Glucocorticoid-induced cell-derived matrix modulates transforming growth factor β2 signaling in human trabecular meshwork cells

Aberrant remodeling of trabecular meshwork (TM) extracellular matrix (ECM) may induce ocular hypertensive phenotypes in human TM (hTM) cells to cause ocular hypertension, via a yet unknown mechanism. Here, we show that, in the absence of exogenous transforming growth factor-beta2 (TGFβ2), compared with control matrices (VehMs), glucocorticoid-induced cell-derived matrices (GIMs) trigger non-Smad TGFβ2 signaling in hTM cells, correlated with overexpression/activity of structural ECM genes (fibronectin, collagen IV, collagen VI, myocilin), matricellular genes (connective tissue growth factor [CTGF], secreted protein, acidic and rich in cysteine), crosslinking genes/enzymes (lysyl oxidase, lysyl oxidase-like 2–4, tissue transglutaminase-2), and ECM turnover genes/enzymes (matrix metalloproteinases-MMP2,14 and their inhibitors-TIMP2). However, in the presence of exogenous TGFβ2, VehMs and GIMs activate Smad and non-Smad TGFβ2 signaling in hTM cells, associated with overexpression of α-smooth muscle actin (α-SMA), and differential upregulation of aforementioned ECM genes/proteins with new ones emerging (collagen-I, thrombospondin-I, plasminogen activator inhibitor, MMP1, 9, ADAMTS4, TIMP1); with GIM-TGFβ2-induced changes being mostly more pronounced. This suggests dual glaucomatous insults potentiate profibrotic signaling/phenotypes. Lastly, we demonstrate type I TGFβ receptor kinase inhibition abrogates VehM-/GIM- and/or TGFβ2-induced upregulation of α-SMA and CTGF. Collectively, pathological TM microenvironments are sufficient to elicit adverse cellular responses that may be ameliorated by targeting TGFβ2 pathway.

GIMs overexpressed α-smooth muscle actin in the presence of TGFβ2 in hTM cells. Given that ocular hypertensive phenotypes induced by TGFβ2 signaling pathways include fibrotic changes in actin cytoskeletal stress fibers 21,33 , we determined whether GIMs could modulate the protein expression of α-smooth muscle actin (αSMA) in seeded hTM cells in the presence or absence of exogenous TGFβ2. We discovered that, compared with hTM cells cultured on VehMs for 7 days, in the absence of exogenous TGFβ2, GIMs trended towards Primary hTM cells were cultured in the presence or absence of 100 nM dexamethasone for 4 weeks in complete growth media. Cells were subsequently removed using 20 mM ammonium hydroxide solution to obtain GIMs and vehicle control matrices (VehMs). Same strain, fresh primary hTM cells were then seeded on these matrices with or without exogenous 5 ng/mL TGFβ2 in 1% fetal bovine serum growth media for 24 hours. Protein was extracted for Western blot analysis. β-Actin was used as an internal control for normalization. Representative cropped blots (top) and densitometric analysis (bottom) of (A) Smad2, (B) Phosphorylated Smad2 (pSmad2), (C) Smad3, (D) Phosphorylated Smad3 (pSmad3), and (E) Smad4. Columns and error bars; means and standard error of mean (SEM). One-way ANOVA with the Tukey pairwise comparisons post hoc test was used for statistical analysis. (n = 4 biological replicates. *p < 0.05, **p < 0.01, ***p < 0.001 for the group of interest versus control, VehM. ## p < 0.01, ### p < 0.001 for the group of interest versus GIM. † p < 0.05, † † † p < 0.001 for comparison between VehM + TGFβ2 and GIM + TGFβ2). hTM, human trabecular meshwork. As indicated by the black horizontal demarcating lines between respective blot images, all membranes were re-probed for β-Actin as a housekeeping protein. Full length blots can be found in Supplementary Figs. S5A-E and S6A-E, respectively. Densitometric analyses was done using ImageJ 1.8.0_112 software (https ://image j.nih.gov/ij/, 1997-2018).

GiMs induced pathological changes in ecM components in htM cells in the presence or
absence of exogenous TGFβ2. TGFβ2 signaling is well-established in ocular hypertension partly because of its potent regulatory role on TM ECM restructuring 16,26,34,52 . Since it is technically challenging to differentiate between existing ECM proteins from the cell-derived matrices with any induced changes in the same genes/ proteins, we primarily focused on mRNA levels.
ECM structural genes. First, we determined changes in genes that regulate structural ECM proteins. Compared with hTM cells seeded on control matrices (VehMs) for 7 days, in the absence of TGFβ2, GIMs significantly overexpressed fibronectin (1.9-fold, p < 0.001); similarly, in the presence of TGFβ2, both VehMs and GIMs exhibited markedly increased fibronectin expression (2.3-fold and 1.9-fold, p < 0.001, respectively) ( Fig. 4A), although no differences were observed between VehM and GIM groups. Further, in the absence of TGFβ2, compared with VehMs, there were no significant changes in collagen I expression in hTM cells on GIM, although it trended upwards. However, with exogenous TGFβ2, GIMs markedly increased collagen I (22.1-fold, p < 0.001) far more potently than on VehMs (6.9-fold, p < 0.001) (Fig. 4B). In addition, in the absence of TGFβ2, compared with VehMs, GIMs significantly overexpressed collagen IV (3.3-fold, p < 0.001) in hTM cells. Similarly, in the presence of TGFβ2, collagen IV was markedly increased by VehMs and GIMs (2.9-fold, p < 0.01 and 8.1-fold, p < 0.001, respectively), with GIMs showing a more exacerbated response (Fig. 4C). Moreover, in the absence    Genes responsible for ECM turnover. To determine whether the GIM-induced overexpression of specific ECM structural genes (Fig. 4), matricellular genes ( Fig. 5) and ECM crosslinking genes ( Fig. 6) in hTM cells could be physiological or pathological, we probed for the gene expression of specific matrix metalloproteinases (MMPs) implicated in aqueous humor outflow and TM ECM degradation. We observed that, in hTM cells, compared with VehMs, MMP1 was significantly reduced by all experimental groups: GIMs, interaction between VehMs and TGFβ2, and GIM-TGFβ2 interaction (-5-fold, -2.5-fold and -10-fold, p < 0.001, respectively); the greatest reduction was observed with GIM-TGFβ2 interaction (Fig. 7A). However, in the absence of TGFβ2, GIMinduced MMP2 was not any different from VehMs in hTM cells; in the presence of TGFβ2, MMP2 was significantly overexpressed by VehMs and GIMs (2.2-fold, p < 0.01 and 3.7-fold, p < 0.001, respectively), with a greater effect seen in the latter group (Fig. 7B). In addition, compared with VehMs, MMP9 was markedly upregulated in hTM cells only when GIMs interacted with TGFβ2 (4.2-fold, p < 0.001); although statistically insignificant an upward trend was observed with VehM-TGFβ2 interaction compared with control VehM (Fig. 7C). Moreover, without exogenous TGFβ2, compared with VehMs, GIMs significantly increased MMP14 (1.8-fold, p < 0.05) in hTM cells; with exogenous TGFβ2, while VehMs had no significant effect on MMP14, GIMs markedly overexpressed it (2.2-fold, p < 0.001) (Fig. 7D). Moreover, in the absence of TGFβ2, GIMs' effect on a disintegrin and metalloproteinase with a thrombospondin motif 4 (ADAMTS4) was not any different from VehMs; with www.nature.com/scientificreports/ exogenous TGFβ2, VehMs and GIMs markedly upregulated ADAMTS4 (3.7-fold and 2.8-fold, p < 0.001, respectively) (Fig. 7E). Next, we determined the expressional levels of tissue inhibitors of matrix metalloproteinases (TIMPs) given their inhibitory role on the function of MMPs in degrading aberrant ECM accumulation. We discovered that, in the absence of exogenous TGFβ2, GIMs had no effect on TIMP1 in hTM cells compared with VehMs. However, in the presence of TGFβ2, both VehMs and GIMs significantly overexpressed TIMP1 (4.7-fold, p < 0.001 and 2.9-fold, p < 0.01, respectively; Fig. 7F), although the effect was greater in VehM than on GIM. Finally, in the absence of exogenous TGFβ2, compared with hTM cells cultured on VehMs, GIMs profoundly increased TIMP2 (1.8-fold, p = 0.001). Similarly, in the presence of TGFβ2, TIMP2 was markedly overexpressed by VehMs and GIMs (two fold and 3.1-fold, p < 0.001, respectively), with a greater effect in cells on GIMs (Fig. 7G).

Enzymatic activities of MMP2, MMP9 and LOX in conditioned media.
To investigate whether the upregulation of MMP2, MMP9 and LOX (ECM crosslinking gene) at the gene level translated into functioning proteins, we harvested and subsequently subjected conditioned media from the respective experimental groups to enzymatic assays. As shown in Fig. 8A, in the absence of exogenous TGFβ2, compared with control VehMs, conditioned media from hTM cells seeded on GIMs for 7 days had significantly increased activity of MMP2 (1.3-fold, p < 0.05); similarly, in the presence of TGFβ2, MMP2′ activity in conditioned media from VehMs and GIMs was significantly elevated (1.5-fold and 1.6-fold, p < 0.001, respectively). However, compared with cells on control VehMs in the absence of TGFβ2, conditioned media from hTM cells cultured in the presence of TGFβ2 on VehMs and GIMs had a marked increase in MMP9′ activity (3.3-fold and 2.8-fold, p = 0.001, respectively) ( Fig. 8B). Further, we observed that, in the absence of exogenous TGFβ2, conditioned media from three (3) out of the four (4) hTM Cell Strains seeded on GIMs for 7 days had markedly increased LOX activity (Fig. 8C, hTM

Figure 5. GIMs overexpress crucial matricellular genes with or without exogenous TGFβ2 in hTM cells.
Primary hTM cells were cultured in the presence or absence of 100 nM dexamethasone for 4 weeks in complete growth media. Cells were subsequently removed using 20 mM ammonium hydroxide solution to obtain GIMs and vehicle control matrices (VehMs). Same strain, fresh primary hTM cells were then seeded on these matrices with or without exogenous 5 ng/mL TGFβ2 in 1% fetal bovine serum growth media for 7 days. RNA was extracted for reverse transcription and qPCR. GAPDH was used as an internal control for normalization. TGFβRI kinase inhibitor repressed GIM-induced overexpression of CTGF. Finally, we investigated whether selective inhibition of type I transforming growth factor beta receptor (TGFβRI) kinase would attenuate GIM-induced overexpression of connective tissue growth factor (CTGF) with or without exogenous TGFβ2. We cultured primary hTM cells on VehMs and GIMs in 1% serum media for 24 hours in the presence or absence of 5 ng/mL TGFβ2 with or without 5 μM TGFβRI kinase inhibitor (LY364947). We then extracted proteins and performed Western blotting to probe for the expression of CTGF. As shown in Fig. 9, in the absence of TGFβRI kinase inhibitor (TGFβRi) or exogenous TGFβ2, compared with control VehMs, GIMs significantly upregulated CTGF (4.2-fold, p < 0.05) in hTM cells. In the presence of TGFβ2 and absence of TGFβRi, both VehMs and GIMs profoundly overexpressed CTGF (10.7-fold and 13.2-fold, p < 0.001, respectively) in hTM cells. Together, this validated CTGF' gene expression data in Fig. 5A. However, in the presence of TGFβRi, VehM-/GIM-and/or TGFβ2-induced overexpression of CTGF was profoundly abrogated to levels similar to that of control matrices (VehMs). The same observation was true for αSMA ( Supplementary Fig. S2).

Discussion
The extracellular matrix is a three-dimensional framework of structural and matricellular proteins and serves as a reservoir for soluble signaling molecules such as cytokines secreted by resident cells. There is an ever-increasing amount of literature that documents that physico-chemical and biophysical properties of the matrix can profoundly influence cell behavior. While it is well recognized that the ECM is altered in disease, how such a matrix affects cell fate is not known. For example, in glaucoma, a number of studies demonstrate elevated amounts of fibronectin [57][58][59][60][61][62] and connective tissue growth factor 63,64 in the TM. In addition, elevated levels of profibrotic cytokines 27,65-69 have been reported in the aqueous humor of glaucomatous patients. It is thus inevitable that resident TM cells are continuously subjected to both secreted factors within the tissue and those circulating and passing through the TM. How TM cells respond to such simultaneous pathologic insults has not been evaluated www.nature.com/scientificreports/ to the best of our knowledge. Therefore, in this study, we report using for the first time an in vitro system, to determine the role that pathologic matrices (for example, GIMs) play in modulating critical signaling pathways (for instance, TGFβ) intrinsically in hTM cells, and under the influence of exogenous profibrotic cytokines (for instance, TGFβ2) added to this culture system. To this effect, we hypothesize that hTM cell responses to GIMs may primarily be driven by: (1) Biochemical composition and topography of GIMs 11 ; (2) Biomechanical properties, given that GIMs are approximately fourfold stiffer than VehMs 11,13 ; and/or (3) differential presentation of ligands to cells via interactions between biochemical and biomechanical factors from GIMs. Previous studies have determined both in vitro and in vivo that treatment with dexamethasone (DEX) results in overexpression/secretion of active TGFβ2 which may subsequently activate the Smad TGFβ pathway to cause changes in ECM proteins 29 . However, whether ECM remodeling downstream of glucocorticoid receptor signaling is independent of canonical TGFβ signaling is unclear. In addition, it was unknown until now whether the ECM deposited after such DEX treatment further determined cell fate. Thus, first, we showed that, in the absence of any exogenous chemicals (for example, DEX) or growth factors (for instance, TGFβ2), GIMs activated non-Smad-dependent TGFβ signaling to result in additional overexpression of ECM genes/proteins in freshly plated hTM cells or conditioned media (Fig. 10, left cellular compartment). Similar to the interaction between VehMs and exogenous TGFβ2, these aberrant ECM changes encompassed increased accumulation/activity of specific ECM structural, matricellular and crosslinking genes/proteins together with dysregulated ECM turnover. Since there was no exogenous DEX, we infer that the responses directed by GIMs in hTM cells are independent of direct influences from activated glucocorticoid receptor (GR) signaling. Specifically, we showed that, with the exception of phospho-JNK (pJNK), GIMs markedly overexpressed phospho-ERK1/2 (pERK1/2), phospho-P38 (pP38) and Rho-associated protein kinase (ROCK) in hTM cells. These findings were associated with significant increases in ECM structural genes like fibronectin, collagen IV, collagen VI and myocilin; matricellular genes involving connective tissue growth factor (CTGF) and secreted protein, acidic and rich in cysteine (SPARC); ECM crosslinking genes including lysyl oxidase (LOX), LOX-like 2 (LOXL2), LOXL3, LOXL4 and tissue transglutaminase 2 (TGM2); and genes responsible for ECM turnover (matrix metalloproteinase 14 [MMP14] and tissue inhibitor of metalloproteinases 2 [TIMP2]). An imbalance in matrix crosslinking versus degradation may in part play a role in matrix stiffening 70,71 .
Our results are consistent with several studies that have implicated non-Smad TGFβ signaling pathways in aberrant ECM restructuring. For instance, TGFβ2-induced activation of P38 is crucial for the induction of collagen I 72 and SPARC 73 in hTM cells. Further, Rho GTPases enhance secretion of bioactive TGFβ2 into the It was somewhat surprising GIM had no effect on pJNK given that a previous study had implicated pJNK in the induction of all four LOX genes in TGFβ2-stimulated hTM cells 37 . The difference between the findings of the two studies may be due to the differences in the type of glaucomatous insult used: GIMs in ours versus exogenous TGFβ2 in theirs. Additionally, since all the phosphorylated forms of the other non-Smads (pERK1/2, pP38, and ROCK) were upregulated by GIMs in hTM cells, having no GIM-induced change in pJNK may be an efficient mechanism of reducing signaling redundancy. Concurrent with these changes were GIM-induced increased deposition of ECM structural/matricellular/ crosslinking genes. Increased accumulation of fibronectin, collagen IV, collagen VI and myocilin have been implicated in endoplasmic reticulum (ER) stress, ocular hypertension and glaucoma 16,23,43,57,[75][76][77] . Upregulation of fibronectin is particularly critical because not only is its synthesis associated with ER stress 23 , but fundamental to the assembly of other ECM proteins like collagen IV, fibrillin and laminin 78 which may contribute to altered tissue biomechanics 8 . CTGF and SPARC have also been documented as culprits in ocular hypertension and glaucoma. For example, while CTGF is elevated in pseudoexfoliation glaucoma 63,64, and its overexpression causes ocular hypertension in mice 79 , SPARC-null mice have a lower intraocular pressure despite the presence of exogenous TGFβ2 35 . Further, LOX, LOXLs and TGM2 do have their place in glaucoma; single-nucleotide polymorphism in LOXL1 has been implicated in exfoliation glaucoma 55,80 . While all five isoforms of LOX (LOX and LOXL1-4) are upregulated in either gremlin-or TGFβ2-induced hTM cells 37,52 , LOX inhibition increases outflow facility in ex vivo perfusion organ cultures 10 . Similarly, overexpressing TGM2 in mice causes ocular hypertension 56 , and knocking it out reduces TGFβ2-induced ocular hypertension 81 .

Figure 9. GIM-and/or TGFB2-induced overexpression of connective tissue growth factor in hTM cells is attenuated by TGFβRI kinase inhibitor.
Primary hTM cells were cultured in the presence or absence of 100 nM dexamethasone for 4 weeks in complete growth media. Cells were subsequently removed using 20 mM ammonium hydroxide solution to obtain GIMs and vehicle control matrices (VehMs). Same strain, fresh primary hTM cells were then seeded on these matrices with or without exogenous 5 ng/ml TGFβ2 and/ or 5 µM type I TGFβ receptor (TGFβRI) kinase inhibitor in 1% fetal bovine serum growth media for 24 hours. Protein was extracted for Western blot analysis. β-Actin was used as a housekeeping protein for normalization. Representative cropped blot (top) and densitometric analysis (bottom) of connective tissue growth factor (CTGF). Columns and error bars; means and standard error of mean (SEM). One-way ANOVA with the Tukey pairwise comparisons post hoc test was used for statistical analysis. (n = 4 biological replicates. *p < 0.05, ***p < 0.001 for the group of interest versus control, VehM. # p < 0.05, ### p < 0.001 for the group of interest versus GIM).TGFβ2, Transforming growth factor β2, TGFβRi, type I TGFβ receptor kinase inhibitor. hTM, human trabecular meshwork. Black horizontal demarcating lines between the blots indicate two images taken from two different parts of the same membrane blot. Full length blots can be found in Supplementary Figs. S5G and S6G. Densitometric analyses was done using ImageJ 1.8.0_112 software (https ://image j.nih.gov/ij/, 1997-2018).

Scientific RepoRtS
| (2020) 10:15641 | https://doi.org/10.1038/s41598-020-72779-w www.nature.com/scientificreports/ Also, there was GIM-stimulated differential modulation of ECM turnover enzymes: MMPs and TIMPs. Because MMP14 and TIMP2 have functional roles in cleaving MMP2 to degrade ECM proteins 82 , one would think GIM-induced overexpression of MMP14 and TIMP2 genes together with increased MMP2 activity in conditioned media would be a homeostatic response to counteract increased ECM synthesis/deposition. However, there is a tipping point in TIMP2′ overexpression beyond which it actually sequesters MMP2 to stall its activity 82 . Additionally, increased activity/expression of MMPs may pathologically cleave and cause dysfunction of key cell surface glycoproteins like prion protein important for maintaining cell-ECM/aqueous homeostasis 83 . Further, the gene expression of other MMPs like MMP1 was significantly downregulated. Moreover, given that crosslinking enzymes covalently crosslink collagen, elastin and fibronectin 84,85 , making them resistant to degradation, GIM-induced increased activity of LOX in conditioned media would hinder MMP2′ activity. Taken together, the combination of GIM-induced activation of non-Smad TGFβ signaling, aberrant ECM synthesis/ deposition, increased crosslinking genes/enzymes and dysregulated ECM turnover in hTM cells may contribute to increased stiffness.
Second, we showed that, in the presence of exogenous TGFβ2, GIMs activated both Smad and non-Smad TGFβ signaling in association with changes in actin cytoskeleton, and more pronounced ECM changes in hTM cells compared with either interaction between VehMs and TGFβ2 or GIM alone (Fig. 10, right cellular compartment), in agreement with previous studies 21, 33 . This suggests that, synergistic interaction between dual glaucomatous insults (here, being a pathologically remodeled GIM and TGFβ2), as is normally the case in vivo, may further heighten induction of glaucomatous phenotypes/profibrotic conditions beyond what just only one glaucomatous stimulus would do. Notably, GIM-TGFβ2 interaction further increased the expression/activity of ECM genes/proteins that had previously been upregulated by either GIM alone or VehM-TGFβ2 interaction, while overexpressing those that were formerly unaffected by these experimental groups. For instance, besides further upregulating collagen IV in hTM cells beyond that of GIM alone or the interaction between VehMs and TGFβ2, interaction between GIM and TGFβ2 profoundly increased collagen I which was unaffected by GIM alone. Further, this GIM-TGFβ2-induced overexpression of collagen I was significantly beyond that of VehM-TGFβ2 interaction. Similar to the other ECM structural genes 16,77,78,86 , Collagen I has been implicated in glaucoma in association with ER stress, fibrosis and reduced outflow facility 23 . Further, while GIM-TGFβ2 combination heightened the overexpression of CTGF and SPARC in hTM cells beyond GIM alone or VehM-TGFβ2 interaction, other matricellular genes like thrombospondin 1 (TSP1) and plasminogen activator inhibitor (PAI), which www.nature.com/scientificreports/ were hitherto unchanged by GIM alone, were increased; and this increase was markedly beyond that of VehM-TGFβ2-interaction. GIM-TGFβ2-induced upregulation of TSP1 is important given its elevation in glaucomatous TM 87,88 , and potential contribution to abnormal growth factor signaling 46 . On the other hand, apart from being elevated in the aqueous humor of glaucoma patients 89 , overexpression of PAI suppresses proteolytic degradation of ECM structural proteins by MMPs leading to increased ECM accumulation 90 . Moreover, GIM-TGFβ2 interaction exaggerated the overexpression of LOX and LOXL2 beyond GIM alone or interaction between VehM and TGFβ2, suggesting more pronounced crosslinking of ECM structural proteins. Therefore, although GIM-TGFβ2 combination increased the expression and/or activity of MMP2, MMP9 and/or a disintegrin and metalloproteinase with a thrombospondin motif 4 (ADAMTS4) beyond that of only GIM or VehM-TGFβ2-interaction, repressed MMP1 together with increased LOX-dependent crosslinking of ECM structural proteins would make their proteolytic degradation function challenging. Further, given that both TIMP1 and TIMP2 (genes that inhibit MMP activity) were concurrently upregulated in hTM cells via GIM-TGFβ2 interaction, ECM turnover would most likely be impaired. In aggregate, GIM-TGFβ2 interaction exaggerated the overexpression of target TGFβ molecules in hTM cells beyond GIM alone or interaction between VehM and TGFβ2 most likely because of the heightened synergistic interaction between activated Smad and non-Smad TGFβ signaling molecules. This would probably compound altered biomechanics associated with reduced outflow facility 9,10 beyond GIM alone or interaction between VehMs and TGFβ2. Since cells interact with the microenvironment and resident proteins through receptor mediated signaling, we next inhibited type I TGFβ receptor (TGFβRI) kinase using a small molecule inhibitor (LY364947) whose effective concentration (5 μM) has been validated in hTM cells by several studies 29,33,36,37,52 . Importantly, at the concentration used to suppress the kinase activity in vitro, the above-mentioned studies have documented successful inhibition of TGFβ signaling, glucocorticoid-induced endoplasmic reticulum stress, formation of cross-linked actin networks (CLANs), and secretion of major ECM and crosslinking proteins. Consistent with these, here, we observed that, treatment of hTM cells with the same inhibitor abrogated GIM-, VehM-TGFβ2-, or GIM-TGFβ2induced overexpression of CTGF which may be associated with dysregulated ECM turnover, increased stiffness and elevated intraocular pressure 63,79,91 . We selected CTGF from all the other ECM TGFβ targets because it has been shown to be a subsequent mediator of TGFβ2 signaling 45 . Moreover, overexpressing CTGF in vivo in mice causes ocular hypertension in association with optic nerve damage 79 . Because TGFβ signaling-mediated ocular hypertensive phenotypes go beyond pathological ECM changes to actin cytoskeletal remodeling 21,33 , we also showed that GIM-and/or TGFβ2-induced overexpression of αSMA was attenuated by TGFβRI kinase inhibitor ( Supplementary Fig. S2), suggesting inhibition of cell contractility or profibrotic phenotype.
Conversely, a previous study found TGFβ2-induced αSMA was unaffected by inhibitors of Smad3, ERK, P38, ROCK and JNK after 24 hours of respective cotreatment 33 . This difference between our study and theirs may be due to the following reasons. First, we inhibited TGFβRI kinase rather than downstream Smad and non-Smad molecules, suggesting that TGFβ receptor inhibition may be a more effective strategy of ameliorating GIMinduced ocular hypertensive phenotypes. Indeed, in vivo, when used at a concentration of 1% (w/v; 36.7 mM), Kasetti and colleagues 29 demonstrated such a strategy decreases glucocorticoid-induced ocular hypertension, thus, setting a precedent for this applicability. Future studies will be essential to determine the relative contribution of Smad and non-Smad molecules to GIM-and/or TGFβ2-induced ocular hypertensive phenotypes including altered stiffness in hTM cells. Second, culturing hTM cells on VehMs and GIMs in our study, as opposed to culturing hTM cells on tissue culture plastics/glass coverslips as was done in the previous study likely results in fairly dramatic differences in gene expression levels and cell signaling.
This study is however not without limitations. The effects of VehM-/GIM-induced changes in hTM cell responses were performed on a relatively small yet acceptable sample size (n = 3-5 donors). In compliance with consensus recommendation for TM cell characterization and culture 92 , only hTM cells that responded to DEX treatment with myocilin overexpression were used in our experiments. Thus, determining changes in cells non-responsive to DEX treatment was not done. Further, all our experiments were performed under in vitro conditions. While this is advantageous in delineating the effect of a pathological matrix (for instance, GIMs) on otherwise healthy cells, it does not entirely recapitulate the scenario in vivo where native cells (that may or may not have developed pathology) continue to persist as ECM remodeling occurs. Thus, extrapolations of our data to what is observed in vivo, or ex vivo, are only suggestive of mechanisms, but should be taken cautiously considering functional changes in outflow resistance and facility were not determined. Clever strategies where anterior segment tissues (healthy and pathologic) are decellularized and subsequently recellularized ex vivo, may be essential to translate our in vitro findings for functional assessments. Developing such models are out of the scope of the current study, although future studies will focus on such methodologies. Also, for the most part, we characterized mRNA expression changes of pathway ECM targets in hTM cells on cell-derived matrices and were therefore, unable to distinguish expression/synthesis of new ECM proteins versus those from the existing matrices. Further studies using specialized techniques such as Stable Isotope Labeling by/with Amino acids in Cell culture (SILAC) may be required to definitively identify target ECM protein changes. Moreover, although DEX was used to simulate the pathological microenvironment, truly glaucomatous cells or matrices were not used in this study. Future studies would be essential to investigate if similar signaling pathways are implicated, although most glaucomatous cells/matrices are hardly free from the effects of previous anti-glaucoma medications that may confound experimental outcomes.
In conclusion, we found that similar to interaction between control VehMs and exogenous TGFβ2, a pathologically remodeled matrix like GIMs, can single-handedly drive pathological changes in hTM cells. These diseased cellular phenotypes are more pronounced upon interaction between this pathologically remodeled matrix and an exogenous glaucomatous stimulus like TGFβ2. Importantly, our results suggest that care must be taken when considering cell-based therapeutic options (for example, stem cell or cell replacement options) [93][94][95][96] . This is because similar to resident pathologic cells 97  www.nature.com/scientificreports/ towards a pathologic phenotype rather than ameliorate symptoms especially in the long term. However, inhibiting TGFβRI kinase reverses all these pathological outcomes irrespective of either stimuli or their interaction, suggesting it may potentially be used as an ocular hypotensive target. Cells that subsequently moved out of the TM were cultured in complete growth media and generally utilized between passages two (2) and six (6). For characterization of primary hTM cells, all cell strains were examined for dexamethasone (DEX)-induced expression of myocilin as recommended ( Supplementary Fig. S3) 98

Generation of vehicle control-and glucocorticoid-induced matrices from cultured tM cells.
Primary hTM cells were cultured in sterile pretreated 60 mm dishes (5,000-10,000 cells per cm 2 ) and amino-silane modified 12 mm glass coverslips 11,99 in a 24-well culture plate (5000-10,000 cells per cm 2 ) using complete growth media. Upon reaching 80-90% confluence, hTM cells were treated with 100 nM dexamethasone (DEX; Sigma-Aldrich Corp., St. Louis, MO, USA) or its vehicle, ethanol (EtOH), for 4 weeks with media changes done on every other day. As previously described, denuding the ECM of its resident cells to obtain vehicle control-(VehMs) or glucocorticoid-induced matrices (GIMs) was done using 20 mM ammonium hydroxide, 0.05% Triton X-100 and deionized water as solvent 11,17 . These cell-derived matrices (CDMs) were then incubated with 50 U/mL DNase I and RNase A for 2 hours, washed thoroughly in Hank's balanced salt solution (HBSS) prior to subsequent immunocytochemistry (to confirm successful removal of cells from their deposited ECM) and cell culture.  29,33,36,37,52 . Subsequently, cells were lysed for Western blotting and reverse transcriptasequantitative polymerase chain reaction (RT-qPCR); their respective conditioned growth media were concurrently harvested and subjected to enzymatic assays to probe activities of specific matrix metalloproteinases (for example, MMP2 and MMP9) and crosslinking enzymes (for instance, lysyl oxidase) ( Supplementary Fig. S4). www.nature.com/scientificreports/ scraped into radioimmunoprecipitation assay (RIPA) buffer (ThermoScientific, Waltham, MA, USA) supplemented with protease and phosphatase inhibitors (Fisher Scientific, Hampton, NH, USA) on ice, and then subsequently centrifuged at 12,000 g for 15 minutes at 4 °C to pellet and get rid of cellular debris. Supernatants were transferred to fresh tubes and quantified via a modified Lowry assay (DC assay; Biorad, Hercules, CA, USA) with BSA as the standard. Protein lysates were subsequently denatured in a 1:10 mixture of 2-mercaptoethanol and 4 × Laemmli buffer by boiling at 100 °C for 5 minutes. After quickly centrifuging proteins at 15,000×g for 30 seconds, equal amount of protein was loaded per well (20 µg) for each sample and ran on denaturing 4-15% gradient polyacrylamide ready-made gels (Biorad); subsequently transferred onto polyvinylidene difluoride (PVDF) membranes by electrophoresis. Membrane blots were blocked in 5% BSA in 1 × tris buffered saline/tween-20 (TBST) for 1 hour. Immunoblots were incubated overnight at 4 °C with specific primary antibodies (Supplementary Table S1) on a rotating shaker. The membrane blot was washed thrice with TBST; each wash lasting for approximately 10 minutes. Subsequent incubation with corresponding HRP-conjugated species-specific secondary antibodies (Supplementary Table S1) for 45 minutes was done, followed by three 10-minute washes with TBST. The protein bands were then visualized using ECL detection reagents (SuperSignal West Femto Maximum Sensitivity Substrate; Life technologies, Grand Island, NY, USA) and imaged with a Bio-Rad ChemiDoc MP imaging system. Respective membrane blots were stripped and probed with β-Actin as a loading control. Data were exported into ImageJ 1.8.0_112 software (https ://image j.nih.gov/ij/, 1997-2018) for densitometric analysis.

Biochemical characterization of cell-derived matrices via immunocytochemistry.
RnA isolation and quantitative real-time pcR. Total RNA was isolated from cells that had been seeded on VehMs and GIMs for 7 days with or without TGFβ2 treatment using an RNA purification kit (Catalog number: 12183025; PureLink RNA Mini kit, Invitrogen, Carlsbad, CA). First-strand cDNA was synthesized using 1 µg of total RNA and the High-Capacity cDNA Reverse Transcription Kit (Catalog number: 4368813; Applied Biosystems, Foster City, CA) following the manufacturer's instructions. Quantitative real-time polymerase chain reaction (qPCR) was performed on 20 ng of the cDNA with specific primers (Supplementary Table S2) and the PowerUp SYBR Green Master Mix kit (Catalog number: A25918; Applied Biosystems, Foster City, CA) in total volumes of 10 µL per reaction using a CFX Connect Real-time System from Bio-Rad Laboratories (Bio-Rad, Hercules, CA, USA). The cycle threshold (Ct) values were obtained from the qPCR equipment and analyzed using the 2 −ΔΔCt method with Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as the housekeeping gene.
Enzymatic activity assay for specific matrix metalloproteinases in conditioned media. MMP enzyme activity was measured using gelatin zymography as per manufacturer's instructions. Briefly, primary hTM cells were cultured on VehMs and GIMs with or without TGFβ2 in 1% FBS growth media for 7 days. Conditioned medium was collected, and total protein was quantitated using the Pierce BCA Protein Assay (Catalog number: 23225, ThermoFisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. Total protein was normalized per sample prior to loading onto 10% Gelatin Zymogram Gels (15 µg/lane) (Catalog number: ZY00100, ThermoFisher Scientific, Waltham, MA, USA). Gels were washed twice in Zymogram Renaturing Buffer (Catalog number: LC2670, ThermoFisher Scientific, Waltham, MA, USA) at room temperature, prior to incubating in Zymogram Developing Buffer (Catalog number: LC2671, ThermoFisher Scientific, Waltham, MA, USA) overnight with shaking at 37ºC. Gels were then washed and stained before imaging using the Odyssey imaging system (Licor Biosciences, Lincoln, NE, USA). Gel bands were exported into ImageJ software for densitometric analysis.
Activity assay for crosslinking enzymes in conditioned media. Lysyl oxidase enzyme activity was measured using the Fluorometric Lysyl Oxidase Activity Assay Kit (Catalog number: K928-100, BioVision, Milpitas, CA) according to the manufacturer's instructions. Briefly, conditioned medium was collected from cultured cells that had been seeded on CDMs for 7 days in 1% serum media in the presence or absence of TGFβ2; and subsequently concentrated tenfold using 10 K MWCO PES Pierce Spin Concentrators (Catalog number: 88512, ThermoFisher Scientific, Waltham, MA). Samples were then normalized to total protein concentration, as described above. Samples were prepared with reaction components and fluorescence was measured (Ex/ Em = 535/587) in kinetic mode every 30 seconds for 60 minutes at 37ºC in a black plate using the SpectraMax iD3 Multi-Mode MicroPlate Reader (Molecular Devices, San Jose, CA). The LOX enzyme activity was measured as the Vmax (in RFU units per second) using the SoftMax Pro 7 Data Acquisition and Analysis Software (Molecular Devices; https ://www.molec ulard evice s.com/produ cts/micro plate -reade rs/acqui sitio n-and-analy sis-softw are/softm ax-pro-softw are).
Statistics. One-way ANOVA followed by Tukey multiple comparisons post hoc test was used for analysis, with P values less than 0.05 considered to be statistically significant. Group-wise statistical comparisons are indicated within each figure legend. All data are presented as mean ± SEM, mostly in bar graphs, representative immunofluorescent micrographs, and blots where applicable.