Cardiovascular disease risk factors induce mesenchymal features and senescence in cardiac endothelial cells

Aging, obesity, hypertension and physical inactivity are major risk factors for endothelial dysfunction and cardiovascular disease (CVD). We applied fluorescence-activated cell sorting (FACS), RNA sequencing and bioinformatic methods to investigate the common effects of CVD risk factors on cardiac endothelial cells (ECs). Aging, obesity and pressure overload all upregulated pathways related to TGF-β signaling and mesenchymal gene expression, inflammation, vascular permeability, oxidative stress, collagen synthesis and cellular senescence, whereas exercise training downregulated most of the same pathways. We identified collagen chaperone SerpinH1/HSP47 to be significantly increased by aging and obesity and repressed by exercise training. Mechanistic studies demonstrated that SERPINH1/HSP47 in human ECs changed cell morphology and increased mesenchymal gene expression, while its silencing inhibited collagen deposition. Our data demonstrate that CVD risk factors significantly remodel the transcriptomic landscape of cardiac ECs to acquire senescence and mesenchymal features. SERPINH1/HSP47 was identified as a potential therapeutic target in ECs.


Introduction 1
and cardiac angiogenesis have been described previously 38 , but exercise-induced 1 molecular changes in ECs have not been characterized. Aging and obesity, on the other 2 hand, are known to contribute to capillary rarefaction and/or dysfunction 10,11,39,40 , and 3 another novel aspect in this study was the comparison of several CVD risk factors to identify 4 common pathways and genes, which could drive the pathogenesis in cardiac disease, and 5 could be considered as potential therapeutic targets. ECs would provide an attractive target 6 for drug development, as they are the first cells to encounter drugs in the bloodstream. 7 8 Dysfunctional endothelium likely contributes to more diseases than any other tissue in the 9 body as it affects all organs. On the other hand, endothelium could act as an important 10 mediator of the health-promoting effects of exercise in a variety of tissues. Our finding that 11 aging, obesity and pressure overload induce mesenchymal gene programs in cardiac ECs 12 adds to the increasing evidence that activated endothelial TGF-β signaling and acquisition 13 of mesenchymal features play an important role in the development of EC dysfunction and 14 cardiac diseases 13,20,41,42 . Importantly, genes related to TGF-β production and cellular 15 aging were repressed by exercise, highlighting the mechanisms behind the potential of 16 exercise training in preventing and delaying the development of CVD. The activation of TGF-17 β signaling pathway has been implicated as a driving force for EndMT 21,[23][24][25][26] . Several 18 studies have recently suggested that EndMT could contribute to the development of various 19 cardiovascular diseases 16 13, 14, 27 , but currently there is a lack of understanding of the causal 20 relationships and mechanisms linking EndMT and CVD 13 . Furthermore, whether the 21 transition from ECs to mesenchymal cells occurs completely in various CVDs is still actively 22 debated in the literature. It has been suggested that pathological EC activation will result in 23 acquired EndMT features e.g. expression of mesenchymal genes, without full 24 transformation from one cell type to another 44 . This is in line with our findings, as only cells 25 with high CD31 expression and with no expression of CD45, CD140a and Ter119 were 1 included in our analyses. Thus, all the analyzed cells were endothelial cells, but in the CVD 2 risk factor groups they demonstrated increased mesenchymal marker expression. Long-3 term lineage tracing of ECs in response to CVD risk factors would provide further knowledge 4 if and to what extent full transformation of ECs to mesenchymal cells occurs in cardiac 5 vasculature. Our results, however, demonstrate that ECs acquire mesenchymal features 6 due to CVD risk factors, which likely results in EC dysfunction even without EndMT. 7 8 To identify possible pathology-driving genes, which would be common for several risk 9 factors, we performed gene overlap analysis using all data sets. Two genes, SerpinH1 and 10 Vwa1, were found to be significantly increased by both aging and obesity and decreased by 11 exercise, suggesting that they could act as common mediators of EC dysfunction. We 12 focused in this study on Serpinh1/Hsp47, as it is a collagen chaperone and has been shown 13 to contribute to tissue fibrosis 31,45 , an important feature of many cardiac diseases. Recently,14 it was demonstrated in a mouse pressure overload model using Hsp47 cell type -specific 15 knockout mice that Hsp47 in myofibroblasts is an important regulator of pathologic cardiac 16 fibrosis 45 . In line with our results in human cardiac ECs, collagen 1 production was 17 decreased in the EC-specific Hsp47 deficient hearts 45 . In human ECs, our results placed 18 SERPINH1/HSP47 downstream of TGF-β and ROS, and demonstrated that its 19 overexpression promoted mesenchymal features in human cardiac EC. Furthermore, 20 SERPINH1/HSP47 was found to be important for extracellular collagen 1 deposition and EC 21 proliferation/migration. Silencing of SERPINH1 also prevented the TGF-β induced 22 appearance of TAGLIN-positive cells in human cardiac EC, which is considered as a marker 23 for EndMT 21,37 . Based on the publicly available single-cell RNA sequencing data and 24 immunohistochemistry of the human heart samples, SERPINH1/HSP47 is abundantly 25 expressed in all cardiac endothelial populations. For further translational impact, the role of 1 endothelial SERPINH1/HSP47 in aged, obese and hypertensive human hearts needs to be 2 determined. 3 4 In conclusion, our data demonstrate that the major CVD risk factors significantly remodel 5 the cardiac EC transcriptome promoting cell senescence, oxidative stress, TGF-β signaling 6 and mesenchymal gene features, whereas exercise training provided opposite and 7 protective effects. SerpinH1/Hsp47 was identified as one of the downstream effectors of 8 TGF-β, which could provide a novel therapeutic target in endothelial cells.  Table 3. For EndMT induction in HCAEC, a previously published method 1 using TGF-β and H2O2 was used 21,37 . The detailed procedures for immunohistochemical stainings, western blotting and real-time The data from the individual experiments were analyzed by student's t test. P<0.05 value 13 was considered statistically significant and P values in the graphs are mentioned as 14 *P<0.05, **P<0.01 and ***P<0.001. The data is shown as mean ± SEM. The GraphPad 15  Computational data a, Cell type a, Writing g, Supplemental text writing g and Principal i.    Data is presented as mean ± SEM. Student's t test was used, *p<0.05, **p<0.01, ***p<0.001. 22 with TGF-β1 (50ng/ml) or H2O2 for five days. In panel A, C, D and F, N=3 biological 11 replicates/group were analyzed. Scale bar 100 μm. Data is presented as mean ± SEM. 12 Student's t test was used, *p<0.05, **p<0.01, ***p<0.001. N=3 biological replicates/group were analyzed. Scale bar 100 μm. Data is presented as 1 mean ± SEM. Student's t test was used, *p<0.05, **p<0.01, ***p<0.001. presented as mean ± SEM. Student's t test was used, *p<0.05, **p<0.01, ***p<0.001.

Exercise Training 15
Ten-week-old C57BL/6J male mice or 19 to 24 months old female mice were trained on a 16 treadmill (LE 8710, Bioseb). The mice were familiarized to the treadmill for three consecutive 17 days with low speed (8-10 cm/s). Progressive training program consisted of 1-1.5 h training 18 bouts five days a week for a total of six weeks with increasing speed, incline and/or duration 19 each week. The following parameters in the treadmill controller were opted, tread inclination: 20 0°-10°; minimum and maximum tread speed: 10cm to 30cm per second; shock grid intensity: 21 0.2 mA. The aged mice were exercise-trained for four weeks and the same procedures were 22 followed during the training program. 23

High Fat feeding 25
Ten-week-old C57BL/6J male mice were fed with standard chow diet or high-fat diet (HFD) 1 containing 60% kcal derived from fat (Research Diets, D12492) for 4 or 14 weeks and used 2 for immunohistochemistry or RNA-seq analysis, respectively. 3 4

Transverse Aortic Constriction Surgery 5
Ten-week-old C57BL/6J male mice were anesthetized with ketamine and xylazine. The mice 6 were placed in supine position and intubated. The skin along the supra-sternal notch to mid 7 sternum was incised to perform sternotomy to expose the aortic arch, right innominate and 8 left common carotid arteries together with the trachea. Ligation of the transverse aorta 9 between the right innominate left common carotid arteries against blunted 27-gauge needle 10 with a 7-0 suture was performed and the needle was gently removed. The sternum and skin 11 were ligated with monofilament polypropylene suture. Mice were placed in a warm chamber 12 to recover, treated with analgesics (0.05mg/kg of Temgesic i.m.) at the time of the surgery 13 and twice a day for following two days. For the control group (sham), all the steps in the 14 surgical procedure were followed, except constricting the aorta. One group was killed two 15 weeks and another group seven weeks after the surgery. Echocardiography was performed 16 once a week during the experiment.

Body Fat Measurement 1
The mice were anesthetized with ketamine and xylazine and the percentage of total body 2 fat was measured using dual energy x-ray absorptiometry (Lunar PIXImus, GE Medical 3 systems). 4 5

Oral Glucose Tolerance Test 6
Mice were fasted for four to five hours before the experiment. Glucose (1g/kg) was 7 administered by oral gavage to mice. Blood from the tail tip was used to measure glucose 8 levels at the following time points (15, 30, 60 and 90 min) using blood glucose meter 9 (Contour, Bayer). 10 11

Immunofluorescent Staining 12
Frozen mouse heart sections (10μm) were cut with cryomicrotome and stained as described The stained micrographs were initially adjusted for threshold, and an area fraction tool was 19 used to quantify the area percentage of the vessels and collagen (Image J software, NIH). 20 21

Human Heart samples 22
Human heart samples were obtained from 4 organ donor hearts, which could not be used 23 for transplantation e.g. due to size or tissue-type mismatch. The collection was approved 24 by institutional ethics committee and The National Authority for Medicolegal Affairs.

Differential gene expression 22
The sequenced reads were analyzed with the following software packages embedded in the 23 Chipster analysis platform 2 (v3.12.2; https://chipster.csc.fi). Trimmomatic tool 3 24 (https://chipster.csc.fi/manual/trimmomatic.html) was used to preprocess Illumina single end 25 reads. The HISAT2 package 4 (https://chipster.csc.fi/manual/hisat2.html) was employed to 1 align the reads to mouse genome GRCm38.90 and the HTSeq count tool 5 2 (https://chipster.csc.fi/manual/htseq-count.html) to quantify the aligned reads per gene. The 3 raw read count table for genes generated utilizing the HTSeq count were used as an input 4 to perform two-dimensional principal component analysis (PCA) and unsupervised 5 hierarchical clustering analysis using DESeq2 Bioconductor package 6 6 (https://chipster.csc.fi/manual/deseq2-pca-heatmap.html). Next, to perform the differential 7 gene expression (DGE) analysis, the DESeq2 Bioconductor package 6 was used. The 8 advantage of DEseq2 tool is sensitive and precise for analyzing the DEG in studies with few 9 biological replicates. To reliably estimate the within group variance, Empirical Bayes 10 shrinkage for dispersion estimation was used and a dispersion value for each gene was 11 estimated through a model fit procedure (refer to the Figure S5A, which illustrates the 12 shrinkage estimation for the experimental conditions). The gene features obtained after the 13 dispersion estimation were used to perform statistical testing. Next, negative binomial 14 generalized linear model was fitted for each gene and Wald test (raw p-value) was 15 calculated to test the significance. Finally, DEseq2 applies Benjamini-Hochberg correction 16 test to control the false discovery rate, FDR (refer to the Figure S5B indicating the 17 distribution of raw and FDR adjusted pvalue for the experimental conditions). In our DEG 18 analysis, we have set the FDR (p adj.) cut-off as less than or equal to 0.05 (FDR/p-adj ≤

Gene Function and Pathway Analysis 23
The gene function and pathway analysis of the DGE were determined by performing 24 statistical overrepresentation test using the PANTHER classification system 7 (V.14.1; 25 http://www.pantherdb.org). The p<0.05 was considered for the further analysis and the data 1 is presented as -log2(pvalue).  Table 3.  9 10

Scratch wound assay 11
The SERPINH1 overexpressed or silenced HCAECs were seeded in the IncuCyte 12 ImageLock 96-well microplate precoated with 0.1% gelatin and cultured in complete EC 13 growth medium. To the confluent cell monolayers, 700 -800 micron scratch wounds were 14 introduced with IncuCyte WoundMaker, the wells were briefly rinsed with and maintained in 15 complete EC growth medium. The kinetics of the cell migration were recorded and 10X 16 phase contrast time-lapse images were acquired using IncuCyte Live-Cell Analysis System. 17 The wound closure region was measured by Edge-detection and thresholding method in 18 Image J software (NIH). The data is presented is as wound closure (%) relative to time. 19 20

EndMT assay 21
The coverslips or six well plates were precoated with 0.1% gelatin for 20min at 37°C, 22 scrambled or SERPINH1 silenced HCAEC were seeded and cultured in complete EC growth 23 medium. The cells were treated with or without 50ng/ml of recombinant human TGF-β (R&D 24 Technologies) and/or 200μM hydrogen peroxide (Acros organics) for five days as described 1 previously 12, 13 . 2 3

Cell Staining 4
The cells grown on the coverslips were fixed with 4% PFA in PBS for 15 min. Blocking was 5 done using donkey immunomix and the cells were stained with primary antibodies and 6 secondary antibodies as indicated in the Supplemental Table 2. DAPI was used to stain the 7 nucleus, and the cells were mounted using Vectashield (Vector labs). The amount of COL1 8 was quantified by adjusting 10X images for threshold and area fraction tool was used to 9 quantify the area percentage of the collagen deposition (Image J software, NIH).  Table 4. 22 The cells were harvested and homogenized in lysis buffer containing 0.5%NP-40 (v/v) and 1 0.5%Triton X-100 (v/v) in PBS, supplemented with protease and phosphatase inhibitors 2 (A32959, Pierce, Thermo Scientific). Protein concentration was determined using a BCA 3 protein assay kit (Pierce, Thermo Scientific). Equal amounts of total protein were resolved 4 in Mini-PROTEAN TGX Precast gels (Bio-Rad) and transferred to PVDF membrane 5 (immobilon-P, Millipore). 5% BSA (wt/vol) and 0.1% Tween 20 (v/v) in TBS was used to 6 block the membranes followed by incubation with primary antibodies (Supplementary 7 Femto Maximum sensitivity substrate (Thermo Scientific). The blots were imaged with 10 Odyssey imager (Li-COR Biosciences) or Chemi Doc imaging system (Bio-Rad) and 11 quantified with Image Studio Lite Software (Li-COR Biosciences).        1   2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  Supplementary Table 5E: Reference list for endothelial and mesenchymal genes 1 indicated in the Figure 4F (TAC (7)