Integrins αvβ5 and αvβ3 promote latent TGF-β1 activation by human cardiac fibroblast contraction

Abstract

Aims

Pathological tissue remodelling by myofibroblast contraction is a hallmark of cardiac fibrosis. Myofibroblasts differentiate from cardiac fibroblasts under the action of transforming growth factor-β1 (TGF-β1), which is secreted into the extracellular matrix as a large latent complex. Integrin-mediated traction forces activate TGF-β1 by inducing a conformational change in the latent complex. The mesenchymal integrins αvβ5 and αvβ3 are expressed in the heart, but their role in the activation of TGF-β1 remains elusive. Here, we test whether targeting αvβ5 and αvβ3 integrins reduces latent TGF-β1 activation by cardiac fibroblasts with the goal to prevent the formation of α-smooth muscle actin (α-SMA)-expressing cardiac myofibroblasts and their contribution to fibrosis.

Methods and results

Using a porcine model of induced right ventricular fibrosis and pro-fibrotic culture conditions, we show that integrins αvβ5 and αvβ3 are up-regulated in myofibroblast-enriched fibrotic lesions and differentiated cultured human cardiac myofibroblasts. Both integrins autonomously contribute to latent TGF-β1 activation and myofibroblast differentiation, as demonstrated by function-blocking peptides and antibodies. Acute blocking of both integrins leads to significantly reduced TGF-β1 activation by cardiac fibroblast contraction and loss of α-SMA expression, which is restored by adding active TGF-β1. Manipulating integrin protein levels in overexpression and shRNA experiments reveals that both integrins can compensate for each other with respect to TGF-β1 activation and induction of α-SMA expression.

Conclusions

Integrins αvβ5 and αvβ3 both control myofibroblast differentiation by activating latent TGF-β1. Pharmacological targeting of mesenchymal integrins is a possible strategy to selectively block TGF-β1 activation by cardiac myofibroblasts and progression of fibrosis in the heart.

Cardiac fibrosis, Cardiac repair, Mechanical stress, Myofibroblast

1. Introduction

Secretion and remodelling of collagenous extracellular matrix (ECM) by fibroblasts is a physiological process to maintain the mechanical integrity of the heart under high load.¹ However, in response to cardiac overload and hypertrophy, local fibroblastic cells are activated to become myofibroblasts.^2,3 Neo-expression of α-smooth muscle actin (α-SMA) in stress fibres is a hallmark of the myofibroblast and renders fibroblastic cells highly contractile.⁴ Persistent activation of myofibroblasts results in the development of pathological contractures (scars) that impair heart pumping and distensibility, contribute to diastolic and systolic dysfunction, and affect myocardial electrical transmission.^2,5,6 The mechanical and chemical environment of the scar perpetuates fibrosis by converting normal fibroblasts into myofibroblasts.² Fibrosis is a major cause for heart failure and death with no effective therapies available.

Two pivotal factors drive myofibroblast activation: mechanical stress (e.g. the stiffness of the remodelled ECM) and active transforming growth factor-β1 (TGF-β1).^4,7 Fibroblasts secrete TGF-β1 as an inactive complex together with its latency-associated pro-peptide (LAP) that is incorporated into the ECM via the latent TGF-β1-binding protein (LTBP-1).⁸ Integrins link the ECM to the contractile cytoskeleton and have been implicated in TGF-β1 activation by binding to an RGD motif in the LAP.^9–12 Integrins αvβ3, αvβ5, and αvβ6 were all shown to activate TGF-β1 by transmitting cell forces to the latent TGF-β1 complex.^8,13,14 If latent TGF-β1 is anchored via LTBP-1 to a mechanically resisting ECM, force application induces a conformational change in the LAP that liberates active TGF-β1.^15,16 The epithelial integrin αvβ6 is best studied in the context of TGF-β1 activation and contributes to lung fibrosis.^17–20 In the heart, which does not possess an epithelium or cells expressing αvβ6 integrin, mesenchymal cells possibly contribute to mechanical activation of latent TGF-β1. The mesenchymal integrins αvβ5 and αvβ3 are expressed in the normal heart and are up-regulated during heart fibrosis.²¹ However, the implication of both integrins in the activation of TGF-β1 by cardiac myofibroblasts remains elusive. We test the potential of targeting αvβ5 and αvβ3 integrins to reduce latent TGF-β1 activation.

2. Methods

2.1 Animal experiments

The Animal Care Committee at the Hospital for Sick Children approved the animal studies in accordance with the Terms of Reference following the Canadian Council on Animal Care Guidelines and federal and provincial regulations/legislations. To induce fibrosis in the right ventricle, 7-day-old Yorkshire piglets (3.9 ± 0.8 kg) underwent staged bilateral pulmonary vein banding (banded, n = 3) or sham operations (sham, n = 3), as described previously.²² First, the banding of left pulmonary and common pulmonary veins was performed via a left fifth intercostal space thoracotomy. One week later, banding of the right upper and middle pulmonary veins was carried out via a right fourth intercostal space thoracotomy. A 1/8 inch wide cotton umbilical tape was fixed around the pulmonary vein with a length equivalent to 1.3 times the pulmonary vein circumference. Sham-operated piglets underwent identical banding procedures, but the band was not left in place. Anaesthesia was performed with atropine (0.01 mg/kg, i.m.), ketamine (20 mg/kg, i.m.), and isoflurane (5%, delivered by mask). Subsequently, intubation was performed and general anaesthesia was maintained with isoflurane (2–3%) under positive pressure ventilation (20 cmH₂O, 30–40 breaths/min). Seven weeks after banding, under general anaesthesia, piglets were exsanguinated from the inferior vena cava and transmural blocks of the right ventricular myocardium were excised en bloc on ice. Heart tissue was processed for western blotting, histology, and immunohistochemistry.

2.2 Cell culture and drugs

Primary cardiac fibroblasts (human cardiac fibroblast, hCF) isolated from right ventricles of human male donors (age 39–42 years) (CC-2904, Lonza, Walkersville, MD, USA) were expanded in complete FGM-3 medium (Lonza) to passage P3 and routinely cultured between passages P3 and P5 in DMEM/F12 (Invitrogen, Burlington, ON, Canada), supplemented with 10% foetal bovine serum (Invitrogen). To augment myofibroblast differentiation, hCFs were treated with 2 ng/mL of TGF-β1 (R&D Systems, Minneapolis, MN, USA) for 5 days. Myofibroblastic cells were de-differentiated by culture on highly compliant (2 kPa soft) silicone substrates¹³ (ExCellness Biotech, Lausanne, Switzerland) or by treatment with TGF-β1 receptor inhibitor SB431542 (Sigma-Aldrich, St Louis, MO, USA). For integrin inhibition experiments, hCFs were cultured for 2 days before treatment with cyclic peptides (0.001–10 µM) antagonizing integrin αvβ5 (Cilengitide, EMD121974), αvβ3 integrin (EMD66203), and scrambled control (EMD135981) (Merck, Kirkland, QC, Canada). Anti-TGFβ1 antibody (AF-101-NA, R&D Systems) was used to neutralize active TGF-β1 (final concentration 10 µg/mL). Stably transfected mink lung epithelial cells (TMLCs) expressing luciferase under the control of the plasminogen activator inhibitor-1 promoter (Dr Daniel Rifkin, New York University, NY, USA) were routinely cultured in DMEM, supplemented with 10% foetal bovine serum.

2.3 Integrin overexpression and downregulation constructs

To overexpress integrins, hCFs were transfected with purified plasmids coding for green fluorescent protein (GFP)-tagged β3² (Dr Bernhard Wehrle-Haller, University of Geneva, Switzerland) and β5 integrin¹ (Dr Dean Sheppard, University of California San Francisco, CA, USA). To downregulate integrin expression, hCFs were transfected with Mission^®shRNA (Sigma-Aldrich) directed against human β3 integrin (NM_000212.x clone 2343s1c1; CCGGCCACGTCTACCTTCACCAATACTCGAGTATTGGTGAAGGTAGA CGTGGTTTTT) and β5 integrin (NM_002213.3 clone 2970s21c1; CCGGAGCTTGTTGTCCCAATGAAATCTCGAGATTTCATTGGGACAACAAGCTTTTTTG). Controls for over-expression and downregulation of integrins were empty GFP vector (pEGFP-N1, Clontech, Mountain View, CA, USA) and non-target shRNA (SHC002; Sigma-Aldrich), respectively. All transfections were performed using the Neon™ electroporation system (Invitrogen) according to the manufacturer's specifications.

2.4 Immunofluorescence and image analysis

For immunofluorescence staining of porcine tissues, sections were deparaffinized in xylene, rehydrated in ethanol, and rinsed in distilled water. Antigens were revealed by boiling the slides in sodium citrate buffer (Dako, Burlington, ON, Canada) at 95–100°C for 40 min. After cooling to room temperature, sections were rinsed in PBS, blocked with 2% goat serum/1% BSA for 40 min, and stained with primary and secondary antibodies. Preceding immunofluorescence staining of cultured cells, samples were simultaneously fixed and permeabilized in 3% paraformaldehyde/0.2% Triton-X 100. Primary and secondary antibodies were applied sequentially for 1 h at room temperature. Primary antibodies directed against α-SMA (α-SM1, Giulio Gabbiani, University of Geneva, Switzerland), α-sarcomeric actin (A2172, Sigma-Aldrich), CD31 (ab28364, Abcam, Cambridge, MA, USA), desmin (M076029, Dako), GFP (ab290, Abcam), integrin β3 (for hCF: Axum-4, Dr Dean Sheppard and for porcine tissues: MAB1974, Millipore, Etobicoke, ON, Canada), integrin β5 (ab15459, Abcam), vimentin (M0725, Dako), and vinculin (V9264, Sigma-Aldrich) were used. Isotype-specific secondary antibodies were TRITC-conjugated goat anti-mouse IgG1 (Jackson Laboratories, Bar Harbor, MA, USA), Alexa647-conjugated IgG2a, and Alexa488-conjugated goat anti-rabbit (Invitrogen). Micrographs were acquired using an upright Zeiss Axio Observer M35 epifluorescence microscope equipped with structured illumination (Apotome) and an Axiocam HR camera (Carl Zeiss, Jena, Germany). All images were assembled using Adobe Photoshop CS4 (Adobe Systems, San Jose, CA, USA). Contrast and brightness were enhanced identically over all images for publication purposes.

To quantify right ventricle cardiomyocyte size and myocardial fibrosis, we used image analysis. Transmural blocks of the right ventricular myocardium were fixed in 10% formalin, divided into sections, and stained with H&E, Masson's trichrome,² or co-immunostained for α-SMA, desmin, vimentin, and DAPI. Randomly selected high-power fields (30 fields per piglet) from right ventricular-free walls were examined for analysis of cellular size, cardiac fibrosis, and myofibroblasts. The diameter of cardiomyocytes was determined by measuring the distance across the cell at its narrowest plane, including the nucleus (Image J software, US National Institutes of Health, Bethesda, MD, USA). The area of myocardial collagen content for each field was quantified (Photoshop CS4) as a percentage of the entire slide and averaged. Myofibroblast content was quantified as the area in low-power-field images from right ventricle sections, occupied by α-SMA⁺/vimentin⁺/desmin⁻ cells. To identify these cells, we used the Thresholding, Binary mask, and Analyze particle functions of ImageJ.

2.5 Western blotting

For western blotting, cell lysates were collected from each experimental condition and equilibrated for protein content. Porcine tissue nuclear lysates were obtained using the NE-PER nuclear extraction kit (Pierce, Rockford, IL, USA) as per manufacturer's recommendations. Samples for integrin overexpression/knock-down experiments were run on 18% gels to overlay signals from endogenous and transfected GFP integrins. Denaturing (α-sarcomeric actin, α-SMA, desmin, lamin B, Smad3, pSmad3, vimentin, and vinculin) and native (integrins αv, β3, and β5) SDS–PAGE followed by immunoblotting were performed. Membranes were probed with primary antibodies against α-SMA (α-SM1, Gabbiani), α-sarcomeric actin (A2172, Sigma-Aldrich), CD31 (ab28364, Abcam), desmin (M076029, Dako), GFP (ab290, Abcam), integrin αv (611012, BD Biosciences, Mississauga, ON, Canada), integrin β3 (for hCF: Axum-4, Dr Dean Sheppard and for porcine tissues: MAB1974, Millipore), integrin β5 (for hCF: ab15459, Abcam and for porcine tissues: ab1926, Millipore), vimentin (M0725, Dako), and vinculin (V9264, Sigma-Aldrich). Horseradish peroxidase-conjugated secondary goat anti-mouse (626520, Invitrogen) and goat anti-rabbit antibodies (656120, Invitrogen) were used, followed by chemiluminescence (Invitrogen). Blots were analysed using an LI-COR Odyssey Fc imaging system (Mandel Scientific, Guelph, ON, Canada) and quantified with the accompanying software (Image Studio v2.1.100, Mandel Scientific). Protein band intensities were normalized to vimentin loading control and ratios calculated to the experimental control run on the same blot.

2.6 Active TGF-β1 measurements

Active TGF-β1 was quantified using TMLCs (Rifkin). hCFs (3 000 cells/cm²) were cultured in 96-well plates for 5 days. For direct co-culture, TMLCs (100 000 cells/cm²) were directly seeded on top of hCFs in serum-free growth medium supplemented with 0.1% BSA and allowed to adhere for 4 h. Co-cultures were incubated with integrin inhibitors for 1 h and thrombin added to induce fibroblast contraction as described previously.³ hCF-TMLC co-cultures were incubated for another 14 h, lysed, and luciferase activity assessed by light production from a luciferin substrate (Promega, Madison, WI, usa) using a luminometer (Centro LB, Berthold Technologies, Bad Wildbad, Germany). To activate all latent TGF-β1 present in the co-culture (total TGF-β1), all cells and ECM were scraped into serum-free growth medium supplemented with 0.1% BSA and heated to 80°C for 10 min. The resulting solutions were cooled down and used to incubate TMLC pre-grown for 4 h in 96 wells (100 000 cells/cm²) for another 14 h; TMLC were then lysed and luciferase activity was assessed as described above. All results were corrected for baseline TMLC luciferase production in the absence of TGF-β1 and are presented as the mean ± standard deviation (SD) of at least three independent experiments.

2.7 Statistical analysis

When applicable, data are presented as mean ± standard deviation (SD). Differences were assessed between groups with an analysis of variance (ANOVA) followed by a post hoc Tukey's multiple comparison test (significance level, P = 0.05). For experiments comparing only two conditions, a two-tailed paired t-test was performed. Differences were considered to be statistically significant and indicated with * when P ≤ 0.05, ** when P ≤ 0.01, and *** when P ≤ 0.005. Error bars represent standard deviation.

3. Results

3.1 Integrins αvβ5 and αvβ3 are upregulated during hCF-to-myofibroblast differentiation in vivo

To evaluate whether expression of the latent TGF-β1-binding integrins αvβ5 and αvβ3 changes during the development of heart fibrosis, we used a porcine model. Pulmonary vein stenosis in Yorkshire pigs leads to pressure overload and hypertrophy of the right ventricle as we demonstrated recently.²² To control for the development of fibrosis after banding the pulmonary veins for 7 weeks, tissue sections of the right ventricular myocardium were stained with H&E to visualize cardiomyocytes and Masson's trichrome for collagen accumulation (Figure 1A). Quantification of cardiomyocyte diameter and the area occupied by myocardial collagen demonstrated cardiomyocyte hypertrophy (1.3-fold increase in banded animals over sham control) and fibrosis (2.8-fold increase after banding) (Figure 1A). In sham-operated animals, vimentin⁺/α-SMA⁻/desmin⁻ fibroblasts occupied the space between desmin⁺/α-SMA⁻/vimentin⁻ cardiomyocytes (Figure 1A and see Supplementary material online, Figure S1A). Vimentin⁺ fibroblasts were negative for CD31 that stained the junctions between vimentin⁺ vascular endothelial cells (see Supplementary material online, Figure S1A and B). Myofibroblasts were absent from the normal heart, and α-SMA expression was restricted to vascular smooth muscle cells, that unlike myofibroblasts also expressed desmin (Figure 1B and see Supplementary material online, Figure S1A). After banding the pulmonary veins for 7 weeks, myofibroblasts accumulated between cardiomyocytes (Figure 1B and see Supplementary material online, Figure S1B). Quantification of the area occupied by α-SMA⁺/vimentin⁺/desmin⁻ myofibroblasts in immunostained sections demonstrated approximately four-fold increase in the right ventricle myocardium of banded compared with sham-operated animals (Figure 1A).

Figure 1

Integrins αvβ5 and αvβ3 are upregulated in a pig model of right ventricular hypertrophy. A Yorkshire pig model of pulmonary vein stenosis for 7 weeks was used to induce right ventricular cardiac hypertrophy and fibrosis. (A) Transmural blocks of the right ventricular myocardium from pulmonary vein-banded and sham-operated animals were sectioned. Sections were stained with H&E to visualize cardiomyocytes, with Masson's trichrome to demonstrate collagen (light blue), and stained for α-SMA (red), desmin (blue), vimentin (vim, green), and nuclei (white) to identify α-SMA⁺/desmin⁻ myofibroblasts. Randomly selected high-power fields from right ventricular-free walls were quantitatively analysed for cardiomyocyte diameter (H&E stain), the area of myocardial collagen content (percentage of blue stain in the image field). Myofibroblast content was quantified in low-power-field images as the as the area occupied by α-SMA⁺/vimentin⁺/desmin⁻ cells. (B) Sections were stained for α-SMA (red), α-sarcomeric actin (α-sarc, stains cardiomyocytes in blue, not shown in high magnification images), and integrins β5 (ab1926, Millipore) and αvβ3 (LM609, Chemicon) (green). Scale bars: 50 µm. (C) Transmural blocks of the right ventricular myocardium were processed for western blotting of total extracts and nuclear fractions. Protein expression was evaluated by quantitative western blotting and normalized to vimentin for total tissue extracts and lamin B for nuclear fractions. Error bars are the SD of the mean (n = 3, ***P ≤ 0.005, **P ≤ 0.01, *P ≤ 0.05, Student's t-test compared with sham-operated control).

We next elucidated whether induction of right ventricular fibrosis and myofibroblast activation in porcine right ventricles possibly causes changes in TGF-β1 signalling and expression of TGF-β1-activating integrins. Expression of β5 integrin was not detected in fibroblasts and β3 integrin was only expressed in vascular smooth muscle cells of sham-operated animal right ventricles, as assessed by immunofluorescence staining (Figure 1B) and western blotting (Figure 1C). However, both integrins were up-regulated in the hypertrophic heart with strongest expression in myofibroblasts. Correlating with increased levels of integrins β3 and β5 and α-SMA, fibrotic right ventricles exhibited higher levels of phosphorylated and total Smad3 in nuclear extracts compared with controls (Figure 1C), indicating enhanced active TGF-β1 signalling.

3.2 Integrin αvβ5 and αvβ3 are up-regulated during human hCF-to-myofibroblast activation in vitro

Next, we investigated changes in αvβ5 and αvβ3 integrin expression accompanying human hCF-to-myofibroblast activation in vitro. Consistent with our in vivo model, we used hCFs from ventricles that were previously shown to exhibit distinct characteristics compared with atrial fibroblasts.²³ hCFs were treated with TGF-β1 to stimulate myofibroblast differentiation, which was suppressed by the TGF receptor inhibitor SB431542 or culture on soft silicone rubber substrates.¹³ Conventional tissue culture spontaneously induced myofibroblast differentiation (α-SMA expression) in 75% of hCFs. Immunostaining (Figure 2A) and quantitative western blot analysis (Figure 2B and C) demonstrated high baseline expression levels of the myofibroblast marker α-SMA, integrins β5, β3, and αv in untreated control hCF. All proteins were upregulated in TGF-β1-treated hCFs (90% α-SMA⁺), and downregulated by treatment with SB431542 (5% α-SMA⁺) and growth on soft substrates (20% α-SMA⁺) (Figure 2A–C). Hence, porcine cardiac myofibroblasts in vivo and human cardiac myofibroblasts in vitro are characterized by high expression levels of the latent TGF-β1-binding integrins αvβ5 and αvβ3.

Figure 2

Integrins are upregulated in human cardiac myofibroblasts. (A) hCFs were cultured in the presence and absence of TGF-β1, treated with TGF-β1 inhibitor (SB431542) and grown on soft silicone substrate for 5 days. Cells were then immunostained for β5 integrin (ab15459, Abcam, red), β3 integrin (Axum-4, green), and α-SMA (blue). Scale bar: 10 µm. (C and D) Protein expression was evaluated by quantitative western blotting, normalized to the vimentin loading control, and presented as the percentage of the untreated control. Error bars represent the SD of the mean (n = 3, **P ≤ 0.01, ANOVA).

3.3 Integrin αvβ5 and αvβ3 compensate for each other in expression and function

To determine whether hCFs can use αvβ5 integrin, αvβ3 integrin, or both to activate TGF-β1, we first transiently transfected hCFs to overexpress GFP-tagged β3 and β5 integrin. The protein products of both constructs localized to and upregulated the respective integrin intensity in vinculin-containing focal adhesions (FAs; Figure 3A and see Supplementary material online, Figure S2). Conversely, transfecting hCFs with shRNA directed against integrins β5 and β3 eliminated the respective integrin in FAs (Figure 3B) compared with controls (Figure 3C). Transfection with GFP-tagged β integrins increased the levels of the overexpressed β integrin subunit by ∼1.6-fold, and shRNA downregulation with specific shRNA decreased the respective expression levels by ∼1.5-fold compared with control as assessed by western blotting (Figure 3D–F). Interestingly, both β subunits were subject to compensatory regulation after manipulating β5 and β3 integrin expression. Overexpression of β3 integrin lead to decreased expression of β5 integrin and vice versa as revealed by immunofluorescence and quantitative western blotting (Figure 3). Even stronger compensatory expression was observed after knock-down of one β subunit, resulting in the increased expression of the other (Figure 3). The expression levels of the αv integrin subunit were not significantly changed in any condition, and cells exhibited overall normal cell morphology and spreading (Figure 3).

Figure 3

Effect of αvβ5 and αvβ3 integrin overexpression and knock-down on FA formation. hCFs were transfected with (A) GFP-tagged constructs to express αvβ5 and αvβ3 integrins, (B) shRNA targeting mRNA of both integrins, and (C) control non-targeting shRNA constructs. After 5 days of culture, cells were co-stained for β5 integrin (ab15459, Abcam, green) and β3 integrin (Axum-4, red), and confocal images were taken. Scale bar: 30 µm. (D) Transfected and control hCFs were extracted, run on 18% gels, transferred for western blotting, and probed with antibodies directed against β5 integrin (ab15459, Abcam, red) and β3 integrin (Axum-4). Band signals of (E) β5 integrin and (F) β3 integrin were analysed from western blots by densitometry, normalized to vimentin loading control, and presented as the percentage of non-transfected cells. Error bars represent the SD of the mean (n = 6, ***P ≤ 0.005, **P ≤ 0.01, *P ≤ 0.05, ANOVA).

Next, we tested whether β5 and β3 integrins also compensate for each other with respect to latent TGF-β1 activation and myofibroblast regulation. Overexpressing both integrins individually significantly increased the amount of active TGF-β1 released by hCFs to approximately four-fold of control, as measured in TMLC co-culture (Figure 4A). Higher levels of active TGF-β1 correlated with significantly higher levels of α-SMA expression (Figure 4B and C), α-SMA incorporation in stress fibres (Figure 4D), and higher levels of total TGF-β1 (see Supplementary material online, Figure S3), indicating a feed-forward loop of myofibroblast activation. Consistent with integrin compensation at the protein expression level, knock-down of β5 and β3 integrins alone only moderately reduced the release of active TGF-β1 to ∼70% (Figure 4A) and α-SMA expression to ∼60% of control (Figure 4B–D). However, simultaneous knock-down of both integrins lead to a significant decrease of TGF-β1 release to ∼50% (Figure 4A) and α-SMA expression to ∼40% of control (Figure 4B and C). Double-integrin knock-down resulted in dramatically reduced stress fibre incorporation of α-SMA (Figure 4D).

Figure 4

Effect of overexpression and knock-down of αvβ5 and αvβ3 integrins on active TGF-β1 and α-SMA expression levels. hCFs were transfected with GFP-tagged constructs to express αvβ5 and αvβ3 integrins and GFP only, and shRNA targeting both integrins separately, simultaneously, and non-targeting shRNA. (A) After 7days of culture, transformed TMLC reporter cells were seeded on top of the hCF to report the levels of active TGF-β1 in the co-culture. Active TGF-β1 levels were normalized to total TGF-β1 obtained by heat-activating all TGF-β1 and presented as the percentage of non-transfected control hCF. (B) Transfected and control hCFs were extracted for western blotting. (C) Expression of α-SMA was quantified, normalized to vimentin loading control, and shown as the percentage of non-transfected control hCF. Error bars represent the SD of the mean from nine (A) and six (C) independent experiments (***P ≤ 0.005, **P ≤ 0.01, *P ≤ 0.05, ANOVA). (D) Transfected and control hCFs were immunostained for expression of α-SMA (green) and nuclei (blue). Scale bar: 50 µm.

3.4 Pharmacological inhibition of integrins αvβ5 and αvβ3 in hCFs reduces TGF-β1 activation and myofibroblast differentiation

These results suggested that strategies targeting cardiac fibroblast integrins to inhibit TGF-β1 activation will need to consider the simultaneous blocking of αvβ5 and αvβ3 integrins. To test whether acute pharmacological blocking of single integrins leads to compensation with respect to TGF-β1 activation by hCF contraction, we inhibited integrin-substrate binding for 1 h. Immunostaining of hCFs delivered the surprising finding that acute blocking of αvβ3 integrin was without effect on the FA localization of αvβ5 integrin and vice versa. This result was different from integrin manipulation at the expression and transcription level. However, function-blocking antibodies directed specifically against αvβ5 and αvβ3 integrins strongly reduced the presence of the respective integrin in FAs compared with controls (Figure 5A). The same effect was achieved using integrin function-blocking cyclic peptides. EMD66203 (anti-αvβ3 pep) has been reported to primarily target αvβ3 integrin and EMD121974 blocks αvβ5 and αvβ3 integrins in endothelial cells.²⁴ In our hands, EMD121974 used at 1 µM concentration primarily affected the FA localization of αvβ5 integrin in hCFs and was accordingly termed anti-αvβ5 in our study (Figure 5A). Acute inhibition of either αvβ3 or αvβ5 integrin did not visibly alter cell morphology and spreading (Figure 5A).

Figure 5

Acute effect of αvβ5 and αvβ3 integrin inhibition on TGF-β1 activation. hCFs were cultured for 5 days before adding peptide inhibitors directed against integrins αvβ5 (EMD121974), αvβ3 (EMD66203), and scrambled control (EMD135981) (1 µM) or blocking antibodies (Ab, 10 µg/mL). (A) Cells were immunostained after 2 h treatment for β5 integrin (ab15459, Abcam, green) and β3 integrin (Axum-4, red). Scale bar = 20 µm. (B) Active TGF-β1 levels produced by hCFs were determined by direct co-culture of luciferase-producing TMLC reporter cells. Reporter cells were allowed to adhere for 4 h before the addition of integrin function-blocking antibodies and -blocking peptides for 2 h. This treatment was followed by 1 h incubation with or without 0.5 U/mL of thrombin. The dotted line indicates basal latent TGF-β1 activation by myofibroblasts; the dashed line demonstrates active TGF-β1 levels after inducing contraction with thrombin in control conditions (BSA and control peptide). Results are shown as the percentage of non-stimulated control (BSA). Error bars represent the SD of the mean (n = 3, **P ≤ 0.01, *P ≤ 0.05, ANOVA).

Next, integrin binding was inhibited for 1 h before stimulating hCF contraction with thrombin in co-cultures with TMLC that report active TGF-β1. Basal levels of active TGF-β1 were upregulated approximately two-fold by inducing hCF contraction using thrombin in control conditions (Figure 5B and see Supplementary material online, Figure S4). This effect was potentiated by adding Mn²⁺ that non-specifically activates integrins. Pre-treating hCFs for 1 h with anti-αvβ5 antibody, anti-αvβ3 peptide, and anti-αvβ5 peptide completely blocked contraction-induced TGF-β1 activation (Figure 5B). The same antagonists also reduced baseline levels of active TGF-β1 in the absence of thrombin. The strongest blocking effect was achieved with anti-αvβ5 peptide, reducing active TGF-β1 to ∼30% of contraction-induced baseline and to ∼20% without thrombin (Figure 5B). None of the effectors influenced TMLC reporter activity in the absence of myofibroblasts (see Supplementary material online, Figure S4). TMLC reporter activity in hCF co-culture was almost completely blocked by anti-TGF-β1 antibodies, demonstrating that TGF-β1 is the main isoform released in these conditions (see Supplementary material online, Figure S4). Taken together, these results demonstrate that integrins αvβ5 and αvβ3 independently activate TGF-β1 and that cardiac myofibroblast contraction enhances the release of latent TGF-β1 from the latent complex.

3.5 Prolonged inhibition of TGF-β1-activating integrins blocks cardiac myofibroblast differentiation

Finally, we tested whether inhibition of integrins αvβ5 and αvβ3 with function-blocking peptides impairs cardiac myofibroblast differentiation by reducing TGF-β1 activation. hCFs were treated with specific blocking peptides in increasing concentrations for 5 days. Treatment with anti-αvβ3 and anti-αvβ5 peptides (>1 µM) led to reduced expression of β3 and β5 integrin in FAs compared with control as assessed by immunofluorescence (Figure 6A). Single integrin blocking did not substantially affect the formation of vinculin-positive FAs or overall cell morphology (Figure 6A). However, combined treatment with both peptides resulted in hCF detachment already at 0.1 µM (unpublished data).

Figure 6

Long-term effect of αvβ5 and αvβ3 integrin inhibition on TGF-β1 activation and myofibroblast differentiation. hCFs were cultured for 5 days in the presence of daily renewed peptide inhibitors directed against integrins αvβ5 (EMD121974), αvβ3 (EMD66203), and scrambled control (EMD135981). (A) Cells were then stained for β5 (ab15459, Abcam) or β3 (Axum-4) integrins (green), the FA protein vinculin (red), and α-SMA (blue, insets and bottom row). Scale bars: 25 µm. (B) After 5 days of treatment, levels of active TGF-β1 were measured using the TMLC luciferase reporter direct co-culture assay. Active TGF-β1 levels were normalized to total TGF-β1 obtained by heat-activating all TGF-β1. (C) The percentage of hCFs expressing α-SMA in stress fibres was quantified by semi-automatic image analysis of immunostaining. Values are expressed as a function of integrin-blocking peptide concentration. (D) hCFs were cultured for 5 days in the presence and absence of TGF-β1 and peptide integrin inhibitors (1 µM) and the percentage of α-SMA⁺ hCFs was quantified as in C. (B–D) Error bars represent the SD of the mean (n = 4, ***P ≤ 0.005, **P ≤ 0.01, *P ≤ 0.05, ANOVA).

Blocking integrins αvβ5 and αvβ3, respectively, resulted in reduced TGF-β1 activation by hCF contraction, whereas levels of total TGF-β1 were not affected (see Supplementary material online, Figure S3). Reduction of active TGF-β1 was 1.4-fold at 1 µM and two-fold at 10 µM compared with the control peptide EMD 135981 (Figure 6B). Concomitantly, the percentage of hCFs expressing α-SMA was decreased to 20% with 1 µM and 10% with 10 µM blocking peptides, as quantified from immunofluorescence pictures (Figure 6A and C). Importantly, addition of exogenous TGF-β1 with simultaneous integrin inhibition was able to restore control levels of α-SMA expression (Figure 6D). Collectively, these results suggest that chemically blocking either β3 or β5 integrin is sufficient to reduce the capacity of hCF to activate TGF-β1 and to maintain myofibroblast differentiation without impacting on cell adhesion.

4. Discussion

Remodelling of the cardiac ECM by fibroblasts adapts the heart to high load. Under chronic overload, this initially protective mechanism can develop into fibrosis. In cardiac fibrosis, myofibroblasts are activated from cardiac fibroblasts and replace functional heart muscle with collagenous ECM.^3,25,26 TGF-β1 is the most potent pro-fibrotic cytokine known to drive myofibroblast differentiation;⁴ inhibition of TGF-β1 prevents late cardiac remodelling in a mouse model of heart fibrosis.²⁷ Global therapeutic inhibition of TGF-β1 is problematic owing to its pleiotrophic character,⁷ e.g. knockout of TGF-β1 leads to the development of multifocal inflammatory disease in mouse models.²⁸ We hypothesized that inhibiting TGF-β1 activation specifically in myofibroblasts will be a safer and more efficient approach to prevent cardiac fibrosis. We have previously shown that cultured lung myofibroblasts mechanically activate TGF-β1 from latent stores in a sufficiently stiff ECM by pulling on LAP via integrins.¹³ These findings established a direct link between the mechanical conditions prevalent in fibrotic organs and the availability of active TGF-β1 for myofibroblast activation. Our current data identify integrins αvβ5 and αvβ3 to activate latent TGF-β1 in hCFs. Blocking these integrins suppresses the development of contractile myofibroblasts and potentially intercepts the vicious cycle of developing fibrosis.

TGF-β1-activating integrins play a fundamental role in the onset and progression of a variety of fibrotic diseases.^9,10 Knock out of integrin subunits that activate TGF-β1, including β6,²⁹ αv,³⁰ and β8³¹ and mutation of the integrin-binding site in LAP,³² all recapitulate aspects of the TGF-β1 knockout mouse.²⁸ Integrin αvβ6 is best studied in the context of TGF-β1 activation and fibrosis.^17–20 The lungs of β6 integrin-knockout mice are protected from bleomycin-induced fibrosis,¹⁸ and controlled delivery of β6 integrin-blocking antibodies reduces animal lung fibrosis.³³ Mice deficient for β6 integrin exhibit attenuated fibrosis in models of kidney fibrosis^34,35 and bile-duct ligation leading to acute biliary fibrosis.³⁶ However, αvβ6 integrin-knockout mice are not protected against carbon tetrachloride-induced liver fibrosis,³⁶ indicating that αvβ6 does not contribute to TGF-β1 activation in certain types of organ fibrosis. Inhibition of αvβ6 integrin will not be effective for fibrosis of the heart that does not contain αvβ6 integrin expressing cells. In the heart, fibroblastic cells represent a substantial fraction of resident non-cardiomyocytes and conceivably contribute to local TGF-β1 production and activation.

We reveal a novel functional role of integrins αvβ5 and αvβ3 in regulating cardiac fibroblast-to-myofibroblast differentiation through activation of TGF-β1. Both integrins are up-regulated during cardiac myofibroblast differentiation in our porcine model of cardiac fibrosis and correlate with the levels of α-SMA expression and myofibroblast differentiation in cultured hCFs. These results are consistent with the finding that αvβ5 integrin expression is up-regulated in rat cardiac fibroblasts after treatment with TGF-β1 and angiotensin II.²¹ Consistent with our results using hCFs, inhibition of αvβ5 integrin was shown to abolish contraction-mediated activation of TGF-β1 in cultured rat lung myofibroblasts^13,37 and airway smooth muscle cells,³⁸ and reduced α-SMA expression in scleroderma myofibroblasts.³⁹ A similar role has been suggested but not proved for αvβ3 integrin⁴⁰ that is expressed in cardiac fibroblasts.⁴¹ Our results demonstrate up-regulation of αvβ3 integrin in cardiac myofibroblasts and its direct implication in TGF-β1 activation and α-SMA expression.

A major finding of our study revealed that αvβ5 and αvβ3 integrins can compensate for each other's function in activating TGF-β1 and promoting myofibroblast differentiation. Compensation possibly explains the absence of a wound healing phenotype in integrin αvβ5 and αvβ3 knockout animals;⁴² wound healing is promoted by TGF-β1-induced myofibroblast actions. A compensatory mechanism has been demonstrated previously where β5 integrin was able to rescue angiogenesis defects caused by mutations in the β3 subunit.⁴³ In an animal model of liver fibrosis, inhibition of αvβ3 integrin was shown to augment instead of reducing collagen.⁴⁴ Other TGF-β1-mediated disease processes appear to be specifically mediated by αvβ5 integrin. For instance, mice deficient for αvβ5 integrin are protected from enhanced vascular permeability in response to ventilation-induced lung injury.⁴⁵ Interestingly, β3/β5 integrin double-knockout mice are not protected from lung fibrosis, indicating that β6 integrin is sufficient in this model to promote TGF-β1 activation.⁴⁶ It remains to be shown whether αvβ5 and αvβ3 integrin single- and double-knockout mice will be protected from cardiac fibrosis or fibrosis models where TGF-β1 activation relies on myofibroblasts. Notably, compensation does not occur when integrin function is blocked pharmacologically. In contrast to β-integrin subunit protein expression changes, integrin function-blocking per se would not dissociate the αβ heterodimer to make the αv integrin available for a new pairing in the same cell. Hence, function-blocking strategies targeting either αvβ3 or αvβ5 integrin may indeed be successful to reduce cardiac fibrosis with a greater effect being expected for conjoint inhibition of both integrins. Notably, long-term inhibition of integrin functions downregulated expression levels not only of α-SMA but also of the respective integrins, possibly by interrupting the feed-forward loop of TGF-β1 activation. Another possible explanation is that non-liganded integrins are internalized and eventually degraded.

The question remains to what extent integrins αvβ5 and αvβ3 directly contribute to TGF-β1 activation and myofibroblast differentiation in hCF. Inhibiting αvβ5 and αvβ3 integrins reduced levels of active TGF-β1 in our hCF cultures to 10–30% of control, depending on the blocking strategy and time; this reduction was sufficient to abolish myofibroblast differentiation almost completely. Reduced cell adhesion did not seem to contribute to these changes since vinculin-positive adhesions remained unaltered and addition of exogenous TGF-β1 rescued the myofibroblast phenotype in integrin-blocking experiments. αv integrins activate TGF-β1 by two different modes of action.⁹ Whereas αvβ8 integrin-mediated TGF-β1 activation depends on proteases,^10,14 αvβ6, αvβ5, and αvβ3 act independently of proteolysis by transmitting cell contraction forces to the ECM-bound latent TGF-β1 complex.^13,20,38 During mechanical TGF-β1 activation, a conformational change occurs in the latent complex that releases active TGF-β1.^15,16 The relative contribution of mechanical vs. protease-mediated TGF-β1 activation in cardiac fibrosis is unknown. However, a number of studies support that mechanical TGF-β1 activation by myofibroblasts dominates in progressive fibrosis.^9,13,16 Our previous study showed that inhibition of proteases reduces TGF-β1 activation by cultured myofibroblasts by only 20%, whereas inhibition of integrin-mediated contraction blocks the remaining 80%.¹³ A contraction-driven mechanism of TGF-β1 activation is particularly effective in the mechanical conditions of the stiff fibrotic scar ECM and the presence of highly contractile myofibroblasts.^4,11 Consistently, mechanical stress triggers cell reactions when the heart is stretched beyond its normal extension, e.g. by pressure overload. Chronic hypertension causes pathological remodelling, increases the proportion of fibroblastic cells over cardiomyocytes, and induces myofibroblast differentiation.⁴⁷

In conclusion, in contrast to globally inhibiting TGF-β1, which often presents uncontrollable side effects,^7,10 inhibition of mesenchymal integrins would specifically suppress the development of the fibrogenic myofibroblast phenotype. Our data suggest that inhibition of integrins αvβ5 and αvβ3 may be a promising strategy to block hCF-to-myofibroblast differentiation in heart fibrosis. In vivo evidence supporting this hypothesis is yet to be produced. It remains to be shown whether our results with cultured hCF can be reproduced in integrin-knockout animal models of heart fibrosis or by administrating integrin-blocking agents to hypertensive or infarcted hearts in animal models. In vivo, the potential side effects of inhibiting αvβ5 and αvβ3 integrins on vascular cells in the fibrotic environment will have to be evaluated.^24,48 In a recent study, pericyte-specific deletion of αv integrin and application of αv integrin inhibitors were shown to effectively suppress TGF-β1 activation and fibrosis of lung, liver, and kidney without causing adverse reactions.⁴⁹

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Funding

This work was supported by the Canadian Institutes of Health Research (CIHR) (grants #286920 and #210820); the Heart and Stroke Foundation Ontario (grant #NA7086), the Collaborative Health Research Programme (CIHR/NSERC) (grant #413783); and the Canada Foundation for Innovation and Ontario Research Fund (CFI/ORF) (grant #26653) to B.H. Animal experiments were supported by the Saving Tiny Hearts Society. C.L. was supported by a fellowship provided by the Ontario Graduate Scholarship Program (OGS) and E.Z. by a CIHR post-doctoral fellowship (grant #246193).

Acknowledgements

We thank Drs Christine Chaponnier, Giulio Gabbiani, Dean Sheppard, Bernhard Wehrle-Haller, and Dan Rifkin for kindly providing antibodies (anti-α-SMA and anti-β3 integrin), integrin plasmid constructs (ITGB3-GFP and ITGB5-GFP), and the TGF-β1 reporter cells. We thank Wei Hui from the Heart Centre Echocardiography Laboratory for performing Echocardiography and Marvin Estrada in the Animal Laboratory for care of the animals. We appreciate the contribution of participating surgeons, Drs Glen Van Arsdell, Osami Honjo, and Edward Hickey and technical support from Yaqin Yana Fu and Stellar Boo.

Conflict of interest: none declared.

Author notes

†

Both authors contributed equally to this study.