A cell atlas of thoracic aortic perivascular adipose tissue: a focus on mechanotransducers

Abstract

Perivascular adipose tissue (PVAT) is increasingly recognized for its function in mechanotransduction. However, major gaps remain in our understanding of the cells present in PVAT, as well as how different cells contribute to mechanotransduction. We hypothesized that snRNA-seq would reveal the expression of mechanotransducers, and test one (PIEZO1) to illustrate the expression and functional agreement between single-nuclei RNA sequencing (snRNA-seq) and physiological measurements. To contrast two brown tissues, subscapular brown adipose tissue (BAT) was also examined. We used snRNA-seq of the thoracic aorta PVAT (taPVAT) and BAT from male Dahl salt-sensitive (Dahl SS) rats to investigate cell-specific expression mechanotransducers. Localization and function of the mechanostransducer PIEZO1 were further examined using immunohistochemistry (IHC) and RNAscope, as well as pharmacological antagonism. Approximately 30,000 nuclei from taPVAT and BAT each were characterized by snRNA-seq, identifying eight major cell types expected and one unexpected (nuclei with oligodendrocyte marker genes). Cell-specific differential gene expression analysis between taPVAT and BAT identified up to 511 genes (adipocytes) with many (≥20%) being unique to individual cell types. Piezo1 was the most highly, widely expressed mechanotransducer. The presence of PIEZO1 in the PVAT but not the adventitia was confirmed by RNAscope and IHC in male and female rats. Importantly, antagonism of PIEZO1 by GsMTX4 impaired the PVAT’s ability to hold tension. Collectively, the cell compositions of taPVAT and BAT are highly similar, and PIEZO1 is likely a mechanotransducer in taPVAT.

NEW & NOTEWORTHY This study describes the atlas of cells in the thoracic aorta perivascular adipose tissue (taPVAT) of the Dahl-SS rat, an important hypertension model. We show that mechanotransducers are widely expressed in these cells. Moreover, PIEZO1 expression is shown to be restricted to the taPVAT and is functionally implicated in stress relaxation. These data will serve as the foundation for future studies investigating the role of taPVAT in this model of hypertensive disease.

INTRODUCTION

Perivascular adipose tissue (PVAT) is a complex and understudied tissue that surrounds almost all blood vessels in the human body. A seminal finding in 1991 by Soltis and Cassis (1) changed the view that PVAT served simply as structural support for the vasculature. Specifically, PVAT was recognized to uptake norepinephrine (2). Further studies identified that PVAT produces contractile and relaxant substances that affect vascular tone, important findings given that this function changes in cardiovascular disease (3). However, this view of PVAT as solely important for secretion of vasoactive factors is limited. For example, PVAT assists in stress relaxation, a form of mechanical support to the artery (4). Here, mechanical support of the artery is considered a new function of PVAT. However, the mechanisms of stress relaxation, including potential mechanotransducers involved and their expression in the constituent cells of PVAT, are not known. As illustrated by this example, there are significant gaps in our understanding of the tissue’s cellular makeup and molecular mechanisms. We are committed to the discovery and understanding of new functions of PVAT, such as mechanotransduction. More specifically, here we interrogate PVAT for cellular contributions of mechanotransduction, an action informed by the creation of a cell atlas of the Dahl salt-sensitive (Dahl SS) rat thoracic aortic PVAT (taPVAT).

This study aims to fill in those gaps by building a comprehensive single-cell atlas of taPVAT using single-nuclei RNA sequencing (snRNA-seq) to investigate the cellular distribution of mechanotransducers in PVAT, and contrast it to subscapular brown adipose tissue (BAT), a non-PVAT adipose tissue of the same type (5). We focus our study on PVAT from the thoracic aorta because the cellular constituents and type of adipocytes in PVAT are location dependent (6). Rat taPVAT most resembles brown adipose tissue. Abdominal aorta PVAT resembles both brown and white adipose tissue and mesenteric PVAT mostly resembles white adipose tissue (5, 7, 8). Bulk microarray studies have found few differences in global gene expression profiles between thoracic PVAT and interscapular BAT in mouse (5). However, this difference has not been examined at the single-cell level, which can reveal cell-to-cell heterogeneity commonly masked by bulk gene expression measures. Far less work such as this has been done in rat models compared with the mouse.

The thoracic aorta does not share its PVAT with other vessels, as occurs in abdominal vessels, such that conclusions can be made with respect to aortic function confidently. The taPVAT also consists of an anterior strip and two lateral strips (right and left) that are attached to the spine. The lateral and anterior strips emerge from different developmental origin in the mouse (9). The present snRNA-seq experiments allow for construction of a cell atlas, giving insight to the cells that populate this complex tissue. Moreover, we can compare findings from taPVAT to that of BAT from the same rat to determine whether PVAT possesses cell types/genes different from those in BAT. If so, this knowledge could be used to consider potential PVAT-specific therapeutic interventions. Our work was carried out in tissues from the Dahl-SS rat, a strain which is particularly important in the study of hypertensive cardiovascular disease. We identify cell types within taPVAT that express mechanotransducers, and whether one of them, highly expressed Piezo1, participates functionally in mechanotransduction of taPVAT.

MATERIALS AND METHODS

Animal Models and Tissue Collection

Dahl-SS male rats were obtained from a colony at Michigan State University (MSU) bred from parents purchased from Charles River Laboratory (snRNA-seq experiments) or purchased from Charles River (RNAscope, immunohistochemistry, and isometric contraction analyses). Animals were on a normal diet (Teklad 22/5 Rodent diet; Madison, WI) and housed in a 12-h:12-h light/dark cycle at 21°C–23°C. Food and drinking water were available ad libitum. Procedures using animals complied with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals (2011) and were approved by the MSU Institutional Animal Care and Use Committee (PROTO202000009). This study was conducted with Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines (essential 10 and recommended) in mind (10).

Before tissue removal, rats were given pentobarbital as a deep anesthetic (80 mg·kg⁻¹ ip). A bilateral pneumothorax was created before vessel dissection. Tissues for snRNA-seq experiments [thoracic aorta PVAT (taPVAT) and subscapular brown adipose tissue (BAT)] were dissected as whole pieces from each animal. For taPVAT, the anterior strip of taPVAT (not attached to spine) was carefully dissected by cutting at the edges on each side by the same person for all replicates to maintain consistency. BAT was removed between the scapula. The taPVAT was dissected from the vessel while BAT was cleaned of adherent white fat (subcutaneous) under a stereomicroscope and in a Silastic-coated dish filled with physiological salt solution (PSS), containing (in mM) 130 NaCl, 4.7 KCl, 1.18 KH₂PO₄, 1.17 MgSO₄·7H₂O, 14.8 NaHCO₃, 5.5 dextrose, 0.03 CaNa₂EDTA, and 1.6 CaCl₂ 1.6 (pH 7.2). A 50 mg piece of each taPVAT and BAT was minced before snap freezing in liquid nitrogen and storage at –80°C. For RNAscope experiments, aorta with complete PVAT and liver were cleaned of blood and formalin (10%)-fixed for 16–32 h, then embedded in paraffin blocks. These same tissues were used for fluorescent microscopy. For isometric contractility experiments, the thoracic aorta was dissected from the aortic arch to the diaphragm and placed in PSS. Chemical sources and catalog numbers are provided in a table of key resources (Supplemental Table S1; all Supplemental materials may be found at https://doi.org/10.6084/m9.figshare.c.7123246.v1).

Nuclei Isolation and Single-Nuclei RNA Sequencing

Nuclei from frozen minced aortic perivascular adipose tissue or subscapular brown adipose tissue (∼ 50 mg) from two rats each were isolated with the Chromium Nuclei Isolation Kit (10× Genomics) according to the manufacturer’s recommended protocol (CG000505, RevA), with a final resuspension in a 50 µL wash and resuspension buffer. Following isolation, nuclei concentration was determined with a Nexcelom Cellometer Vision and Viastain acridine orange and propidium iodide (AOPI) staining solution. The 10× Chromium Next GEM Single Cell 3′ Reagent Kit v 3.1 (dual index) was used to generate libraries representing one rat sample each with a targeted cell recovery of 10,000 according to the manufacturer’s recommended protocol (CG000315, Rev D). Following gel bead-in-emulsion (GEM) generation, barcoding, cleanup, and amplification, the generated cDNA was quantified and purity determined with an Agilent 4200 TapeStation using a D5000 ScreenTape assay. To generate the single nuclei 3′ gene expression library, 25% of the total cDNA obtained underwent fragmentation, end repair, addition of 3′ nontemplated nucleotide (A-tailing), and adaptor ligation, followed by cleanup with SPRIselect reagent. Individual 10× sample index names from the Dual Index Plate TT Set A were recorded and added to the appropriate sample and then amplified according to the manufacturer’s protocol. A final cleanup with SPRIselect reagent was done, with samples eluted into a 35.5 µL buffer EB. The average fragment size following library construction was determined with an Agilent 4200 TapeStation using a D1000 ScreenTape assay.

Library Quantification

Quantification of library samples was done with the KAPA Library Quantification Kit (Roche, Indianapolis, IN). Library samples were diluted (1:100, 1:1,000, 1:5,000, and 1:10,000) in DNA dilution buffer, containing 10 mM Tris·HCl (pH, 8.0 − 8.5) and 0.05% Tween 20. In triplicate, 4 µL of each diluted sample and kit-supplied DNA Standards were pipetted into a standard 96-well PCR plate containing 6 µL KAPA SYBR FAST qPCR Master Mix. No Template Controls containing 4 µL water were run in duplicate. An adhesive film to prevent sample evaporation was placed over the top of the plate and the plate was centrifuged at 1,000 rpm for 1 min. The following conditions were run in a QuantStudio 6 Flex Real-Time PCR System: 1 cycle of 95°C 5 min; 35 cycles of 95°C 30 s, 60°C 45 s. A standard melt curve was run to ensure adapter dimers were not present. Calculations were performed in the KAPA Library Quantification Data Analysis Template to determine the undiluted library concentration. Libraries were submitted to Novogene for 150 bp paired-end sequencing on a NovaSeq6000 at a target depth of 50,000 reads/nuclei.

Quality Control and Preprocessing

Initial read quality control was performed using FastQC v0.11.7 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were aligned to the rat reference genome Rnor_6.0 (assembly GCA_000001895.4) using the 10× CellRanger v7.1.0 pipeline (11). Scanpy (12) and other commonly used Python packages (i.e., numpy, pandas, matplotlib) were then used to do additional quality control, preprocessing, analyzing, and visualization of the data. Any nuclei that had less than 200 detected genes or genes that were detected in less than three cells were discarded. Cells that were found to be expressing greater than 10% mitochondrial genes were also discarded. Ambient RNA and doublet removal were done using SoupX (13) and scDblFinder (14), respectively. To reduce technical variation between samples, batch correction with the Python package single-cell variational inference (scVI) was performed (15). Principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) were used to project the high-dimensional data into a two-dimensional (2-D) space. After dimensionality reduction, Leiden unsupervised clustering method was used to group the cells into distinct subpopulations based on their gene expression patterns.

Marker Identification, Cell Annotation, and Differential Expression

Leiden clustering was performed at resolutions of 0.1 to 1.5 to identify clearly distinct clusters resulting in a total of 20 clusters at a resolution of 1.0. To support cluster annotation, cell type marker lists were obtained from previously published human and mouse white adipose tissue snRNA-seq (16) and human brown adipose tissue scRNA-seq (17). The marker lists were sorted by fold-change and the top 50 (or all markers when fewer than 50) were used to calculate gene module scores with a control size matching the marker length and a total of 500 bins. In addition, the Wilcoxon Rank Sum test was used to identify the top markers sorted by a fold-change relative to all other clusters combined. Notably, markers identified using this approach are considered only within this data set and may identify markers for different cell types present in other tissues. Collectively, these results were used to manually annotate the clusters based on internal expertise as well as cross-referencing to literature. To simplify cell annotation and differential expression, the largest cluster likely representing subtypes or different cell states of adipocytes was collapsed into an overarching adipocyte annotation. Reclustering of only adipocyte stem and progenitor cells (ASPCs; Dcn, Fbn1, Pdgfra) was performed to highlight the differences within the ASPC group.

Recent studies suggest that the use of pseudobulk and pseudoreplicates for differential expression analysis of single-cell transcriptomic data can reduce false discoveries (18, 19). A total of three pseudoreplicates for each adipose tissue were randomly sampled from the integrated data and aggregated into individual pseudobulk samples. PyDESeq2 (20) was used to perform differential expression analysis using tissue as an experimental design factor for each cell type independently. Genes were considered differentially expressed when the adjusted P value ≤ 0.05.

Gene Set Enrichment Analyses

Rat-specific genes implicated in mechanotransduction were identified in the Gene Ontology (GO) web tool AmiGO (https://amigo.geneontology.org/amigo) by looking up known mechanotransducers and related terms. A total of three GO terms were identified: GO:0008381 Mechanosensitive monoatomic ion channel activity, GO:0140135 Mechanosensitive monoatomic cation channel activity, and GO:0050982 Detection of mechanical stimulus. The terms represent evolving systematically identified genes associated with a range of levels of evidence from experimentally inferred to structural similarity (21). Exact definitions for each of these terms at the time of writing are included in Supplemental Table S2 and span channels that respond to mechanical stress to events implicated in the detection of mechanical stimuli. When gene sets were analyzed collectively, known mechanotransduction genes Ddr2 and Itgb1 were also included as they were not present in these lists. Scores for each of these ontologies were calculated as described in the cluster annotation method and shown as dot plots. Gene set enrichment analysis was performed on preranked genes for each cell types, ranked according to expression level, and then analyzed using GSEAPy (22).

RNAscope

Sections (5-μm thick) from formalin-fixed, paraffin-embedded thoracic aorta + PVAT were cut by the MSU Investigative Histopathology laboratory, mounted on Superfrost Plus microscope slides (Cat. No. 12-550-15, Thermo Scientific, Waltham, MA), and stored at 4°C after air drying. To deparaffinize, slides were baked 1 h in a 60°C oven and then treated by a series of washes in xylene and 100% ethanol. Slides were incubated for 10 min in hydrogen peroxide (proprietary grade) to permeabilize. RNA retrieval was performed first by boiling slides in the Target Retrieval reagent (Cat. No. 322000; Advanced Cell Diagnostics, Hayward, CA) for 8 min and then by incubating slides with Protease Plus solution (Cat. No. 322330, Advanced Cell Diagnostics) for 15 min. For in situ detection of Piezo1 mRNA, the RNAscope 2.5 HD Assay, RED kit (Cat. No. 322350, Advanced Cell Diagnostics) was used according to the manufacturer’s protocol. Slides were incubated with a probe for rat Piezo1 (Cat. No. 587591, Advanced Cell Diagnostics), the housekeeping gene peptidylprolyl isomerase B (Ppib) for a positive control for mRNA detection, or the bacterial gene diaminopimelate (DapB) for an mRNA species that should not be present as a negative control (Cat. Nos. 313921 and 310043, respectively; Advanced Cell Diagnostics) for 2 h at 40°C. To amplify the probe signal, a series of amplification reagents was run (Amp 1: 30 min at 40°C, Amp 2: 15 min at 40°C, Amp 3: 30 min at 40°C, Amp 4: 15 min at 40°C, Amp 5: 30 min at RT, Amp 6: 15 min at RT, “Red” A + B reagent: 10 min at RT; Cat. No. 322360, Advanced Cell Diagnostics). Slides were counterstained with hematoxylin for 2 min and 0.02% ammonia water for 10 s at room temperature then mounted with EcoMount (Cat. No. EM897L, BioCare Medical, Pacheco, CA) and a glass coverslip. Images were taken with a Nikon Digital Sight DS-Qil camera and Nikon NIS Elements BR 4.6 software on a Nikon TE2000 inverted microscope. All background correction was applied to the full image and was consistent across positive and negative controls and images were brightened or contrasted as a whole, never in part. Positive staining was determined by red punctate dots in the cell.

Immunofluorescence

Sections (5-µm thick) from formalin-fixed, paraffin-embedded Dahl-SS rat thoracic aorta + PVAT and kidney (positive control for PIEZO1) were cut by the MSU Investigative Histopathology laboratory, mounted on Superfrost Plus microscope slides (Thermo Scientific, Cat. No. 12-550-15) and were air-dried and stored at room temperature (RT). To deparaffinize, slides were washed two times with Histochoice Clearing Agent (VWR, Cat. No. H103) and four times with isopropanol (VWR, Cat. No. 9084-03) and two times with distilled water for 3 min each wash. Antigen retrieval was performed by boiling slides for 30 s in Antigen Unmasking Solution (Vector, Cat. No. H3301). Slides were rinsed in distilled water and air-dried. To contain the blocking serum, primary and secondary antibody solutions, circles were drawn around the sections with an ImmEdge Hydrophobic pen (Vector, Cat. No. H-4000). Slides were incubated at RT with 1.5% normal goat serum (Vector, Cat. No. S-1000) in Dulbecco’s phosphate-buffered saline (PBS) (Sigma-Aldrich, Cat. No. D8537) blocking solution (BS) for 1 h. Positive control sections were incubated with 1:600 primary antibody anti-Piezo1 (Alomone, Cat. No. APC-087) in 1.5% normal goat serum BS, and negative control sections were incubated with BS overnight at 4°C. The primary antibody and BS were removed from the sections and the slides were rinsed in Dulbecco’s PBS three times for 5 min each rinse. The slides were incubated in 1:1,000 secondary antibody AlexaFluor 488 (Invitrogen, ThermoFisher, Cat. No. A11008) for 1 h at RT. The secondary antibody was removed, and slides were rinsed three times with Dulbecco’s PBS for 5 min and allowed to dry thoroughly at RT. Vectashield with DAPI (Vector, Cat. No. H-1500) was applied to the sections and coverslips were mounted to the slides. The slides were allowed to dry and harden at RT. Images were acquired at 360, 488, and 544 nm on a Nikon Eclipse Ti-S microscope, using a ×10 objective, Nikon DS-Qi1MC camera, and NIS elements BR 4.6 software.

Isometric Contractility

For creating separated rings of the thoracic aorta and its surrounding PVAT, a section of the thoracic aorta + PVAT stood on its end in the silastic dish filled with PSS. Two insect pins were inserted into the lumen of the aorta to hold the vessel open. While gently holding the PVAT layer away from the adventitia with fine forceps, small vannas scissors were used to sever connections around the circumference of the vessel between the PVAT and vessel/adventitia. The ring was flipped vertically, and this process was repeated. The separated ring of PVAT was then lifted off the vessel. All tissues [PVAT alone or aorta alone] were placed onto two L-shaped stainless-steel rings. Rings were mounted in warmed (37°C) and aerated (95% O₂-5% CO₂) Radnoti tissue baths (10 mL volume) on Grass isometric transducers (FT03; Grass Instruments, Quincy, MA) connected to an eight-channel PowerLab C through an Octet Bridge (ADInstruments, Colorado Springs, CO). Sample type (aortic or PVAT ring) and exposure to vehicle or inhibitor were randomized daily in one of eight different tissue baths. Care was taken to apply no tension on the tissue before initiation of the experiment.

All rings started at a tension of 0 g, equilibrating in the warm buffer for 1 h before experimentation. During this hour, the buffer was exchanged every 15 min. To demonstrate similar potential, all rings were challenged before the addition of an inhibitor with a passive tension application of 2 g applied over ∼15 s, applied through the clockwise turning of a rack and pinion. Tissues were allowed to relax to this stretch for 30 min; the tension achieved at the end of this period was recorded. Tissues were then challenged with a maximum concentration of phenylephrine (PE, 10⁻⁵ M). This response plateaued and was recorded. Tissues were washed identically and repeatedly for 30 min to achieve a stable baseline. This first 2-g challenge was considered a viability test and the potential of the tissues dedicated to vehicle or GsMTx4 incubation was equal.

The same tissues were next incubated for 1 h with either vehicle (water) or the PIEZO1 antagonist GsMTx4 (Med Chem Express, Cat. No. HY-P1410; 5 mM). Another 2 g of passive tension (for a total of 4 g) was applied at the end of this hour, tissues were relaxed over 30 min, and the tension was relaxed to recorded. Tissues were challenged with PE (10⁻⁵ M) again and washed to baseline. A subset of tissues was then reincubated with vehicle/GsMTx4 for 45 min without washing. A 4-g passive tension was applied (total of 8 g). Thirty minutes, with no washing, were allotted for tissues to relax and tension was relaxed to recorded. Tissues were again challenged with PE. Tissues were weighed once the experiment was completed. LabChart 8.1.25 (ADInstruments, Colorado Springs, CO) was used to capture tissue tension/responses through a Mac Mini computer connected to a monitor. LabChart’s capture of actual tracings and quantitative responses are shown. The magnitude of the response relaxed to at the end of the 30 min period was recorded for the responses at 2 g applied, 4 g total applied, and 8 g total applied. Data are reported as means ± SE for the number of tissues reported. Data were graphed in GraphPad Prism 9. A one-way ANOVA was used at each magnitude of total stretch applied (2, 4, 8) to determine statistical differences between groups. This test, using Bartletts, also verified statistically equivalent variances in groups. P value < 0.05 indicates significant differences identified by Tukey’s post hoc test.

RESULTS

Characterization of Adipose Tissue Cell Populations

To characterize the cell types of the anterior taPVAT and BAT, we used single-nuclei RNA sequencing (snRNA-seq) as outlined in Fig. 1 (see materials and methods). A total of 29,703 (taPVAT) and 28,387 (BAT) nuclei passed quality control and ambient RNA removal using SoupX (13) for taPVAT and BAT, respectively (Fig. 2, A and B). The median number of unique genes and unique molecular identifiers (transcripts), respectively, identified in each nucleus were 1,393 and 2,029 for taPVAT and 1,638 and 2,398 for BAT. Leiden clustering found 20 clusters (Supplemental Fig. S1) that were manually annotated based on marker genes and comparison with previously published adipose tissue data sets (16, 17, 23, 24) identifying eight major cell types; each represented in similar proportions except for mesothelial cells primarily being present in taPVAT (0.4%, 119 nuclei) compared with BAT (0.004%; 1 nucleus) (Fig. 2C).

Figure 1.

Procedure overview used to derive transcript data to inform cell-based clusters. A: nuclear isolation from thoracic aorta perivascular adipose tissue (taPVAT) and brown adipose tissue (BAT) from male Dahl salt-sensitive (Dahl SS) rat. B: use of Chromium Next GEM Single Cell 3′ kit to produce amplified cDNA. C: construction of a 3′ gene expression library. D: sequencing and analyses of cell-based clusters (see materials and methods for more details). All panels of the workflow overview figure were created using a licensed version of BioRender.com, and some parts of the diagram were modified from the 10× Chromium Nuclei Isolation Kit and Chromium Next GEM Single Cell 3′ Kits user guide.

Figure 2.

Characterization of the male Dahl salt-sensitive (Dahl SS) rat thoracic aorta perivascular adipose tissue (taPVAT) and brown adipose tissue (BAT) adipose tissue depots. A: uniform manifold approximation and projection (UMAP) visualization of nuclei from the taPVAT and BAT following integration using single-cell variational inference (scVI) and manual annotation of cell types. B: UMAP visualization for nuclei from each individual adipose tissue depot. C: relative proportion of each cell type for each adipose tissue depot. Proportions are shown on a log scale for comparison of high-abundance and low-abundance cell types. Bars represent mean values, whereas individual points represent proportions in individual samples. D: top 5 marker genes for each cell type (listed at top) identified common to both taPVAT and BAT.

The largest clusters (Leiden clusters: 0–5, 8–3, and 17) were collectively annotated as adipocytes for broad comparison, though clear subtypes are present (e.g., Acaca- and Acly-enriched cluster; Supplemental Fig. S2). As expected, adipocytes represent ≥ 80% of captured nuclei in both taPVAT and BAT (Fig. 2C). The remaining clusters were annotated according to the highest gene module score for previously identified marker genes in mouse and human white adipose tissue (16) or human BAT (17), as well as examination of the top marker genes in each Leiden cluster (Fig. 2D and Supplemental Fig. S1). Endothelial cells (ECs) clearly resembled those of other published data sets and were marked by Rasip1, which is involved in the formation of EC junctions (25). The immune cell cluster is represented by markers of several immune cell types including monocytes, natural killer (NK) cells, T cells, and B cells (Supplemental Figs. S1 and S3). Smooth muscle cells (SMCs) and pericytes clustered together, marked by Myh11, which was previously used to trace the differentiation of SMCs and SMC-like cells into beige adipocytes (26). In agreement, SMCs and pericytes clustered closely to ASPCs. One population, mesothelial cells, were only identified in taPVAT. Unlike BAT, as a visceral tissue taPVAT has a clear mesothelial layer, which is difficult to omit during tissue collection. A distinct population of adipocytes was also identified expressing several signaling markers often found in neural cells (e.g., Ptprs, Kcnn3, and Ca3) but clearly resembling adipocytes. Another cluster also sharing several markers with neural cells such as oligodendrocytes and Schwann cells (Zfp536 and Sox10) could not be found in other published adipose tissue data sets.

BAT and taPVAT Cell Types Are Highly Similar but Exhibit Tissue Depot-Specific Gene Expression

To examine whether individual cell types of taPVAT closely resemble the cell types in BAT, differential expression (DE) analysis was performed (see materials and methods). Collectively, the number of DE genes ranged between 552 (adipocytes) and 16 (neuronal-like cells) across all cell types except for mesothelial cells, which could not be compared because of their absence in BAT (Fig. 3A). Most DE genes were cell type-specific from 372 in adipocytes to 39 in SMCs and pericytes while immune cells and neuronal-like cells had ≤11 cell type-specific DE genes. Consequently, despite the overall similarities between taPVAT and BAT adipocytes, there may be some subtle functional differences associated with their localization in the Dahl-SS rat. Closer examination of the five DE genes in each cell type with the largest fold-change for taPVAT enriched (induced) and BAT enriched (repressed) individually shows that the biggest differences are represented by only 47 genes (Fig. 3B). Among taPVAT-enriched genes are lipid metabolism-related genes including Scd and Acaca. Gene set enrichment analysis (GSEA) was used to determine the top 15 taPVAT and BAT cell-specific functions (Fig. 3C). Although BAT seems to be most enriched in transcriptional regulation, as well as cholesterol and carbohydrate metabolism, taPVAT was largely enriched in ion transport, signaling, and chemokine functions. These findings may point to taPVAT playing a greater role in signaling in response to cues from the thoracic aorta.

Figure 3.

Comparison of thoracic aorta perivascular adipose tissue (taPVAT) and brown adipose tissue (BAT) adipose tissue depot gene expression. A: UpSet plot of differentially expressed genes (|fold-change| ≥ 2, adjusted P value ≤ 0.05) for individual cell types between taPVAT and BAT. Total number of differentially expressed genes for each cell type is shown as horizontal bars on the left and the intersecting list of differentially expressed genes for the cell types markers by a black dot is shown as vertical bar. B: heatmap of the top 5 taPVAT and BAT-enriched genes. C: top 10 enriched functional groups determined using GSEApy (see materials and methods) where a positive normalized enrichment score (NES; top) represents taPVAT-enriched functions and negative NES (bottom) represents BAT-enriched functions.

PIEZO1 Is a Functional Mechanotransducer in taPVAT

We use snRNA-seq to characterize cell populations in two adipose tissue depots considered to be similar in phenotype (i.e., brown fat) but which are expected to be at least somewhat functionally distinct due to taPVAT being constantly exposed to mechanical forces from blood flow in the aorta, the largest conduit vessel. Therefore, we examined the expression of mechanotransduction genes in both adipose tissue depots (Fig. 4A). Rat genes functionally associated with mechanosensing were identified in gene ontology representing three broad categories: mechanosensitive ion channels, mechanosensitive cation channels, and genes implicated in sensing mechanical stimuli. Known but currently unassigned genes implicated in mechanosensing such as Ddr2 and Itgb1 were also included because of their actions as receptors for collagen. GSEA of each cell type found that mesothelial cells were enriched in genes implicated in the detection of mechanical stimuli while the adipocyte_2 cluster as well as endothelial cells were enriched in mechanosensitive ion/cation channels (Fig. 4B). Among the mechanosensing genes showing the largest fold-change between taPVAT and BAT were the acid-sensing ion channel Asic2 and mechanosensitive ion channel Piezo2 (Fig. 4C). However, both Asic2 and Piezo2 were lowly expressed across all clusters with Piezo2 not falling within the top expressed mechanotransduction-related genes (Fig. 4A). Conversely, some mechanosensing-related genes were expressed in specific cell types and differentially expressed between taPVAT and BAT including Ddr2 (ASPCs; BAT enriched), Cxcl12 (endothelial cells; taPVAT enriched), Ano1 (SMCs and pericytes; BAT enriched), and Fyn (ASPCs; BAT enriched) (Fig. 4, A and D). Two mechanosensing genes were highly expressed homogeneously. Piezo1 was found in all cell types and was highly expressed throughout the reported clusters. Similarly, Slc12a2 was highly expressed in all cell types except the adipocyte_2 population.

Figure 4.

Analysis of mechanotransduction-related gene expression. A: dot plot of the most highly expressed mechanotransduction genes in thoracic aorta perivascular adipose tissue (taPVAT) and brown adipose tissue (BAT). Dot size represents percentage of nuclei expressing the gene, and color intensity represents expression level. The mesothelium of BAT was removed for visualization as it only included 1 nucleus. B: cell-specific enrichment of mechanotransduction-related Gene Ontology (GO) terms (see materials and methods). C: uniform manifold approximation and projection (UMAP) visualization of expression for Ddr2, Slc12a2, Piezo1, and Piezo2. D: heatmap of fold-change for differentially expressed mechanotransduction-related genes (|fold-change| ≥ 2, adjusted P value ≤ 0.05).

We chose to follow the finding that Piezo1 was highly expressed in many clusters with the assumption that this relatively newly understood mechanotransducer could play a role in the process of stress relaxation. We first used RNAscope to validate the presence of Piezo1 mRNA in the PVAT (anterior and lateral) and minimal expression in the media of the isolated thoracic aorta from the male Dahl-SS rat (Fig. 5A). The same pattern of expression can be seen in female Dahl-SS rats (Supplemental Fig. S4). Similarly, PIEZO1 protein, identified immunohistochemically, was present in the taPVAT (anterior and lateral) but not media of the thoracic aorta; these sections were taken from the same aorta as was used for RNAscope experiments (Fig. 5B). Collectively, RNAscope and immunohistochemistry agree on the spatially resolved expression of PIEZO1. The physiological/functional importance of PIEZO1 was studied in the isolated PVAT ring and aortic ring from the Dahl-SS male rat, measuring isometric tension as the primary outcome. In Fig. 5C, the first challenge of 2 g is labeled “no inhibitor” as neither vehicle nor inhibitor was present for this challenge. The finding that, at this initial 2 g stretch, both stress relaxation and PE-induced contraction were similar in the tissues destined for exposure to vehicle or GsMTx4 underscores that the potential of these two groups of tissues was similar in the endpoint measured. This step allows us to have confidence in the finding that the PIEZO1 antagonist GsMTx4 caused a profound loss of ability of the tissue to hold tension during stress relaxation in the isolated PVAT but not in the aorta (Fig. 5C). Collectively, these findings are consistent with the snRNA-seq data in locating Piezo1 mRNA to taPVAT, with PIEZO1 playing a functional role mediating the response to the mechanical intervention of stretch.

Figure 5.

The mechanotransducer Piezo1 is expressed in and functionally serves thoracic aorta perivascular adipose tissue (taPVAT). A: Brightfield of Piezo1 mRNA ZZ-probe detection (RNAscope) in the thoracic aorta of the Dahl salt-sensitive (Dahl SS) male rat. Arrows indicate positive dot detection. P, PVAT (anterior); A, adventitia; M, media. Horizontal bar indicates 50 µm. Photomicrograph represents a ×20 magnification (1st full image) overview of the aorta and taPVAT. Box drawn around a region of interest shows ×40 magnification (2nd full image). To the right are tissue-specific positive (+) controls for housekeeping gene peptidylprolyl isomerase B (PPIB) and negative (−) controls for bacterial gene diaminopimelate (DapB) are shown. Images represent 3 male rats. B: fluorescent immunohistochemical detection of PVAT. Images of sections incubated with (1st full image) and without primary (2nd full image) PIEZO1 antibody. P, PVAT (anterior); M, media; L, lumen; E, endothelium. Kidney nephron tubule epithelium is shown on the far right stained for PIEZO1 in the presence (top) and absence (bottom) of PIEZO1 primary antibody. Representative of 3 male rats. C, left: representative tracing of the response of the aorta (top) or its surrounding ring of PVAT (bottom) to a 4-g passive tension addition in the absence (black) or presence of PIEZO1 inhibitor GsMTx4 (5 µM). C, right: quantifies tension relaxed to after application of a baseline tension of 2 g (no inhibitors present but legend indicates which group tissues will be in); after addition of 2 g tension to achieve 4 g total (after 1-h incubation with vehicle/inhibitor); and after addition of 4 g tension to achieve 8 g total (after reincubation with vehicle/inhibitor). Measurements were repeated on the same tissue without the inhibitor and after addition 4 and 8 g of tension. Bars are means ± SE with individual scattered values. *P < 0.05, statistically significant differences within group members as marked by lines.

ASPC Subpopulations Express Different Levels of Piezo1

Recent work has implicated PIEZO1 in the adipogenesis potential of ASPCs (27). Moreover, ASPCs consist of multiple subpopulations with distinct functional roles (28–30). As expected, these cells expressed elevated levels of common markers including Dcn, Fbn1, Cd34, and Pdgfra (Fig. 6A). We reintegrated and clustered the ASPCs, which were represented by 744 nuclei in taPVAT and 770 nuclei in BAT, resulting in three Leiden clusters (Fig. 6B). Among the top markers was Pi16 (cluster 1) previously found to represent a nonproliferating population (29), whereas Bmper, elevated in cells which differentiate into brown adipocytes (29), was high in cluster 2. The proportion of cells did not widely differ between taPVAT and BAT (Supplemental Fig. S5). More importantly, when evaluating the expression of Piezo1 in these subpopulations, it was found to be primarily expressed in Pi16^high ASPCs and much lower in Bmper^high nuclei (Fig. 6D).

Figure 6.

Adipocyte stem and progenitor cells subtypes in perivascular adipose tissue (PVAT) from Dahl salt-sensitive (Dahl SS). A: expression dot plot of ASPC marker genes Dcn, Fbn1, Cd34, and Pdgfra in individual cell types for thoracic aorta perivascular adipose tissue (taPVAT) and brown adipose tissue (BAT) combined. Dot size represents percentage of genes expressing the gene, whereas color represents mean expression. B: uniform manifold approximation and projection (UMAP) visualization of reintegrated adipocyte stem and progenitor cells (ASPCs) as described in materials and methods. Nuclei were reclustered using Leiden clustering at a resolution of 0.1, identifying a total of 3 subpopulations. C: top 5 markers genes for each ASPC Leiden cluster with a minimum |fold-change| of 2. Only 2 genes met the threshold criteria for cluster 0. D: median expression and distribution of expression for previously identified markers Bmper, Pi16, and Gdf10 are shown as violin plots for each Leiden cluster identified in B. Piezo1 expression is also shown for each Leiden cluster.

DISCUSSION

Cell Atlas of the Rat taPVAT and BAT

To our knowledge, this work is the first cell atlas of a PVAT in rats, a species specifically important to the field of hypertension. Eight major cell types were represented in taPVAT: two adipocyte subtypes, endothelial cells, immune cells, and neuronal-like cells; SMCs and pericytes, mesothelial cells, and ASPCs. The presence of most cell types was expected, and that of immune cells is consistent with previous work that identified T cells, B cells, macrophages, neutrophils, and mast cells in taPVAT of the Dahl-SS male and female rats (31). Although this study did not break down the immune cell cluster for more comprehensive analysis, further subclustering (Supplemental Fig. S3) shows that monocytes, T cells, and B cells were all present in both taPVAT and BAT. Smooth muscle and endothelial cells are likely from the microvasculature of the taPVAT as we ensured that experimental samples contained no tunica media. In this way, neither smooth muscle cells nor endothelial cells from the inner tunicas would be present. Two cell type clusters could not be clearly matched to previously identified mouse and human data sets; adipocytes_2 and neuronal-like cells. The presence of neuronal cells in taPVAT is controversial; no study has demonstrated definitive innervation of adipocytes in PVAT. Analysis of single-cell thoracic aorta between E18 and P3 identified a population of Schwann cells marked by the expression of Mpz, which were not present in adult samples (28). Our data showed low expression of Mpz, but significant expression of several other markers was found in Schwann cells and oligodendrocytes (28, 32). Adipocytes_2 were notable in their expression of the ion channel Kcnn3, which may be reflective of stem cell-like characteristics (33).

Similarities of taPVAT and BAT

Previous bulk gene expression analysis of taPVAT and BAT in mice has shown that both brown adipose tissue depots are highly similar (5). Consistent with this, our data identified genes enriched in PVAT, Cfh, and C7, both immunomodulatory genes. Similarly, the BAT-enriched gene Tbx15, a putative master regulator of adipose tissue genes (34), was identified. However, the total number of DE genes was comparatively greater in our study. This might be attributed to the difference in species, the use of next-generation sequencing technology, and/or the ability to examine cell-specific differences. Nevertheless, snRNA-seq of the two brown adipose tissue depots in Dahl-SS rats show strong similarities in cellular composition and overall gene expression. Interestingly, both taPVAT and BAT expressed many of the same mechanotransducers at comparable levels including Piezo1 and Slc12a2 (aka NKCC1). It was not the goal of this work to understand differences in mechanotransducer expression between BAT and taPVAT. Nonetheless, it is notable that BAT expressed many of the same mechanotransducer transcripts found in taPVAT. It remains unclear why BAT would express mechanotransducers, though evidence points to a link between the stiffness of brown adipocytes and their thermogenic function (35).

Dahl-SS Rat Is an Important Model for Cardiovascular Disease

The Dahl-SS rat has been a mainstay model in hypertension research since its original discovery by Dahl et al. (36). This rat strain, fed a high (4% or above) salt diet, develops elevated blood pressure. The Medical College of Wisconsin (MCW) has led important efforts to understand this rat genetically in the hopes of identifying a gene/genes responsible for hypertension (37). Importantly, both researchers at MCW and MSU find that a high-fat (HF, 60% kCal lard) diet, independent of salt, creates a hypertension in the Dahl SS when fed the diet from weaning (38, 39). The HF diet-fed hypertension occurs in both males and females, making this a valuable model. It is for these reasons that the Dahl-SS rat was our choice of model in the present study. Our work is also important because a vast majority of cell work in adipose tissues (of all kinds) has been done in the mouse and human but not the rat. This work thus lays a foundational study for an adipose tissue cell atlas in a rat.

Piezo1 as a Functional Mechanotransducer in PVAT

SnRNA-seq identified multiple recognized mechanotransducers in taPVAT of the male Dahl-SS rat. Although not exhaustive, major classes of mechanotransducers were interrogated. We focused on membrane-limited mechanotransducers (ion, cation channels). Of these gene ontologies, Piezo1 (40) was the most highly and homogeneously expressed in cells populating taPVAT followed by Slc12a2. We focused on PIEZO1 because of increasing evidence of its role in taPVAT mechanostransduction, though future studies should investigate the role of Slc12a2. Piezo1 mRNA expression was observed in taPVAT, as was PIEZO1 protein. Importantly, inhibition of PIEZO1 function by GsMTx4 (41) caused a loss of the ability of taPVAT to maintain tone after a stretch challenge. GsMTx4 differs from Dooku1, a more widely used antagonist, in that GsMTx4 can inhibit the endogenous function of PIEZO1. By contrast, Dooku1 inhibits those actions stimulated by the agonist Yoda1 (42). GsMTx4 also has the potential to inhibit transient receptor potential (TRP) channels, but these were largely lowly expressed or not detected suggesting limited activity in taPVAT (Supplemental Fig. S6). Notably, neither RNAscope nor immunohistochemistry detected Piezo1 mRNA or protein, respectively, in the media of the same vessel. This lack of measurable Piezo1 explains the overall lack of effect of GsMTx4 on the tone of the isolated aorta.

PVAT thus becomes important as a site for PIEZO1’s role in aortic function. This includes the clinical measure of aortic stiffness as measured by pulse wave velocity, with elevated aortic stiffness serving as an independent risk factor for cardiovascular disease (43). Precisely how Piezo1 affects the maintenance of tone is not yet understood but suggests that this is an active process dependent on cellular involvement, and thus involving more than noncellular elements (e.g., collagen). Similarly, the functional role of PIEZO1 in BAT is unknown but is a provocative finding. We recognize that we have not delved into the many other mechanotransducers that could function in PVAT, including the focal adhesion kinases, cytoskeletal proteins, and other adhesion receptors (44). Specifically, the present data support future investigation of Slc12a2 (Na⁺-K⁺-2Cl⁻ transporter, NKCC1) in PVAT. Nonetheless, using Piezo1, the present work exemplifies a way in which snRNA-seq and physiological/pharmacological studies can be complementary.

Consistent with previous studies in other adipose tissue depots identifying ASPC subpopulations (28–30), this study also found subpopulations expressing similar markers identified in BAT (29) including Bmper^high and Pi16^high populations while Gdf10 was lowly expressed in all clusters. The presence of the cells in taPVAT demonstrates the similarities between ASPC from taPVAT and BAT. In agreement with previous studies showing that mechanoactivation of PIEZO1 reduces the potential of ASPCs to become adipocytes (27), higher expression of Piezo1 was seen in Pi16^high ASPCs compared with Bmper^high ASPCs. The relationship between the mechanosensing function of PIEZO1 and adipogenesis of ASPCs warrants further investigation.

PVAT, especially around the thoracic aorta, is being recognized as a tissue that works in concert with the rest of the vessel to determine aortic stiffness (45, 46). Thoracic aortic stiffness, measured mechanically, was lower when PVAT was included (47). Advanced glycation end products (48) and interleukin-6 (49) both contributed to arterial stiffness in the mouse at least in part from actions within PVAT. Finally, the peroxisome proliferator-activated receptor-γ (PPARγ) activator pioglitazone improved the PVAT microenvironment of the ob/ob mouse to potentially reduce aortic stiffness (50). These findings are consistent with the serendipitous discovery in the smooth muscle cell-specific knockout mouse. In this KO mouse, PVAT did not develop around the thoracic aorta and the KO mouse had a higher pulse wave velocity, a surrogate of aortic stiffness, than the WT (51). Altogether, these findings support that PVAT, now known to possess mechanotransducers, should be considered mechanistically in the determination of arterial stiffness.

Limitations of the Study

We recognize experimental limitations of the present work. The anterior strip of PVAT was used for snRNA-seq. The lateral strips, those which hold the aorta to the spine, were not included in the snRNA-seq analyses but were present in all remaining experiments. Lineage tracing in the mouse found that the anterior PVAT adipocytes were derived from smooth muscle 22a⁺ (SM22 a⁺) progenitors, whereas the lateral PVAT possessed both SM22 a⁺ and Myf5⁺ cells (9). It is unknown whether this different developmental program occurs in the rat. Along these lines, reference data sets and functional annotations were derived primarily from mouse/human resources given that these resources are limited in the rat.

Tissues from males were the focus given the intention to understand whether this study was feasible. However, we previously published that the aorta from the female Dahl-SS rat does demonstrate PVAT-assisted relaxation (52). This endpoint was a more severely diminished stress relaxation response to HF diet in male rats compared with females (52). Whether this is also observed in naïve rats, thus making the male and female intrinsically different, remains to be determined.

Similarly, the present work studied brown/brown-like adipose tissue, as opposed to white, given that the PVAT around the thoracic aorta is brown/brown-like. Future experiments in white PVAT fat (mesenteric) will be important to carry out and are in process. We also recognize the limitations of using single nuclei versus single cell bias (53). However, nuclei versus cell isolation was a deliberate choice given the fragility of isolated adipocytes and the difficulty of working in a high lipid environment, as well as with buoyant, large cells. SnRNA-seq provided an effective means to determine the cell types that populate the PVAT of the thoracic aorta. In this study, we also use stress relaxation as a reproducible and quantifiable measurement of the viscoelastic nature of taPVAT (4). Alternatively, the stress-strain curve could have been used. Furthermore, these studies used immunohistochemistry and RNAscope to provide support for PIEZO1 expression in the PVAT which is in agreement with earlier work (54). Verification by Western blot could provide additional evidence but has been challenging because of the significant size of the protein (286KDa) and antibody specificity used in Western analyses. However, this evidence was not necessary given the qualitative nature of the question asked: Was PIEZO1 present? Finally, the other highly expressed mechanotransducer Slc12a2 (solute carrier also known as Na⁺-K⁺-2Cl⁻ cotransporter NKCC1) is an important target for future studies.

Perspectives and Significance

Characterization of the Dahl-SS rat PVAT cell types is expected to play a fundamental role in developing our understanding of how PVAT responds in hypertension models. We show that the taPVAT and BAT of the male Dahl-SS rat are comprised of eight dominant cell types. The chief cell type difference between these two brown fats was a greater representation of mesothelial cells in taPVAT versus BAT. In taPVAT, most cell types possessed mechanotransducer genes, with Piezo1 and Slc12a2 being the most highly represented transcripts. With the use of isolated taPVAT, PIEZO1 was functionally important for maintaining tone in the stretch-challenged tissue. Collectively, this work defines the cells that reside in PVAT; that these cells are more similar than different to those in BAT; and that cells of taPVAT possess (functional) mechanotransducers that contribute to stress relaxation. Two goals were fulfilled with this study. First, we now have a good sense of the major types of cells in taPVAT. Such knowledge helps us better understand the potential physiology of this tissue. Second, we now know that taPVAT is more similar to BAT in its cellular and gene composition than it is different. This provides a challenge in developing PVAT-specific therapies based on the cell or gene.

DATA AVAILABILITY

Data and code are shared with consideration of the findable, accessible, interoperable, and reusable (FAIR) guiding principles (55). Raw and processed snRNA-seq data are publicly available on the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) at Accession ID GSE244451 and for visualization and querying on the Broad Single-Cell Portal (SCP; https://portals.broadinstitute.org/single_cell) with the Accession ID SCP2384. Analysis code is available on GitHub (https://github.com/naultran/BAT_taPVAT_snRNAseq).

SUPPLEMENTAL DATA

Supplemental Tables S1 and S2 and Supplemental Figs. S1–S6: https://doi.org/10.6084/m9.figshare.c.7123246.v1.

GRANTS

This work was supported by National Institutes of Health Grants P01HL152951 (to J.M.T., S.W.W., L.T., G.A.C., C.R., C.J.R., L.L., S.B., and R.N.) and T32GM142521 via the Department of Pharmacology and Toxicology at Michigan State University (to E.W.).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.M.T., L.T., E.W., L.L., and R.N. analyzed data; J.M.T., S.W.W., L.T., G.A.C., C.R., C.J.R., S.B., and R.N. interpreted results of experiments; J.M.T., L.T., E.W., L.L., and R.N. prepared figures; J.M.T. and R.N. drafted manuscript; J.M.T., S.W.W., L.T., G.A.C., C.R., C.J.R., E.W., L.L., S.B., and R.N. edited and revised manuscript; J.M.T., S.W.W., L.T., G.A.C., C.R., C.J.R., E.W., L.L., S.B. and R.N. approved final version of manuscript.