Suppression of TGF-β1 signaling by Matrigel via FAK signaling in cultured human trabecular meshwork cells

The trabecular meshwork (TM) is composed of TM cells and beams of the extracellular matrix, together contributing to aqueous humor (AH) outflow resistance. Herein, we validated that our culture system on 2D Matrigel expressed putative TM markers and myocilin, of which the latter was upregulated by dexamethasone. Continuous passage of these cells on 2D Matrigel resulted in a gradual loss of expression of these markers. However, such a loss was restored by seeding cells in 3D Matrigel where expression of TM markers was further upregulated upon continuous passage. In contrast, TM cells seeded on fibronectin, collagen I/IV, or laminin lost expression of these markers and turned into myofibroblasts with expression of αSMA, which were dose-dependently upregulated by TGF-β1/TGF-β2. TM cells in 3D Matrigel also expressed TGF-β1/TGF-β3 despite challenge of TGF-β1. The maintenance of TM phenotype by 3D Matrigel was linked to inhibition of canonical TGF-β signaling and activation of pFAK-pSrc-pP190RhoGAP-P120RasGAP signaling. These findings indicate that basement membrane matrix with low rigidity plays an active role in maintaining TM phenotype in the presence of TGF-β1 and shed light on its physiological role. Furthermore, abnormal matrices may perpetuate the pathological TM phenotype when the level of TGF-β2 is elevated in glaucoma patients.


3D Matrigel maintains TM phenotype and TGF-β expression.
Similar to what we have reported 27 , TM cell morphology changed from more cuboidal cells to spindle cells after serial passage on 2D Matrigel in MESCM + 5% FBS from P1 to P7 (Fig. 1A). Also consistent with our report 27 , seeding of P2 TM cells from 2D Matrigel back to 3D Matrigel resulted in cell aggregation, which was maintained after continuous passage in 3D Matrigel to P7 (Fig. 1A). Serial passage on 2D Matrigel significantly diminished the transcript expression of severe different TM markers such as AQP1, CHI3L1, MGP, AnkG, TIMP3, ADRA2A, and CRYAB when compared to that from P0 TM cells (Fig. 1B). However, such a loss could be regained by seeding P2 cells in 3D Matrigel and expression of all seven TM markers except ADRA2A was further upregulated by continuous passage to P7 in 3D Matrigel (Fig. 1B). As reported 36 , the TM cell phenotype is highlighted by significant upregulation of MYOC mRNA and protein by dexamethasone, a TM phenotype not shared by neighboring cells (reviewed in 28,29 ). Before DEX treatment, our confluent TM cells were spindle on 2D Matrigel in MESCM + 5% FBS. They became more flatted after DEX treatment (Fig. 1C). Significant upregulation of MYOC mRNA and protein by DEX was noted (Fig. 1D,E). Therefore, these data collectively allowed us to conclude that TM cells expanded in 3D Matrigel maintained the TM cell phenotype. For the first time, we noted that the above TM phenotype was accompanied by expression of TGF-β1, TGF-β2, and TGF-β3. Serial passage on 2D Matrigel resulted in a significant loss of expression of TGF-β1, TGF-β2, and TGF-β3 transcripts (Fig. 1F) and proteins (Fig. 1G). Expression of TGF-β1 and TGF-β3, but not TGF-β2, was restored in 3D Matrigel (Fig. 1F,G).

3D Matrigel maintains TM phenotype despite exogenous TGF-β1.
To determine whether the aforementioned finding was unique to 3D Matrigel, P3 TM cells cultured on 2D Matrigel were seeded on different substrates including fibronectin, collagen I, collagen IV, laminin, 2D or 3D Matrigel. After 24 h, all TM cells attached and spread well except those on laminin, of which some were round and detached, whereas cells in 3D Matrigel uniquely formed aggregates ( Fig. 2A, left column). After switching to a serum-free medium with or without exogenous 10 ng/ml TGF-β1 for 24 h, TM cells on fibronectin, collagen I, collagen IV, laminin or 2D Matrigel became spindle ( Fig. 2A, middle column) and enlarged at 72 h after TGF-β1 treatment ( Fig. 2A, right column). In contrast, TM cells remained aggregates in 3D Matrigel ( Fig. 2A, bottom panel). These results strongly suggested that the TM cell morphology was uniquely maintained in 3D Matrigel despite being challenged by exogenous TGF-β1. Compared to the control cultured on fibronectin, the transcript expression of AnkG, CHI3L1, MGP, AQP1 TIMP3, ADRA2A and CRYAB was significantly upregulated in TM cells cultured in 3D Matrigel without exogenous TGF-β1 ( Fig. 2B-H, blue bars) while such upregulated transcript levels were maintained or further upregulated with exogenous TGF-β1 ( Fig. 2B-H, orange bars). Among all other substates, laminin upregulated transcript expression of CHI3L1, MGP and ADRA2A; collagen IV and 2D Matrigel upregulated transcript expression of ADRA2A without TGF-β1 ( Fig. 2B-H, blue bars). After TGF-β1 treatment, laminin upregulated transcript expression of AnkG, CHI3L1, MGP and AQP1, collagen IV upregulated transcript expression of MGP, AQP1 and ADRA2A, while collagen I only upregulated transcript expression of ADRA2A ( Fig. 2B-H, orange bars). Immunostaining disclosed strong cytoplasmic staining of CHI3L1 and MGP in cells in 3D Matrigel with or without TGF-β1 but weak or no expression in cells on all other substrates (Fig. 2I). Therefore, 3D Matrigel uniquely maintained TM phenotype despite being challenged by exogenous TGF-β1.
TGF-β1 and TGF-β2 dose-dependently upregulate αSMA expression and stress fibers on fibronectin but not in 3D Matrigel. To further demonstrate the aforementioned difference between 3D Matrigel and other substrates in the response to exogenous TGF-β, we performed a dose-response experiment by seeding P3 TM cells on FN or in 3D Matrigel with or without 1 to 10 ng/ml of TGF-β1 or TGF-β2 for 72 h. As expected, TM cells on fibronectin were enlarged after TGF-β1/2 treatment in a dose-dependent manner while those maintained aggregates in 3D Matrigel (Fig. 3A). On fibronectin, there was an increasing positive immunofluorescence staining of αSMA in TM cells in a dose-dependent manner after being treated with TGF-β1 or TGF-β2 and stress fibers became apparent after treated with 10 ng/ml of TGF-β1 or TGF-β2 (Fig. 3B, top panel). In contrast, TM cells cultured in 3D Matrigel did not express αSMA or stress fibers (Fig. 3B, bottom panel), confirming that 3D Matrigel was unique in withstanding the challenge of exogenous TGF-β1 and TGF-β2.
3D Matrigel suppresses myofibroblast differentiation by downregulating smad2/3 signaling. Because both TGF-β1 and TGF-β2 exhibited comparable dose-dependent responses in promoting myofibroblast differentiation as evidenced by expression of αSMA and stress fibers (Fig. 3), we thus selected 10 ng/ml of TGF-β1 for the remaining experiments. Compared to TM cells cultured on fibronectin, addition of TGF-β1 further upregulated expression of TGF-β1 and TGF-β3 transcripts ( Fig. 4A-C) and proteins ( Fig. 4D-F) and such upregulation was more prominent in TM cells cultured in 3D Matrigel. Among all other substrates, laminin and 2D Matrigel also upregulated expression of TGF-β3 transcript and protein (Fig. 4C,F). Expression of TGF-β2 transcript and protein was not affected by exogenous TGF-β1 on all substrates (Fig. 4B,E). Interestingly, cytoplasmic staining of fibronectin, αSMA, and nuclear pSmad2/3 staining were pronounced in cells cultured on www.nature.com/scientificreports/ fibronectin, collagen I, collagen IV and laminin following addition of TGF-β1 (Fig. 4G). TM cells showed mildly increased cytoplasmic staining of fibronectin, weak staining of αSMA, but negative nuclear pSmad2/3 staining when cultured on 2D Matrigel (Fig. 4G, fifth column). In contrast, cell aggregates in 3D Matrigel exhibited no staining of fibronectin and αSMA and cytoplasmic staining of pSmad2/3 (Fig. 4G, right column). Collectively, these results strongly suggested that canonical TGF-β signaling was activated in TM cells cultured on fibronectin, collagen I, collagen IV and laminin after addition of TGF-β1 and such signaling was accompanied by deposi- Folds of transcript expression of TM markers were measured by qRT-PCR using the expression level of P0 TM cells set as 1 (B, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001). Upon confluence, P3 TM cells on 2D Matrigel in MESCM + 5% FBS were treated with 100 nM dexamethasone (DEX) for 10 days. Cell morphology was analyzed by phasecontrast microscopy at Day 0, 2, 5 and 10 after DEX treatment (C, scale bar: 100 µm). qRT-PCR was used for quantitation of transcript levels of myocilin (MYOC) at Day 0, 2, 5 and 10 using the expression level by the cells cultured at Day 0 set as 1 (D, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001). Western blotting was used for MYOC protein level at Day 0, 2, 5 and 10 using β-tubulin as the loading control (E). In serially passaged cells, folds of transcript expression of TGF-βs were measured by qRT-PCR using the expression level of P0 TM cells set as 1 (F, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001). Levels of TGF-βs proteins in culture media were measured by respective ELISA (G, n = 3, *P < 0.05, **P < 0.01 and ***P < 0.001). www.nature.com/scientificreports/ tion of fibronectin and myofibroblast differentiation. In contrast, canonical TGF-β signaling was not activated in TM cells cultured on 2D or in 3D Matrigel despite addition of TGF-β1 and was not accompanied by fibronectin deposition and myofibroblast differentiation.

FAK inhibitor 14 downregulates TM markers, upregulates expression of matrix markers and αSMA, and activates canonical TGF-β signaling in 3D Matrigel.
To further evaluate the causal relationship mediated by the FAK-Src signaling, P3 cells cultured on 2D Matrigel in MESCM + 5% FBS were seeded in 3D Matrigel and then were added with 3 μM of FAK inhibitor 14 39 for 24 h before being switched to DMEM/ F12/ITS with 10 ng/ml TGF-β1 for 24 h. Addition of FAK inhibitor 14 did not change formed aggregates after addition of TGF-β1 (Fig. 6A). However, the upregulated transcript expression of CHI3L1, MGP, AQP1, TIMP3 and CRYAB by TGF-β1 in 3D Matrigel was significantly downregulated by FAK inhibitor 14 (Fig. 6B). The upregulated transcript expression of fibronectin, collagen I A1, and Laminin A1 by TGF-β1 in 3D Matrigel was further upregulated by FAK inhibitor 14 (Fig. 6C). Furthermore, addition of FAK inhibitor 14 promoted cytoplasmic staining of fibronectin and αSMA as well as nuclear staining of pSmad2/3 in TM aggregates cultured www.nature.com/scientificreports/ in 3D Matrigel in the presence of TGF-β1 (Fig. 6D). Therefore, blockade of FAK by its inhibitor 14 resulted in downregulation of TM markers, upregulation of matrix markers including fibronectin and αSMA, and activation of canonical Smad2/3-mediated TGF-β signaling in TM cells cultured in 3D Matrigel in the presence of TGF-β1.

Discussion
TM cells lie on beams of ECM, which consists of fibronectin, collagen IV, laminin and other substrates 5,40,41 . Glaucoma TM ECM is characterized by increased deposition of matrix components such as fibronectin, collagens type I, III, V, VI, XI, XII and XIV 42,43 . Following the binding of TGF-βs with TGF-βRI and TGF-βRII, the canonical signaling pathway leads to nuclear translocation of pSmad2/3 that triggers transcriptional activation of many downstream genes including fibronectin and αSMA that are found in pathological fibrosis in many tissues 44,45 (also depicted in Fig. 7A). Besides canonical Smad-mediated signaling, TGF-β also elicits non-canonical signaling such as RhoA-ROCK signaling leading to stress fiber formation and cytoskeletal reorganization 46 (depicted in Fig. 7A). Herein, our study showed that myofibroblast differentiation, fibronectin deposition and nuclear staining of pSmad2/3 were accompanied by the loss of expression of TM markers when TGF-β1 was added to TM cells cultured on fibronectin, collagen I, collagen IV or laminin (Fig. 4). Therefore, we conclude that an elevated TGF-β level will elicit canonical TGF-β signaling when TM cells are exposed to these scar-prone matrices, mimicking the pathological state of glaucoma as reported by others by culturing TM cells on plastic with or without laminin [47][48][49] , fibronectin or collagens 50 . Because altered ECM may act as a fibrogenic niche leading to tissue fibrosis 51 , we surmise that altered ECM plays a "pathological" role in perpetuating TM cells to respond to TGF-β from AH to adopt an abnormal phenotype with increased cellular contractility and deposition of altered ECM 23,24,52 .
Our study further validated that cells expanded in 2D Matrigel and reseeded in 3D Matrigel were indeed TM cells as evidenced by expression and further upregulation of all seven putative TM markers, i.e., MGP, CHI3L1, AQP1, AnkG, TIMP3, ADRA2A and CRYAB in cell aggregates after continuous passage (Fig. 1A,B). Furthermore, these TM cells also exhibited another TM feature, i.e., significant upregulation of MYOC by DEX (Fig. 1C-E) as reported 27 . Besides expression of these putative TM markers, our study also demonstrated for the first time that these TM cells expressed all three isoforms of TGF-β (Fig. 1), suggesting that expression of TGF-βs is another key feature of the normal TM phenotype. The expression of TGF-β3 by TM cells is particularly intriguing as TGF-β3 differs from TGF-β1 and TGF-β2 in exerting an anti-scarring action [53][54][55] and promoting deposition of normal ECM by corneal stromal cells 56,57 . Further studies are needed to determine the role of TM www.nature.com/scientificreports/ cells in producing TGF-βs present in AH and whether expression of TGF-β3 is an important feature to maintain TM phenotype. Unlike other substrates, TM cells cultured in 3D Matrigel maintained the normal but not "pathological" phenotype even in the presence of a high concentration of TGF-β1, which was comparable to TGF-β2 as judged by a dose-response relationship (Fig. 3) by suppression of canonical Smad-mediated TGF-β signaling (Fig. 4, also depicted in Fig. 7B). Matrigel, a basement membrane matrix extracted from Engelbreth-Holm-Swarm mouse sarcoma consisting of laminin, collagen IV, entactin/nidogen, heparan sulfate proteoglycans and growth factors such as TGF-β and EGF 58 , has matrix stiffness (rigidity) of 50 Pa (3D, 50% Matrigel) 59 , which is significantly lower than that of collagen I (more than 6 kPa) 60,61 , fibronectin (up to 40 kPa) 62 , and laminin (up to 250 Pa) 63 . Besides matrix rigidity, our study showed that one action mechanism leading to the suppression of canonical Smad-mediated TGF-β signaling in 3D Matrigel could be activation of FAK-mediated signaling (Fig. 5). Addition of a small FAK inhibitor resulted in pathological manifestation characterized by activation of canonical Smad-mediated TGF-β signaling, downregulation of TM markers, and upregulation of αSMA and matrix marker in TM cells in 3D Matrigel (Fig. 6). FAK is a non-receptor cytoplasmic protein tyrosine kinase and activated when cells bind to ECM proteins through integrin receptors [64][65][66][67][68] during cell adhesion on different substrates in cooperation with various growth factors including TGF-β [69][70][71][72][73][74][75] .
FAK can be activate by TGF-β1 at sites of integrin/matrix engagement 76 . TGF-β1 induces FAK activation in a time and dose dependent manner that precedes expression of αSMA and collagen deposition in liver fibrosis 77 . Activation of FAK is through phosphorylation on Y397 as a high-affinity binding site for the SH2 domain of Src family kinases and leads to the recruitment and activation of Src via Y416 78,79 . The downstream signaling of pFAK(Y397)-pSrc(Y416) may lead to αSMA cytoskeleton reorganization, profibrotic gene expression and matrix deposition through MEKK1-JNK 75 , PI3K-Hippo 80 , and RhoA-ROCK 76,81 . Therefore, TGF-β1 may activate RhoA-ROCK through non-canonical signaling with or without involvement of FAK (Fig. 7). However, FAK serves as a "switch" for multiple signaling outputs (for reviews see 64,82,83 ). For example, FAK is a critical component of a pathway leading to signals that either positively or negatively modulate the assembly and breakdown of adhesions at the leading and/or trailing edges of migrating cells 84 . The downstream signaling of pFAK(Y397)-pSrc(Y416) for TM cells in 3D Matrigel uniquely activated a complex of pP190RhoGAP (Y1105) and P120RasGAP (Fig. 5). This scenario has been reported in C3H10T1/2 murine fibroblasts 85 , chicken embryo fibroblasts, and vascular endothelial cells leading to cytoskeletal rearrangement via disassembly of actin stress fibers 85,86 to facilitate cell migration 85,87,88 , cell-cell contact 89 , and endothelial cell polarity and barrier function 90,91 . In 3D Matrigel, activation of downstream p190RhoGAP and p120RasGAP complex resulted in inhibition of RhoA activity as evidenced by S188 phosphorylation (Fig. 5), a hallmark of suppressed RhoA signaling 37,38 . Therefore, we conclude that activation of pFAK-pSrc-pP190RhoGAP-P120RasGAP signaling in 3D Matrigel uniquely suppresses both RhoA and TGF-β Smad-mediated signaling to maintain TM morphology and phenotype and prevent differentiation into contractility-prone myofibroblasts even under exogenous challenge of TGF-β1 (Fig. 7B). Further studies are needed to determine whether in vivo TM ECM possesses the biochemical and physical properties similar to 3D Matrigel and plays an active role in maintaining normal TM phenotype and withstanding the challenge of an elevated level of TGF-β2 in AH. Isolation, expansion and treatment of TM Cells on 2D Matrigel. Human donor corneas from deidentified cadaver sources were provided by Florida tissue banks and handled in a HIPAA-compliant manner in accordance with the tenets of the Declaration of Helsinki. This study was reviewed and approved by Tissue Tech Inc, Miami, Florida, USA. TM cells were isolated and expanded as previously reported 27 . In brief, human corneoscleral rims stored at 4 °C for less than 7 days were obtained from Florida Lions Eye Bank (Miami, FL). Under a dissecting microscope, the corneoscleral rim was stripped off the iris by forceps and was cut through the inner edge of Schwalbe's line delineated by pigmentation as reported by Keller et al. 29 . After rinsing with MESCM, the Descemet's membrane was stripped off. The remaining TM tissue was removed and digested with 2 mg/ml collagenase A at 37 °C for 16 h (overnight) in MESCM, which is made of Dulbecco's Modified Eagle's Medium (DMEM)/F-12 nutrient mixture (F-12) (1:1) with 10% knockout serum, 4 ng/ml bFGF, 10 ng/ml LIF, 5 ng/ml sodium selenite supplement, 5 mg/ml transferrin, 5 mg/ml insulin, 1.25 μg/ml amphotericin B, 50 μg/ ml gentamicin and 5% FBS 92 . The resultant cells were washed once in MESCM + 5% FBS and seeded at a density . 3D Matrigel suppresses myofibroblast differentiation by downregulating Smad2/3 Signaling. P3 cells cultured on 2D Matrigel in MESCM + 5% FBS were seeded on fibronectin, collagen I, collagen IV, laminin, 2D Matrigel, or 3D Matrigel for 24 h before being switched to DMEM/F12/ITS for 24 h with or without 10 ng/ml TGF-β1 for another 24 h. Expression levels of TGF-βs transcripts were measured by qRT-PCR at 24 h using the level expressed by cells cultured on fibronectin without (blue) or with (orange) TGF-β1 set as 1, respectively (A-C, n = 3, *P < 0.05 and # P < 0.05). Expression levels of TGF-βs proteins in culture media measured by ELISA were compared to cells cultured on fibronectin without (blue) or with (orange) (D-F, n = 3, *P < 0.05 or # P < 0.05). TM cells were subjected to immunostaining of pSmad2/3 at 24 h and fibronectin and αSMA at 72 h (G, nuclear counterstained by Hoechst 33342, scale bar: 20 µm). www.nature.com/scientificreports/ of 2 × 10 3 per cm 2 for expansion on plastic coated with 2D Matrigel (5% diluted Matrigel) after 1 h incubation at 37 °C in MESCM + 5% FBS. Upon 80-90% confluence, TM cells were passaged serially at a density of 5 × 10 3 per cm 2 in MESCM + 5% FBS. Passage 3 (P3) TM cells were cultured until confluence and were maintained another 5 days to obtain stable monolayers before being used for experiments. These cells were treated with or without 100 nM dexamethasone (DEX) for up to 10 days to assess expression of myocillin 36 .
Quantitative real-time PCR (qRT-PCR). These steps are similar to our previous report 27 . In brief, total RNAs were extracted from TM cells in different groups using a RNeasy kit and reverse-transcribed using a High Capacity Reverse Transcription kit. cDNA of each group was amplified by qRT-PCR using specific primerprobe mixtures and DNA polymerase in a real-time PCR system (QuantStudio; ThermoFisher). The PCR profile consisted of 10 min of initial activation at 95 °C, followed by 40 cycles of 15 s denaturation at 95 °C, and 1 min annealing and extension at 60 °C.
Immunofluorescence staining. The staining procedures are similar to our previous report 27 . In brief, after being treated with or without TGF-β1 for 15 min, 24 h or 72 h in different groups, single TM cells after trypsin/EDTA digestion or 3D clusters were harvested and prepared by cytospin at 1000 rpm for 4 min (StatSpin, Inc., Norwood, MA). The cytospin slides or cells cultured on cover slips were fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.2% Triton X-100 in PBS for 30 min, and blocked with 2% BSA in PBS for 1 h before being incubated with specific primary antibodies for 16 h (overnight) at 4 °C. After washing with PBS, samples were incubated with respective secondary antibodies for 1 h at room temperature. After another wash with PBS, the nucleus was counterstained with Hoechst 33342 before being analyzed with confocal microscope (LSM700; Carl Zeiss, Thornhood, NY).  www.nature.com/scientificreports/