Discoidin Domain Receptor 2 signaling through PIK3C2α in fibroblasts promotes lung fibrosis

Abstract

Pulmonary fibrosis, especially Idiopathic Pulmonary Fibrosis (IPF), portends significant morbidity and mortality and currently therapeutic options are suboptimal. We have previously shown that type I collagen signaling through discoidin domain receptor 2 (DDR2), a receptor tyrosine kinase expressed by fibroblasts, is critical for regulation of fibroblast apoptosis and progressive fibrosis. However, the downstream signaling pathways for DDR2 remain poorly defined and could also be attractive potential targets for therapy. A recent phosphoproteomic approach indicated that PIK3C2α, a poorly studied member of the PI3 kinase family, could be a downstream mediator of DDR2 signaling. We hypothesized that collagen I/DDR2 signaling through PIK3C2α regulates fibroblast activity during progressive fibrosis. To test this hypothesis, we found that primary murine fibroblasts and IPF-derived fibroblasts stimulated with endogenous or exogenous type I collagen led to formation of a DDR2/PIK3C2α complex resulting in phosphorylation of PIK3C2α. Fibroblasts treated with an inhibitor of PIK3C2α or with deletion of PIK3C2α had fewer markers of activation after stimulation with TGFβ and more apoptosis after stimulation with a Fas-activating antibody. Finally, mice with fibroblast-specific deletion of PIK3C2α had less fibrosis after bleomycin than did littermate control mice with intact expression of PIK3Cα. Collectively, these data support that collagen/DDR2/PIK3C2α signaling is critical for fibroblast function during progressive fibrosis, making this pathway a potential target for anti-fibrotic therapy.

Introduction

Progressive fibrosis is a devastating condition that can be caused by environmental exposure, chronic inflammation or from primary conditions such as Idiopathic pulmonary fibrosis (IPF) [1,2]. Diverse inciting events can initiate pro-fibrotic signaling pathways, leading to accumulation of activated fibroblasts and deposition of fibrotic extracellular matrix proteins, especially type I collagen [3]. Some of these pathways may involve activation of a feed forward positive feedback loop that, once initiated, lead to further pro-fibrotic activation and progressive fibrosis. Therefore, these pathways may be of particular interest as therapeutic targets against fibrosis. We have previously reported that type I collagen, while often considered an end product of fibrosis, is produced early after injury [4] and can potentially promote further activation of pro-fibrotic cellular phenotypes through several mechanisms. Crossed-linked type I collagen fibers form a much stiffer matrix than the normal lung basement membrane, and this stiff matrix is thought to promote fibroblast activation [5–8]. Type I collagen can also act as a ligand by binding cell surface receptors initiating intracellular signal leading to pro-fibrotic cell behavior [9]. We have previously shown that two type I collagen receptors, α2β1 integrin and discoidin domain receptor 2 (DDR2), are important for fibrosis [10,11]. DDR2 is of particular interest because it is highly expressed on fibroblasts which accumulate in large numbers in the fibrotic tissue [12]. DDR2 is involved in cell survival and fibroblast activation in resonse to TGFβ [11,12]. Thus, collagen I deposition by activated DDR2-expressing fibroblasts can lead to further autocrine fibroblast activation and resistance to apoptosis leading to progressive fibrosis. Several groups have shown that DDR2 deficient mice are protected from fibrosis in animal models and several nonspecific inhibitors with activity against DDR2 have also been shown to protect in animal models of fibrosis raising the possibility that type I collagen/DDR2 signaling may represents an attractive target for therapy against fibrosis [11–16].

DDR2 is a receptor tyrosine kinase activated by fibrillar collagens. The downstream events of DDR2 signaling however remain poorly defined [17]. We recently showed that DDR2 regulates cell survival through a PDK1/Akt signaling pathway [11]. However, the immediate downstream events after DDR2 phosphorylation remained unknown. A recent phosphoproteomic approach identified SHP-2 as a downstream target signaling molecule phosphorylated by DDR2 [18]. Several other putative targets were identified but have yet to be validated, including NCK, Lyn and PIK3C2α. For the current report, PIK3C2α was of particular interest as it has also been reported to regulate PDK1/Akt signaling [19].

PIK3C2α is a member of the phosphoinositide 3-kinase (PI3 kinase) family of enzymes that mediate intracellular signaling through phosphorylation of various phosphatidylinositol molecules [20]. There has been considerable interest in the function of PI3 kinase signaling in the context of fibrosis and other pathologies [21]. PI3 kinase signaling regulates cell proliferation, survival, activation and migration. There is emerging data that there is complex crosstalk among the PI3 kinase family members [22]. The signaling pathways regulated by PI3 kinases can tightly regulate signaling through other PI3 kinases. In some cases cells can compensate for loss or inhibition of one of the PI3 kinases through signaling from other PI3 kinases. PI3 kinase signaling is important in multiple cells types involved in fibrosis including fibroblasts, epithelial cells and macrophages [23]. A number of PI3 kinase inhibitors are in preclinical or early phase clinical trials against pulmonary fibrosis [23,24]. The PI3 kinase family consists of three classes based on structure and substrate specificity [25]. The most utilized PI3 kinase inhibitors, such as wortmannin, primarily target class I PI3 kinases and have little activity against class II and class III PI3 kinase family members. Thus, the functions of class II and class III PI3 kinase family members in fibrosis have not been well studied. Recently, novel PI3 kinase inhibitors with activity against class II and class III PI3 kinase family members have been developed and have been proposed as potential treatments for cancer and other pathologies [20]. PIK3C2α is one of three class II PI3 kinase family members. PIK3C2α is involved in generation of PI(3,4)P2 and PI(3)P. It contains a clathrin-binding domain and a C2 domain [25]. In endothelial cells PIK3C2α has been shown to regulate VEGF and TGFβ signaling [26,27] and PIK3C2α activation has also been shown to regulate cell survival through a PDK1/Akt pathway [19].

In this report we found that type I collagen induces formation of a DDR2/PIK3C2α complex in fibroblasts leading to phosphorylation of PIK3C2α. Fibroblasts with inhibition or deletion of PIK3C2α had impaired activation after TGFβ and had exaggerated apoptosis. Mice with fibroblast specific deletion of PIK3C2α were protected from fibrosis. Finally, we show that IPF derived fibroblasts had greater activation of DDR2/PIK3C2α signaling.

Materials and methods

Reagents.

DDR2 antibody (Cat. # MABT322, clone 2B12.1) was from Millipore, Temecula, CA, USA. Type I collagen antibody (Cat. # PA5-29569) was from Thermo Fisher Scientific, Rockford, IL, USA. Antibody to phospho-PDK1 (Tyr376) (Cat. # 12481) was from Signalway Antibody, Greenbelt, MD, USA. Antibodies to PDK1 (Cat. # 3062), GAPDH (Cat. # 2118, clone 14C10), Akt (Cat. # 4691, clone C67E7), phospho-Akt (Thr308) (Cat. # 4056, clone 244F9), Smad2 (Cat. # 5339, clone D43B4), phospho-Smad2 (Cat # 3108, clone 138D4) and phospho-tyrosine (Cat. # 9411, clone P-Tyr-100) were from from Cell Signaling Technology, Danvers, MA, USA. The antibody to PI(3,4)P2 (Cat. # Z-PO34b) was from Echelon Biosciences, Salt Lake City, UT, USA. The Caspase Glo 3/7 kit was from Promega, Madison, WI. TGFβ was from R&D Systems, Minneapolis, MN, USA. The Fas-activating antibody (Cat. # 554254, clone Jo2) was from BD Biosciences, Franklin Lakes, NJ, USA. PIK90 was from Selleckchem, Houston, TX, USA. PITCOIN1 was from Glixx Laboratories, Hopkinton, MA, USA. Biotin-conjugated antibodies against mouse CD45 (Cat. #553078) and CD16/32 (Cat # 553143) were from BD Biosciences. Streptavidin-coated magnetic beads and the magnetic particle separator were from Thermo Fisher Scientific. WRG-28 was purchased from Aobious, Gloucester, MA, USA. Protein A and Protein G conjugated agarose beads were from Thermo Fisher Scientific. NuPage LDS sample buffer was from Invitrogen. All other reagents were from Sigma Pharmaceuticals, St. Louis, MO, USA.

Mice.

Mice were housed in specific pathogen-free conditions until the day of sacrifice. All animal experiments were approved by the University of Michigan Animal Care and Use Committee. Floxed Col1a1, floxed DDR2, floxed PIK3C2α and Col1a2-CreERT mice, all in a C57Bl/6 background, have been described previously [4,10,28–30]. Bleomycin or saline was delivered to six- to eight-week old mice by an oropharyngeal route at a dose of 2.8 U/kg as described previously [10]. Mice were euthanized at the timepoints indicated and lungs harvested for analysis.

For mice carrying the Col1a2-CreERT transgenes, recombination of the floxed allele was achieved by i.p. injections of tamoxifen (1 mg) as described previously [31] at the time points indicated. Littermate control mice lacking either the floxed allele or the Col1a2-CreERT transgene were also treated with tamoxifen and were used as controls.

In some experiments, mice were treated with daily doses of WRG-28 (10 mg/kg) versus vehicle control by i.p. injection for one week starting on day ten after bleomycin as described [32].

Hydroxyproline assay.

Lung collagen content was determined using a hydroxyproline assay as described previously [28,33,34]. In brief, lungs were isolated from mice sacrificed at the timpoints indicated. Lungs were homogenized and heated overnight in 12 M HCl at 120 °C. The samples were cooled to room temperature and then aliquots of each sample were neutralized with citrate acetate buffer. The samples were then incubated with chloramine T at room temperature. After 20 min, Erlich’s solution was added and the samples were incubated at 65 °C for 15 min. The Absorbance at 540 nm was measured and the hydroxyproline concentration was calculated against a standard curve.

Lung histology.

Mouse lung histology was assessed as described previously [34,35]. In brief, mice were sacrificed at the timepoints indicated. After exposing the thoracic cavity and trachea, the trachea was cannulated. The lungs were inflated with 10% neutral buffered formalin (NBF) at a pressure of 25 cm H₂O. The lungs were removed and fixed in NBF overnight. The lungs were then embedded in paraffin wax, sectioned and stained with Masson’s trichrome by the McClinchey Histology Lab (Stockbridge, MI, USA). The lung sections were visualized on a Nikon E-800 microcope and images were captured using NIS Elements software.

Cell isolation and culture.

Human IPF or normal lung fibroblasts and lung tissue were obtained from the University of Michigan biorepository core facility. Primary mouse lung fibroblasts were isolated as described previously [36]. In brief, mice were sacrified and lungs were lavaged and perfused with PBS. Lungs were removed and minced, then maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS and cultured in a 5% CO₂, 37 °C incubator to allow fibroblast outgrowth from the minced lungs. Cells and minced lungs were resuspended with trypsin and filtered through a 20-µm nylon filter to remove lung material. For in vitro experiments, fibroblasts were seeded onto tissue culture plates in DMEM 10% FBS and allowed to attach overnight. Cells were then rinsed with PBS and covered with serum-free DMEM and serum starved for 24 h. The medium was then changed to fresh serum-free DMEM containing 0.5 µM PIK90, 10 µM PITCOIN1 or vehicle control. After 1 h, TGFβ (4 ng/ml) was added to some wells. This dose of TGFβ was chosen based on prior dose response curves indicating a robust fibroblast response to TGFβ at this dose[37,38]. In some experiments, type I collagen, 50 µg/ml, was added to some wells. In some experiments, fibroblasts were treated with adenovirus (5 pfu/cell) expressing Cre recombinase (AdCre) or GFP (AdGFP) or LacZ (AdLacZ) for one week to ensure adequate time for recombination of the floxed gene and turnover of the existing protein prior to performing the experiment. Adenoviruses were purchased from the University of Iowa Vector core.

Alveolar epithelial cells (AECs) were isolated as described previously [36,39]. In brief, mice were sacrificed and lungs were lavaged and perfused with PBS. Lungs were filled with dispase followed by low melting temperature agarose. Lungs were removed and digested at room temperature for 45 min in dispase. Lungs were mechanically teased apart and then sequentially filtered through 70-µm, 40-µm and 20-µm filters. The crude cell suspension was incubated with biotin-conjugated anti-CD16/32 and anti-CD45 antibodies, followed by streptavidin-coated magnetic beads to remove leukocytes expressing these markers using a magnetic particle separator. AECs were further negatively selected by incubating on a tissue culture plate. Bronchoalveolar lavage (BAL) cells were isolated from mice as described previously [36]. In brief, mice were sacrificed and the lungs were sequentially lavaged with 1 ml of PBS for a total of ten lavage aliquots. Cells were collected from the pooled lavage by centrifugation. AECs and BAL cells were immediated lysed and analyzed for expression of PIK3C2α.

Active caspase 3/7 assay.

Apoptosis was assessed by measuring the amount of active caspase 3/7 activity as previously described[11,33,39]. In brief, fibroblasts were seeded and cultured overnight in serum-containing DMEM. In some experiments cells were then rinsed and treated with PIK90 (0.5 µM ), PITCOIN1 (10 µM) or vehicle control in serum-free medium. After 1 h, Fas-activating antibody (0.5 µg/ml) and CHX (5 µg/ml) (versus isotype control antibody and vehicle control) were added to the cells to induced apoptosis. After an additional 24 h, cells were washed twice with PBS and then lysed in RIPA buffer. The levels of active caspase 3/7 were measured using the Caspase Glo 3/7 assay following the manufacturer’s instructions. In brief, 5 µl of each sample was mixed with 25 µl of freshly prepared Caspase-Glo reagent in a 96 well solid white plate and allowed to incubated for one hour while protected from light. The amount of luminescence was quantified using a Veritas Microplate Luminometer using a 0.1 s integration time and 0 s delay time for each well. The values were normalized to total protein concentration. The values are expressed as relative levels compared to control conditions.

Immunoblotting and Immunoprecipitation

Immunoblotting and immunoprecipitaion of cell and lung lysates were performed as described previously [34]. For immunoblotting, cells were washed twice with PBS then lysed with RIPA buffer. Equal amounts of protein from each sample were incubated with NuPage 4x LDS Sample Buffer (Invitrogen) and heated to 70 °C for 10 min. Samples were separated by SDS-PAGE and transferred to nitrocellulose membranes. The membranes were blocked with 5% milk poweder in PBS with 0.5% Tween for 1 h at room temperature. Membranes were then incubated overnight with the appropriate primary antibody (1:1000 dilution in blocking buffer) at 4 °C overnight. The membranes were then washed three times with blocking buffer and then incubated with the appropriate HRP-conjugated secondary antibody (1:5000 dilution in blocking buffer). The membranes were then washed three times with blocking buffer followed by three additional washes with PBS. The membranes were then incubated with ECL solution and the immunoblot images were captured using an Amersham Bioimager 600RGB imager. The images shown are representative of at least six independent experiments. Densitometry was quantified using ImageJ by measuring the mean gray value of equally sized regions of interest for each band. Relative densitometry values are expresses as fold difference compared to the average of the control conditions. For immunoprecipitaton, cells were washed twice with PBS then lysed in RIPA. Equal amounts of protein for each sample were first precleared with protein A and protein G-conjugated agarose beads for 1 h at 4 °C. The beads were removed by centrifugation and the precleared supernatants were incubated with the appropriate primary antibody (1:100 dilution in RIPA) for 2 h at 4 °C. The samples were then incubated with protein A and protein G conjugated beads overnight at 4 °C. The beads were then washed three times with RIPA to remove proteins not bound the the beads. The precipitated proteins were then eluted by incubating the beads in LDS sample buffer and heating to 70 °C for 10 min. The beads were removed by centrifugation and the supernatants were then analyzed by immunobloting as described above. Levels of phosho-PIK3C2α were determined by immunoprecipitation using a phospho-tyrosine antibody (1:100 dilution in RIPA) followed by immunoblotting for PIK3C2α as described above.

Immunofluorescence

Cells were stained as described previously [4,35,40]. In brief, cells were cultured on glass chamber slides overnight. Cells were then rinsed with PBS and fixed with methanol for 15 min at 4 °C. Cells were then blocked with blocking buffer containing 5% normal goat serum, 1% BSA in PBS for 1 h at room temperature. Cells were then stained with anti-PI(3,4)P2 antibody (1:100 dilution in blocking buffer) overnight at 4 °C. Cells were rinsed three times with PBS and then incubated with goat anti-mouse IgG secondary antibody conjugated with AlexaFluor488 (1:200 dilution in blocking buffer) for 1 h at room temperature. Cells were rinsed three times with PBS and once with deionized distilled water and mounted with ProLong Gold containing DAPIand coverslipped. Slides were allowed to cure in overnight at room temperature protected from the light. Cells were visualized uaing a Nikon E-800 microcope and images were captured using NIS Elements software. Intensity of fluorescence was quantified with ImageJ.

Statistical analyses

Data are presented as dot plots with the median values indicated. Statistically significant differences between two groups were determined using two-tailed unpaired Student’s t-tests. Differences among more than two groups were determined by 2-way ANOVA with Tukey’s post hoc test for multiple comparisons. P-values of less than 0.05 were considered statistically significant.

Results

Collagen I/DDR2 dependent phosphorylation of PIK3C2α

PIK3C2α activity has been shown to be regulated by phosphorylation by receptor tyrosine kinases including EGFR and Insulin Receptor[25]. DDR2 is a receptor tyrosine kinase expressed on fibroblasts and activated by collagens[4,17,41]. We wanted first to determine if Collagen I-mediated active DDR2 can interact with PIK3C2α in fibroblasts. We used primary lung fibroblasts isolated from floxed Col1a1 mice and treated with adenovirus expressing Cre recombinase (AdCre) to first remove endogenous production of type I collagen, the primary collagen produced by lung fibroblasts. Collagen I-deficient fibroblasts were then seeded and treated with or without exogenously added type I collagen. After 1 h cells were lysed and immunoprecipitated for DDR2 (or IgG isotype control). DDR2 immunoprecipitates from collagen treated cells demonstrated co-precipitation with PIK3C2, indicating formation of a complex that includes DDR2 and PIK3C2α which is induced by type I collagen (Figure 1A and supplementary material, Figure S1). Notably, PIK3C2β did not co-precipitate with DDR2. To determine if type I collagen/DDR2 signaling results in phosphorylation of PIK3C2α we utilized floxed DDR2 and floxed Col1a1 mice. Primary lung fibroblasts isolated from floxed DDR2 mice were treated with AdCre (versus control adenovirus expressing GFP, AdGFP). After one week, cells were analyzed for PIK3C2α phosphorylation. As expected, floxed DDR2 cells treated with AdCre had a dramatic decrease in expression of DDR2. Total levels of PIK3C2α were unchanged but there was a dramatic decrease in levels of phosphorylated PIK3C2α (p-PIK3C2α) (Figure 1B and supplementary material, Figure S1). Next, we treated primary lung fibroblasts from floxed Col1a1 mice with AdCre or AdGFP as before. After one week, some floxed Col1a1 cells that were treated with AdCre were treated with exogenous type I collagen. Cells were lysed and analyzed by immunoblotting. Floxed Col1a1 cells treated with AdCre demonstrated diminished levels of p-PIK3C2α compared to AdGFP treated cells. Treating cells with exogenous type I collagen restored levels of p-PIK3C2α (Figure 1C and supplementary material, Figure S1). Collectively, these results indicate that type I collagen (either endogenously produced or exogenously added) mediates formation of a complex containing DDR2/PIK3C2α (although not necessarily directly complexed together) leading to PIK3C2α phosphorylation.

Figure 1. Collagen/DDR2 signaling phosphorylates PIK3C2α.

(A) Co-immunoprecipitation for DDR2 and PIK3C2α. Primary lung fibroblasts from floxed Col1a1 mice pre-treated with adenovirus expressing Cre recombinase (AdCre) were stimulated with or without exogenous collagen I (col) and immunoprecipitated for DDR2 (versus isotype control IgG). Cell lysate and immunoprecipitate were analyzed for DDR2, PIK3C2α and PIK3C2β by immunoblotting. (B) Floxed DDR2 fibroblasts treated with AdCre (versus control adenovirus expressing GFP, AdGFP) have reduced levels of p-PIK3C2α. (C) Floxed Col1a1 fibroblasts treated with AdCre (versus AdGFP control) have less PIK3C2α phosphorylation. Addition of exogenous col restores phospho-PIK3C2α levels.

PIK3C2α regulates fibroblast apoptosis and activation of the PDK1/Akt survival pathway

We have previously shown that DDR2 signaling regulates fibroblast apoptosis through activation of the PDK1/Akt survival pathway[11]. However, we could not identify formation of a DDR2/PDK1 complex (not shown) suggesting that DDR2 does not directly phosphorylate PDK1. To determine if DDR2-dependent activation of PIK3C2α regulates fibroblast apoptosis we utilized two novel inhibitors with activity against PIK3C2α. PIK90 is a potent inhibitor of PIK3C2α with a reported IC₅₀ of 0.05 µM however it also has inhibitory activity against other PI3 kinase family members[42,43]. PITCOIN1 is a novel inhibitor with excellent specificity for class II PI3 kinase family members, with greatest inhibitory effect against PIK3C2α (with a reported EC₅₀ of 5 µM for PIK3C2α-dependent plasma membrane tubulation and a dramatic decrease in PI(3,4)P2 formation and endocytosis at a concentration of 20 µM) with only modest effect against PIK3C2β and PIK3C2γ and minimal effect against other PI3 kinase family members or a panel of other kinases[44]. We treated IPF human lung fibroblasts with PIK90 (0.5 µM), PITCOIN1 (10 µM) or vehicle control. After 2 h, cells were lysed and analyzed for p-PDK1 and p-Akt. Inhibition of PIK3C2α led to a dramatic decrease in phosphorylation of both PDK1 and Akt without affecting total levels of these proteins (Figure 2A and supplementary material, Figure S2). We then treated IPF fibroblasts with PIK90 or PITCOIN1 as before. After two hours, cells were treated with Fas activating antibody (Fas) and cyclohexamide (CHX) to induce apoptosis. After an additional 24 h cells were lysed and analyzed for activation of caspase 3/7, a common final pathway in cellular apoptosis. As expected, Fas/CXH induced robust activation of caspase 3/7 in fibroblasts (Figure 2B). Fibroblasts treated with PIK90 or PITCOIN1 demonstrated a dramatic increase in caspase 3/7 activation.

Figure 2. PIK3C2α regulates fibroblast apoptosis.

(A) Fibroblasts treated with PIK90 or PITCOIN1 (PIT1) have reduced phosphorylation of Akt and PDK1 (B) Fibroblasts treated with PIK90 or PIT1 have greater activation of caspase 3/7 in response to Fas activating antibody (Fas) and cyclohexamide (CHX) compared to fibroblasts treated with controls. N=6 per group, *p<0.05, ***p<0.001. (C) Floxed PIK3C2α fibroblasts treated with AdCre have reduced expression of PIK3C2α and reduced activation Akt and PDK1. (D) Floxed PIK3C2α fibroblasts treated with AdCre and greater apoptosis in response to Fas and CHX compared to floxed PIK3C2α fibroblasts treated with control AdGFP. N=6 per group, ****p<0.0001.

While PIK90 is a potent inhibitor and PITCOIN1 is a specific inhibitor of PIK3C2α, we cannot rule out off-target effects by these inhibitors. To confirm the importance of PIK3C2α specifically we utilized floxed PIK3C2α mice. Primary lung fibroblasts were isolated from floxed PIK3C2α mice and were treated with AdCre versus control AdGFP. After one week, cells were lysed and analyzed for p-PDK1 and p-Akt as above. Similar to the result using the PIK3C2α inhibitors, loss of expression of PIK3C2α resulted in a dramatic reduction in PDK1 and Akt phosphorylation (Figure 2C and supplementary material, Figure S2). Next, AdCre and AdGFP treated floxed PIK3C2α lung fibroblasts were treated with Fas and CXH as above. After 24 h inhibitor, levels of active caspase 3/7 were assessed. Again, consistent with the results using a PIK3C2α inhibitor, loss of expression of PIK3C2α resulted in an increase in fibroblast apoptosis (Figure 2D). Collectively these data indicate that PIK3C2α regulates activation of PDK1/Akt, a prominent pro-survival signaling pathway, as well as the overall fibroblast apoptotic response.

Fibroblast activation in response to TGFβ is regulated by PIK3C2α

TGFβ is a prominent pro-fibrotic cytokine and can induce fibroblast activation resulting in upregulation of type I collagen and α-smooth muscle actin (α-SMA). IPF fibroblasts were treated with PIK90 or PITCOIN1 for 2 h as above, and then stimulated with TGFβ. After 24 h, cells were analyzed for expression of collagen I and α-SMA by immunoblotting. Inhibition of PIK3C2α with PIK90 or PITCOIN1 led to a significantly reduced upregulation of both collagen I and α-SMA compared to fibroblasts that had been treated with vehicle control (Figure 3A,B). As before, to confirm the importance of PIK3C2α we utilized floxed PIK3C2α fibroblasts. Floxed PIK3C2α cells were treated with AdCre and AdGFP as before. After one week, cells were stimulated with TGFβ. After an additional 24 h, cells were analyzed for expression of collagen I and α-SMA. First, we again confirmed robust loss of PIK3C2α expression in cells treated with AdCre. Also, similar to the results using the PIK3C2α inhibitors, fibroblasts with deletion of PIK3C2α demonstrated a dramatic reduction in TGFβ-induced expression of collagen I and α-SMA (Figure 3C,D). To confirm the importance of collagen I signaling in fibroblast activation, floxed PIK3C2α fibroblasts that had been treated with AdCre were treated with exogenously added collagen I and then stimulated with TGFβ as before. As expected, collagen-treated, PIK3C2α-depleted cells did show increased levels of collagen, but they did not have increased expression of collagen or α-SMA in response to TGFβ (supplementary material, Figure S3). Similarly, the addition of collagen to PIK3C2α-depleted cells did not affect the apoptotic response to CHX/Fas (supplementary material, Figure S3). Collectively, these results indicate that PIK3C2α is a regulator of collagen-mediated fibroblast behavior.

Figure 3. PIK3C2α regulates fibroblast activation in response to TGFβ.

(A–C) Fibroblasts treated with PIK90 or PITCOIN1 (PIT1) have less collagen I and α-SMA after TGFβ by (A) immunoblotting and quantified by densitometry for (B) collagen and (C) α-SMA. N=6 per group, ****p<0.0001. (D–F). Floxed PIK3C2α fibroblasts treated with AdCre have reduced expression of PIK3C2α and reduced expression of collagen I and α-SMA after TGFβ by (D) immunoblotting and quantified by densitometry for (E) collagen and (F) α=SMA. N=5 per group, *p<0.05, ***p<0.001.

PIK3C2α regulates fibroblast PI(3,4)P2 and Smad2 phosphorylation in response to TGFβ.

TGFβ binding to its receptor activates canonical intracellular signaling through Smad-dependent and Smad-independent pathways, leading to transcription of TGFβ-responsive genes [45,46]. In endothelial cells, a detailed mechanism by which PIK3C2α regulates TGFβ signaling has been established and involves PIK3C2α-dependent production of PI(3,4)P2, clathrin-dependent internalization of the TGFβ receptor, an interaction between SARA and Smad2/3, Smad phosphorylation [26,27]. To determine if PIK3C2α might regulate TGFβ signaling in a similar fashion, we focused on early and late events in this pathway. Floxed PIK3C2α fibroblasts were treated with AdCre versus a control adenovirus expressing LacZ (AdLacZ). After one week, cells were reseeded on glass chamber slides overnight. Cells were then fixed and analyzed for levels of PI(3,4)P2 by immunofluorescence. Floxed PIK3C2α cells treated with AdCre demonstrated a significant reduction in PI(3,4)P2 staining compared to floxed PIK3C2α cells treated with AdLacZ (Figure 4A–C). Next, we determined the effect of Cre-mediated removal of PIK3C2α on Smad2 phosphorylation in response to TGFβ. Thirty minutes after TGFβ, control fibroblasts demonstrated a marked increase in phospho-Smad2 while in PIK3C2α depleted fibroblasts the induction of Smad2 phosphorylation was dramatically reduced (Figure 4D and supplementary material, Figure S4).

Figure 4. PIK3C2α regulates fibroblast PI(3,4)P2 and Smad2 phosphorylation.

Immunofluorescence staining of floxed PIK3C2α lung fibroblasts pre-treated with (A) AdLacZ and (B) AdCre for cellular PI(3,4)P2. (C) Quantification of PI(3,4)P2 staining. N=8 per group, p<0.05. (D) Levels of Smad2 phosphorylation 30 min after TGFβ treatment in floxed PIK3C2α lung fibroblasts that been treated with AdCre or AdGFP.

Mice with fibroblast-specific deletion of PIK3C2α are protected from fibrosis

To determine if PIK3C2α signaling within fibroblasts is critical for lung fibrosis we generated mice with fibroblast specific deletion of PIK3C2α, by crossing floxed PIK3C2α mice with transgenic mice expressing CreERT regulated by a fibroblast specific portion of the Col1a2 promoter (Col1a2-CreERT). Six-week old Col1a2-CreERT/floxed PIK3C2α mice or littermate control mice lacking either the Col1a2-CreERT transgene or the floxed PIK3C2α allele were treated with every other day with doses of tamoxifen i.p. for one week (for a total of three doses). Fibroblast-specific loss of PIK3C2α expression in Col1a2-CreERT/floxed PIK3C2α mice was confirmed by immunoblotting of fibroblasts, while expression of PIK3C2α in AECs and BAL cells were similar to cells isolated from control mice (Figure 5A and supplementary material, Figure S5). After an additional week, mice were injured with bleomycin or given normal saline as a control. Three weeks after bleomycin injury mice were sacrificed and the extent of fibrosis analyzed by histology (Figure 5B,C) and by the hydroxyproline assay (Figure 5D). Mice with fibroblast-specific deletion of PIK3C2α demonstrated a marked reduction in bleomycin induced fibrosis compared to littermate control mice treated with bleomycin.

Figure 5. Fibroblast-specific deletion of PIK3C2α attenuates fibrosis.

(A) Fibroblasts, alveolar epithelial cells (AECs) and bronchoalveolar lavage (BAL) cells from Col1a2-CreERT/floxed PIK3C2α mice (ΔPIK3C2α ) treated with tamoxifen versus littermate control mice lacking one of the transgenes were analyzed for levels of PIK3C2α by immunoblotting. (B–D) Col1a2-CreERT/floxed PIK3C2α and littermate control mice were treated with tamoxifen starting two weeks prior to bleomycin injury. (B) Littermate control mice had greater fibrosis after bleomycin compared to (C) Col1a2-CreERT/floxed PIK3C2α mice as assessed using trichrome staining and a hydroxyproline assay. (E) Hydroxyproline assay of mouse tissue three weeks after bleomycin. Mice were treated with tamoxifen starting on day 7 after bleomycin, Col1a2-CreERT/floxed PIK3C2α had attenuated fibrosis. N=8 mice per group, ****p<0.0001.

We wanted next to determine if deletion of PIK3C2α after bleomycin injury, during the fibrogenic phase, would attenuate fibrosis. Col1a2-CreERT/floxed PIK3C2α mice, versus littermate controls lacking one or more of the transgenes, were injured with bleomycin or given saline. After seven days mice were treated with doses every other day of tamoxifen i.p. for one week. Three weeks after bleomycin injury, the amount of fibrosis was assessed using the hydroxyproline assay as before. Mice with fibroblast specific deletion of PIK3C2α after the onset of fibrogenesis demonstrated a marked reduction in bleomycin induced fibrosis compared to littermate control mice treated with bleomycin (Figure 5E).

We recently reported that DDR2 is a highly expressed fibroblast marker and that WRG-28, a novel inhibitor of DDR2, can attenuate bleomycin induced fibrosis[11,32]. To determine if inhibition of DDR2 coupled with fibroblast-specific deletion of PIK3C2α further reduces fibrosis after bleomycin, Col1a2-CreERT/floxed PIK3C2α mice, versus control mice, were treated with tamoxifen followed by bleomycin as before. Mice were then treated with daily i.p. doses of WRG-28 (10 mg/kg) starting on day ten and continued for one week. Twenty-one days after bleomycin, mice were analyzed for fibrosis as before. While both WRG-28 and deletion of PIK3C2α blocked fibrosis, the combination of the two strategies did not lead to a further reduction in fibrosis (supplementary material, Figure S6).

Fibroblasts derived from patients with IPF have greater activation of DDR2/PIK3C2α signaling.

We and others have shown that IPF fibroblasts are resistant to apoptosis and have greater activation in response to TGFβ[47,48]. We have also previously shown that IPF fibroblasts have greater activation of DDR2/PDK1/Akt signaling [11]. To determine if they also have greater activation of PIK3C2α we first measured levels of PIK3C2α phosphorylation. As expected, IPF fibroblasts demonstrated a statistically significantly higher levels of p-PIK3C2α compared to normal primary human lung fibroblasts, while total levels of PIK3C2α were similar between IPF and normal fibroblast (Figure 6A,B). We extended this analysis to lung tissue from patients with IPF and normal human lung tissue. Again, as expected, IPF lung tissue demonstrated marked upregulation of PIK3C2α compared to normal lung tissue (supplementary material, Figure S7). We and others have shown that IPF lung tissue has markedly increased levels of DDR2 compared to normal lung tissue [11,12,49] making assessment of a complex containing DDR2 and PIK3C2α difficult, however, total levels of DDR2 were similar between normal and IPF fibroblasts[11]. IPF and normal human lung fibroblast lysates were precipitated for DDR2 and then immunoblotted for PIK3C2α to assess for the extent of DDR2/PIK3C2α co-immunoprecipitation. Similar to the results for PIK3C2α phosphorylation, we found more PIK3C2α associated with DDR2 in IPF fibroblasts compared to normal human fibroblasts (Figure 6C,D). Collectively, these results support a mechanism in which collagen/DDR2/PIK3C2α signaling regulates fibroblast activation and survival during fibrosis.

Figure 6. IPF fibroblasts have greater activation of PIK3C2α.

(A) The amounts of PIK3C2α and p-PIK3C2α present in IPF and normal human fibroblasts was determined immunoblotting. (B) Densitometry quantification of p-PIK3C2α. N=8 per group, *p<0.05. (C) The amount of PIK3C2α associated with DDR2 was compared between IPF and normal human lung fibroblasts by co-immunoprecipitation. (D) Densitometry quantification of PIK3C2α co-precipitating with DDR2. N=8 per group, ***p<0.001.

Discussion

IPF and other fibrotic diseases are characterized in part by accumulation of fibroblasts and deposition of excessive extracellular matrix proteins [50]. There has been considerable interest in matrix proteins as regulators of progressive fibrosis [51,52]. Type I collagen is often considered to be an end product of fibrosis, but we and others have shown that collagen production begins early after injury [4]. Like most matrix proteins, type I collagen binds to specific cell surface signaling receptors resulting in changes in cell phenotype[9,17]. We and others have shown that several collagen receptors, including α2β1 integrin and DDR2 promote a pro-fibrotic fibroblast phenotype [10,11]. Thus, collagen I deposition could fuel further activation and recruitment of fibroblasts leading to accelerated collagen I deposition in an autocrine, positive feedback fashion. Disruption of this feed-forward loop thus represents a potentially attractive target for therapy. One possible approach is to target DDR2 itself. However, the downstream signaling events remain poorly defined and could also represent potentially attractive targets. Here we found that DDR2/PIK3C2α signaling regulates fibroblast acquisition of several important pro-fibrotic behaviors, including activation into myofibroblasts, production of type I collagen and resistance to apoptosis [53–55]. Cultured IPF-derived fibroblasts are known to retain some of their profibrotic characteristics even in culture conditions including resistance to apoptosis and excess production of collagen [47,48]. The excess production of collagen itself may account for some of the resistance to apoptosis in part through a DDR2/PIK3C2α pathway described in this report.

Interestingly, patients with homozygous loss-of-function mutations in PIK3C2α have been identified from three separate families and they are characterized by short stature and skeletal abnormalities reminiscent of what has been reported for DDR2-null mice [56,57]. Fibroblasts isolated from these patients exhibit decreased levels of PI(3,4)P2 similar to our findings using fibroblasts derived from floxed PIK3C2α mice. DDR2-null and PIK3C2α-deficient mice have both been reported to have severe vascular defects [12,29]. The extent to which these similar phenotypes are due to a direct functional interaction between DDR2 and PIK3C2α would be an interesting focus for further investigation. In this report we found that collagen drives DDR2 and PIK3C2α to co-precipitate in a complex, but it remains unknown if this is because they are directly complexed together or if there are other critical proteins needed for the functional activity of this complex.

The PI3 kinase family has been studied extensively in the context of fibrosis [21,58]. However, mechanistic detail on the function of class II and class III family members is only recently emerging. We found that PIK3C2α, one of three class two PI3 kinases, is a critical regulator of fibroblast activation in response to TGFβ (Figure 3) and survival against inducers of apoptosis (Figure 2). There is a growing understanding that there is a complex signaling network among the PI3 kinases and crosstalk with other signaling molecules including receptor tyrosine kinases and serine/threonine kinases [22,25,59]. In some cases, deletion or inhibition of one PI3 kinase may be compensated by increased activity of other PI3 kinases. This has potential implications as a mechanism of resistance in clinical trials targeting specific PI3 kinases and it is possible that a strong clinical effect will require inhibition of several PI3 kinases for optimal therapeutic effect. In our current study we did not determine if there was upregulation of other PI3 kinase activity in response to inhibition or deletion of PIK3C2α. A limitation to our experiments it that, to ensure optimal recombination the floxed PIK3C2α allele and degradation of the PIK3C2α protein, cells were cultured for one week after treated with AdCre which could have led to phenotypic changes to the cell during this time-period preceding the experimental treatment. Conversely, compensatory activity by other PI3 kinases could take longer than the timeframe of our experiments. However, the observed changes in fibroblast behavior were similar when cells were treated with PIK90, an inhibitor with activity against a broad spectrum of PI3 kinases, and with PITCOIN1, which has greater specificity against class II PI3 kinases [44]. This suggests that class II PI3 kinases might have a unique function making it an attractive target for specific therapy. Inhibition of DDR2 and fibroblast specific deletion of PIK3C2α led to a similar, incomplete inhibition of fibrosis (supplementary material, Figure S6). Interestingly, the combination of DDR2 and deletion of PIK3C2α did not lead to further reduction in fibrosis (supplementary material, Figure S6) which could suggest the importance of a compensatory or alternate pathway leading to fibrosis. At present, most PI3 kinase inhibitors have activity against multiple PI3 kinase family members [20]. The role of other members of the class II and class III PI3 kinase family in regulating fibrosis is poorly defined. Class I members of the PI3 kinase family have already been implicated in promoting fibrosis [23]. Inhibition of several PI3 kinase pathways may be necessary to inhibit fibrosis. However, given the importance of PI3 kinase signaling during homeostasis [21] it is unclear if a more narrow approach might actually be superior to broad inhibition of PI3 kinase activity with fewer off target effects.

One unique feature of class II PI3 kinases is the clathrin-binding domain [20,60]. Importantly, signaling through a number of receptor kinases requires clathrin-mediated endocytosis and local production of PIP3 and PI(3,4)P2. Although beyond the scope of this report, in future studies, it will be interesting to further define a detailed mechanism by which PIK3C2α regulates TGFβ signaling. Data from two reports demonstrate an elegant mechanism by which PIK3C2α, through its regulation of membrane PI(3,4)P2 formation, regulates clathrin-mediated TGFβ receptor internalization into early endosomes containing SARA (Smad anchor for receptor activation). This led to impaired association between SARA and Smad2 and Smad3, decreased Smad phosphorylation and decreased Smad nuclear localization [26,27]. Notably, SARA contains a PI(3)P-binding FYVE domain [61]. It is unclear if PIK3C2α dependent production of PI(3)P is important for recruitment of SARA into this complex or if this requires involvement from other PI3 kinases [20,26,29]. It will also be interesting to explore a potential role for PIK3C2α in non-Smad signaling and non-canonical TGFβ signaling as well.

PIK3C2α is of particular interest because it has been shown to regulate internalization and activation of the VEGF receptor in addition to the TGFβ receptor[26,27,29]. Importantly, nintedanib, one of two approved anti-fibrotic therapies, is a non-specific tyrosine kinase inhibitor thought to function in part through inhibition of VEGF signaling[62–64]. Other tyrosine kinase receptors inhibited by nintedanib include the FGF receptors and the PDGF receptor. In future studies it would be interesting to investigate if PIK3C2α regulates internalization and activation of these and other receptors. Interestingly, we have found that nintedanib is also a potent inhibitor of DDR2 [32], thus the mechanism by which nintedanib inhibits fibrosis remains incompletely understood and could include multiple steps of the DDR2/PIK3C2α pathway.

In summary, we have identified a type I collagen/DDR2/PIK3C2α signaling pathway which is critical for several pro-fibrotic fibroblast functions. DDR2 and PIK3C2α are thus attractive potential targets for therapy.

Supplementary Material

Supinfo

Figure S1. Relative densitometry for immunoblots in Figure 1

Figure S2. Relative densitometry for immunoblots in Figure 2

Figure S3. Floxed PIK3C2α fibroblasts treated with AdCre and exogenously added collagen

Figure S4. Relative densitometry for Figure 4D

Figure S5. Relative densitometry for Figure 5A

Figure S6. Inhibition of DDR2 in addition to fibroblast-specific deletion of PIK3C2α does not further reduce fibrosis in response to bleomycin

Figure S7. IPF lung tissue have greater activation of PIK3C2α

Data availability statement

All data are included in the manuscript and supplementary material.