Single-Molecule Spatial Transcriptomics of Human Thoracic Aortic Aneurysms Uncovers Calcification-Related CARTPT Expressing Smooth Muscle Cells

Abstract

Background:

Although single-cell RNA-sequencing is commonly applied to dissect the heterogeneity in human tissues, it involves the preparation of single-cell suspensions via cell dissociation, causing loss of spatial information. In this study, we employed high-resolution single-cell transcriptome imaging to reveal rare smooth muscle cell (SMC) types in human Thoracic Aortic Aneurysm (TAA) tissue samples.

Methods:

Single-molecule spatial distribution of transcripts from 140 genes were analyzed in fresh-frozen human TAA samples with region and sex-matched controls. In vitro studies and tissue staining were performed to examine human CARTPT (CART prepropeptide) regulation and function.

Results:

We captured thousands of cells per sample including a spatially-distinct CARTPT expressing SMC subtype enriched in male TAA samples. Immunoassays confirmed human CART enrichment in male TAA tissue and truncated CARTPT secretion into cell culture medium. Oxidized low-density lipoprotein, a cardiovascular risk factor, induced CARTPT expression, while CARTPT overexpression in human aortic SMCs increased the expression of key osteochondrogenic transcription factors and reduced contractile gene expression. Recombinant human CART treatment of human SMCs further confirmed this phenotype. Alizarin red staining revealed calcium deposition in male TAA samples showing similar localization with human CART staining.

Conclusions:

Here we demonstrate the feasibility of single-molecule imaging in uncovering rare SMC subtypes in the diseased human aorta, a difficult tissue to dissociate. We identified a spatially-distinct CARTPT expressing SMC subtype enriched in male human TAA samples. Our functional studies suggest that human CART promotes osteochondrogenic switch of aortic SMCs, potentially leading to medial calcification of the thoracic aorta.

Graphical Abstract:

INTRODUCTION

Thoracic aortic aneurysm (TAA) is a major cause of morbidity and mortality in the United States. Surgical repair is still the standard TAA treatment due to the inadequate understanding of human TAA pathogenesis and lack of alternative therapeutics. The human aorta harbors multiple cell types, and vascular smooth muscle cells (SMCs) of different developmental origins are the predominant cell type forming the vascular wall. SMC defects are the major determinant of aortic diseases and regional TAA susceptibility has been linked to lineage-specific SMC dysfunction.^1–7 Given the significant genetic and regional heterogeneity of human TAAs, high-resolution approaches are required to reveal the SMC defects underlying TAA initiation and progression.^7,8

Single-cell RNA-sequencing (scRNA-seq) is frequently employed to resolve cell heterogeneity in healthy and diseased human tissue. However, scRNA-seq involves the preparation of single-cell suspensions which causes gene expression changes and loss of spatial information. The human aorta is particularly difficult to dissociate due to its high extracellular matrix content. Despite this challenge, a number of single-cell genomics studies provided key insights into the SMC subtypes in the aneurysmal human aorta.^9–11 Spatial transcriptomics is a powerful alternative to scRNA-seq as it enables gene expression measurements in native tissue context. Particularly, in situ hybridization (ISH)-based spatial profiling strategies including MERFISH (Multiplexed error-robust fluorescence ISH) offer near single-molecule sensitivity and high spatial resolution in fresh-frozen tissue preparations, representing an ideal tool to study human aortic heterogeneity.¹² To date, high-resolution, multiplexed spatial profiling technologies have not been applied to investigate human TAA, leaving the spatial remodeling in human TAA tissue including disease-causing and protective changes unexplored.

In this study, we demonstrate the feasibility of MERFISH profiling of the human aorta for the first time, revealing the spatial organization of rare SMC subtypes in human TAA. This platform enabled us to uncover a spatially-distinct SMC subtype marked by CARTPT (CART prepropeptide) expression. We present functional evidence that human CART (cocaine- and amphetamine-regulated transcript) promotes the expression of osteochondrogenic switch markers in human SMCs and is likely a potentiator of medial calcification in male human TAA samples.

METHODS

All experiments were performed according to the protocols (HUM00035836, HUM00096079, HUM00054585 and HUM00052866) approved by the Institutional Review Board at the University of Michigan. The study materials can be made available from the corresponding authors on reasonable request. Single-cell MERSCOPE (MERFISH-based spatial transcriptomics platform) data is deposited to the Gene Expression Omnibus database under the accession number GSE241870. Detailed descriptions of the reagents are in the Major Resources Table.

MERFISH Profiling and Single-Cell Data Analysis

Surgically resected fresh ascending aorta samples from two control subjects with no aortic aneurysms (denoted as ♂Control and ♀Control) and three TAA subjects (denoted as ♂TAA1, ♂TAA2 and ♀TAA1) with ascending aneurysms were immediately embedded and frozen in Optimal cutting temperature compound (OCT, Fisher Catalog No. 23–730-571). All tissue samples were extracted from the greater curvature of the ascending aorta for a proper comparison. Age, sex, race, aneurysm size, disease condition and smoking status of each subject are indicated in Table S1. RNA quality of each sample was assessed using Agilent Bioanalyzer (RNA integrity number>7) prior to processing the samples. OCT-embedded ascending aorta samples were cut into 10 μm sections using a cryostat and adhered onto the imageable area of Vizgen slides. Immediately after removing the slides from the cryostat, they were fixed in pre-heated 4% paraformaldehyde for 30 minutes at 47°C. Subsequent slide processing was performed according to Fresh Frozen Tissue preparation protocol by Vizgen. Hybridization of custom MERFISH probes targeting 140 genes, gel embedding and tissue clearing were performed by Vizgen according to the MERSCOPE User Guide. Each sample was stained with both DAPI and PolyT staining reagents.

Cell segmentation boundaries were determined using the default MERFISH approach.¹² The Vizgen MERSCOPE data (one intact tissue segment per sample) and cell boundaries were inputted to Seurat 4.2 (R package) using LoadVizgen function.¹³ Cells with less than 20 molecules were discarded from the subsequent analysis. SCTransform and ScaleData functions of Seurat were used to normalize and scale the data respectively. Unsupervised clustering (Louvain algorithm) and dimensionality reduction were performed using Seurat to identify spatially-resolved cell types on the merged dataset of two control and three TAA samples. FeaturePlot and ImageFeaturePlot functions were executed to visualize the data on a two-dimensional embedding. ImageDimPlot function was used to visualize the localization of individual molecules. Cluster markers were identified using PrepSCTFindMarkers and FindAllMarkers functions respectively. Differential gene expression comparing control and TAA samples was performed using FindMarkers function, which uses Wilcoxon Rank Sum Test. p-value adjustment is performed using Bonferroni correction.

The MERSCOPE data was compared to human TAA single-cell RNA-sequencing (scRNA-seq) dataset generated by Li et al.⁹ The scRNA-seq expression matrices were downloaded from Gene Expression Omnibus dataset under GSE155468 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE155468). Three controls (ages 61 to 63 years old), three male TAA samples (ages 56 to 69 years old) and three female TAA samples (ages 67 to 75 years old) were inputted to Seurat 4.2 (R package). We excluded TAA subjects with bicuspid aortic valve to be consistent with our MERSCOPE TAA samples. The single cells with more than 1000 molecule count and less than 10% mitochondrial content were kept for the subsequent analysis. The resulting samples were merged using IntegrateData function and clustered using Seurat. Cluster markers were identified using FindAllMarkers function and SMC clusters were identified based on significant MYH11 enrichment (Fold-change (FC)>2.0, adjusted p-value (p-adj)<0.001) in the merged dataset.

Immunohistochemistry, Alizarin Red and Von Kossa Stainings

Immunohistochemical (IHC) analysis was conducted using Abcam ab64264 kit. Paraffin blocks were prepared from two control (non-aneurysmal) subjects independent of the MERSCOPE control subjects, four male TAA subjects (including ♂TAA1 and ♂TAA2) and four female TAA subjects (including ♀TAA1). The blocks were sectioned to a thickness of 5 μm, and then deparaffinized. After rehydration, the samples underwent antigen retrieval using citrate buffer in a pressure cooker for 15 minutes. After cooling completely, the tissue sections were washed four times in Tris-buffered saline (TBS) and treated with a Hydrogen Peroxide blocking solution for ten minutes. Following two washes with TBS, the tissue sections were incubated at room temperature for ten minutes in Protein blocking solution. Subsequently, the tissue sections were incubated in CART antibody (Cell Signaling Technology, Catalog No. 14437, 1:75) in 1% Bovine serum albumin (BSA) overnight at 4°C. After the primary antibody incubation, the sections underwent four TBS washes and were then incubated with a biotinylated secondary antibody for 30 minutes. The tissue sections were then washed four additional times in TBS and Streptavidin peroxidase was applied to the sections for 30 minutes. Following four TBS washes, the slides were incubated with DAB Chromogen and DAB substrate (1:30) for 1 minute. A counterstain of Hematoxylin was applied, and the slides were imaged using a KEYENCE microscope (BZ-X810).

For Alizarin red staining, the paraffin-embedded tissue specimens were sectioned into 4 μm thick slices. After deparaffinization and hydration, the samples were covered in an Alizarin Red staining solution (Sigma, Catalog No. TMS-008-C) and incubated for 30 minutes at room temperature. Subsequently, the slides were washed in Phosphate-Buffered Saline (PBS) before dehydrating in Acetone and then in an Acetone-Xylene (1:1) solution for 40 seconds in each. The slides were then soaked in xylene for a minute twice before mounting. For Von Kossa staining, paraffin blocks were sectioned to a thickness of 5 μm, deparaffinized and rehydrated. The sections were incubated in 5% Silver nitrate solution (Sigma, Catalog No. S8157) for 1 hour under Ultraviolet light. Following two washes in Milli-Q water (three minutes each), the sections were covered in 2% Sodium Thiosulphate (Sigma, Catalog No. 72049) for 5 minutes. Following two additional washes in Milli-Q water, the sections were counterstained with Eosin (Fisher 7111L), dehydrated and cleared. The sections were imaged using a KEYENCE microscope (BZ-X810) and analyzed with ImageJ.

CARTPT Induction, Overexpression and Recombinant Human CART Treatment

Primary aortic SMCs were isolated from the thoracic aorta as previously described.^14,15 Cells were cultured in Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12; Gibco, Catalog No. 11330–057), supplemented with 20% Fetal Bovine Serum (FBS) and 1% Penicillin-Streptomycin (PS) prior to the cell culture experiments. To ensure consistency in cellular behavior and integrity, primary SMCs from the same passage were used in non-treatment and treatment conditions. For the induction of CARTPT expression, a healthy male donor (with no aortic aneurysm) primary SMCs (passage 4) were cultured in DMEM/F12, supplemented with 10% FBS and 1% PS, until 70–80% confluence was reached. Cells were then stimulated for 16 hours using either a vehicle control, 10 ng/mL recombinant human TNF-alpha (rhTNFα; Fisher, Catalog No. 210-TA-100), or 25 μg/mL oxidized Low-Density Lipoprotein (oxLDL). The oxLDL was freshly prepared through copper catalysis of native human plasma LDL (Fisher, Catalog No. L3486) for 4 hours at 37°C, followed by dialysis against PBS. Subsequently, mRNA was extracted for further analysis. CARTPT overexpression experiments were performed by culturing the healthy donor primary SMCs (passage 4) overnight in DMEM/F12, supplemented with 10% FBS and 1% PS, until approximately 60% confluence was achieved. Cells were then washed with PBS and transfected with either 1 μg/mL of control pcDNA 3.1, pRP[Exp]-CMV>EGFP-EF1A>human CARTPT (Vectorbuilder, used for RT-qPCR) or p-CMV3-CARTPT-HA (Sino Biological, Catalog No. HG13240-CY, used for Co-IP) vectors. Transfection was carried out for an additional 48 hours in serum-reduced OPTI-MEM (Gibco, Catalog No. 31985070) using Lipofectamine 2000 (Invitrogen, Catalog No. 1668019) following the manufacturer’s protocol. Post-transfection, DNase digested mRNA and protein were extracted, and conditioned media were collected for further analysis. For recombinant CART treatment, the healthy donor primary SMCs were cultured overnight in DMEM/F12 supplemented with 10% FBS and 1% PS until approximately 60% confluence was achieved. Subsequently, the cells were treated with either recombinant human CART (rhCART; R&D, Catalog No. 10974-CO-050) or control 4mM HCl for 0 to 48 hours, following a decreasing order. Upon completion of the treatment, mRNA and protein were harvested.

RNA Extraction and Quantitative Real-Time PCR (RT-qPCR)

For RNA extraction, we used the RNeasy Mini Kit to directly isolate mRNA from the cells. The isolated mRNA was reverse-transcribed into complementary DNA (cDNA) using the SuperScript III kit (Fisher, Catalog No. 18080051), as per the manufacturer’s instructions, and utilizing random hexamers to prime the reaction. To quantify the relative abundance of cDNA, the BioRad Real-Time PCR Detection System was employed, using the SYBR Green Fast qPCR Mix (Abclonal, Catalog No. RK21203) for fluorescence detection. Gene expression levels were quantified using the 2^−ΔΔCt method, with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) serving as the internal control for normalization. The primer sequences used for each target gene in the qPCR analyses are listed in the Major Resources Table. All RT-qPCRs were conducted on primary SMCs from independent rounds of treatments, and matching control and experimental samples were collected from each round of treatment.

Western Blotting and Co-Immunoprecipitation

Cells were lysed using Radioimmunoprecipitation Assay (RIPA) buffer (Fisher, Catalog No. 89901), supplemented with protease inhibitor cocktail (Roche, Catalog No. 11873580001) and PhosSTOP phosphatase inhibitors (Roche, Catalog No. 4906845001). The resultant protein extracts were subjected to Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) for resolution, and then transferred onto nitrocellulose membranes. Membranes were blocked with 5% non-fat milk for one hour at room temperature. Subsequently, they were incubated with the 1:1000 diluted p-ERK (CST, Catalog No. 4370) and ERK (CST, Catalog No. 9102) primary antibodies overnight at 4°C, followed by a 1-hour incubation at room temperature with the fluorescence-conjugated secondary antibody (Li-Cor Bioscience, 1:4000). The membrane was then scanned using the LI-COR DLx Odyssey imaging system. Protein bands were quantified using the LI-COR Empiria Studio software. For co-immunoprecipitation studies, conditioned media were centrifuged at 500g for 10 minutes to remove cellular debris. The supernatants were then immunoprecipitated using Protein G Magnetic Beads (CST, Catalog No. 7024) in conjunction with either IgG control or HA-targeting antibody (CST, Catalog No. 3724), following the manufacturer’s protocol. Eluted samples were further concentrated by evaporating at 65°C. Western blotting was then conducted using primary antibody against CART (CST, Catalog No. 14437). All western blots were conducted on primary SMCs from independent rounds of treatments, and matching control and experimental samples were collected from each round of treatment.

Statistical Analysis for Cell Culture Experiments

The cell culture experiments were performed on at least six replicates and the quantitative cell culture data were presented as mean ± standard deviation. When analyzing two datasets that are normally distributed (Shapiro-Wilk normality test), we performed two-tailed unpaired t test. If the variances are significantly different (F-test), we used two-tailed unpaired t test with Welch’s correction. When analyzing two datasets that are not normally distributed, we used nonparametric two-tailed Mann-Whitney test. When analyzing more than two groups that are not normally distributed, we performed nonparametric Kruskal-Wallis test with Dunn’s multiple comparisons test. Individual p values are presented in the Figures up to the significance level p-value<0.0001 and p-value≤ 0.05 was considered statistically significant. The statistical analyses were performed using GraphPad Prism 9 Software.

RESULTS

Single-molecule spatial transcriptomics of human TAA tissue

We analyzed the single-molecule spatial distribution of transcripts from 140 aortic cell heterogeneity and aneurysm related genes plus signaling pathways on two control samples (♂Control, ♀Control) without aortic aneurysm and three TAA samples with ascending aneurysm (Ascending aorta measurements: ♂TAA1 = 5.0 cm, ♂TAA2 = 5.6 cm, ♀TAA1 = 6.0 cm) using the MERSCOPE platform (Figure 1A, Table S1).^{7–9,11,16–31} The complete list of genes in the panel are listed in Supplemental Excel Sheet 1. OCT-embedded ascending aorta samples were cut into 10 μm sections and processed using optimized conditions for human aortic tissue. Of note, when the aorta is sliced into smaller pieces, it bends towards adventitia, positioning adventitia in the inner curve and intima in the outer curve (Figure 1A). Overall transcript/molecule detection efficiency per field of view was consistent between different samples (Figure 1B). Cell segmentation was able to capture thousands of single cells per sample; (♂Control: 6753 cells, ♀Control: 10172 cells, ♂TAA1: 19767 cells, ♂TAA2: 16430 cells, ♀TAA1: 15882 cells (Figure 1A).¹² The segmented cells had similar individual cell volume distribution across different samples (Figure 1C). MERFISH-derived single-cell expression profiles were analyzed using Seurat 4.2 (R package) extension for MERSCOPE system.¹³

Figure 1. Single-molecule spatial transcriptomics of aneurysmal human ascending aorta.

A) Experimental design showing MERSCOPE profiling of non-aneurysmal (Control) and aneurysmal (TAA) human ascending aorta. The tissue samples were extracted from the greater curvature. B) Violin plots showing transcript/molecule counts per field of view (fov) in each sample. Each fov is 200 um × 200 um. C) The distribution of individual cell volumes in each sample after cell segmentation. D) Uniform Manifold Approximation and Projection (UMAP) visualization showing major cell clusters (SMC: Smooth muscle cell clusters constituting the aortic media in grey; Fibroblast: Fibroblast cluster in red; Immune: Immune cell cluster in purple; EC: Endothelial cell cluster in green). Individual samples are highlighted in dark green on the merged dataset. E) Expression of top cluster markers on UMAP visualizations. F) Spatially resolved normalized expression of cluster markers in each centroid across different samples. G) Spatial distribution of Fibroblast and EC clusters in the aortic wall. Individual transcript positions are highlighted using enlarged circles. Dashed boxes indicate the focused areas. H) Expression of top MERSCOPE cluster markers on UMAP visualizations of Li et al. dataset (three control and six TAA samples).⁹

Unsupervised clustering revealed three distinct cell clusters in addition to SMC clusters constituting the aortic media, and the major clusters contained contributions from each sample (Figure 1D). SMC clusters were further subclustered in Figure 2. Fibroblast cluster was enriched in CCDC80 (coiled-coil domain containing 80), PODN (podocan) and SERPINF1 (serpin family F member1) (FC>2, p-adj<0.001, Supplemental Excel Sheet 2); Immune cell cluster in MS4A7 (membrane spanning 4-domains A7) and FCER1G (Fc epsilon receptor Ig) (FC>2, p-adj<0.001, Supplemental Excel Sheet 2); and Endothelial cell (EC) cluster in PECAM1 (platelet/endothelial cell adhesion molecule 1) and VWF (von Willebrand factor) (FC>2, p-adj<0.001, Supplemental Excel Sheet 2) (Figure 1E). We also examined the spatially resolved expression of cluster markers showing that SMC markers dominate the aortic media while Fibroblast, Immune cell and Endothelial cell cluster markers were located in the adventitia and intima layers (Figure 1F). Endothelial cells were present both in the aortic intima and adventitial microvessels (Figure 1G). Among the fibroblast markers, individual CCDC80 molecules were more restricted to the adventitia, while PODN molecules also showed localization in the intima, implying additional fibroblast heterogeneity consistent with the previous reports (Figure 1G).⁹ Next, we analyzed a previously published dataset of human TAA single-cell RNA-sequencing.⁹ Our major cluster markers for SMCs (MYH11 (myosin heavy chain 11), CARMN (cardiac mesoderm enhancer-associated non-coding RNA), CNN1 (calponin 1)), fibroblasts (CCDC80, PODN, SERPINF1), endothelial cells (PECAM1, VWF) and immune cells (MS4A7, FCER1G) labeled different cell groups in the merged dataset while the markers for the same cell cluster showed co-localization to the same cell groups demonstrating the relevance of our cluster markers (Figure 1H).

Figure 2. MERSCOPE Profiling reveals rare SMC subtypes.

A) UMAP visualization showing SMC subclusters. Expression levels of SMC subcluster markers highlighted in green. B) Violin plots showing expression levels of enriched markers in SMC subclusters. C) Abundance of SMC3 and SMC4 in different samples normalized to the total SMC number in each sample. D) Spatially resolved normalized expression of SMC3 and SMC4 subcluster markers in SMCs across different samples. E) Expression levels of top MERSCOPE SMC subcluster markers in SMC clusters extracted from Li et al. dataset.⁹ F) Dot plots showing CARTPT expressing cell abundance and CARTPT expression levels among all cells or SMCs across three control and six TAA scRNA-seq samples.

MERSCOPE profiling reveals rare SMC subtypes

SMC defects are the main potentiators of TAA initiation and progression. To dissect SMC heterogeneity in the control and TAA samples, we performed SMC subclustering analysis, revealing five SMC subtypes (Figure 2A, Supplemental Excel Sheet 3). All SMC subtypes expressed SMC markers such as MYH11, CNN1, CARMN and ELN (elastin) at variable levels (Figure 2B). SMC1 did not have any distinct markers within the 140 gene panel (Figure 2B). SMC2 was significantly enriched in PLN (phospholamban) and LGR6 (leucine rich repeat containing G protein-coupled receptor 6), while SMC5 had only one exclusive marker HAS1 (hyaluronan synthase 1) (Figure 2B and Supplemental Excel Sheet 3). Interestingly, we identified two rare SMC subtypes: SMC3 enriched in NR4A3 (nuclear receptor subfamily 4 group A member 3), ADAMTS4 (ADAM metallopeptidase with thrombospondin type 1 motif 4) and DACT1 (dishevelled binding antagonist of beta catenin 1); and SMC4 enriched in CARTPT (CART prepropeptide) and TNFRSF11B (TNF receptor superfamily member 11b) (p-adj<0.001, Supplemental Excel Sheet 3). SMC3 and SMC4 also showed differential distribution among Control and TAA samples (Figure 2C). Interestingly, SMC4 was expanded in the TAA samples showing a male bias. Furthermore, differential gene expression analysis revealed signaling receptors PLN and LGR6 were enriched in the Control SMCs, while NOTCH3 (notch receptor 3) and TNFRSF11B were higher in the TAA SMCs implying significant signaling changes in the TAA tissue (Figure S1A, Supplemental Excel Sheet 4).

Next, we plotted the spatially resolved expression of SMC3 and SMC4 markers on SMCs only across the control and TAA samples (Figure 2D). This revealed a restricted distribution of SMC3 markers such that NR4A3 and ADAMTS4 appeared more localized to the SMCs neighboring the adventitia (Figure 2D). Of note, the top SMC3 markers such as NR4A3 and ADAMTS4 are significantly enriched in fibroblasts implying non-SMC origins of SMC3 (Supplemental Excel Sheet 2). Surprisingly, the most exclusive SMC4 marker; CARTPT transcript showed a spatially distinct distribution, labeling a continuous layer of SMCs in the inner media of male TAA samples (Figures 2D and S1B). The expression of Fibroblast, Immune cell and Endothelial cell cluster markers was very low in CARTPT expressing SMC4 (Figure S1C). The other SMC4 marker; TNFRSF11B, is a more pervasive marker compared with CARTPT (Figure 2B) but still showed enrichment in the inner media of male TAA samples (Figure 2D).

To confirm the relevance of SMC3 and SMC4 in the human TAA scRNA-seq study, we examined their marker expression in SMCs from the Li et al. dataset.⁹ CARTPT showed sparse labeling but overlapped with the more pervasive TNFRSF11B (Figure 2E) as well as another SMC4 marker; CYTL1 (cytokine like 1) (Supplemental Excel Sheet 3). Similarly, NR4A3 and ADAMTS4 appeared co-expressed in similar SMC populations (Figure 2E). More importantly, CARTPT expression showed a clear enrichment in three male TAA samples compared with three control and three female TAA samples consistent with our findings (Figure 2F). Additionally, we plotted the expression of CARTPT and more pervasive SMC4 markers (TNFRSF11B and CYTL1) on all single cells in the MERSCOPE and human scRNA-seq datasets. SMC4 markers had similar expression patterns with CARTPT being the most exclusive marker in both datasets (Figures S1D and S1E). Cumulatively, the human scRNA-seq data confirms the sex and TAA-based expression of CARTPT in the human aorta.

Human CART is a secreted protein enriched in male human TAA tissue

Next, we investigated the spatial distribution of SMC3 and SMC4 among all SMCs across different samples. As expected, SMC3 were positioned near the adventitia with an enrichment in the female aneurysm sample, while SMC4 formed a continuous layer in the inner media (Figure 3A). Given the male bias and spatially-distinct distribution of CARTPT, we focused on this gene and performed immunohistochemistry to confirm our findings at the protein level. Immunostainings on two control, four male TAA and four female TAA samples revealed a clear enrichment of human CART in male TAA media with a distinguishable CART+ layer (Figures 3B and S2A). Surprisingly, human CART is localized in the middle of the media and did not completely overlap with CARTPT mRNA expression positioned closer to intima suggesting the post-translational modification of CARTPT protein (Figures 2D and 3B).

Figure 3. CARTPT expressing cells form a spatially-distinct layer in male TAA samples.

A) Spatial distribution of SMC3 (red) and SMC4 (green) in the aortic media. SMC1, SMC2 and SMC5 were colored in grey. B) Immunohistochemical stainings showing human CART labeling in male TAA samples. Left-to-right TAA subjects ascending aorta measurements: male TAA (5.0 cm, same as ♂TAA1), male TAA (5.6 cm, same as ♂TAA2), male TAA (5.3 cm), male TAA (4.8 cm), female TAA (6.0 cm, same as ♀TAA1), female TAA (5.6 cm), female TAA (5.7 cm), female TAA (5.0 cm). C) Western blot showing cleaved human CART secretion in cell culture medium. Cell culture medium and lysate were collected from primary SMCs two days after transfection with p-CMV3-human CARTPT-HA. Human CART immunoprecipitation in cell culture medium was performed using an anti-HA antibody. D) CARTPT expression values (TPM: transcript per million) in age-matched male and female human aorta samples downloaded from ADEIP Database (Mann-Whitney test).

CARTPT protein has several cleavage sites in its N-terminal and produces active peptides after site-specific cleavage.^32–35 Cleaved CART peptide can be released in the bloodstream in response to hypotensive stress.³⁶ Human CART peptides are also released in the cell culture medium.^32,35 Next, we asked whether human CART can be secreted by primary human aortic SMCs. To test this, primary SMCs were transfected with a vector containing human CARTPT with a C-terminal HA tag. Conditioned media was collected and subjected to co-immunoprecipitation (co-IP) using Protein G Magnetic Beads in conjunction with either IgG control or HA-targeting antibody. Western blot showed a larger band in cell lysate corresponding to the full-length CARTPT with an HA tag while we detected a smaller band in the cell culture medium co-immunoprecipitated with HA antibody suggesting that cleaved CART was released into the medium (Figure 3C).

To gain more insight into CARTPT regulation, we examined its expression among human aorta samples in ADEIP (An integrated platform of age-dependent expression and immune profiles across human tissues) and GTEx (The Genotype-Tissue Expression Project) databases.³⁷ The ADEIP data confirmed the male enrichment of CARTPT in different age groups and near lack of CARTPT expression in female subjects (Figure 3D). We also noticed an age-dependent trend in CARTPT expression implying its regulation by vascular aging-related processes (Figure 3D). The GTEx data also confirmed the male enrichment of CARTPT in the human aorta and further suggests CARTPT enrichment in human aorta compared with most other tissues in the body (Figure S2B).

Human CART promotes the expression of osteochondrogenic transcription factors in human primary SMCs

CARTPT expression is sensitive to stress and has been implicated in a stress-response system in mammalian brain.^32,38,39 To address its regulation in human aortic primary SMCs, we stimulated the cells using either a vehicle control of 10 ng/mL recombinant human TNFα (rhTNFα) or 25 μg/mL oxLDL. Interestingly, oxLDL treatment significantly increased CARTPT expression while rhTNFα treatment did not alter CARTPT level suggesting that a cardiovascular risk factor can induce CARTPT expression (Figure 4A). Next, we investigated the effects of human CART on SMC function. We transfected human aortic primary SMCs with human CARTPT vector, which increased CARTPT expression (Figure 4B). Canonical SMC contractility markers (MYH11 and SM22α (transgelin)) were significantly downregulated only after 48 hours (Figure 4B). Given the upregulation of CARTPT mRNA and protein in male TAA tissue (Figures 2D and 3B), we also measured SMC phenotype switch markers including elastolytic enzyme MMP2 (matrix metallopeptidase 2) and key osteochondrogenic differentiation transcription factors; RUNX2 (RUNX family transcription factor 2) and MSX2 (msh homeobox 2).^40–45 All three markers were robustly upregulated in response to human CARTPT overexpression (Figure 4B).

Figure 4. Human CART promotes the expression of osteochondrogenic transcription factors in SMCs.

A) Relative CARTPT expression in untreated (n=12), oxidized LDL (oxLDL)-treated (n=12) and human TNFα-treated (n=11) primary SMCs (Kruskal-Wallis test with Dunn’s multiple comparisons test). B) Relative expression of SMC markers, osteochondrogenic transcription factors and MMP2 after CARTPT overexpression in primary SMCs (n=6; unpaired t test with Welch’s correction for CARTPT and SM22α (TAGLN); unpaired t test for MYH11 and RUNX2; Mann-Whitney test for MSX2 and MMP2). C) Relative expression of SMC markers and osteochondrogenic transcription factors after recombinant human CART (rhCART) treatment of primary SMCs for 48 hours (n=6, unpaired t test with Welch’s correction for RUNX2, MSX2 and SM22α; unpaired t test for MYH11). D) Western blot showing ERK phosphorylation after recombinant human CART treatment for 24 hours (n=3). E) Alizarin red staining showing variable medial calcium deposition in male TAA samples.

Given that biologically active human CART is a secreted protein, we treated the primary SMCs with recombinant human CART (rhCART). Consistent with the overexpression experiment, the SMC contractility marker SM22α was reduced and MYH11 showed a downregulation trend following 48 hours of rhCART treatment (Figure 4C). The expression of osteochondrogenic differentiation transcription factors RUNX2 and MSX2 were significantly upregulated (Figure 4C). Lastly, we performed RT-qPCRs to measure fibroblast marker expression (CCDC80 and SERPINF1) after CARTPT over-expression and rhCART treatment. CCDC80 and SERPINF1 levels were either reduced or unaltered in both conditions suggesting that human CART does not promote fibroblast differentiation (Figure S3A). CART signaling has been shown to activate ERK signaling in different studies.^46–49 To further validate the effectiveness of rhCART treatment on human aortic primary SMCs, we treated the cells with rhCART for 24 hours and performed Western blots. Immunoblots revealed a robust elevation in phosphorylated-ERK (p-ERK) levels following rhCART treatment (Figure 4D).

Based on these results, human CART signaling appears to decrease SMC contractility markers and promote osteochondrogenic differentiation, which may eventually lead to calcium deposition in aorta. We examined calcification in our TAA samples by Alizarin red staining, commonly used to label calcium deposits.⁵⁰ The staining revealed variable levels of calcium deposition in male human TAA aortic media suggesting that male TAA samples were starting to calcify at the time of surgical resection (Figure 4E). Of note, two of the male TAA samples had significantly higher calcium deposition (Figure 4E). Strikingly, Alizarin red and human CART staining of male TAA samples label overlapping areas implying the link between human CART and medial calcification (Figures 3B and 4E). Additionally, to confirm the calcification in the male TAA samples, we performed Von Kossa stainings, which labels mineralization and calcification in tissues.⁵⁰ This staining further validated the variable levels of mineralization/calcification in the aortic media of male TAA samples compared with the female TAA samples (Figure S3B).

DISCUSSION

This study demonstrates the feasibility of single-cell transcriptome imaging in uncovering aortic heterogeneity, leading to the discovery of a spatially-distinct CARTPT-expressing SMC subtype in male samples likely critical for medial calcification of the thoracic aorta. Unusually high extracellular matrix composition of the human aorta prevents reliable tissue dissociation. Suboptimal cell dissociation can induce gene expression changes in individual cells and alter compositional profiles of different cell populations.⁷ Although currently limited by the throughput, here we show that ISH-based spatial profiling strategies including MERFISH represent a powerful alternative to identify tissue-resident cell types in the human aorta and reveal their heterogeneity.

TAAs are more common in men but TAA growth and outcomes are worse in women for unknown reasons.^51,52 Similarly, abdominal aortic aneurysm (AAA) is less frequent in female population but grows more aggressively than male AAA with a higher risk of aortic rupture.^53,54 Our data reveals a robust enrichment of CARTPT expression in the male samples (Figures 2C, 2D and 3B), which we confirmed using other single-cell and bulk human aorta gene expression datasets (Figures 2F, 3D and S2B). We propose that CART signaling promotes osteochondrogenic switch of aortic SMCs in male TAA tissue, eventually leading to medial calcification (Figures 4A–4E and S3B). Although thoracic and abdominal aortic calcification have been associated with higher cardiovascular and overall mortality, it is not clear whether calcification directly contributes to these adverse outcomes or is a consequence of underlying cardiovascular complications.⁵⁵ In fact, some studies suggest that aortic calcification stabilizes the aortic wall and slows down aortic aneurysm expansion, likely due to arterial stiffening.^56,57 It remains to be determined whether CART signaling induced osteochondrogenic differentiation contributes to TAA formation or is protective against aggressive aortic dilation, dissection and rupture in male TAA patients.

In several vascular disease models including aortic aneurysms, a subset of preexisting SMCs can change their identity and undergo clonal expansion.⁵⁸ Osteoblast-like SMC differentiation has also been documented in the aging aorta.^59–62 NR4A3-enriched SMC3 are positioned near the adventitia, while CARTPT-enriched SMC4 are more abundant in the inner media suggesting that SMC3 and SMC4 likely originate from different aortic cell pools (Figure 3A). Our MERSCOPE data and the analysis of a previous human TAA scRNA-seq study demonstrate the emergence and expansion of CARTPT+ SMCs in male TAA tissue; however, the in vivo molecular triggers underlying this sex discrepancy are unknown. Although genetic predisposition plays a dominant role in TAA compared with AAA, a significant portion of TAA patients do not have a family history of the disease.⁸ Causes of non-familial TAA remain to be determined.⁶³ Histologically determined atherosclerosis is observed in sporadic TAA cases and is associated with increased aortic media degeneration.⁶⁴ Our data suggests that oxLDL can induce CARTPT expression implying an atherosclerotic process in CARTPT regulation (Figure 4A). However, this finding does not suggest a link between atherosclerosis and male TAA as we currently do not know whether CARTPT expressing SMCs contribute to TAA formation.

A limitation of MERSCOPE profiling approach is the number of genes that can be assayed. One hundred forty genes are undoubtedly insufficient to capture aortic heterogeneity, and inclusion of additional genes will likely reveal the full complexity of end-stage TAA tissue using the method described here. In addition, we were limited by our sample size. The analysis of the large ADEIP database shows CARTPT expression in a small number of aged female samples, albeit with lower levels and frequency compared with aged male samples (Figure 3D). Inclusion of female TAA subjects with medial calcification and male TAA subjects without medial calcification could also shed light on sex differences in CARTPT regulation. Despite these limitations, our study is a first step towards a high-resolution spatial cell ontology of the human aorta, and will be instrumental in defining actionable cell types and signaling pathways in human TAA tissue.

HIGHLIGHTS.

• Single-molecule spatial transcriptomics is a powerful method to begin to characterize tissue-resident human aortic cell heterogeneity.

• We identified a spatially-distinct CARTPT-expressing smooth muscle cell subtype predominantly found in male Thoracic aortic aneurysm tissue.

• Functional studies suggest that human CART promotes osteochondrogenic differentiation of aortic smooth muscle cells, potentially leading to medial calcification of the thoracic aorta.

NON-STANDARD ABBREVIATIONS AND ACRONYMS

MERFISH

Multiplexed error-robust fluorescence in situ hybridization

TAA

Thoracic Aortic Aneurysm

SMC

Smooth muscle cell

CARTPT

CART prepropeptide