Divergent effects of canonical and non‐canonical TGF‐β signalling on mixed contractile‐synthetic smooth muscle cell phenotype in human Marfan syndrome aortic root aneurysms

Abstract Aortic root aneurysm formation is a cardinal feature of Marfan syndrome (MFS) and likely TGF‐β driven via Smad (canonical) and ERK (non‐canonical) signalling. The current study assesses human MFS vascular smooth muscle cell (SMC) phenotype, focusing on individual contributions by Smad and ERK, with Notch3 signalling identified as a novel compensatory mechanism against TGF‐β‐driven pathology. Although significant ERK activation and mixed contractile gene expression patterns were observed by traditional analysis, this did not directly correlate with the anatomic site of the aneurysm. Smooth muscle cell phenotypic changes were TGF‐β‐dependent and opposed by ERK in vitro, implicating the canonical Smad pathway. Bulk SMC RNA sequencing after ERK inhibition showed that ERK modulates cell proliferation, apoptosis, inflammation, and Notch signalling via Notch3 in MFS. Reversing Notch3 overexpression with siRNA demonstrated that Notch3 promotes several protective remodelling pathways, including increased SMC proliferation, decreased apoptosis and reduced matrix metalloproteinase activity, in vitro. In conclusion, in human MFS aortic SMCs: (a) ERK activation is enhanced but not specific to the site of aneurysm formation; (b) ERK opposes TGF‐β‐dependent negative effects on SMC phenotype; (c) multiple distinct SMC subtypes contribute to a ‘mixed’ contractile‐synthetic phenotype in MFS aortic aneurysm; and (d) ERK drives Notch3 overexpression, a potential pathway for tissue remodelling in response to aneurysm formation.

TGF-β signalling systemically, MFS patients classically develop focal aortic root aneurysms with preserved dimension of the adjacent ascending aorta. 5 We generally hypothesize that distinct embryonic origins of thoracic aortic segments 6 (second heart field, neural crest, or paraxial mesoderm) may govern variable aortic smooth muscle cell (SMC) response to enhanced TGF-β and explain this predisposition to aortic root dilatation, although hemodynamic and other biomechanical factors have been proposed as mediators of region-specific changes as well. 7 Paradoxically, TGF-β signalling seems to have a variable role depending on aortic developmental stage: critical for early normal aortic wall development, 8 while pathologic when enhanced during late aneurysm formation. 9 Multiple studies 10 Intriguingly, we report that ERK drives Notch3 overexpression, a potential protective pathway for tissue remodelling in response to MFS aneurysm formation. Four Notch receptors (Notch 1-4) have been described in humans and represent large transmembrane proteins that bind ligands expressed on adjacent cells. 21 Because Notch plays a key role in neural crest migration and SMC differentiation during ascending aorta/aortic arch development, abnormal signalling may predispose to aneurysm formation. 22 Although Notch signalling has not been studied in MFS, Notch1 gene mutations have been reported in patients with bicuspid aortic valves (BAV) and BAV aortopathy. Notch 1-4 mRNA levels were significantly decreased in BAV aortic specimens compared to normal tricuspid aortic valve aortas. 23 Similarly, reduced Notch 1 and 3 gene expression levels were reported in human abdominal aortic aneurysm samples. 24 We hypothesize the Notch pathway incites productive tissue remodelling in response to MFS aneurysm formation and affords a provocative avenue for therapeutic intervention via forced Notch3 overexpression.

| Human studies
The Stanford Institutional Review Board (IRB) approved experiments involving human specimens. All patients included in this study gave informed consent for tissue banking and participation in human subject studies during elective cardiac surgery cases. Blanket research consent was obtained from surrogate decision-makers for all included organ donor controls by the referring organ procurement organization.

| Tissue handling
Fresh surgical specimens were collected within 30 minutes of excision, dissected to remove adventitial tissue, and snap-frozen in liquid nitrogen. For molecular assays, tissue samples were thawed and remaining adventitial and intimal layers removed. The tissue was snapfrozen prior to lysis for downstream analysis.

| Protein isolation and processing
Isolated medial aortic tissue was suspended in RIPA lysis buffer (MilliporeSigma, St. Louis, MO) supplemented with pan-protease and phosphatase inhibitor cocktail (MilliporeSigma) and disrupted with a rotor/stator homogenizer, snap-frozen and homogenized again.
Cultured SMCs monolayers were treated with Trypsin (TrypleE, Gibco), cells were pelleted in a microcentrifuge, washed in PBS and lysed with RIPA buffer. Lysates were allowed to dissociate on a rotator at 4°C for 60 minutes, then centrifuged to pellet insoluble tissue debris. The supernatant was collected and subjected to protein content quantification by BCA assay kit (ThermoFisher Scientific).

| Wes semi-quantitative protein immunoblotting
Protein lysates from tissue and in vitro cell culture lines were processed for use on Wes Simple Western assays according to manufacturer protocols (Protein Simple). Samples were mixed with Simple Western Sample Master Mix (80 mmol/L DTT, 2× sample buffer, 2× fluorescence standard) and denatured. The Simple Western kit plate was loaded with denatured samples, primary antibody, HRP-conjugated anti-rabbit antibody, luminol-peroxide substrate and wash buffers. The proprietary capillary-based separation system was utilized to automatically load, separate, immobilize and immunoprobe protein lysates for proteins of interest using HRPmediated chemiluminescence. The chemiluminescent signal was detected using the system's built-in CCD camera and analysed for signal intensity using accompanying Compass software. Band intensity was used to generate a traditional Western blot lane. Primary antibodies were titrated to determine optimal protein and primary antibody concentration. Antibodies for pERK1/2 (1:100 dilution, rabbit monoclonal, #4695), pSMAD2/3 (1:50 dilution, rabbit monoclonal, #8828), vinculin (1:150 dilution, rabbit polyclonal #4650) and Notch3 (1:100 dilution, rabbit monoclonal #5276) were purchased from Cell Signaling Technology.

| RNA isolation and processing
Total RNA was extracted from frozen tissue using TRIzol (ThermoFisher Scientific). Cultured SMCs were trypsinized, pelleted in a microcentrifuge, washed in PBS and processed for RNA using the GeneJET RNA Purification Kit (ThermoFisher, Scientific). RNA was purified using the GeneJET RNA Cleanup and Concentration micro kit (ThermoFisher Scientific). RNA concentration was calculated using a NanoDrop spectrophotometer and quality determined by analysing the A 260/280 ratio for values > 1.9.
The aortic tissue was minced and the dish placed in an incubator overnight at 37°C and 5% CO 2 . Foetal bovine serum (Lonza) was added after incubation to arrest digestion. The contents were collected and centrifuged; the supernatant was removed and the pel-

| SMC culture in vitro assays
Human aortic SMC lines were used for in vitro assays between pas-  Libraries were sequenced on an Illumina HiSeq 4000 instrument.

| mRNA sequencing and analysis
Total RNA pools from SMC lines treated with either (1) PD98059 or (2) vehicle control were prepared for mRNA sequencing using All samples were checked for quality using a Bioanalyzer prior to library preparation. Single-end reads were aligned to Hg19 genome using TopHat2 software (http://ccb.jhu.edu/softw are/topha t/index. shtml ) with default settings. HTSeq was then used to determine read counts supporting expression of genes from RefSeq Hg19 (the htseq-count function). These values were log-transformed and normalized as counts per million reads as input for QuSage (http:// bioco nduct or.org/packa ges/relea se/bioc/html/qusage.html). The two-way paired group response for Donor control or MFS patient samples was used to determine gene sets with patient-specific responses to PD98059 treatment. Single-gene-level data for ERK inhibitor responses were generated by cross-referencing the differentially expressed genes for each sample group to create an MFS aortic root-specific gene set consisting of 228 genes. The DAVID bioinformatics database was used to cluster these genes by biologic function and screened for pathways of interest.
Washed cells were imaged using fluorescence to confirm transfection success. Cells were cultured in hSMC media for 48 hours prior to further assays.

| MTT proliferation assay
Smooth muscle cells were seeded in a 96-well optical plate after siRNA transfection and given (a) standard or (b) serum-free hSMC media for 48 hours. The MTT assay (Promega) was performed according to the manufacturer's protocol and the absorbance read at 570 nm.

| Annexin V Apoptosis Assay (Fluorescence-Activated Cell Sorting)
Transfected SMCs were trypsinized and counted after 48 hours.

| Total elastin quantification assay
Total cell lysates after siRNA transection and 48 hours incubation were isolated from 200 000 cells using RIPA Buffer (MilliporeSigma).
Total bound elastin was quantified using the Fastin Elastin Kit (Biocolor; Carrickfergus, UK) using manufacturer protocols. A standard curve was generated with provided purified elastin for quantification of total sample elastin.

| Gelatin zymography
Conditioned media from siRNA-transfected SMCs incubated in hSMC media for 48 hours was collected and concentrated using Amicon Ultra-4 centrifugal filter tubes (Ultracel-30K) (MilliporeSigma). The pro and active forms of MMP-2 were evaluated with commercially available Gelatin Zymography Kit (Cosmobio) using 10 μL of concentrated supernatant according to manufacturer protocols. Band density was calculated using ImageJ.

| ERK signalling is enhanced in both aneurysmal and normal calibre aortic specimens from MFS patients
Marfan syndrome patients with focal aortic root aneurysms in MFS patients ( Figure 1C). Interestingly, ERK phosphorylation was similarly elevated in both aortic regions (n = 6) from MFS patients (1.06-fold, P = .80), indicating that ERK activation does not correlate with aneurysm formation. Importantly, we did not find any statistically significant differences in the tested contractile genes with paired comparisons of MFS aneurysmal aortic root vs the same patient's non-dilated ascending aorta.

| Analysis of SMC phenotype demonstrates
Synthetic genes related to ECM remodelling were also tested.
Collagen isoforms 1 and 3 (COL1A1 and COL3A1) were both significantly increased in MFS aortic root compared to non-dilated donor control aortic root (5.6-and 5.3-fold, P < .01), while elastin (ELN) expression was not significantly different (3.6-fold, P = .13, Figure 2B). Of these genes, only COL1A1 demonstrated a significant difference in expression level between the aneurysmal MFS aortic root and the same patient's non-dilated ascending aorta (2.25-fold higher in aortic root, P = .03). MMP2 expression was 5.7-fold higher in MFS aortic root compared to donor (P < .01, Figure 2C); there was no significant difference in MMP9 or TIMP1 expression. TIMP2 was significantly enhanced in MFS aortic root (4.2-fold, P = .04). There were no statistically significant differences between MFS aortic root and non-dilated ascending aorta for these genes.
Taken together, these results demonstrate that in MFS aortic root aneurysm specimens, both contractile and synthetic (ECM remodelling) gene pathways are activated suggesting a 'mixed' tissue phenotype.

| Smad and ERK signalling compete for regulation of MFS SMC contractile and ECM genes, in vitro
To

| ERK affects critical gene pathways specific to MFS aortic root SMCs including Notch signalling
To better understand the role ERK signalling plays during aortic root aneurysm development, we performed transcriptome-wide RNA se-  Table S1). These data imply complex contributions of ERK to SMC turnover in the aortic wall and a complex response related to cell-cell adhesion, with genes related to these pathways both significantly up-and downregulated by ERK inhibition.
Having evaluated individual gene regulation in the MFS aortic root only, we next used QuSAGE 34 (Quantitative Set Analysis for Gene Expression) to systematically assess for changes in core biologic processes in response to ERK. Predefined KEGG gene set responses were assessed for statistical significance and hierarchically clustered ( Figure 5A, complete data set in Table S2), revealing differential responses to ERK blockade between groups. To study ERK signalling in processes related to SMC function and aortic aneurysm biology, we specifically evaluated cell cycle, apoptosis and SMC contraction ( Figure 5B). All sample groups demonstrated and ligands as a driver for this significant response ( Figure 5D).  Figure 6A).

F I G U R E 4
RNA-sequencing transcriptome-wide assessment with ERK inhibition. A, Volcano plots for whole-genome sequencing data depicting expression fold change with ERK inhibition relative to control and quantification of genes with significantly altered expression with ERK blockade at cut-off P < .05. n = 5. B, Workflow for identification of gene set with ERK responsive genes specificity to MFS aortic root. C, DAVID analysis of MFS Root-specific gene set (228 genes) stratified by biologic processes modulated with ERK blockade. RNA-sequencing data acquired from n = 5 donor and MFS aortic specimens We next performed RT-PCR on primary aortic SMC cell lines treated with PD98059 or vehicle control for 24 hours (n = 5). Marfan syndrome aortic root SMCs demonstrated a significant reduction in Notch3 expression with ERK blockade, whereas Notch1 and 2 did not significantly change ( Figure 6B). Therefore, these in vitro studies suggest that focal Notch3 overexpression in the aneurysmal MFS aortic root may be ERK-dependent.

| Notch3 enhances aortic root SMC proliferation, reduces apoptosis and modulates ECM turnover, in vitro
To better understand the role Notch3 may play during aneurysm formation, MFS aortic root SMC lines (n = 4 patients) were treated with a Notch3-specific siRNA or scrambled control (Scr), then tested for F I G U R E 5 QuSAGE analysis of aortic RNA-sequencing data identifies Notch signalling as pathway of interest. A, Hierarchical map of 186 predefined KEGG gene sets clustered by -log P value following QuSAGE analysis; sign denoted positive or negative based on direction of gene set activity change. B, Individual probability density function (PDF) charts developed for pathways of interest relevant to VSMC phenotype. C, PDF plot for Notch signalling pathway and D, summary plot of individual gene activity within Notch signalling pathway following ERK inhibition. *P < .05, #P < .01 SMC-specific gene expression and functional phenotypic changes.
Cell-based assays were used to further characterize the effects of Notch3 on aortic root SMCs. Cell proliferation was assessed via MTT assay. Notch3 siRNA inhibition resulted in 25% MTT reduction in both starved and normal media (P < .05 for both conditions), suggesting Notch3 naturally increases cell proliferation ( Figure 6E).
Elastin and tropoelastin synthesis were quantified with a cell lysate-based assay and demonstrated modest but significant increase in ELN production with Notch3 silencing (24% increase, P = .013, Figure 6H). These in vitro studies reveal that Notch3 likely reduces aortic root SMC contractile protein expression while enhancing cell proliferation, resisting apoptotic stimuli and decreasing MMP activity in MFS aortic root SMCs. While these changes suggest a largely protective role for Notch3 with regards to aneurysm biology, Notch3 may act as an inhibitor of ELN expression in vitro.

| D ISCUSS I ON
The pathophysiology of aneurysm development in MFS is a complex, multi-factorial process. TGF-β is central to this process; however, its effects appear highly dependent on developmental stage. The nuance of spatiotemporal TGF-β signalling leading to aortic root aneurysms in MFS has precluded the development of novel therapies to prevent or slow aneurysm growth. The novel findings of this F I G U R E 7 Proposed role of TGFβ-dependent ERK/Smad signalling in MFS aortic root aneurysms. SMCs in the aortic root and ascending aorta receive enhanced TGF-β signalling. While Smad signalling is more enhanced in the aneurysmal aortic root compared to the non-dilated ascending aorta, ERK activation is uniformly increased in both regions. Normal quiescent contractile SMCs may adopt increased contractile/ synthetic gene expression leading to heterogeneity and 'mixed' phenotype in the aortic wall, likely driven by Smad signalling. ERK dampens Smad-dependent changes and seems to play a role in SMC turnover by enhancing cell cyclerelated genes and apoptosis. NOTCH3 is focally enhanced by ERK in aneurysmal MFS aortic root SMCs and may represent a protective response against further aneurysm growth study include the following: (a) ERK signalling is not synergistic with canonical TGF-β signalling with regard to SMC phenotype in MFS; (b) SMCs receive a complex combination of TGF-β-induced signals ultimately producing a mixed synthetic-contractile phenotype that likely disrupts aortic homeostasis; and (c) ERK-dependent Notch3 overexpression may be an intrinsic protective response against aneurysm growth. This novel potential tissue remodelling pathway is a provocative avenue for therapeutic intervention via forced early Notch3 Further characterization of cell lines and primary surgical tissue from additional patients will provide clarity to the roles of these distinct cell populations in aortic aneurysms.

ACK N OWLED G EM ENTS
The writers would like to acknowledge Ivy Sanders Schneider for graphic figure development, Meredith Weglarz for FACS technical assistance, the Stanford Shared FACS Facility (SSFF) and the Stanford Functional Genomics Facility (SFGF).

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
AJP and MPF designed the project and wrote the manuscript, AJP, TK, JT, IP, KJ and GB performed experiments and analysed data, AR performed analysis of computational data, PC provided critical assistance acquiring donor tissue, YT, JC and CI assisted with data analysis and critical paper assessment, and GB assisted with pathologic tissue analysis.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.