Spatial Gene Expression of Human Coronary Arteries Revealed the Molecular Features of Diffuse Intimal Thickening in Explanted Hearts

1. Introduction

The intimal layer of aortas in commonly used mouse models of atherosclerosis contains endothelial cells lying on top of the internal elastic lamina. In contrast, humans exhibit diffuse intimal thickening (DIT) before atherosclerosis develops, especially in atherosclerosis-prone arteries such as the coronary arteries. In DIT, the intimal layer thickens with enriched smooth muscle cells (SMCs) and extracellular matrix [1]. DIT is universal in both sexes and progression is seen with increasing age [2]. DIT were observed as early as 36 weeks of gestation and these regions are prone to atherosclerosis development in later life [3]. The American Heart Association defines DIT as a normal preclinical stage with well-organized intimal SMCs in the absence of lipid deposition or robust infiltration of immune cells [4].

Why the intimal layer of DIT regions is particularly prone to atherogenesis remains largely unknown. The intimal and medial layers of DIT regions are both enriched with SMCs, which are a major source of proteoglycans and collagens, forming the extracellular matrix in the vessel wall. One of these proteoglycans enriched in the intima of DIT, biglycan, has a high binding capacity for low-density lipoprotein [5], which could promote the retention of extracellular lipids and SMC-foam cells during early atherogenesis [6]. Intimal SMCs can undergo phenotypic changes and adopt various phenotypes in mouse and human atherosclerotic lesions [7]. Human SMCs in the intima of DIT express fewer myosin heavy chains than the fully differentiated contractile SMCs in the medial layer, suggesting a dedifferentiated phenotype [1]. Our previous studies using single cell RNA-sequencing have found a unique dedifferentiated SMC population in mouse atherosclerotic lesions, which expresses stem cell marker Sca1 [8,9]. Pseudotime trajectory analysis suggested that Sca1⁺ SMCs are in transition from contractile to proinflammatory phenotypes [8,10], playing a key role in the onset of atherosclerosis in mouse models. It is tempting to assume that similar stem-like SMCs exist in the intima of DIT in humans, priming these regions for atherogenesis upon exposure to risk factors such as cholesterol and inflammation. However, the Sca1 gene is not present in humans. Therefore, a thorough molecular profile of SMCs in the DIT may reveal their earliest phenotypic changes and risk factors that are essential for atherogenesis.

Knowledge of the molecular features of DIT is limited, as previous studies relied on microscopic observation and immunostaining of coronary arteries from autopsies. Determining the molecular features in a specific region requires high-throughput RNA sequencing in the spatial context. Spatial transcriptomics of coronary arteries would allow for the gene expression of intimal and medial layers to be observed separately, which is especially useful to compare the molecular features between intimal and medial SMCs in DIT. However, only one successful application in human coronary arteries was reported so far [11], suggesting that spatial transcriptomics is still at its infant stage in answering clinically relevant questions of coronary artery disease [12]. Of note, human coronary arteries from autopsies are easily accessible compared to those collected from freshly explanted hearts. RNA degradation in autopsy samples may introduce bias in the analysis and further complicate data interpretation. Before more studies pursue spatial transcriptomics to understand the etiology of coronary artery disease, the potential pitfall of using autopsy samples needs to be investigated.

In the present study, we applied the spatial gene expression technology Visium to obtain whole transcriptomic profiles of the intimal and medial layers of coronary arteries from formalin-fixed paraffin-embedded (FFPE) tissue sections. Visium uses DV200 (i.e., the percentage of fragments >200 nucleotides) to select FFPE samples of high RNA integrity and the minimum requirement is 30% [13]. Data quality between autopsies and explanted hearts was compared. Freshly isolated coronary arteries from heart transplant patients were selected to reveal the molecular features of thickened intima, which cannot be studied using commonly used mouse models or single-cell RNA sequencing of dissociated blood vessels.

2. Materials and Methods

Human Samples

Human coronary lesions were obtained from autopsies or explanted hearts donated by heart transplant patients. Using the modified American Heart Association classification, a pathologist assessed coronary arteries after Movat pentachrome staining. Regions of DIT or pathologic intimal thickening (PIT) were selected [14] from heart transplant patients (n=7) and autopsies (n=7).

Immunofluorescence Staining and Confocal Microscopy

To distinguish DIT and PIT, we stained the FFPE tissue sections with SMC lineage marker myosin heavy chain 11 (MYH11, Abcam, CAT#ab82541) and pan-leukocyte marker CD45 (Biolegend, CAT#368502). Briefly, 5 µm FFPE tissue sections were dewaxed, rehydrated, and boiled in 10 mM sodium citrate, 0.05% Tween buffer (pH 6.0) for 30 min using an autoclave for antigen retrieval. After blocking with 10% goat serum for 60 minutes, sections were incubated with the primary antibodies at 4˚C overnight, washed and then incubated with Alexa Fluor secondary antibodies (Invitrogen CAT#A11008 and #A21236) at room temperature for 1 h, and imaged under a Zeiss LSM880 inverted confocal microscope.

RNA Extraction and Quality Tests

Eight successive 10 µm thick FFPE sections were collected into RNase-free tubes immediately after sectioning and RNA was extracted following the manufacturer's protocol (Qiagen, CAT#73504). Total RNA was condensed by lyophilization. RNA concentration was measured by a Qubit Fluorometer to calculate the yield of RNA from each tissue section. RNA quality was determined by DV200 using an RNA 6000 Pico kit (Agilent, CAT# 5067-1513) on a Bioanalyzer. DIT samples with a DV200 ≥30% were sectioned, mounted on Visium SD slides (10xGenomics), and stained by H&E. Library preparation were performed according to the manufacturer’s protocol and sequenced at a minimum of 25,000 reads per capture spot.

Spatial Gene Expression Analysis

Visium data was processed using Seurat package (v5.0.1) [15]. To assess the RNA quality of each sample, [16] Harmony[17] was used to correct for batch effects. Unsupervised clustering was performed using Harmony-corrected components using the Louvain algorithm. Clustered capture spots were visualized with the Uniform Manifold Approximation and Projections (UMAP) algorithm, using the first 30 harmony-corrected components. To view the distribution of clusters in their spatial context, clustered capture spots were overlaid on top of H&E images for each tissue section. For differential gene expression analysis, specific capture spots within our regions of interest were selected using location tags, collected interactively using the SpatialDimPlot function from the Seurat R-library

Differential Gene Expression Analysis

We identified capture spots from the intimal and medial layers of the blood vessels (from the apical side to the external elastic lamina) by overlaying the Visium capture spots on the corresponding H&E image. We first fit a linear mixed model with LIMMA-VOOM (v3.50.3) [18] to determine differentially expressed genes (DEGs) of the coronary arteries (intimal and medial layers combined) between autopsy and explanted heart samples using pseudo-bulk gene-counts. Then, for coronary arteries from the explanted heart samples only, we used LIMMA-VOOM to determine DEGs between the intima (below the endothelium to the internal elastic lamina) and media (below the internal elastic lamina to the external elastic lamina) adjusting for patient-specific random effects. A false discovery rate of 0.01 was used to correct for multiple testing. DEGs were ranked by log₂-fold change and visualized by volcano plots using ggplot2 (v3.5.1). DEGs were assessed for gene ontology (Cellular Component 2023) or pathways enrichment (BioPlanet 2019) through the web-based platform Enrichr [19].

Comparison to SMCs in Single-Cell RNA Sequencing Studies

To compare the similarity of SMCs in DIT and in atherosclerotic lesions, DEGs in either the intimal or medial SMC-enriched capture spots were mapped against the top 100 specific markers of Sca1⁺ SMCs from atherosclerotic lesions of Apoe^−/− mice⁸, Ldlr^−/−Apob^100/100 mice, and the 8 SMC sub-clusters found in human carotid atherosclerotic leisions [9] to find overlapped genes. Detailed analysis methods to obtain specific markers have been described previously [8,9]. The average counts of representative signature genes in the intimal and medial SMC-enriched capture spots were visualized by heatmap using ggplot2.

Statistical Analysis

For statistical analyses performed using GraphPad Prism, the normality of distribution was assessed using the D'Agostino & Pearson normality tests (α=0.05). Normally distributed data were presented as mean ± SEM and non-normally distributed data were presented as median and interquartile range. Normally distributed data with equal variances were analyzed using unpaired t-tests. Data that did not meet parametric assumptions were analyzed using the Wilcoxon test. Statistical significance was defined as p-value <0.05.

3. Results

Movat staining showed that the intima is thicker than the media (Figure 1A) in DIT regions. SMCs expressing contractile protein marker MYH11 were enriched in both intima and media, whereas CD45⁺ leukocytes were rarely observed (Figure 1A and Supplemental Figure 1), confirming that the selected DIT regions met the pathology definition. When the layout of Visium SD capture spots was overlaid on top of the confocal images, it was estimated that RNA extracted from one capture spot (Figure 1A, white circle) can originate from more than one nucleus (blue) and from different vessel layers (white arrow).

The characteristics of sample donors are summarized in Supplemental Table 1. No significant differences in RNA fragmentation, in terms of DV200 value, were observed between autopsy and explanted heart samples (Figure 1B, left panel). However, the yield of RNA from autopsy samples was roughly half of that from the explanted heart samples, even though the autopsy samples were stored in FFPE blocks for less than 2 years, suggesting RNA degradation due to postmortem events (Figure 1B, right panel). For DIT samples with a DV200 above 30%, autopsy samples had a much higher background signal from genomic DNA, as indicated by the estimated portion of UMI counts from genomic DNA (Figure 1C). The mitochondrial gene transcripts were enriched in the autopsy samples, especially in the regions of intima and media (Figure 1D). One sample, autopsy sample A3, had only 8.4% of transcripts mapped to probes and failed sequencing quality control (Figure 1C). This sample was ruled out from further analysis.

At an FDR of 1%, there were 336 genes more enriched in the intimal and medial layers of DIT from explanted hearts (Figure 1E). Gene ontology analysis (Figure 1F) showed that these genes encode proteins that are involved in cellular components like “Focal Adhesion” and “Cell-Substrate Junction” (e.g., ITGB1, JAK2, ACTG1, VIM, ACTB), “Vesicle” (ANXA1, UBB, ACTB, MYO1D), and “Cell-Cell Junction” (e.g., CCN2, ASPN, SERPINE1, ELN) (Supplemental Table 2). The top overrepresented genes in autopsy samples are olfactory receptors (e.g., OR8D1, OR6C2, OR6T1, OR2B3, OR2S2, OR6C75) (Figure 1E).

Capture spots at the apical layer of the intima could be a mix of endothelial cells (ECs) and SMCs (yellow spots, Figure 2B, panel a). This was confirmed by the expression of both ECs marker PECAM1 and SMCs marker MYH11 (Figure 2B, panel b). After these capture spots were ruled out, the remaining spots located in the subendothelial part of the intima (blue spots, intima-subendothelium) (Figure 2B, panel c) were compared with those in the media (red spots) (Figure 2B, panel e). Both the blue and red capture spots were enriched with SMCs expressing MYH11 but not with ECs expressing PECAM1 (Figure 2B, panel d and f), suggesting that they contain predominantly intimal and medial SMCs, respectively. Differential gene expression analysis (Figure 2C) revealed that the intimal SMC-dominated blue capture spots were enriched with genes in the pathways of TGF-β regulation of extracellular matrix (e.g., VCAN, LTBP2), syndecan-1 pathway and collagen biosynthesis (e.g., COL8A1, COL16A1), extracellular matrix organization (e.g., MMP2), the RAGE pathway (e.g., TNFRSF11B), PDGF genes and receptors (e.g., PDGFRA), and glycosaminoglycan metabolism (e.g., VCAN, BGN), as compared to the red capture spots, which were occupied mainly by medial SMCs (Figure 2D, Supplemental Table 3 and 4). SMCs in the media were enriched with genes for muscle contraction (e.g., ACTG2, ACTC1, MYH11), suggesting a contractile phenotype. Interestingly, complement components C3 and C7 in the alternative complement pathway were overrepresented in the media.

Some signature genes of Sca1⁺ SMCs, including VCAN, BGN, MMP2, TNFRSF11B, COL8A1, EFEMP1, and LTBP2, were more enriched in the intimal SMC-enriched blue capture spots as compared to medial SMC-enriched red capture spots (Figure 2E, panel a, and Supplemental Table 5). However, these intima SMC-dominated capture spots did not turn on genes in the inflammatory response (e.g., VCAM1, LUM, DCN, SPP1, C3, C4B, CXCL12, and SERPINA3), as Sca1⁺ SMCs did (Supplemental Table 6). DEGs in intimal SMC-enriched capture spots overlapped the most with signature genes of sub-cluster 5 from human carotid lesions (Supplemental Table 5). They are involved in the pathway of TGF-β regulation of extracellular matrix (e.g., VCAN, EFEMP1) (Figure 2E, panel b). However, these intimal SMC-enriched capture spots lack the osteogenic and inflammatory features (e.g., LUM, DCN, VCAM1, C3, C7) presented in dedifferentiated SMCs clusters 6-8 in human carotid lesions (Figure 2E, panel b, Supplemental Table 6). Capture spots in the media were enriched with MYH11, CNN1, ACTA2, ACTG2, and LMOD1, sharing similar features with contractile SMCs in both human and mouse atherosclerotic lesions (Figure 2E).

4. Discussion

Spatial gene expression technology is a rapidly-developing technique for obtaining transcriptome profiles in the tissue context. Compared to single-cell RNA sequencing, archived FFPE tissue blocks can be analyzed. The application of spatial gene expression technology on human specimens has thus far focused on tumor biopsies, which are easily accessible from live patients. To study vascular biology, particularly coronary arteries, postmortem hearts are easily accessible, widely used, and well-accepted for pathology studies. The discovery of DIT and its relationship with atherosclerosis are the results of autopsy studies [3]. However, in the era of next-generation ‘omics’, the use of postmortem tissues for RNA sequencing is debatable.

Our study is the first to benchmark RNA quality and spatial sequencing performance of postmortem coronary arteries. Forensic studies have indicated that postmortem RNA degradation is tissue-specific and gene-specific [20]. RNA was found to be relatively more stable in the brain and cardiac muscle than in the aorta [20]. Sequencing quality assessment metrics, including estimated UMI from genomic DNA and reads mapped to probes, are also tissue-specific [21]. Therefore, coronary arteries from heart transplant patients are needed as a control to assess RNA quality prior to postmortem degradation. Our results have shown that even if autopsy samples have met the minimal requirement of the DV200 RNA integrity test [13], this test does not accurately indicate RNA degradation in autopsy samples. It only predicts that the length of RNA fragments is suitable for library construction [22]. The estimated UMI from genomic DNA and mitochondrial transcripts is high in some autopsy samples because DNA and mitochondrial RNA are less sensitive to postmortem degradation than mRNA [23]. Therefore, we do not recommend solely relying on DV200 for quality control. Measure of RNA yield per tissue section could be an additional criterion to rule out samples with degraded RNA before proceeding to sequencing. If only postmortem blood vessels are available, samples with both high RNA yield and DV200 should be prioritized.

Despite RNA degradation, the transcriptome profiles of autopsy tissues may still be valuable to identify gross differences between diseased and healthy populations. When the clinical characteristics of donors are considered, heart failure of explanted heart samples is a cofounding factor for data interpretation. The overrepresented olfactory receptors in autopsy samples (donors without clinically diagnosed heart failure) aligns with impaired olfactory function and dysregulated expression in the hearts of heart failure patients [24]. However, to distinguish subtle differences between phenotypes (e.g., SMCs in the intima and media), our results have shown that RNA degradation could jeopardize the resolution of spatial gene expression. Hence, it is important to critically assess the value of postmortem tissues based on the research questions.

Due to limited access to human samples and in vitro models, it is still unknown how DIT is formed. One theory is that DIT is an adaptive response to mechanical stresses [4]. Well preserved mRNA from explanted hearts highlights the heterogeneity of SMC-enriched capture spots, suggesting that intimal SMCs are phenotypically different from medial SMCs. One limitation of the Visium SD platform is that it does not have single-cell resolution to directly identify capture spots with a pure SMC population. We assumed that the MYH11⁺ capture spots in the subendothelial and the medial layers contain predominantly SMCs because CD45 staining is rarely seen in DIT, ruling out the possibility of leukocytes contributing to the transcriptome profiles of these capture spots. Although pericytes and fibroblasts can express SMC contractile markers, pericytes are distributed in the microvessels [25] and fibroblasts are in the adventitia of non-atherosclerotic coronary arteries [26]. We used gene ontology analyses to predict the functional status of intimal SMCs, as the outcome of stimuli in the microenvironment of intima. It suggested that SMCs in the intima may have been exposed to TGF-β, an important cytokine that stimulates the production of extracellular matrix by SMCs, including versican (encoded by VCAN) and biglycan (encoded by BGN) [27,28]. In the current study, gene expression analysis indicated that VCAN and BGN were enriched in the subendothelial regions of the intima, which is consistent with their protein expression in DIT [29] and at least part of the phenotypic outcome of TGF-β activation. Increased shear stress generated by the blood flow stimulates the secretion of active TGF-β from ECs [30]. Together with platelet-derived growth factor, TGF-β attract SMCs to establish and enlarge arteries during vascular development [27]. However, we do not think that shear stress-induced secretion of TGF-β from ECs can simply explain DIT formation. Rodent aortas face much greater shear stress than humans [31], but do not have DIT. The molecular features of DIT seen in this study (from patients over 40 years old) are the results of decades of tissue remodeling, and aging itself can contribute to DIT formation after birth. Increased MMP2 in the intima has also been observed in the thickened intima of older monkeys as compared to the thin intima of the young, both without atherosclerosis [32]. MMP2 can activate TGF-β production by SMCs in vitro, increasing their expression of collagen and proteoglycans [33]. Coronary arteries from young donors would be a better fit to investigate how DIT was originally formed during embryonic development. DIT in heart failure patients, especially those from patients with ischemic cardiomyopathy, are of particular interest because they remain at a pre-clinical stage when other parts of the coronary arteries that are exposed to the same genetic and circulating risk factors have developed lesions. We can infer the proatherogenic features of SMCs and risk factors in the lesion microenvironment by comparing SMCs in these DIT to those from the atherosclerotic lesions.

Intimal SMCs are hypothesized to derive from migration and clonal expansion of a subset of medial SMCs, like Sca1⁺ SMCs in mouse atherosclerotic lesions [34,35]. Without cell lineage tracing tools and specific phenotypic markers for intimal SMCs, we cannot directly test the origin, stimulus of phenotypic transition, and function of intimal SMCs in humans. Nevertheless, our results suggested that Sca1⁺ SMCs from mouse lesions reflect a phenotype in response to inflammation in the lesions and human intimal SMCs in the DIT represent the phenotypic outcome of no or low inflammation activation. Cellular response to cytokine stimulus and cytokine-mediated signaling pathway are the top 2 gene ontology features of Sca1⁺ SMCs from the lesions of Ldlr^−/−Apob^100/100 mice [9], which were not found in human SMCs in the intima of DIT. To predict the functional differences between intimal and medial SMCs in the DIT, interpretation based on individual DEGs is a traditional challenge [36]. For example, VCAN and FBLN1 increased the proliferation and migration of SMCs in vitro [37,38], but they were enriched in the intima and media, respectively, in the current study. The relationship between gene ontology pathways and cellular function of Sca1⁺ SMCs from Apoe^−/− mouse lesions has been tested: activation of the complement cascade led to the secretion of complement component C3, which promotes proliferation and inflammation [8]. Since activation of the complement cascade is absent in intimal SMCs, we do not expect that intimal SMCs in DIT will be hyperproliferative and proinflammatory as Sca1⁺ SMCs. Moreover, trajectory analysis of sub-clusters from human carotid lesions indicated that sub-cluster 5, sharing the most featured markers with intimal SMCs in the DIT, is distinct from other dedifferentiated SMCs that were considered as chondrogenic, osteogenic, or proinflammatory [9]. All of our evidence above supports the theory of Nakashima et al.: intimal SMCs in DIT are in a stable status and they are not adequate to drive disease progression [6].

Intimal SMCs in DIT could be in a primed state that can readily transdifferentiate to different phenotypes and inflammation can be the key stimulus for them to switch on more proatherogenic features. Such a group of primed SMCs was found in healthy mouse aortas, and the authors suggested that inflammation drives the initial phenotypic changes in atherogenesis [39]. Proteoglycans in the extracellular matrix play a significant role in modulating cell response to inflammation. LUM and DCN, highly expressed by the osteogenic and inflammatory SMCs in advanced human carotid lesions [9], modulate the bioactivity of cytokines, contributing to inflammation and calcification of atherosclerotic lesions [40-42]. In the current study, the lack of these “proatherogenic” proteoglycans in SMCs from DIT regions suggests that the interplay between SMC phenotypes and the extracellular matrix is important for disease progression. Intimal SMCs may pave the way for foam cell formation by secreting BGN [34], but the formation of SMC-foam cells will not necessarily activate cell death or chronic inflammation within human lesions [43]. Studying how inflammation affects SMC phenotype and its role in tissue remodeling will help explain why while everyone has DIT, those with chronic inflammatory disorders have a higher risk of developing coronary artery disease, independent of traditional risk factors for cardiovascular disease [44].