely expressed in normal endothelial cells 63 , but still critical for endothelial function and -regeneration 64 .Song et al. 62 , also suggested DKK3 to affect erectile dysfunction partially through pericytes and in general, pericytes play a vital role in angio-/vasculogenesis 65 .Whether DKK3 protein in the CM was produced by PCs expanding in vitro, or induced in ECs remains unanswered.However, the presence of various cell types in the CD31+ enriched fraction may have been advantageous for the angiogenic effect as several studies suggest synergistic regenerative effects from cell mixtures.Using a rat in vivo bone regeneration model, Sass an coworkers 66 showed that CD31+ enriched cells from peripheral blood (PBMC) performed significantly better than further fractionated CD31+ CD14− or CD31+ CD14+ PBMCs.Revisiting ADRC (SVF) clinical data, Kilinic and co-workers 67 , reported improved efficacy using a 2:1 ratio of adipose-derived stromal/stem cells to adipose-tissue derived endothelial progenitor cells (EPC)s.Defined as CD34high CD45− CD31+ CD146 (MCAM)+ 67 these EPCs likely correspond to cells in our EC clusters, especially C7.C7 cells express the highest levels of ENG (CD105) and CD200 corresponding to recently reported AEPCs 29 .
In theory, cells with CD31/PECAM1 expression may have a homing advantage since CD31/PECAM1 engage in trans-homophilic interactions 71 .Interestingly, several non-endothelial clusters including cluster 10 and 11 also express high CD31 levels.These clusters are annotated as immune cells, suggesting CD31 plays a less uncharacterized role in these cells.
Collectively, our data support the notion that a potent angiogenic ADRC subpopulation exists, which may underlie the superior beneficial therapeutic effects as compared to cultured ASC counterparts.The CD31+ ADRC subset can be enriched and adjusted to meet GMP-compliance for direct autologous use or ex vivo expansion.In this regard the populations identified herein by scRNA-seq may serve as reference for future refinement of ADRC isolation-and expansion methodologies.

Methods
All methods were carried out in accordance with relevant guidelines and regulations, and are reported in accordance with ARRIVE guidelines.
Patient samples and data.Clinical outcome-and flow cytometry data used for the retrospective correlation analyses herein, were obtained in a previously described clinical phase 1 safety study for treating erectile dysfunction (ED) 2,3 .Clinical efficacy was addressed by evaluation of IIEF-5 and EHS scores, and the ADRC phenotype assessed by flow cytometry using markers for CD31, CD34, CD73 and CD90 3 .
Lipoaspirates for ADRC isolation (n = 10) were obtained from placebo patients enrolled in two separate phase 2 RCTs using ADRCs for cell therapy of ED (unpublished; awaiting final data analyses) and lymphedema (submitted).All patients gave written informed consent before participation.Animals.8-week-old C57BL6 mice (male and female) were obtained from Janvier Labs (Le Genest-Saint-Isle, France), and kept in an animal facility, with controlled temperatures, a 12-h light/dark cycle.For collection of penile and aortic tissue, animals were sacrificed with CO 2 .All experimental protocols were approved by The Regional Committees on Health Research Ethics for Southern Denmark.

Isolation of adipose derived regenerative cells (ADRC).
ADRCs were isolated from human lipoaspirates as described in detail previously 3,72 .
Magnetic cell sorting enrichment of CD31+ ADRCs.Enrichment of CD31 positive and -negative ADRC subpopulations was accomplished by magnetic cell sorting (MACS) using MS separation columns (Miltenyi Biotech) under RNase-free conditions.Initially, the isolated ADRC was subjected to red cell lysis (RBC buffer, Miltenyi Biotech, cat no.130-094-183).Magnetic labelling was performed using a CD31 microbead kit (Miltenyi Biotech) after which cells were sorted on an OctoMACS™ Separator (Miltenyi Biotech).Hereby, CD31+ and CD31− ADRCs were obtained based on a positive or negative selection for the CD31 marker.
Corpus cavernosum explant co-culture assay.Mouse corpus cavernosum was prepared according to Ghatak et al. 73 .Briefly, following termination, the penile tissue from 6 to 8-week-old C57BL6 mice (Janvier, France) was dissected.For the ex vivo assay, the tissue was cut into three/four 1 mm pieces, and the explants were plated on growth factor reduced Matrigel (Corning, cat no.356231).Following polymerization, explants were supplemented with 1 ml EBM and 2.5% FBS and incubated at 37 °C with 5% CO 2 48 .After 24 h, media was replaced with 1.5 ml fresh EBM with 2.5% FBS.Next, Nunc™ polycarbonate Cell Culture Inserts with 0.4 µm pore size (Thermo Scientific catalog no.140620) were placed above the explant, and 20,000 cells (ADRCs, CD31− or CD31+) were seeded in 500 µl EBM with 2.5% FBS (3 wells pr.cell type).The plate was then incubated at 37 °C with 5% CO 2 for 15 days without medium exchange.
Structurally distinct regions of tubular sprouting from corpus cavernosum explants were comparable to the regions previously described for the aortic ring assay 37 .The tubular network was quantified in defined, fixed areas (see Supplementary Fig. 2c) using the ImageJ software (Fiji) with an angiogenesis plugin.The angiogenesis evaluation was based on number of nodes, total length of tubes, number of meshes/mm 2 and mesh coverage.
Aortic ring ex vivo assay.The aortic ring assay was set up as previously described 74 .Briefly, the thoracic aorta was dissected from 8 to 12 weeks old C57BL6 mice into 0.5-0.7 mm wide rings, and serum-starved in EBM for 24 h.For indirect co-culture assay, one aortic ring per well of a 24-well Nunc™ Carrier Plate was embedded between two 40 µl drops of Matrigel with 700 µl EBM containing 2.5% FBS.Next, Nunc™ polycarbonate Cell Culture Inserts with 0.4 µm pore size were placed above the aortic rings and 50,000 cells per insert (ADRCs, CD31+ or CD31− ADRCs) were seeded in 500 µl EBM with 2.5% FBS (8 wells pr.cell type).To test CM effects, Matrigel embedded aortic rings were cultured in 500 µl CM from ADRC's, CD31+ or CD31− ADRCs.The plate was cultured at 37 °C for 8 days, at which time, pictures were acquired with phase contrast microscopy (5 × magnification) and analyzed using the angiogenesis plug-in in ImageJ.
Stable isotope labeling of protein samples with TMT-10 plex.Proteins from CM were isolated by transferring the supernatant to five equivalents of ice-cold acetone.Proteins were reduced using 5 mM dithiothreitol (DTT), followed by 15 mM iodoacetamide blocking before trypsination overnight at 37 °C at a protein:trypsin (Promega, Madison, WI, USA) ratio of 50:1 w/w. 10 µg of the tryptic digest was labeled with a 10-plex TMT-kit (Thermo Scientific), resuspended in anhydrous ethanol, and a 40 µg sample was labeled according to the scheme in Supplementary Table 1.Labeled samples were pooled in equal ratios, dried in a vacuum centrifuge, re-dissolved in 50 µl trifluoroacetic acid solution (0.1%), purified, loaded on a reverse phase microcolumn (equal w/w amounts of Poros R2 and Oligo R3 material) and fractionated by high pH liquid chromatography as described 75 .
The Fractions were analyzed by RP-nanoLC-MS/MS on an Orbitrap Eclipse mass spectrometer (Thermo Fisher Scientific) equipped with a nano HPLC interface (Dionex UltiMate 3000 nano HPLC) as described 75 .Raw data files were quantified using Proteome Discoverer version 2.4 (Thermo Scientific) as previously described using human and bovine database searches 76 .
Isolation of primary murine cavernous pericytes.Penile tissue from mice was prepared and embedded in Matrigel as described for the Corpus Cavernosum explant assay.The Matrigel drop and the tissue was plated in a 60 mm petri-dish.After Matrigel polymerization, the petridish was supplemented with EBM containing 20% FBS.The medium was changed every 4 days.Cells sprouted from the corpus cavernosum explants (4 per dish), and after approximately 2 weeks, they became confluent.Only cells migrating out of the Matrigel onto the plastic surface of the petri-dish, were further sub-cultivated.These migrated pericyte cells were trypsinized and seeded at 10,000 cells/cm 2 density in EBM with 20% FBS for further experiments.

Migration in vitro assays (with conditioned media).
To test the angiogenic effect of conditioned medium from CD31+ ADRCs, we adopted the wound healing assay using pericyte migration.Primary pericytes were seeded in culture inserts (ibidi culture-insert 2 well, ibidi GmbH, Martinsried, Germany) at a density of 25,000 cells per well.After allowing cells to attach overnight, we removed the culture inserts creating a cell-free gap and washed the cells with sterile PBS to remove non-adherent cells.We then provided 300 μl of CM from ADRC, CD31+ , or CD31− ADRCs, or EBM with 2.5% FBS as a control.Images of cell-free gaps were taken immediately after removing inserts with a bright field microscope at 5 × magnification.We monitored the gap for 24 h after culturing cells in respective CM, at which point images of the gaps were captured again.The cells migration ability was evaluated by the area of the gap they had covered in 24 h using ImageJ.
RNA purification and RT-qPCR.RNA was extracted from samples of ADRCs, CD31+ and CD31− ADRCs, which were obtained from the aortic ring co-culture assay, the CM production set-up, and from the in vivo Matrigel cell culture.The samples were homogenized, and extraction of the total RNA was performed using the Tri Reagent® protocol (ThermoFisher Scientific).The RNA quantity and purity was assessed by nanodrop measurements (Nanodrop® Technologies).The mRNA was reversed transcribed using a High-Capacity cDNA kit (Applied Biosystems, Thermo Fisher Scientific) and the RT-qPCR reaction was performed using

Figure 1 .
Figure 1.Patient-reported outcomes following cell therapeutic treatment for radical prostatectomy-related erectile dysfunction, correlate positively with the number of injected CD31+ ADRCs.Fifteen urine-continent patients suffering from erectile dysfunction following radical prostatectomy were treated with a single intracavernousal injection of autologous ADRCs and the patient-related outcomes (i.e., recovery of erectile function) were evaluated according to the IIEF-5 score 12-months post-treatment 2 .The individual IIEF-5 scores are plotted against: (a) the corresponding total numbers of injected ADRCs and the numbers of the (b) CD73+ subpopulation, (c) CD90+ subpopulation, (d) CD31+ subpopulation as well as (e) CD34+ subpopulation of ADRCs, respectively, and analyzed for correlation using Spearman r correlation test.p-values are shown in the figure panels.Only the number of injected CD31+ ADRCs significantly correlated with higher, improved IIEF-5 scores.

Figure 2 .
Figure2.Test of potential paracrine angiogenic effects of human ADRCs, CD31-and CD31+ ADRCs in ex vivo co-culture assays with mouse corpus cavernosum explants.(a) Representative pictures of the sprouting from mouse corpus cavernosum explants after 15 days of co-culture with 20,000 ADRCs, CD31−, and CD31+ ADRCs, respectively.Two structurally distinct regions can be identified by visual inspection: (1) in proximity to the corpus cavernosum explants an unstructured area is characterized by high cell numbers but low structural organization; (2) more distally, a well-developed area is characterized with higher structural organization of tubes and establishment of mesh-like structures.The border between these areas is indicated with a stippled line.Quantification of the sprouting was done in three fixed-sized squares, one of these is indicated by a solid-lined box and shown at higher magnification in the lower line of figure panels.(b) Quantification and statistical analyses of the sprouting from mouse corpus cavernosum explants, evaluated based on number of nodes/mm 2 , total length (mm)/mm 2 , number of meshes/mm 2 , and mesh coverage i.e., mesh area per area analyzed.Data is based on five experiments, using cells from three male donors and two female donors.Each experiment consisted of triplicates for each tested cell type.Statistical analysis: For each experiment and condition, outliers were identified by the Rout method before normal distribution was confirmed using D' Agostino-Pearson or Kolmogorov-Smirnov normality tests as appropriate.The means of each of the 4 conditions (Negative control, ADRC, CD31− and CD31+ ADRCs) for each experiment were calculated and subsequently compared using one-way ANOVA.Depiction of data: Each data point represents a mean from one experiment.The box represents the mean of 5 means ± standard deviation (SD).Statistically significant p-values are shown in the figure panels.

Figure 3 .
Figure 3. Test of potential paracrine angiogenic effects of human ADRCs, CD31− and CD31+ ADRCs in ex vivo co-culture assays with mouse aortic ring explants.(a) Representative pictures of the sprouting from mouse aortic ring explants after 8 days of co-culture with 50,000 ADRCs, CD31− and CD31+ ADRCs, respectively.(b) Quantification and statistical analyses of sprouting from aortic rings evaluated based on the total branch length (mm), number of branch points, number of meshes and total mesh area (mm 2 ).Data is based on four experiments, using cells from three male donors and one female donor.Each experiment consisted of 6-8 replicates for each tested cell type.For each experiment and condition, outliers were identified by the Rout method before normal distribution was confirmed using D' Agostino-Pearson or Kolmogorov-Smirnov normality tests as appropriate.The means of each of the 4 conditions (Negative control, ADRC, CD31− and CD31+ ADRCs) for each experiment were calculated and subsequently compared using one-way ANOVA.Depiction of data: Each data point represents a mean from one experiment.The box represents the mean of 4 means ± standard deviation (SD).Statistically significant p-values are shown in the figure panels.

Figure 4 .Figure 5 .
Figure 4. Mass spectrometry reveals significantly different secretomes of CD31+ and CD31− ADRCs.ADRCs, and CD31− and CD31+ ADRCs, respectively, were cultured for 15 days and the isolated proteins from the corresponding conditioned media were analyzed by RP-nanoLC-MS/MS analysis.Data is based on cells from three donors (one male and two females).(a) Heat map showing 14 significantly upregulated and 4 significantly downregulated proteins in the CD31+ ADRC vs CD31− ADRC conditioned media, as identified by individual t-tests with a p-value of 0.05.Upregulated and downregulated proteins are indicated by red and blue colors, respectively.A volcano-plot-representation of the differentially expressed proteins can be seen in Supplementary Fig. 5.(b) Relative mRNA levels of ANGPT2, ANXA2, DKK3, and VIM in ADRC, and CD31− and CD31+ ADRCs after 8 days of co-culture with mouse aortic ring explants confirming the upregulation of these transcripts under the conditions in the ex vivo assay.The data was based on cells from one donor, and eight replicates for each of the three populations.To obtain sufficient material, two replicates were pooled in relation to RNA extraction and RT-qPCR performed (in technical triplicates) on the resulting 4 replicates per population.The mRNA expression was normalized to the expression of the reference genes B2M and TBP, based on the geNorm analysis performed in qBase+ (CV = 0.066, M = 0.191).Statistical analyses were performed using ordinary one-way ANOVA.Statistically significant p-values are shown in the figure panel.

Figure 6 .
Figure 6.Single-cell RNA sequencing of 24,403 CD31+ ADRCs from 4 donors.(a) Uniform manifold approximation and projection (UMAP) 2D visualization of 24,403 CD31+ enriched ADRCs in 31 clusters.(b) Expression of the CD31-encoding PECAM1-gene in cells visualized in a 2D UMAP plot.(c) Percentage of total cells in each of the 31 clusters visualized for each of the four patient samples (Pt.1-4) and for all four samples combined.(d) Dot plot showing marker-based assignment of the 31 clusters to four major groups: endothelial cells (EC)s (markers CLDN5 and VWF), immune cells (IC)s (PTPRC, CD74, and CD14), perivascular mural cells (PC)s (RGS5, ACTA2, and TAGLN), and adipose stem and progenitor cells (AC)s (CFD, PDGFRA, and DCN).Color saturation of a dot indicates the average gene expression level in positive cells, while dot size reflects the percentage of cells in each cluster expressing the gene.(e) Violin plot of PECAM1 expression levels in cells of the 31 clusters.(f) Dot plot of marker genes revealing the molecular identities of clusters 1-13 and 29, which were selected based on high average PECAM1 expression and/or high percentages of PECAM1 expressing cells.

Figure 7 .
Figure 7. Several clusters of CD31+ selected ADRCs express genes encoding the 14 proteins that were significantly upregulated in the conditioned media from cultured CD31+ selected ADRCs.(a) Dot plot showing expression of the genes encoding the 14 proteins that were significantly upregulated in the conditioned media from cultured CD31+ selected ADRCs.Color saturation of a dot indicates the average gene expression level in positive cells, while dot size reflects the percentage of cells in each cluster expressing the gene.Note that particularly clusters 5 (presumptive postcapillary venule endothelial cells) and 7 (presumptive immature angiogenic endothelial cells) express high numbers of the genes, 9 and 11 genes, respectively, including the 4 genes encoding the secreted proteins ANGPT2, ANAX2, ST3GAL1 and VIM.(b) Number of genes, encoding proteins that were significantly upregulated in the conditioned media from cultured CD31+ selected ADRCs, expressed in EC clusters 1-9 compared to the remaining combined clusters 10-31 using the Mann-Whitney test.(c) Number of genes expressed in EC clusters 1-9 compared to the groups of IC clusters (10-24), PC clusters (25-27), and AC clusters (28-31), respectively using one-way ANOVA followed by the Dunnett test comparing every mean to the EC mean.The bars in panels b and c represent means.AC adipose stem and progenitor cells, EC endothelial cells, IC immune cells, PC perivascular mural cells.