Comprehensive transcriptomics and proteomics analysis of neointima formation in human saphenous vein: implications for bypass graft disease

Abstract

Introduction:

Human saphenous veins (SV) are widely used as grafts in coronary artery bypass (CABG) surgery but often fail due to neointima formation. Little is known, however, regarding the cellular, transcriptomic and proteomic dynamics of neointima formation in human veins. Here, we performed transcriptomics and proteomics analysis in an ex vivo tissue culture model of neointima formation in human SV procured for CABG surgery.

Methods and results:

Histological examination demonstrated significant elastin degradation and neointima formation (indicated by increased neointima area and neointima/media ratio) in SV subjected to tissue culture. Analysis of data from 72 patients suggest that the progression of SV remodeling and neointima formation differs according to sex and body mass index, which negatively associated with neointima formation in males only. RNA sequencing demonstrated upregulation of pro-inflammatory and proliferation-related genes during neointima formation and identified novel processes, including increased cellular stress and DNA damage responses, reflecting tissue trauma associated with vein harvesting. Proteomic analysis identified upregulated extracellular matrix-related and coagulation/thrombosis proteins and downregulated metabolic proteins. Spatial transcriptomics, used to infer regionally enriched gene expression, suggested dynamic alterations in fibroblast and vascular smooth muscle cell (VSMC) states during neointima formation. Specifically, we identified the emergence of HES1⁺ and MMP2⁺/MMP14⁺ expression in VSMCs and fibroblasts, respectively, during neointima formation. Furthermore, our data suggest that MIR647, identified through screening, maintains VSMC contractile gene expression.

Conclusion

Our findings suggest dynamic transcriptomic and proteomic changes during neointima formation in human veins and provide useful mechanistic information for the pathogenesis of SV graft disease.

New & Noteworthy

Using multi-omics and spatial transcriptomics, we uncover dynamic molecular and cellular changes driving neointima proliferation in human saphenous veins, the most common conduit for bypass surgery. Our study highlights sex- and BMI-associated differences, novel fibroblast and smooth muscle cell states, and a role for microRNA-647 in preserving vascular contractile phenotype. These findings provide new insight into the mechanisms of vein graft failure and may guide future strategies to improve coronary bypass outcomes.

Graphical Abstract

Introduction

Human saphenous veins (SV) are the most commonly employed conduits in coronary artery bypass grafting (CABG) surgery (estimated 400,000 each year in the US)(1). Despite their widespread use, over 50% of all SV grafts fail within 10 years of implantation(2–4). Early graft failure mainly results from acute thrombosis caused by technical errors, vessel trauma, or hypercoagulable states, while intermediate and late failures are driven by intimal hyperplasia (e.g., the accumulation of smooth muscle cells and extracellular matrix) and subsequent development of atherosclerosis. These pathological changes are initiated by the exposure of the venous graft to the high-pressure arterial circulation, leading to endothelial injury, inflammation, and vascular remodeling. Clinical studies have demonstrated a positive association between SV graft failure and mortality in patients undergoing CABG(5). However, treatment of SV graft disease is clinically challenging and often requires complex interventional procedures or re-do CABG surgery, which is associated with high morbidity and mortality. To date, no medical therapies have been identified to effectively mitigate the process of neointima formation in SV. Prior research has identified a number of novel biochemical and molecular pathways that may contribute to SV graft failure, but thus far, the findings from these studies have not been successfully translated into clinical practice, due to it being a multifactorial process involving patient-specific factors, inflammation, thrombosis, and endothelial and VSMC dysfunction. This illustrates the critical need to better understand the mechanisms of neointima formation in SV in order to identify strategies to prevent SV graft failure.

The process of surgically harvesting the SV and preparing them for surgery imposes tissue trauma that is thought to serve as a trigger for neointima formation(6,7). Vascular smooth muscle cells (VSMC), the major contractile cells of the vascular wall, undergo phenotype switching in response to vascular injury, leading to dedifferentiation, migration, aberrant proliferation and loss of contractile function(8). Adventitial fibroblasts likewise have been reported to convert to myofibroblasts and migrate toward the subendothelial space following vascular injury, contributing to neointima formation(9). Fibroblasts and VSMC isolated from SV have been studied in vitro and may provide insight into mechanisms and clinical factors associated with SV graft failure. Furthermore, unlike arteries, fibroblasts in human SV are not confined to the adventitia but are dispersed throughout the wall, and many veins exhibit pre-existing intimal hyperplasia at the time of grafting (10,11). In this regard, Kenagy et al reported that proliferative responses of VSMC and fibroblasts isolated from SV correlated with SV graft stenosis in patients undergoing bypass surgery for peripheral artery disease(12), suggesting that the propensity to develop SV graft disease may in part be related to intrinsic properties of the VSMC and fibroblasts contained therein.

Large clinical studies have demonstrated sex disparities wherein fewer CABG surgeries are performed in women, who also experience higher perioperative mortality and worse long-term outcomes following CABG surgery, than men(13). Additionally, an “obesity paradox” has been reported wherein overweight to moderately obese patients exhibit reduced perioperative mortality and better intermediate-term survival rates as compared to their lean counterparts(14). The role of SV graft failure in mediating these disparities is unknown, and the influence of sex and body mass index (BMI, an indicator of obesity) on neointima formation in human SV has not been systematically investigated.

Recently, multi-omics approaches have been employed to investigate the molecular mechanisms and cellular dynamics of neointima formation(15); however, such studies cannot be readily conducted in human blood vessels. Here, we took advantage of an established tissue culture model of neointima formation in human SV(16–19), wherein segments of SV harvested during CABG surgery are placed in tissue culture containing 30% fetal bovine serum, which induces neointima formation over a ~7 day period. Utilizing transcriptomics and proteomics analyses, we detected dynamic molecular and cellular changes associated with neointima formation in human SV. Furthermore, spatial transcriptomics revealed global upregulation of MMP2 and MMP14 in fibroblast-enriched capture spots and HES1 in VSMC-enriched capture spots, patterns not previously associated with neointima formation in rodent models. In addition, our data suggest that the progression of SV remodeling and neointima formation may differ according to sex and BMI. Finally, we demonstrate the model’s relevance by showing that MIR647 modulates VSMC contractile gene expression. Taken together, these findings provide novel insight into mechanisms of neointima formation in human blood vessels and may be pertinent to SV graft disease pathogenesis and the clinical factors influencing outcomes after CABG surgery.

Methods

Ethical statement.

Collection of the human SV samples and related experiments were designated as exempted from human subjects research by the Institutional Review Board (IRB) at Augusta University and performed in compliance with the Declaration of Helsinki. All experiments were carried out in accordance with institutional biosafety and chemical safety guidelines and regulations.

Human SV ex vivo culture.

Collection and culturing of human SV method was adapted from our previous protocol(16). In brief, freshly collected SV samples were obtained from patients undergoing CABG and immediately transported to the laboratory. The SV segments were placed in sterile phosphate-buffered saline (PBS) and transversely cut into 3 mm segments, some of which were stored at −80° C for subsequent analyses (Day 0). Remaining segments were cultured for 7 days (Day 7) in RPMI 1640 containing NaHCO₃ 2 g/L, penicillin/streptomycin (100 IU/100 μg per ml), L-glutamine 4 ml/L, and 30% fetal bovine serum (FBS) at 37°C in 5% CO₂. Media was replaced every other day, and then, segments were stored at −80° C for subsequent analyses. All SV samples were collected from patients with the endovascular harvesting technique (EVH).

Bulk RNA-sequencing (seq) in SV and data analysis.

Total RNA from six biological replicates of Day 0 and Day 7 human saphenous veins was DNase-treated and assessed for quality using the Agilent TapeStation. All samples had RNA integrity numbers (RIN) ≥7.0 and passed downstream library QC, consistent with expected variability for human surgical tissue. Paired samples from each donor were submitted to Genewiz and Integrated Genomics Core at Augusta University for library preparation using ribosomal RNA depletion. Sequencing was performed at a depth of 20 million reads per replicate using Illumina HiSeq 2500 system, using a 150 bp paired-end protocol (Illumina). Raw reads were trimmed using cutadapter version 4.1, and the processed reads were mapped to the GRCh38/hg38 genome assembly by using STAR version 2.7.0(20). DESeq2 version 1.22, which utilizes the Wald test to assess significance followed by the Benjamini-Hochberg (BH) procedure, was used to generate normalized read counts and conduct differential gene expression analysis with a false discovery rate (FDR) threshold of less than 0.05 using R studio version 494 (https://posit.co/download/rstudio-desktop/) and public server of usegalaxy.org(21,22). Principal component analysis (PCA) and volcano plot was created through ggplot2. Gene ontology analysis and gene-network graph were performed and created through clusterProfiler version 3.0.4(23).

Liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis.

Human saphenous vein segments were thoroughly washed in ice-cold PBS to remove all serum and blood contaminants, then homogenized in RIPA buffer with protease/phosphatase inhibitors to extract both cytosolic and extracellular matrix (ECM) proteins. Protein lysates from three biological replicates of Day 0 and Day 7 SV were submitted to the Proteomics Core at Augusta University. Protein digestion and mass spectrometry were performed as previously described(24). Briefly, protein lysates were separated through Ultimate 3000 nano-UPLC system (Thermo Scientific) and run on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Scientific). Raw data were analyzed using Proteome Discoverer (v1.4, Thermo Scientific) and searched against the UniProt human protein database. Peptide-spectrum match (PSM) values, a total number of identified peptide spectra matched for the protein, were log-transformed to normalize the dataset, and missing values were imputed through Amica version 3.0.1(25). After normalization of the dataset, differential protein expression analysis was performed with threshold set at FDR of less than 0.05. Gene ontology analysis and gene-network graph were performed and created through clusterProfiler version 3.0.4(23). For targeted proteomics, equal protein amounts from the same lysates were analyzed using stable isotope–labeled peptides as previously described. The labeled peptides were quantified to identify differential protein expression across experimental conditions. Two-sample t-test was used for each protein, using an initial p-value cutoff of 0.05. For proteins where p-values could not be calculated due to missing values, we selected the DE proteins identified in at least three times in one group (with a coefficient of variation, CV≤0.6) but not identified in the other group as described (26).

Visium spatial transcriptomics.

Spatial Transcriptomics data was generated according to the manufacturer’s protocol (10x Genomics). Briefly, 5-μm FFPE sections from three biological replicates of Day 0 and Day 7 SV with DV200 >50% were mounted onto Visium FFPE slides, deparaffinized, and stained with H&E prior to imaging. Following brightfield image acquisition, probe hybridization, ligation, extension, and library construction were carried out using the Visium FFPE Gene Expression workflow. Libraries were sequenced on a NovaSeq 6000 system (Illumina). The intima, media, and adventitia were demarcated using H&E and VVG staining. Demultiplexed FASTQ files were processed into spot-level count matrices using SpaceRanger v1.3.1 (10x Genomics) and mapped to GRCh38. Probe hybridization and library construction were performed using the Visium Human Transcriptome Probe Set v2.0 as part of the FFPE workflow. Spot annotations were performed in Loupe Browser 5 (10x Genomics).

For downstream analysis, spots with <500 UMIs or >5% mitochondrial reads were removed as low-quality. Remaining spots were normalized using SCTransform (Seurat v4.3) (27,28). Dimensionality reduction was performed with PCA on the top 2,000 variable genes, followed by shared nearest neighbor (SNN) clustering at a resolution of 0.6. To infer cell-type identities, we used the Seurat label transfer algorithm referencing publicly available human arterial single-cell RNA-seq datasets (GEO: GSE155468 and GSE131778(29,30)). Because single-cell datasets for human saphenous veins are not yet available, arterial references were selected based on vascular similarity, and this limitation is acknowledged. Cell-type proportion analysis was performed to quantitatively assess changes in cellular composition. All samples were merged, ~5000 high-quality spots were clustered, and cellular identities were assigned using published single-cell signatures together with spatial tissue morphology(31,32). Differentially expressed genes were defined using the Benjamini–Hochberg FDR <0.05 threshold, and top DEGs were ranked by absolute fold-change. Gene ontology analysis followed previously published methods, and pseudotime trajectory analysis was performed using Monocle3(33). For comparative gene expression between vessel layers, barcoded spots were overlaid on H&E images, assigned to intima, media, or adventitia, and analyzed separately within Seurat.

Quantitative PCR (qPCR).

Total RNA was extracted from human SV with QIAzol Lysis Reagent, and purified with miRNeasy Mini Kit (Qiagen). Real time quantification of mRNA levels of the genes of interest was performed using Brilliant II SYBR Green QPCR Master Mix (Agilent Technologies) per manufacturer’s instructions. Normalized Ct values were subjected to statistical analysis and fold change was calculated by ΔΔ Ct method as described previously, normalized against 18S(34,35). miRCURY LNA SYBR Green PCR Kit (Qiagen) and appropriate miRCURY LNA MIRNA PCR primers were used to quantify MIRNA expression. Synthetic spike-in RNA or 18s were used as respective controls. Primer sequences are listed in Supplemental Table 1.

Isolation of human saphenous vein smooth muscle cells (HSVSMCs).

Saphenous vein segments were rinsed with sterile PBS, the adventitia was carefully removed, and the vein was cut longitudinally and pinned. The intimal layer was gently scraped with a surgical scalpel to remove endothelial cells. The medial layer was peeled and minced into 1 mm² pieces before being plated on a 12-well plate at 37°C in Smooth Muscle Cell Growth Medium (SmGm-2, Lonza) with 20% fetal bovine serum and 1% penicillin-streptomycin.

Culture and transfection of HSVSMCs.

HSVSMCs were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37°C in a 5% CO₂ humidified incubator standardize conditions across experimental groups and to follow reagent-specific manufacturer recommendations. Cells were seeded in 6-well plates and grown to 60–70% confluence before starvation overnight with 0.2% FBS, followed by transfection with RNAiMAX (Qiagen) and 25 uM of MIR-647 mimic, or negative control (MedChemExpress) according to the manufacturer’s protocol. For loss-of-function studies, 10 uM of siMIR-647 or scrambled siRNA (Qiagen) were used. Cells were collected 24–48 hours post-transfection.

Western blotting.

Protein extraction and Western blotting were performed as described previously(34). Antibodies used are listed in Supplemental Table 2.

Immunohistochemistry and histology.

The SV tissues were fixed in 10% neutral buffered formalin for less than 24 hours, dehydrated with 70% ethanol followed by a series of graded alcohol immersions, cleared with xylene, and embedded in paraffin wax. Paraffin-embedded SV tissues were sectioned (5 μm) transversely for histology, immunohistochemistry, and spatial transcriptomics to preserve the circumferential vessel architecture, and then, stained with hematoxylin and eosin (H&E), Verhoeff van Gieson (VVG), α-smooth muscle actin (SMA), VE-cadherin, and MMP2, and DAB Substrate (Vector Labs) kits were used for visualization as previously described(34). Quantification of intima, media, adventitia and total vessel area of SV were measured as previously described(36). Luminal area was quantified by manual tracing in ImageJ, and when a linear metric was required, lumen diameter and wall thickness were defined as the maximum luminal distance and maximum perpendicular wall thickness, respectively. Briefly, quantification was analyzed with ImageJ software (https://imagej.net) using freehand selection tool on VVG stained slides. The number of elastin breaks was counted in three sections per tissue sample to quantify elastin degradation. Elastin integrity was evaluated on VVG-stained sections using a semi-quantitative scale from 0 to 4, where 0 indicated intact elastic lamellae, 1 mild waviness, 2 moderate fragmentation, 3 severe fragmentation, and 4 complete loss of elastic structure. Scoring was performed independently by two blinded observers. To quantify the immunostaining data, stained area in the intimal and medial region was analyzed using ImageJ software. Antibodies used are listed in Supplemental Table 2.

Data availability.

All transcriptomic datasets generated in this study have been deposited in the Gene Expression Omnibus (GEO) under accession numbers GSE310013 (bulk RNA-seq) and GSE310351 (spatial transcriptomics). RNA-seq and spatial transcriptomics quality-control metrics and sequencing statistics are provided in Supplemental Tables 4–6. Full raw proteomics datasets are provided in the Supplemental Files. All additional data supporting the findings of this study are available from the corresponding author upon reasonable request.

Statistical Analysis.

Sequencing datasets for transcriptomics and proteomics were analyzed as described above. All experiments were repeated in at least 3 independent experiments, and statistical analysis was conducted with GraphPad Prism 10.0. Quantitative results were presented as mean±SEM with individual data points additionally visualized using violin plots or paired line plots where indicated. In vivo data with sample size > 6 were tested for normal distribution by the Shapiro-Wilk test. A two-tailed t test was performed when all groups were normally distributed, and Mann-Whitney test was performed for one or more groups were not normally distributed. A t test with Welch’s correction was performed for unpaired comparison with non-equal variances. A one-way ANOVA was performed in comparisons for more than two groups with equal variance, and the Brown-Forsythe test was used with unequal variance. A one-way ANOVA followed by a Dunnett’s test was performed when two or more groups were compared with the same control. Two-way ANOVA followed by Tukey correction test was used for multiple comparisons with two or more variances. P value < 0.05 was considered statistically significant. Details of statistical analysis are included in the figure legends for each individual experiment.

Results

Human SV incubation model.

Human SV segments were obtained in patients undergoing CABG surgery as previously reported(16). The experimental workflow is depicted in Figure 1A. Veins were either processed immediately for investigation (Day 0) or maintained in a cell culture incubator ex vivo for 7 days (Day 7), after which they were processed for investigation using identical procedures. Histological examination (H&E staining) confirmed significant neointima formation in SV at Day 7 compared to Day 0 (Figure 1B, left panel). Minimal tissue growth or intimal thickening was observed when SVs were cultured under serum-free medium (not shown).

Figure 1. Human saphenous vein (SV) explants undergo neointima formation ex vivo.

A, Schematic of ex vivo model of neointima formation in human SV. B, Representative image of H&E (left) and VVG (right) staining in unincubated (Day 0) and incubated (Day 7) human SV. Scale bar: 500 μm (4x) and 200 μm (10x). C, Representative images of α-Smooth Muscle Actin (αSMA; red) staining in Day 0 and Day 7 SV. Scale bar: 200 μm (10x) and 20 μm (40x). D, Representative images of VE-cadherin (brown) staining in Day 0 and Day 7 SV. Scale bar: 200 μm (10x) and 20 μm (40x).E-G, Quantification of elastin fragmentation score (E, from VVG staining, n=8), αSMA positive area in the intima (F, n=3), and VE-cadherin staining along the luminal lining (G, n=3). Data are mean±SEM by paired t-test (n=3).I, intima; M, media, A adventitia.

Key aspects of neointima formation in the ex vivo incubation model.

The ex vivo model of neointima formation in human SV was further characterized histologically. We first performed VVG staining to detect elastin fibers in human SV. VVG staining demonstrated that the internal elastic lamina was fully intact at Day 0, whereas significant disruption and fragmentation was apparent at Day 7 (Figure 1B, right panel, quantified in 1E).

To assess VSMC accumulation in the intima, a key aspect of neointima formation, we performed α-SMA staining on Day 0 and Day 7 SV. As expected, SV at Day 0 demonstrated little α-SMA positive staining within the intima, while α-SMA staining was readily apparent in the proliferated neointima at Day 7 (Figure 1C, 1F). Additionally, VE-cadherin immunostaining demonstrated the consistent presence of endothelial cells (EC) along the luminal surface in both Day 0 and Day 7 SV, suggesting preserved EC integrity in this model (Figure 1D, quantified in 1G). To further characterize structural remodeling, we quantified wall geometry, ECM, and cellularity. Masson’s trichrome staining demonstrated a parallel increase in collagen-rich ECM, including greater trichrome-positive area within the intima (Supplemental Figure S1). Wall thickness and the wall-to-lumen ratio increased significantly from Day 0 to Day 7, whereas luminal area showed no significant change despite an increase in total vessel area, suggesting wall expansion rather than tissue shrinkage (Supplemental Figure S2). Quantitative nuclear counts showed higher total and intimal nuclei, with increased nuclei density in the intima (and more modestly in the media), while adventitial nuclear density remained relatively stable (Supplemental Figure S3). Together, these data support that ex vivo culture induces localized cellular accumulation and matrix deposition within the wall, consistent with early neointima formation.

Neointima quantification and separation by sex.

Next, we extended the ex vivo incubation model of neointima formation to include 72 paired SV samples from individual patients. Demographic and clinical data from the cohort are listed in Supplemental Table 3. Quantification of the neointima area and neointima-to-media ratio demonstrate statistically significant increases at Day 7 compared to Day 0 (Figure 2A). Paired before (Day 0)-and-after (Day 7) results from the same patients are shown in Figure 2B; while there was variability in the magnitude of the response, the majority of SV from individual patients exhibited neointima formation at Day 7.

Figure 2. Sex differences in neointima formation in relation to BMI in human SV.

A, Quantification of neointima area and neointima-to-media ratio of Day 0 and Day 7 human SV (paired). Data are full distribution of values with medians and interquartile ranges indicated by Wilcoxon signed-rank test (n=72). B, Before-after line graphs of neointima area and neointima-to-media ratio in SV at Day 0 and Day 7 (paired). Data are byWilcoxon signed-rank test (n=72). C,D Quantification of neointima area and neointima-to-media ratio of Day 0 (C) and Day 7 (D) human SV, separated by sex. Data are mean±SEM by Mann-Whitney test (Male: n=52; female: n=20). E, Quantification of total vessel area of Day 0 (left) and Day 7 (right) human SV, separated by sex. Data are mean±SEM by Mann-Whitney test (Male: n=52; female: n=20). F-H, Scatterplot of BMI in relation to the change of neointima area between Day 0 and Day 7 SV in the entire cohort (n=72) and in males (G, n=52) and females (H, n=20).

To investigate whether the variability in neointima formation could reflect differences in individual vein segments, as opposed to patient-specific differences, we performed experiments using four distinct segments of SV harvested from the same patients, with two segments examined at Day 0 and two at Day 7. Under these conditions, the neointima area was highly consistent in distinct sections of SV obtained from the same patients at both Day 0 and Day 7 (Figure S4A), with majority of the patient samples exhibiting < 15% variability (Supplemental Figure S4B&C). This suggests that the variability of neointima formation in the ex vivo model is most likely related to inherent differences in the patients from which the SV were harvested.

SV tend to be smaller in diameter in females than in males(37). Moreover, SV graft disease in females was reported to differ pathologically from that in males, with the former exhibiting a more prominent cellular fibrous tissue component(38). Thus, we analyzed the SV data separately by sex. As expected, far more specimens were collected from males (n=52) than females (n=20). Interestingly, at Day 0, SV from females had significantly less neointima compared to males, but this difference was diminished by Day 7 (Figure 2C, 2D). We also compared total vessel area from female versus male patients at Day 0 and Day 7. While no significant differences were noted at Day 0, total SV area was significantly smaller in women compared to men at Day 7 (Figure 2E).

Overweight and mildly obese individuals have been reported to exhibit improved intermediate term survival after CABG(39). Therefore, we separated the SV data by BMI. At both Day 0 and 7, we did not detect an association between SV neointima area and BMI (not shown). We then quantified the change in neointima between Day 0 and Day 7 (an indicator of neointima formation), which showed a trend towards a negative association in the entire cohort (Figure 2F). Notably, when we separated the BMI data by sex, SV from males exhibited a significant negative association between change in neointima and BMI, while SV from females trended towards a positive association, although the numbers were small and statistically non-significant. We also examined other clinical factors previously linked to graft failure, including diabetes mellitus and smoking(13,40), but did not observe significant associations with neointima formation in our cohort (Supplemental Figure S5&6). Together, these findings suggest that mechanisms of SV remodeling and neointima formation may differ according to sex and BMI.

Bulk RNA-seq in neointima formation in human SV.

Workflow for multi-omics experiment is depicted in Figure 3A. To examine the global transcriptomic changes occurring during neointima formation, bulk RNA-seq was first performed in patient-matched Day 0 and Day 7 SV. PCA plot depicted divergent and independent clustering of Day 0 and Day 7 SV, indicating distinct gene expression profiles between the two time points (Figure 3B).

Figure 3. Bulk RNA-seq in neointima formation in human SV.

A, Schematic of workflow for the multi-omic study. B, PCA plot from bulk RNA-sequencing depicting Day 0 (blue) and Day 7 (red) SV. C, Volcano plot comparing the differentially expressed genes between Day 0 and Day 7 SV. Each dot represents significantly increased (red) or decreased (blue) genes in Day 7 compared to Day 0 SV. D, Heatmap of selected genes which were highly differentially expressed between Day 0 and Day 7 SV. E-G, Top GO terms of all (E), upregulated (F) and downregulated (G) DE genes (Fold Change <0.5 OR >2, FDR < 0.01).

A total of 36,969 genes were detected, with 2,416 upregulated genes and 1,967 downregulated genes in Day 7 SV compared to Day 0 SV (Fold Change <0.5 OR >2, FDR < 0.01, Figure 3C). As expected, pro-inflammatory (TNF, NFKB1, NFKB2, IL1B, IL6) and proliferation-related (CCNB1, CDK1, CDKN1A, MCM2) genes were significantly upregulated in Day 7 compared to Day 0 SV. Additionally, smooth muscle differentiation-related genes (ITGA8, CNN1, ACTA2, TAGLN, LMOD1) and extracellular matrix forming genes (FBLN1, DCN, ELN, COL3A1) were significantly downregulated in Day 7 compared to Day 0 SV (Figure 3D). While endothelial monolayer remained intact in Day 0 and Day 7 SV, endothelial marker genes (VWF, PECAM1, ICAM1, EMCN) were significantly downregulated (not shown). To interrogate the lesser known molecular and cellular mechanisms associated with neointima formation, pathways and gene ontology (GO) enrichment analyses were performed. GO analysis of upregulated genes in Day 7 SV revealed enrichment of RNA-related processes, intrinsic apoptotic signaling, and protein ubiquitination (Figure 3E&F). In contrast, GO analysis of downregulated genes demonstrated enrichment of pathways related to actin modulation, cytoskeletal organization, and second messenger signaling (Figure 3E&G).Pathway associated with extracellular matrix degradation was not detected in GO analysis although some matrix metalloprotease-related genes were significantly increased (TIMP1, MMPs).

To identify the most differentially regulated GO processes, we constructed a gene-plot network, which provides a visual representation of the interactions and relationships among genes, aiding in the comprehensive analysis of complex biological systems. This analysis suggested increased nuclear transport and ribosome biogenesis in Day 7 SV (Supplemental Figure S7A). Taken together with the greater transcriptional diversity and abundance of mapped reads from Day 7 SV, these data suggest that higher levels of both transcription and translation occur at this time during neointima formation. Consistent with the GO pathway analysis, the gene network visualization illustrated downregulation of actin-cytoskeleton modulating genes, muscle contraction genes, and second messenger signaling genes at Day 7 (Supplemental Figure S7B).

Proteomic profiling of neointima formation in human SV.

Next, we performed unlabeled LC/MS proteomics on SV lysates. A total of 3380 proteins were detected, with 55 upregulated proteins and 74 downregulated proteins in Day 7 SV compared to Day 0 SV (Fold Change <0.5 OR >2, p-value < 0.05). Given that a limited number of proteins accounted for the majority of differential expressions, we applied a less stringent p-value threshold to prioritize biological relevance. Similar to the RNA seq data, PCA plot of the proteomics samples demonstrated distinct clustering of proteins in Day 0 versus Day 7 SV, although samples within the same group showed a wide degree variability (Figure 4A), as was likewise observed with respect to neointima formation (Figure 2B). Interestingly, the top upregulated proteins in Day 7 SV represented a mixture of both ECM-producing and ECM-degrading proteins (COL4A1, MMP2, FBN1, EMILIN1), while the top downregulated proteins represented proteins which regulate metabolic processes, including mitochondrial metabolism (IDH2, ACO2, HADHA, ALDH2) (Figure 4B).

Figure 4. Proteomic profiling of neointima formation.

A, PCA plot from untargeted LC/MS proteomics depicting the Day 0 (blue) and Day 7 (red) SV. B, Volcano plot comparing the differentially expressed proteins between Day 0 and Day 7 SV. C-E, Top GO terms of all (C), upregulated (D) and downregulated (E) DE proteins (Fold Change <0.5 OR >2, p-value < 0.05).

Protein Ontology (PRO) and protein-plot network analyses identified extracellular matrix organization and negative regulation of proteolysis/peptidase activity to be amongst the most significant terms, consistent with increased ECM-producing proteins in Day 7 SV (Figure 4C–E). Interestingly, other PRO terms included blood coagulation and hemostasis, suggesting that changes in protein expression occurring during the early phase of SV remodeling may promote thrombosis (Figure 4D). Top downregulated PRO analysis terms included small molecule catabolic process, tricarboxylic acid cycle, and fatty acid beta-oxidation, suggesting significant metabolic alterations during neointima formation (Figure 4E). Additional top upregulated (Supplemental Figure S8A) and downregulated (Supplemental Figure S8B) terms included decrease of B cell receptor signaling pathway, in particular IgG heavy chains (IGHA1, IGHG1, IGHG2, IGHG3, IGHG4). Using this unlabeled proteomics approach, expression of most VSMC contractile proteins was not significantly reduced, and expression of one contractile protein, MYH11, appeared be increased in Day 7 SV.

To interrogate common cellular and metabolic processes characterizing neointima formation in SV, we compared bulk RNA-seq and proteomics datasets. 416 differentially expressed genes were filtered out due to low expression or expression of multiple isoforms of the same protein. Fewer altered proteins were captured compared to the wide array of altered genes, partly due to the high VSMC-specific peptide representation and greater variability associated with the proteomics analysis. Out of the 125 differentially expressed proteins, 46 (36.8%) were also differentially expressed in their respective transcripts, indicating significant overlap. Only 5 gene transcripts were shared between upregulated proteins and genes (10% of total upregulated proteins), while 23 gene-transcripts were shared between downregulated proteins and genes (30% of total downregulated proteins) (Supplemental Figure S9A). Notable cellular processes identified from both bulk RNA-seq and proteomics datasets included wound healing, blood coagulation, collagen-containing extracellular matrix and muscle contractile fiber (Supplemental Figure S9B).

Spatial transcriptomic analysis in human SV.

Given the transcriptional and proteomic changes observed during neointima formation, we next applied spatial transcriptomics using 1^st generation Visium platform to evaluate regional gene expression patterns within the vessel wall in Day 0 and Day 7 SVs. Representative spatial capture images are shown in Figure 5A. Merged UMAP analysis of barcoded spots revealed eight transcriptomic clusters, with distinct spatial separation between Day 0 and Day 7 and limited intermixing of captured regions (Figure 5B). Five transcriptionally distinct putative clusters were observed in Day 0 SV, each exhibiting discrete spatial localization and gene expression patterns. Barcoded spots enriched for contractile VSMC-associated transcripts (e.g., MYH11, CNN1) were predominantly localized to the media, though they presented a minority of the transcriptomic signal, with fibroblast-like expression signatures (e.g., FBLN1, DCN) dominating both the media and adventitia at Day 0. By Day 7, VSMC contractile marker signals were largely absent, replaced by regionally distinct expression patterns consistent with synthetic and osteogenic-like phenotypes. Fibroblast-associated regions exhibited a shift from secretory to matrix-degrading signatures, including upregulation of MMP2, MMP14, and TIMP1. Endothelial marker expression was reduced over time, and no discrete endothelial-rich barcoded regions were observed at either time point (Figure 5B–D, Supplemental Figure S10A-D).

Figure 5. Spatial transcriptomics analysis of human SV during neointima formation.

A, Representative image of H&E (top) and 10X Visium spatial transcriptomics (bottom) in Day 0 and Day 7 SV. Scale bar: 500 μm. B, Uniform Manifold Approximation and Projection (UMAP) plot of barcoded spots identified by Seurat (top), and UMAP plot split by condition (Day 0 vs Day 7 SV, bottom). C, Percentage change of each cluster during neointima formation and the contribution of each cluster to Day 0 and Day 7 SV. D, Dotplot of top marker genes associated with each cluster for Day 0 and Day 7 SV. E, Gene Ontology (GO) enrichment for differentially expressed genes within each cluster.

Given the limited resolution of the Visium platform, we performed focused pseudotime trajectory analysis of putative VSMCs and fibroblasts to better resolve cell type–specific gene expression trends. Pseudotime trajectory analysis of the merged UMAP revealed a progression from a less transdifferentiated VSMC cluster (C0) to a more transdifferentiated VSMC cluster (C1), although no direct trajectory toward the HES1⁺ VSMC-enriched cluster (C4) was observed. Regarding putative fibroblasts, secretory populations (C5–6) demonstrated a transition toward clusters with elevated transcription of MMP2 and MMP14 (C7) (Supplemental Figure S11). GO enrichment analysis revealed that fibroblast- and VSMC-like clusters were associated with coagulation cascade, glycolysis and oxidative phosphorylation pathways (Figure 5E). These consistent findings across platforms support spatially organized transcriptional shifts amongst VSMC and fibroblast populations during neointima formation.

Differential gene expression analysis across SV layers demonstrates coordinated cellular processes during neointima formation

Having detected dynamic changes in transcripts within specific layers of the blood vessel wall during neointima formation (Figure 5A, Supplemental Figure S10), we next compared gene expression profiles across the intima, media and adventitia. After manual annotation of the barcoded spots illustrated in Figure 6A, 105 genes were found to be differentially expressed across various layers at Day 7 compared to Day 0. Interestingly, 36 genes were differentially expressed across all three layers, while 62 DE genes were shared between the intima and the media, suggesting that many cellular processes occur in tandem in the intima and the media (Figure 6B). Consistently downregulated genes across all layers at Day 7 compared to Day 0 included MYH11, CNN1, ACTA2, MYL9, DCN, and TALGN. Additionally, top GO terms shared across all layers included smooth muscle contraction (Figure 6C), which further highlights the ubiquitous importance of VSMC phenotypic changes during neointima formation. Consistently upregulated genes across all layers included SOD2, GADD45B, PPP1R15A, IL6, CXCL8, TIMP1, and MMP2 (Figure 6C). While cellular senescence, DNA damage response and response to reactive oxygen species were highly represented in all three layers, processes of collagen degradation and integrin cell surface interactions were represented in intima and the media, but not the adventitia (Figure 6D, Supplemental Figure S12). KEGG analysis on the intima at Day 7 identified significantly upregulated cellular senescence, AGE-RACE signaling, transcriptional regulation in cancer, and IL-17 signaling pathways (Figure 6E), while significantly downregulated processes included focal adhesion, motor proteins, vascular smooth muscle contractions and proteoglycans in cancer (Figure 6F).

Figure 6. Comparative gene expression in intima, media and adventitia layers during neointima formation in human SV.

A, Spatial dimensional plots projected onto the H&E images of Day 0 (left) and Day 7 (right) SV, and manually annotated into the respective SV layers. Scale bar: 500 μm. B, Venn diagram showing the overlap between the DEGs of the three layers. C, Merged volcano plot of significantly DEGs of Day 7 and Day 0 SV, in which each annotated layer of the Day 7 SV was compared to that of Day 0 SV. D, Reactome pathways of the layers of Day 7 SV compared to those of Day 0 SV. KEGG pathway of the intimal layer of Day 7 and Day 0 SV, based on the upregulated (E) and downregulated (F) genes at Day 7 SV.

Global VSMC transcriptional changes during neointima formation in human SV

To assess global VSMC transcriptional changes during neointima formation in SV, we first examined VSMC contractile genes through spatial transcriptomics (Figure 7A) and present the results in violin plots (Figure 7B). These findings confirm reduced VSMC contractile gene expression at the whole tissue level. Several VSMC-enriched contractile proteins were not significantly altered in the untargeted proteomics dataset (Figure 4B), likely due to their high abundance and limited unique tryptic peptide representation, which reduces detectability by LC-MS/MS despite robust transcript-level decreases. To address this limitation, we performed targeted proteomics analysis of VSMC contractile proteins, including calponin 1 and transgelin, which demonstrated that the area under the curve was significantly less in Day 7 SV than in Day 0 SV, while β-actin levels were similar (Figure 7C). qPCR and Western blot confirmed significant decrease in these VSMC contractile genes (Figure 7D) proteins (Figure 7E&F) at Day 7. Interestingly, spatial transcriptomics data also showed that extracellular matrix-forming genes (e.g., FBLN1, DCN, and COL3A1) were downregulated (Figure 8A&C), while extracellular matrix-degrading genes (e.g., MMP2, MMP14) were upregulated (Figure 8B&D), suggesting transcriptional changes in gene expression and phenotype switching in fibroblasts that potentially facilitate neointima formation. Immunostaining of αSMA did not co-localize with PDGFRB, a fibroblast marker highly activated during fibrosis and extracellular matrix deposition, in SV at Day 0, suggesting that the medial layer is primarily composed of both VSMCs and fibroblasts (Figure 9A). Both αSMA and PDGFRB staining positivity were lower in Day 7 SV compared to Day 0 SV (Figure 9B). Immunofluorescence data showed predominant adventitial staining of MMP2 at Day 0, which was further increased at Day 3 and Day 7. (Figure 9C–E). MMP2 staining was also increased in the media and intima significantly increased by Day 3 and Day 7, respectively; however, whether this reflects local upregulation or migration of pre-existing MMP2⁺/MMP14⁺ cells cannot be determined from our current data.

Figure 7. VSMC in human SV undergo de-differentiation during neointima formation.

A, Visium spatial transcriptomics feature plot of VSMC contractile markers (ACTA2, CNN1, MYH11, TAGLN) in Day 0 (upper) and Day 7 (lower) SV. Scale bar: 500 μm. B, Violin plot of VSMC contractile genes (ACTA2, CNN1, MYH11, TAGLN) in Day 0 and Day 7 SV (n=3). C, Targeted proteomics analysis of VSMC contractile proteins (Calponin 1, Transgelin) and β-actin in Day 0 and Day 7 SV. Upper and lower panels represent distinct peptide transitions used for quantification of each target protein. Data are mean±SEM by Student’s t-test (n=3). D, qRT-PCR of representative VSMC contractile genes (ACTA2, CNN1, TAGLN) in RNA from Day 0 and Day 7 SV. Data are mean±SEM by Student’s t-test (n=7). E, Western blot analysis of LMOD1, CNN1, and αSMA in Day 0 and Day 7 SV. F, Mean densitometric analysis for LMOD1, CNN1, and αSMA normalized to the total protein. Data are mean±SEM by Student’s t-test (n=3).

Figure 8. Transcriptional changes in genes regulating the extracellular matrix (ECM), and analysis of key VSMC, fibroblast and extracellular matrix markers during neointima formation in human SV.

E, Representative immunofluorescence images of ±SMA and PDGFRB in Day 0 (left) and Day 7 (right) SV. Scale bar: 20 μm. F, Quantification of immunofluorescence images of aSMA (left) and PDGFRB (right) in whole vessels. Data are meanαSEM by two-tailed unpaired t test (n=4–8). G, Representative immunofluorescence images of MMP2 in Day 0 (left), Day 3 (middle), and Day 7 (right) SV separated by intima, media and adventitia. Scale bar: 100 μm. H, Representative immunofluorescence inset images of MMP2 in Day 0 (left), Day 3 (middle), and Day 7 (right) SV. Scale bar: 10 μm. I, Quantification of immunofluorescence images, separated into whole vessel, intima, media and adventitia. Data are mean±SEM by One-way ANOVA followed by Dunnett’s test (n=3-4).

Figure 9. MIR647 is downregulated during neointima formation and regulates VSMC contractility genes and proliferation.

Top upregulated (A) and downregulated (B) miRNAs in Day 7 SV compared to Day 0 SV. C, RT-qPCR of miR647 in Day 0 and Day 7 SV. Data are mean±SEM by two-tailed paired t test (n=4). D-E, RT-qPCR of miR647 after treatment of TGFβ (10 ng/ml), TNFα (10 ng/ml) or PDGF-BB (30 ng/ml) for 24 hours in HSVSMCs compared to their respective vehicle control. Data are meanαSEM by two­way ANOVA with Tukey’s correction (n=3) or two-tailed paired I test (n=3). F, RT-qPCR of miR647 after treatment of control or miR647 mimic (25 μM). Data are mean±SEM by two-tailed unpaired t test (n=4). G, Manual cell counting of HSVSMCs transfected with control or miR647 mimic under basal and PDGF-BB treated conditions. Data are meanαSEM by two-tailed unpaired t test (n=4). H, RT-qPCR of VSMC contractility markers CNN1, LMOD1, MYOCD, TAGLN and ITGAB in HSVSMCs transfected with control or miR647 mimic. Data are mean±SEM by two-tailed unpaired I test (n=4). I, RT-qPCR quantification of miR647 after siRNA-mediated knockdown in HSVSMCs. Data are mean±SEM and analyzed by two-tailed unpaired t test (n=4). J, Manual cell counts of HSVSMCs transfected with scramble control or siMIR647. Data are mean±SEM by two-tailed unpaired t test (n=4). K, RT-qPCR of VSMC contractility-associaled genes ACTA2, LMOD1, and TAGLN in HSVSMCs transfected with scramble control or siMIR647. Data are mean±SEM and analyzed by two-tailed unpaired t test (n=4). Statistical significance was denoted as follows: ^*,# for p < 0.05, **, ^## for p < 0.01, and ***, ^### for p < 0.001.

MIR647 is downregulated during neointima formation and regulates VSMC differentiation genes and proliferation

To explore novel mechanisms of neointima formation, we focused on the top DE microRNAs (miRNAs) detected between Day 0 and Day 7 of SV incubation (Figure 10A&B). While many miRNAs identified were novel with no known annotations, several miRNAs previously associated with VSMC function were also detected, including MIR126 and MIR145. Excluding MIR145, MIR647 was the most significantly downregulated miRNA in Day 7 SV compared to Day 0 SV, and we validated its expression via qPCR (Figure 10C). In human saphenous vein (HSV) SMCs, MIR647 expression significantly increased following treatment with TGFβ (10 ng/mL) compared to vehicle control, whereas treatments with TNFα (10 ng/mL) or PDGF-BB (30 ng/mL) did not significantly affect MIR647 expression (Figure 10D). Transfection with MIR647 mimic significantly increased the expression of MIR647 by more than two-fold in HSVSMCs (Figure 10E). Overexpression of MIR647 significantly altered cell morphology, with increased spindle-shaped cells compared to the negative control mimic (not shown), in conjunction with reduced cell numbers (~35%) under both basal and PDGF-BB-stimulated growth conditions (Figure 10F). Additionally, MIR647 overexpression led to a significant increase in the expression of VSMC differentiation markers, including CNN1, LMOD1, MYOCD, TAGLN, and ITGA8, suggesting that MIR647 regulates VSMC phenotype proliferation (Figure 10G&H). To directly test whether MIR647 has a distinct function in venous SMCs compared with prior reports in aortic SMCs, we performed complementary loss-of-function studies in primary human saphenous vein SMCs. siRNA-mediated knockdown of MIR647 significantly accelerated cell growth under basal conditions (Figure 10J) and reduced expression of key contractile markers, including ACTA2, and TAGLN (Figure 10K). These findings contrast with the modest antiproliferative effects observed in MIR647-overexpressing cells (Figure 10G) and indicate that MIR647 supports maintenance of the contractile phenotype in HSVSMCs. Together, the gain- and loss-of-function data demonstrate that MIR647 exerts a context-dependent regulatory role in venous SMC differentiation and growth.

Figure 10. MIR647 is downregulated during neointima formation and regulates VSMC contractility genes and proliferation.

Top upregulated (A) and downregulated (B) miRNAs in Day 7 SV compared to Day 0 SV. C, RT-qPCR of miR647 in Day 0 and Day 7 SV. Data are mean±SEM by two-tailed paired t test (n=4). D-E, RT-qPCR of miR647 after treatment of TGFβ (10 ng/mL), TNFα (10 ng/mL) or PDGF-BB (30 ng/mL) for 24 hours in HSVSMCs compared to their respective vehicle control. Data are mean±SEM by two-way ANOVA with Tukey’s correction (n=3) or two-tailed paired t test (n=3). F, RT-qPCR of miR647 after treatment of control or miR647 mimic (25 μM). Data are mean±SEM by two-tailed unpaired t test (n=4). G, Manual cell counting of HSVSMCs transfected with control or miR647 mimic under basal and PDGF-BB treated conditions. Data are mean±SEM by two-tailed unpaired t test (n=4). H, RT-qPCR of VSMC contractility markers CNN1, LMOD1, MYOCD, TAGLN and ITGA8 in HSVSMCs transfected with control or miR647 mimic. Data are mean±SEM by two-tailed unpaired t test (n=4). I, RT-qPCR quantification of miR647 after siRNA-mediated knockdown in HSVSMCs. Data are mean±SEM and analyzed by two-tailed unpaired t test (n=4). J, Manual cell counts of HSVSMCs transfected with scramble control or siMIR647. Data are mean±SEM by two-tailed unpaired t test (n=4). K, RT-qPCR of VSMC contractility-associated genes ACTA2, LMOD1, and TAGLN in HSVSMCs transfected with scramble control or siMIR647. Data are mean±SEM and analyzed by two-tailed unpaired t test (n=4). Statistical significance was denoted as follows: *, ^# for p < 0.05, **, ^## for p < 0.01, and ***, ^### for p < 0.001.

Discussion

Saphenous veins remain the most commonly used conduit for CABG surgery, yet 50% of vein grafts are reported to fail within 10 years of implantation(2,3). While early vein graft failure is typically associated with technical factors and/or thrombosis, most SV fail over time due to progressive neointima formation(41–43). The process of neointima formation in SV is thought to be initiated by injury to the vessel wall sustained during surgical harvesting, pressurization and preparation for implantation(44). Consequently, when SV segments harvested for CABG grafting are maintained in tissue culture, they form a VSMC-rich neointima. Here, we performed a detailed characterization of neointima formation in human SV, and we employed multi-omics to investigate cellular dynamics and molecular mechanisms therein. These findings may provide novel insight into the pathogenesis of SV graft disease and could help to illuminate two poorly understood disparities (sex and obesity) in patients undergoing CABG surgery.

A total of 72 paired samples from individual patients were studied on the day of harvesting (Day 0) or after 7 days (Day 7) in tissue culture to elicit neointima formation. In general, neointima formation in this model was characterized by pronounced accumulation of VSMC and degradation of elastin fibers, while the intact endothelial monolayer became dysfunctional. However, we detected considerable variability in the extent of neointima present in the SV in this study, both at Day 0 and Day 7. Not all veins developed comparable intimal thickening during culture, and variable responses were observed across donors. All viable specimens were included in the analyses, with exclusion limited to samples showing non-viability on initial staining or gross morphological damage at harvest. This variability is illustrated by paired analyses of each vein before and after culture, highlighting donor-specific differences while confirming an overall increase in intimal thickness across the cohort. Interestingly, separate SV segments obtained from the same patients displayed minimum differences in neointima formation at Day 0 and Day 7, suggesting that the variable results are not due to technical or experimental artifacts, but instead reflect patient-specific differences.

We further investigated clinical factors that may contribute to the variability in neointima formation. Our findings indicate that the variability might be linked, at least in part, to sex-related differences. We observed that SV from males exhibited a larger intima and intima/media ratio than females at Day 0. This sex difference in neointima size was less apparent at Day 7, however. Additionally, the total vessel area was similar in men and women at Day 0 but significantly smaller in women at Day 7. It should be noted that women were under-represented in this study, as has been the case with other studies involving CABG surgery, in which women typically represent 20–30% of the CABG population(45). Nevertheless, these findings raise the possibility that early responses to injury and adaptive remodeling of SV grafts might differ in women compared with men, possibly reflecting hormonal, immunologic, or adipose-mediated mechanisms(46–48). Women have been reported to experience a higher rate of myocardial infarction and re-vascularization after CABG surgery(49); whether our findings might contribute to such sex-dependent differences in outcomes after CABG surgery remains to be determined.

Traditional atherosclerosis risk factors, including smoking, hypertension, diabetes and hyperlipidemia, have been associated with SV graft failure in humans(50). However, factors which may protect against SV graft failure have not been fully elucidated. Neointima formation was not significantly associated with smoking status, left ventricular ejection fraction, or diabetes in our cohort, due to limited sample size and confounders. Ongoing studies with larger sample sizes aim to clarify these associations. However, we observed a trend towards a negative association between BMI and neointima formation in our cohort. Breaking down the data by sex, we found that the association was significant in men, but not in women, in whom the data trended in the opposite direction (i.e., towards a positive association between BMI and neointima formation). Aside from the fact that this may represent another sex difference in SV remodeling, our findings may help to explain an apparent “obesity paradox” that has been reported in studies of CABG, wherein overweight to moderately obese individuals tend to exhibit better outcomes and intermediate-term survival rates compared to their leaner counterparts(14,51). This benefit appears to be lost over time, likely due to the adverse health burden imposed by obesity and its associated risk factors (i.e., hypertension, hyperlipidemia, insulin resistance, etc.)(52). It should also be noted that morbid obesity is an independent risk factor for adverse outcomes following CABG; however, in this study, none of our patients exhibited a BMI of > 40 and thus were not classified as morbidly obese. While few longitudinal studies have directly reported bypass graft patency data, a recent meta-analysis by An et al reported that patients with BMI between 25 and 40 were significantly less likely to experience graft failure compared to the patients with normal BMI(53). Although the mechanisms whereby obesity impacts graft patency remain unclear, our findings suggest that differences in neointima formation could be contributory.

In this study, we performed a multi-omics analysis of human SV to probe the molecular and cellular mechanisms involved in neointima formation. We show that the process of neointima formation is highly dynamic, with ~5000 genes exhibiting changes in expression at Day 7 as compared to Day 0 (FDR < 0.05). Bulk RNA-seq analysis demonstrated a number of upregulated genes and pathways during neointima formation, many of which (i.e., inflammation, cell proliferation) have been reported in other model systems(54,55). In addition, we detected downregulation of genes associated with VSMC differentiation and EC homeostasis, which has also been reported previously(56). Gene ontology analysis demonstrated that during neointima formation, pathways associated with intrinsic apoptosis and response to DNA damage stimulus are highly upregulated. These processes occur in tandem with increased transcription and nucleocytoplasmic transport, with high representation of ribonucleoprotein biogenesis and non-coding RNA processes. These findings are indicative of vascular cell stress responses, including DNA damage responses, which may have their origins in the trauma associated with surgical harvesting and could impact the processes of adaptation and neointima formation. Gene ontology analysis of the downregulated genes was almost exclusively associated with actin-modulation and muscle contraction processes, which reaffirm the importance of VSMC phenotypic changes that occur during neointima formation. These findings suggest that while our model recapitulates certain aspects of in vivo animal models, it generates novel insight into the complex molecular and cellular mechanisms which underpin neointima formation in SV.

Our data from proteomics, transcriptomics and histological analysis indicates dynamic changes in expressed proteins, with large alterations between the Day 0 and Day 7 SV. This modest transcriptome-to-proteome overlap is consistent with the lower sensitivity and narrower dynamic range of unbiased LC–MS/MS compared with RNA sequencing. Actin-modulating and contractile proteins were the most highly represented in the dataset, yet untargeted proteomic analysis did not show significant changes between Day 0 and Day 7 SV. Further validation through targeted proteomics of the contractile proteins demonstrated significant downregulation, yet not as dramatic as the changes in mRNA expression. This finding implies temporal divergence between the transcriptomic and proteomic signatures during the progression of neointima formation, from its initial to later phases. Common altered pathways included blood coagulation and thrombosis, extracellular matrix deposition processes, and VSMC contractile processes. Blood coagulation and thrombosis, which are believed to contribute to the early phase of vein graft failure(7–9), were the top processes represented in the protein oncology analysis. Extracellular matrix-producing proteins were also significantly increased at Day 7. While we did not see a strong representation of cytokines or other inflammatory signaling proteins, we detected significant alterations of proteins which regulate mitochondrial metabolism, which has been reported to occur in the presence of cellular stress responses and DNA damage(57,58). These alterations suggest that the vascular cells are driven towards glycolysis from oxidative phosphorylation, with increased mitochondrial fragmentation and impairment, which are hallmarks of mitochondrial dysfunction. Given the high transcriptomic and proteomic signatures of VSMC, this suggests that perturbations in mitochondrial metabolism may accompany VSMC phenotypic changes during neointima formation in SV. We note that our proteomics data indicate alterations in the expression levels of proteins associated with mitochondrial metabolism. However, direct functional assessments, such as measurements of metabolic flux or mitochondrial dynamics, would be necessary to define metabolic reprogramming or dysfunction in mitochondria.

Spatial transcriptomics enables the inference of regionally enriched gene expression within intact tissue architecture, providing insight into layer-specific transcriptional trends rather than precise single-cell resolution. Our spatial transcriptomic data show that, Day 7 SV exhibited many more clusters which had significantly altered gene expression profiles while Day 0 SV exhibited few clusters with similar gene expression. The putative cellular composition of the SV at both Day 0 and Day 7 was predicted to be a mixture of fibroblasts and VSMC, without representation of EC or pericytes. Day 0 SV were primarily composed of signatures resembling the mixture of healthy, contractile and transdifferentating VSMC and secretory fibroblasts. Surprisingly, more signatures representing fibroblasts than VSMC were detected in the media at Day 0. Additionally, the intima of the Day 0 SV was exclusively composed of transdifferentiating VSMC-like signatures, which suggests that initial processes of neointima formation, including VSMC fate switching and migration, are already active at the time of SV implantation. While the intima was still composed of mostly fibroblasts- and VSMC-like signatures in Day 7 SV, the gene expression profiles shared almost no overlap with Day 0. Although EC and Endothelial-to-Mesenchymal transition (EndoMT) markers were localized in the intima of SV and differentially expressed between Day 0 and Day 7 SV, these genes were expressed at very low levels in all samples.

The gene signatures which indicated high secretory activity in putative fibroblasts were significantly elevated at Day 0, with high expression of ELN, DCN, COL3A1, compared to Day 7. Concomitantly, PDGFRβ staining, indicative of secretory fibroblasts, was significantly higher at Day 0 compared to Day 7. Notably, gene signatures responsible for extracellular matrix degradation, including MMP2, MMP14 and TIMP1, were significantly increased at Day 7. The vascular cells which gain the expression of these genes at Day 7 were localized in the outer media and the neointima and may contribute to the significant fragmentation of internal elastic lamina detected at Day 7.

Neointima is dominated by highly dysregulated VSMC which appear to have transdifferentiated into many fates, including previously reported synthetic, PCNA+ VSMC. We also report novel signatures within these presumably dysregulated VSMC which primarily reside within the neointima, including retinoid-induced genes such as AKAP12 consistent with osteogenic signature, and HES1⁺, a Notch-dependent transcription factor known to regulate VSMC phenotype switching and vascular remodeling(59,60). While the HES1⁺ VSMC-enriched cluster exhibited reduced contractile marker expression, it did not display strong enrichment for canonical synthetic or inflammatory markers. It remains unclear whether these HES1⁺ cells represent a distinct VSMC state or a transitional phenotype, and whether they play a functional role in neointima formation in SV. Further in-depth study is required to determine the mechanism of HES1 in human veins. Recent work examining failed vein grafts in a canine model also highlights the importance of injury in driving many processes, including cell proliferation and migration of fibroblasts and VSMC, which in turn triggers transdifferentiation of these maladaptive cell populations(61). These findings highlight the important conserved pathways regulating vein graft failure and may be important to identify critical targets for early intervention to minimize vein graft disease. It is important to note that there was a possibility of misidentifying true VSMC subpopulations from our omics data. Furthermore, other potential populations such as myofibroblasts may upregulate VSMC markers expression. Phenotypic modulation of VSMCs often involves transition toward myofibroblast-like states, with loss of contractile markers (MYH11, TAGLN) and gain of matrix-remodeling genes (COL1A1, MMP2). Because VSMCs and myofibroblasts exist along a continuum rather than as distinct lineages, our spatial transcriptomic findings showing co-expression of VSMC and fibroblast markers likely reflect this transdifferentiation process. Therefore, it is required to acknowledge this plasticity and the limitations of transcriptomic inference in future studies.

We also identified differentially expressed miRNAs during neointima formation in human SV. The most highly downregulated miRNA was MIR145, which has been shown to directly modulate neointima formation and VSMC contractility through its target genes KLF5 and myocardin(62). Notably, MIR647 miR647 has been reported to affect the contractility and proliferation of gastric cancer cells [consistent with our novel findings in HSVSMCs (Figure 10)] through interaction with SRF(63). Our data in HSVSMCs suggest that a similar interaction between MIR647 and SRF may occur in HSVSMCs, although this will require further investigation. Conversely, it was also reported that MIR647 promotes proliferation and migration of aortic VSMC, suggesting the distinct mechanisms of action of MIR647 in different VSMC populations between aorta and vein, which warrants further investigation (64).

While this study presents a comprehensive multi-omics analyses of the molecular and cellular mechanisms driving neointima formation in human saphenous veins, several notable limitations should be acknowledged. First, our ex vivo human SV culture model does not incorporate hemodynamic parameters such as pressure or flow, and therefore cannot fully recapitulate the physiological environment of human vein graft disease. Given the small size of the vein pieces, we were unable to apply pressure and flow, which are important determinants of both adaptation and neointima formation(65,66). The abrupt changes in pressure and flow when SV are grafted into the arterial circulation may induce early endothelial responses, including endothelial dysfunction and the subsequent endothelial-to-mesenchymal transition, both of which are believed to be key drivers in the progression of neointima formation(67,68). Second, our system is devoid of humoral and cellular components of the immune system, which are important factors involved in SV graft disease. Third, the progression of neointima formation is accelerated in our model compared to in vivo, likely due to the high concentration of FBS in the culture media. Fourth, the cohort size was insufficient to permit a complete characterization of neointima formation and its association with clinical variables, and we only studied two points in time (Day 0 and Day 7), which may provide an incomplete understanding of the cellular and molecular dynamics of neointima formation. Time-lapse imaging of intact human veins, which could help resolve dynamic processes such as migration versus proliferation, is technically infeasible due to tissue thickness and long-term culture constraints; however, our paired Day 0-Day 7 design partially mitigates inter-sample variability by allowing each vein to serve as its own control. Additionally, while the findings with regard to sex and BMI were statistically significant, caution must be exercised with regard to clinical relevance given the model limitations and moderate numbers of patients included in this study. Fifth, the SV segments were collected using endoscopic vein harvesting technique, and the processes underlying neointima formation may differ in veins harvested using the no-touch technique, which is believed to lead to significantly less vein graft occlusion(69). However, the no-touch technique is associated with increased risk of wound infection and scarring, and it has not been widely adopted by cardiac surgeons worldwide despite being recommended in the European Revascularization Guidelines (Class of Recommendation IIa, Level of Evidence B)(69,70). Sixth, weak representation of EC and pericyte signatures was observed in all of the sequencing studies, including spatial transcriptomics. This likely reflects the injury to the EC during surgical preparation of the vein grafts(71,72), along with the limited resolution inherent to the 1^st generation of spatial transcriptomics technology, which prevents the definitive identification of rare cell populations. Additionally, our approach relied on independent single-cell RNA-seq datasets for cell type annotation, which may have influenced the detection sensitivity for certain cell populations. In particular, spatial transcriptomic annotation relied on human arterial single-cell RNA-seq reference datasets rather than venous references(73), as available human saphenous vein datasets are enriched for endothelial cells and do not capture the VSMC phenotypic states associated with vein graft remodeling. Future studies employing complementary approaches and higher-resolution spatial platforms may help resolve these limitations. Finally, we acknowledge that our multi-omics data did not fully leverage the potential of cross-platform analyses. There was a low degree of overlap (10–30%) between proteomics and transcriptomics data since a small number of proteins were detected by LC-MS/MS compared to genes by RNA-seq. In addition, unbiased LC–MS/MS failed to quantify several key SMC markers; however, targeted proteomics using 2–3 unique peptides per protein confirmed their downregulation, in accordance with the results from Western blot. Future studies incorporating methods such as correlation analyses and network-based approaches (e.g., weighted gene co-expression network analysis) to compare mRNA and miRNA with protein expression profiles, and statistical approaches such as regression analysis on unpaired observations(74), are required to effectively capture cross-modal relationships for the integration of multi-omics data.

In conclusion, we have interrogated the process of neointima formation in human SV using multi-omics analyses. We detected significant sex differences in both the neointima and vessel area, as well as a negative association between BMI and neointima formation that was observed only in men. Our molecular and proteomic analyses point towards induction of vascular cell stress and metabolic responses that may be initiated by trauma associated with surgical procurement, as well as increases in coagulation and thrombosis that could contribute to early graft occlusion. Finally, using spatial transcriptomics, we have detected novel signatures within regions enriched in VSMCs and fibroblasts that are dynamically activated and may contribute to the ECM degradation and neointima formation characteristic of this model.

Supplementary Material

Supplemental Material

Supplemental Figures S1-S12, Supplemental Tables 1–6, and proteomics dataset: https://doi.org/10.6084/m9.figshare.30895781

Abbreviations

BMI

body mass index

CABG

coronary artery bypass grafting

DEG

differentially expressed gene

ECM

extracellular matrix

GO

gene ontology

miR

microRNA

SMA

smooth muscle actin

SV

saphenous vein

VSMC

vascular smooth muscle cells