Transforming growth factor β2 (TGFβ2) signaling plays a key role in glucocorticoid-induced ocular hypertension

Elevation of intraocular pressure (IOP) is a serious adverse effect of glucocorticoid (GC) therapy. Increased extracellular matrix (ECM) accumulation and endoplasmic reticulum (ER) stress in the trabecular meshwork (TM) is associated with GC-induced IOP elevation. However, the molecular mechanisms by which GCs induce ECM accumulation and ER stress in the TM have not been determined. Here, we show that a potent GC, dexamethasone (Dex), activates transforming growth factor β (TGFβ) signaling, leading to GC-induced ECM accumulation, ER stress, and IOP elevation. Dex increased both the precursor and bioactive forms of TGFβ2 in conditioned medium and activated TGFβ-induced SMAD signaling in primary human TM cells. Dex also activated TGFβ2 in the aqueous humor and TM of a mouse model of Dex-induced ocular hypertension. We further show that Smad3−/− mice are protected from Dex-induced ocular hypertension, ER stress, and ECM accumulation. Moreover, treating WT mice with a selective TGFβ receptor kinase I inhibitor, LY364947, significantly decreased Dex-induced ocular hypertension. Of note, knockdown of the ER stress–induced activating transcription factor 4 (ATF4), or C/EBP homologous protein (CHOP), completely prevented Dex-induced TGFβ2 activation and ECM accumulation in TM cells. These observations suggested that chronic ER stress promotes Dex-induced ocular hypertension via TGFβ signaling. Our results indicate that TGFβ2 signaling plays a central role in GC-induced ocular hypertension and provides therapeutic targets for GC-induced ocular hypertension.

Glaucoma is the second leading cause of irreversible blindness worldwide. In 2013, an estimated 64.3 million people were affected by glaucoma globally, a number projected to increase to 76 million by 2020 and to 111.8 million by 2040 (1). Primary open-angle glaucoma (POAG) 2 is the most common type of glaucoma, accounting for 74% of all glaucoma cases (2). Elevated intraocular pressure (IOP) is an important and modifiable risk factor for the development and progression of POAG (3). Glucocorticoids (GC) have proven to be vital agents for the treatment of a wide range of disorders including various ocular diseases involving inflammation. Although administration of GC has several benefits, topical or systemic GC can lead to ocular hypertension in about 30 -50% of patients depending on the route of administration, and sustained GC treatment can lead to secondary iatrogenic open-angle glaucoma if not withdrawn (4 -6). Although GC-induced glaucoma is a form of secondary iatrogenic open-angle glaucoma, its clinical presentations are similar in many ways to POAG (5). Steroid responders are at high risk of developing POAG (6), and almost all POAG patients are considered to be steroid responders (7).
It is known that GC-induced ocular hypertension is caused by increased resistance to aqueous humor outflow at the trabecular meshwork (TM) tissue (5,(7)(8)(9). However, the molecular pathological mechanisms of how GC treatment leads to increased outflow resistance at the TM are still unclear. Several studies have shown that GC-induced ocular hypertension is associated with increased accumulation of ECM in the TM, particularly in the juxtacanalicular connective tissue region and the inner wall endothelium of Schlemm's canal (10,11). In addition, GC-induced ocular hypertension is associated with other morphological changes including decreased intratrabecular spaces (7), fingerprint-like depositions in the uveal meshwork (12,13), fibrillar material in the juxtacanalicular connective tissue region (14), and the presence of ␣-smooth muscle actin positive myofibroblasts in the Schlemm's cannal region (8). GCs modulate the expression and secretion of various proteins including myocilin (15), fibronectin (16,17), collagen (17), laminin (18), and elastin (19) in TM cells treated with a potent GC, dexamethasone (Dex). Dex also alters the expression of proteolytic enzymes, which regulate ECM turnover in the TM (20 -22). Apart from humans, GC-induced ocular hypertension has been reported in eight other species (23). This phenomenon has also been observed in ex vivo human and bovine anterior perfusion culture (7,24). Previously, we developed a mouse model of GC-induced ocular hypertension by administering topical Dex eye drops three times daily for several weeks (25). We also recently developed a rapid and sustained GC-induced ocular hypertension mouse model by administrating periocular injections of Dex acetate once a week (9). Using these mouse models and primary human TM cells, we have demonstrated the role of ER stress and increased ECM accumulation in GCinduced ocular hypertension. However, it is not understood how Dex leads to ER stress in the TM.
Transforming growth factor ␤2 (TGF␤2) is a major regulator of the ECM in the TM. TGF␤2 is known to be involved in the pathogenesis of POAG (26). TGF␤2 levels have been shown to be elevated in the aqueous humor of eyes from POAG patients (27,28). Treatment of primary human TM cells with recombinant TGF␤2 has been shown to increase the synthesis and deposition of ECM (29,30) and induce ECM cross-linking enzymes in TM cells (31)(32)(33). The binding of active TGF␤2 to its receptor leads to phosphorylation of SMAD2/3 and activation of the SMAD signaling pathway (34). Adenoviral injections of bioactive TGF␤2 elevate IOP in a SMAD3-dependent manner and also induce ECM deposition in mouse TM tissues (35,36). In addition, exogenous TGF␤2 elevates IOP ex vivo in the human anterior segment perfusion system (29). Although the profibrotic effects of TGF␤2 and GC in TM are well-established, the link between GC and TGF␤2 signaling in the TM has not yet been studied. Several studies have shown inhibitory cross-talk between GC and TGF␤2 signaling in some but not all cell types (37)(38)(39)(40). GCs inhibit TGF␤ signaling in multiple cell types either by reducing the bioavailability of TGF␤ or by regulating the SMAD signaling pathway (37,38,40). A synergistic effect of GC and TGF␤ signaling was observed in ovarian cancer cells (39). A recent study showed similarities in proteomic changes between GC-and TGF␤2-treated TM cells, suggesting that a similar regulation is employed by both of these pathways (41). We hypothesize that cross-talk between GC and TGF␤2 signaling in the TM plays an important role in inducing ECM accumulation and ER stress, elevating IOP. In the present study, we explored this cross-talk between GC and TGF␤2 signaling and its role in the regulation of ECM, ER stress, and Dex-induced IOP elevation.

Dex increases TGF␤2 in primary human TM cells
We hypothesized that Dex-induced ECM accumulation and ER stress are regulated by TGF␤2 signaling. Inactive TGF␤2 (ϳ54 kDa) is secreted as a precursor form that associates with latency-associated peptide. The dimeric, mature, and biologically active forms of TGF␤2 (25 kDa) are created with cleavage of the latency-associated peptide (42). We first examined whether treatment of primary human TM cells with Dex alters TGF␤2 and ECM proteins (Fig. 1). Primary human TM cells were treated with either vehicle (Veh) or Dex (100 nM). TGF␤2 and ECM proteins were examined in the conditioned medium and in cellular lysates. Western blot analysis of the conditioned medium demonstrates that Dex increases both pro and active forms of TGF␤2 along with fibronectin (Fig. 1A). Densitometric analysis demonstrates that Dex increases both the active form of TGF␤2 and fibronectin by more than 3-fold over vehicle-treated cells (Fig. 1B). ELISA revealed that Dex significantly increases both the total and active forms of TGF␤2 in the conditioned medium (Fig. 1, C and D). Dex also increases fibronectin, collagen IV, and the active form of TGF␤2 in the cellular lysates (Fig. 1E). Interestingly, Dex induction of ECM and TGF␤2 is also associated with increased ER stress as determined by ATF4 and CHOP analysis in primary human TM cells (Fig. 1E). Dex treatment caused a 1.3-fold increase in TGF␤2 mRNA levels and a 2-fold increase in fibronectin mRNA levels (Fig. 1F). Although Dex induction of fibronectin mRNA expression was not statistically significant, Dex-induced fibronectin protein is evident in Fig. 1, A and E, indicating that primary TM cells responded well to Dex treatment. Together, these data indicate that Dex increases secretion and activation of TGF␤2, which correlates with Dex-induced ECM and ER stress induction in human TM cells.

Dex activates the SMAD signaling pathway in human TM cells
Previous studies have shown that TGF␤2 activates the canonical SMAD signaling pathway, which regulates ECM in TM cells (43). Because Dex increases the active form of TGF␤2, we next examined whether Dex also leads to activation of the SMAD signaling pathway in human TM cells. GTM3 cells were treated at various time points with vehicle or Dex (100 nM) under serum-free conditions. Western blot analysis of cellular lysates revealed that Dex treatment increases phosphorylation of SMAD3, starting from 60 min of treatment, compared with vehicle treatment, which correlated well with the timing of induction of TGF␤2 and the ER stress marker GRP78 ( Fig. 2A). Cytoplasmic to nuclear translocation of the phosphorylated SMAD complex is a critical step in TGF␤-mediated signal transduction. Therefore, we examined whether Dex increases the nuclear translocation of SMAD3. As shown in Fig. 2B, Western blot analysis demonstrated a prominent increase in pSMAD3 and SMAD3 in the nuclear fractions. GTM3 cells treated with recombinant TGF␤2 (5 ng/ml) alone for 60 min were used as a positive control. The presence of lamin A/C and the absence of GAPDH in nuclear fractions indicated the relative purity of the nuclear fractions. Immunostaining for SMAD4 revealed that Dex dramatically increases nuclear levels of SMAD4 over control (Fig. 2C). The SMAD reporter assay further demonstrated a more than 2-fold increase in SMAD-luciferase activity after 16 h of Dex treatment (Fig.  2D). These data indicate that Dex increases the active form of TGF␤2 and activates the downstream SMAD signaling pathway.

Inhibition of TGF␤ signaling reduces Dex-induced ECM accumulation and ER stress in human TM cells
We next investigated whether inhibition of TGF␤ signaling via SMAD3 knockdown abrogates Dex-induced ECM changes and ER stress. SMAD3 knockdown was achieved by various approaches: with CRISPR (CR)-Cas9 targeting of SMAD3 or TGF␤ signaling promotes GC-induced ocular hypertension using a selective chemical inhibitor of SMAD3 phosphorylation (SIS3) or a selective TGF␤ receptor kinase I inhibitor (LY364947). Primary human TM cells (n ϭ 3 cell strains) were treated with Dex for 3 days, and immunostaining was performed in nonpermeabilized (no Triton added) cells to stain for extracellular fibronectin. Dex increased extracellular fibronectin staining (Fig. 3A), which was completely blocked by SIS3 and LY364947. Western blot analysis further demonstrated that Dex-induced fibronectin and ER stress markers, including GRP78, GRP94, and CHOP, are reduced when GTM3 cells are treatedwithSIS3,LY364947,orCR-SMAD3.Reducingthephosphorylation of SMAD3 in GTM3 cells treated with SIS3, LY364947, or CR-SMAD3 further supported inhibition of TGF␤ signaling (Fig. 3B). Densitometric analysis of Western blots for fibronectin demonstrated a significant increase in fibronectin after Dex treatment, which was completely blocked by co-treatment with SIS3, LY364947, or SMAD3 knockdown in GTM3 cells (Fig. S1) and in primary human TM cells (Fig.  3C). Inhibition of TGF␤-induced SMAD signaling was further examined using a SMAD reporter assay. GTM3 cells were transfected with SMAD-luciferase constructs and treated with vehicle or Dex with or without TGF␤-signaling inhibitors (LY364947, SIS3, or CR-SMAD3). As shown in Fig. 3D, Dex significantly increased SMAD-luciferase activity, which was significantly reduced in the presence of TGF␤-signaling inhib-itors. These results clearly demonstrate that inhibition of TGF␤ signaling prevents Dex-induced ECM accumulation and ER stress in TM cells.
We recently demonstrated that Dex-induced intracellular fibronectin and type I collagen interact and co-localize with KDEL antibody, which recognizes the ER stress markers GRP78 and GRP94 (44). Here, we examined whether inhibition of TGF␤ signaling prevents Dex-induced co-localization of fibronectin and type I collagen with KDEL in primary human TM cells (Fig. 4). Human primary TM cells were stained with fibronectin and KDEL antibody (Triton-permeabilized). Co-localization of fibronectin ( Fig. 4) or type I collagen (( Fig. S2) with KDEL antibody demonstrated that Dex increases the co-localization of fibronectin or type I collagen with KDEL, which was completely blocked by inhibition of TGF␤ signaling using either SIS3 or LY364947 in human primary TM cells. Together, these data indicate that inhibition of TGF␤ signaling reduces Dex-induced ECM accumulation and ER stress in human primary TM cells.

Increased TGF␤2 in the aqueous humor of Dex-treated mice
We next explored whether Dex regulates TGF␤2 signaling in vivo. C57 mice were treated with topical vehicle or Dex eye drops daily for several weeks, and TGF␤2 levels were examined in the aqueous humor samples before and after IOP elevation TGF␤ signaling promotes GC-induced ocular hypertension (Fig. 5). Western blotting (Fig. 5, A, C, and E) and densitometric analyses (Fig. 5, B, D, and F) of aqueous humor samples demonstrated a significant induction of the active form of TGF␤2 starting from 3 weeks of Dex treatment. Interestingly, the active form of both TGF␤2 (Fig. 5, A and B) and fibronectin (Fig. S3A) showed little change prior to IOP elevation at 1 week of treatment. However, active TGF␤2 and fibronectin levels were significantly elevated upon 3 and 8 weeks of Dex treatment (Figs. 5, C-F, and S3, B and C). We further examined TGF␤2 levels in the aqueous humor by ELISA. C57 mice were treated with weekly periocular injections of vehicle or Dex for 2 weeks, and aqueous humor samples were subjected to ELISA for TGF␤2. Dex-treated mice demonstrate significant increases in both total and active TGF␤2 compared with vehicle-treated mice (Fig. 6, A and B). Western blot analysis of anterior segment tissues from 3-week vehicle and Dex-treated mice demonstrated that Dex treatment dramatically increases the active form of TGF␤2 in anterior segment tissues (Fig. 6C). Furthermore, immunostaining for TGF␤2 showed little staining in vehicle-treated mice, whereas TGF␤2 appeared to increase in the TM and ciliary body of Dex-treated mice (Fig. 6D). We also determined whether Dex induces Smad signaling in vivo by Western blot analysis of anterior segment tissue lysates. Dextreated mice had increased pSmad3 compared with vehicletreated mice, further indicating activation of TGF␤ signaling in TM (Fig. 6E).

Figure 2. Dex treatment activates SMAD signaling.
A, GTM3 cells treated with Veh or Dex (100 nM) for 0.5, 1, 12 and 24 h under serum-free conditions. Cellular lysates were examined for the timing of TGF␤-SMAD signaling activation. Increases in SMAD3 phosphorylation and TGF␤2 was observed starting at 1 h after Dex treatment. Increased GRP78 and fibronectin was observed after 12 h of Dex treatment (n ϭ 2 replicates). B, GTM3 cells were treated with Veh or Dex for 1, 2, and 12 h under serum-free conditions. Cytoplasmic and nuclear fractions were subjected to Western blot analyses of SMAD-signaling proteins. GTM3 cells treated with recombinant TGF␤2 (5 ng/ml) was used as a positive control for the activation of SMAD signaling. GAPDH and lamin A/C were used to demonstrate the relative purity of cytoplasmic and nuclear fractions, respectively (n ϭ 2 replicates). C, SMAD4 localization was examined by immunostaining in GTM3 cells treated with Veh or Dex for 60 min. Increased nuclear localization of SMAD4 in Dex-treated cells indicates activation of SMAD signaling. n ϭ 2 replicates. D, GTM3 cells transfected with SMAD-luciferase reporter construct for 24 h and treated with Veh or Dex for 16 h under serum-free conditions. "Relative luciferase units" represents firefly luciferase intensities normalized to Renilla activity. A more than 2-fold increase in SMAD reporter activity was observed in Dex-treated GTM3 cells compared with Veh-treated cells (unpaired t test, n ϭ 3).

TGF␤ signaling promotes GC-induced ocular hypertension Smad3 ؊/؊ mice are protected from Dex-induced ocular hypertension, ER stress, and abnormal ECM accumulation
To further test whether TGF␤ signaling regulates Dex-induced ocular hypertension, we utilized Smad3 Ϫ/Ϫ mice. A previous study shows that TGF␤2-induced ECM remodeling in the TM and IOP elevation depends on SMAD3 (36). Weekly bilateral periocular injections of vehicle or Dex acetate suspension were given to 3-month-old Smad3 Ϫ/Ϫ and WT littermates. Conscious IOPs were measured as described previously (9). A significant increase in IOP was observed in WT mice after 1, 2, and 3 weeks of Dex treatment compared with vehicle-treated WT mice (Fig. 7A). Notably, Dex did not significantly increase IOP in Smad3 Ϫ/Ϫ mice compared with WT mice treated with Dex. Interestingly, IOPs in Smad3 Ϫ/Ϫ mice were similar to WT mice treated with vehicle, suggesting complete protection from Dex-induced ocular hypertension. Smad3 ϩ/Ϫ (heterozygous) mice responded with elevated IOP measurements similar to WT mice with Dex treatment (data not shown). We further examined whether Smad3 Ϫ/Ϫ mice were also protected from Dex-induced ECM accumulation and ER stress in the anterior segment lysates (Fig. 7, B and C). Western blotting (Fig. 7B) and its densitometric analysis (Fig. 7C) clearly demonstrated that Dex significantly increases fibronectin and ER stress markers including GRP78, GRP94, ATF4, and CHOP, which are significantly reduced in Smad3 Ϫ/Ϫ mice. These data demonstrate that inhibition of TGF␤2 signaling via the genetic loss of Smad3 prevents Dex-induced abnormal ECM, ER stress, and ocular hypertension in mice.

Inhibition of TGF␤ signaling via treatment with LY364947 reduces Dex-induced ocular hypertension in mice
We further examined whether inhibition of TGF␤2 signaling blocks Dex-induced IOP elevation via treating mice with topical ocular eye drops of LY364947 (selective TGF␤ receptor kinase I inhibitor). C57BL/6J mice were treated with weekly bilateral periocular injections of vehicle or Dex acetate suspen-

Figure 3. Inhibition of TGF␤ signaling attenuates Dex-induced elevation of ECM and ER stress.
A, primary human TM cells treated with Dex with or without TGF␤-signaling inhibitors (LY364947 is a TGF␤ receptor I kinase inhibitor, and SIS3 is a selective inhibitor of SMAD3 phosphorylation) for 72 h and stained for fibronectin without Triton permeabilization. Dex increased extracellular fibronectin, which was blocked by TGF␤-signaling inhibitors (n ϭ 3 cell strains). Scale bar ϭ 50 m. B, Western blot analysis of fibronectin, ER stress markers (GRP78, GRP94, and CHOP), and TGF␤-signaling proteins (pSMAD3 and total SMAD3) in GTM3 cells treated with Veh or Dex in the presence or absence of LY364947, SIS3, and CR-SMAD3 (CRISPR-Cas9 targeting SMAD3). C, densitometric analysis of fibronectin Western blotting in human primary TM cells treated with Dex and TGF␤2-signaling inhibitors (n ϭ 3, one-way ANOVA; *, p Ͻ 0.05; **, p Ͻ 0.01). D, GTM3 cells were transfected with SMAD-luciferase constructs and treated with Veh or Dex with or without TGF␤-signaling inhibitors (LY364947, SIS3, and CR-SMAD3). Dex-induced SMAD reporter activity was significantly reduced in GTM3 cells treated with inhibitors of TGF␤ signaling (n ϭ 3 replicates, one-way ANOVA; ***, p Ͻ 0.001).

TGF␤ signaling promotes GC-induced ocular hypertension
sion as described previously (9). After 1 week of Dex injection, conscious IOP revealed that Dex treatment elevated IOP significantly compared with vehicle-injected mice (data not shown). Topical ocular LY364947 eye drops (1%) were given to the right eyes, and the contralateral left eyes were given vehicle eye drops twice daily after 1 week of periocular injections (Fig. 8). The left eyes, treated with periocular Dex and vehicle control eye drops, showed significantly elevated IOP after 2 and 3 weeks of treatment compared with periocular vehicle injected mice. The right eyes, treated with periocular Dex acetate suspension and LY364947 eye drops, demonstrated a significant reduction in IOP compared with the contralateral left eyes treated with periocular Dex acetate suspension and vehicle eyes drops. We did not observe any significant change in IOP in mice treated with periocular vehicle suspension and LY364947 eye drops compared with mice treated with periocular vehicle suspension and vehicle control eye drops. These data indicate that inhibition of TGF␤2 signaling blocks Dex-induced IOP elevation.

ATF4 and CHOP is involved in Dex-induced TGF␤2 activation and ECM accumulation in TM cells
As shown previously (25) and in Fig. 1, Dex-induced TGF␤2 activation is associated with increased ATF4 and CHOP in TM cells. Moreover, Chop knockout mice are protected from Dexinduced ocular hypertension (25). Therefore, we examined whether Dex-induced TGF␤2 activation is regulated by ATF4 and CHOP in human TM cells. First, we examined whether loss of ATF4 or CHOP inhibits TGF␤2 signaling, thus preventing an abnormal ECM accumulation and induction of ER stress (Fig.   9A). GTM3 cells were transfected with CRISPR-Cas9 targeting CHOP or ATF4 for 24 h and then treated with vehicle or Dex for an additional 48 h. Western blot analysis demonstrated that Dex increases precursor and active TGF␤2, fibronectin, and ER stress markers including GRP78, ATF4, and CHOP, which were reduced in cells pretreated with CR-ATF4 or CR-CHOP resulting in knockdown of these proteins. Western blotting data confirmed that CR-ATF4 or CR-CHOP treatment reduces protein levels of ATF4 and CHOP in TM cells transfected with these plasmids and treated with Dex. We further confirmed that CRISPR-Cas9 targeting ATF4 reduces ATF4 levels in control GTM cells, although control Cas9 plasmids did not alter ATF4 levels (Fig. S4). We next examined whether ATF4 knockdown prevents Dex-induced TGF␤2 in the conditioned medium. GTM3 cells were transduced with adenovirus 5 expressing CRISPR-Cas9 targeting ATF4 or a dominant-negative inhibitor of ATF4 (45) followed by treatment with vehicle or Dex for 48 h. Western blot analysis of the conditioned medium demonstrated that ATF4 knockdown decreases Dex-induced expression of pro and active TGF␤2 levels (Fig. 9B). Coomassie staining demonstrated relatively similar protein loading of the conditioned medium (Fig. S5). We also observed that knockdown of ATF4 or CHOP reduced both active and total TGF␤2 levels in the conditioned medium of control GTM cells (Fig. S6).
We assessed whether knockdown of ATF4 and CHOP would reduce the Dex-induced SMAD signaling pathway using a SMAD-luciferase assay (Fig. 9C). GTM3 cells were co-transfected with SMAD-luciferase constructs and plasmids expressing CRISPR-Cas9 targeting ATF4 or CHOP for 24 h followed by treatment with vehicle or Dex for an additional 24 h. A SMAD reporter assay revealed that Dex significantly increased SMADluciferase activity, which was significantly reduced by genetic knockdown of ATF4 or CHOP. We also observed that control Cas9 plasmids did not alter SMAD-luciferase activity significantly in control GTM3 cells compared with cells treated with Lipofectamine alone (Fig. S7). In addition, CRISPR-Cas9 targeting ATF4 or CHOP did not reduce SMAD-luciferase activity significantly in control GTM3 cells compared with control Cas9 treatment (Fig. S7). We next examined whether overexpression of ATF4 alone was sufficient to induce TGF␤2 in TM cells without Dex treatment (Fig. 9D). GTM3 cells were transduced with adenovirus-expressing ATF4 (45), and the conditioned medium was examined for TGF␤2. Western blot analysis demonstrated that overexpression of ATF4 alone is sufficient to increase the precursor and active forms of TGF␤2 in TM cells. Western blot analysis of cell lysates confirmed that GTM3 cells transduced with Ad5-ATF4 overexpressed ATF4 compared with Ad5-empty-treated cells (Fig. S8). Consistent with this finding, the overexpression of ATF4 increased SMAD-luciferase activity in a manner similar to Dex treatment. Interestingly, co-treatment with Dex and ATF4 further enhanced SMAD activity, suggesting synergistic interactions (Fig. 9E). These data support the idea that ATF4 and CHOP are involved in the activation of TGF␤2 signaling in Dex-induced ocular hypertension.

TGF␤ signaling promotes GC-induced ocular hypertension
Discussion GC-induced IOP elevation is associated with increased ECM accumulation and ER stress in the TM. However, it is not understood how GC induces increased ECM accumulation and ER stress in the TM. Here, we have demonstrated that Dexinduced IOP elevation, ECM accumulation, and ER stress are mediated by TGF␤2 signaling in the TM. Dex induces the secretion and activation of TGF␤2, which triggers canonical SMAD signaling to increase ECM synthesis and deposition, inducing ER stress in primary human TM cells and in mouse TM and resulting in elevated IOP. The inhibition of TGF␤ signaling prevents Dex-induced ECM deposition and ER stress and rescues Dex-induced IOP elevation. These findings indicate that the activation of TGF␤2 signaling is responsible for ECM remodeling, ER stress, and IOP elevation in GC-induced glaucoma.
TGF␤2 is secreted in the latent form and remains inactive with latent TGF␤-binding proteins and latency-associated peptide, and the active form of TGF␤2 is required for downstream signal transduction (46). The active form of TGF␤2 is increased in the aqueous humor of POAG patients (47). The overexpression of WT TGF␤2 (biologically inactive) alone is not enough to elevate IOP in mice, but expression of a bioactivated form of TGF␤2 in the TM is required to elevate IOP in mice (35). Consistent with this, we observed that Dex increased the active form of TGF␤2 levels in conditioned medium and aqueous humor collected from Dex-treated mice, suggesting that Dex induces extracellular activation of TGF␤2, which is important for the activation of downstream TGF␤2 signaling. We also observed that the activation of SMAD signaling and the timing of activation of TGF␤2 in the aqueous humor in Dex-treated mice correlated well with IOP elevation, indicating that the activation of TGF␤2 in the aqueous humor is associated with IOP elevation. Importantly, inhibiting TGF␤ signaling completely protects Dex-induced ECM accumulation, ER stress, and IOP elevation.
Dex increased TGF␤2 levels as well as downstream SMAD proteins within 60 min of treatment in cultured TM cells, suggesting that the effects of Dex on TGF␤2 are post-translational. In addition, the timing of TGF␤2 induction by Dex correlates well with the timing of fibronectin and ER stress induction. Therefore, Dex-induced fibrosis and ER stress are most likely regulated by the TGF␤2 signaling pathway. In support of this hypothesis, we observed that inhibition of TGF␤ signaling at the TGF␤ receptor level or SMAD3 level prevents Dex-induced fibrosis and ER stress in primary human TM cells. More importantly, the loss of SMAD3, which is required for TGF␤2-induced IOP elevation, protects from Dex-induced IOP elevation, ECM accumulation, and induction of ER stress. Inhibition of TGF␤ signaling via LY364947 significantly reduces Dex-in- Figure 7. Smad3 ؊/؊ mice are protected from Dex-induced ocular hypertension, ER stress, and abnormal ECM accumulation. Three-month-old WT (Smad3 ϩ/ϩ ) and Smad3 Ϫ/Ϫ littermates were given periocular injections of Veh or Dex suspension for 3 weeks. A, Dex treatment significantly elevated IOP in WT mice, whereas IOP in Dex-treated Smad3 Ϫ/Ϫ mice was similar to Veh-treated WT mice, indicating that Smad3 Ϫ/Ϫ mice are protected from Dex-induced ocular hypertension (n ϭ 8 -10 in each group, one-way ANOVA). Western blotting (B) and densitometric analysis (C) of anterior segment lysates of above described mice revealed that Dex induced ECM (fibronectin) and ER stress (GRP78, GRP94, CHOP and ATF4) in WT mice, which were significantly decreased in Smad3 Ϫ/Ϫ mice treated with Dex (n ϭ 4 in each group, two-way ANOVA). *, p Ͻ 0.05; **, p Ͻ 0.01; and ***, p Ͻ 0.001 versus Veh-treated Smad3 ϩ/ϩ mice; #, p Ͻ 0.05; ##, p Ͻ 0.01; and ###, p Ͻ 0.001 versus Dex-treated Smad3 ϩ/ϩ mice).

TGF␤ signaling promotes GC-induced ocular hypertension
duced ocular hypertension. The TGF␤-signaling inhibitors utilized in this study are not exclusively selective for TGF␤2 signal transduction, as these inhibitors can also block TGF␤1 signaling. Because TGF␤2 is a major isoform in the TM and the fact that Dex activates the secretion of TGF␤2, it is likely that the effects observed in this study are attributable to TGF␤2 rather than TGF␤1. Our analysis of SMAD signaling using a SMADluciferase reporter assay, which measures several SMADs, demonstrates that Dex increases total SMAD transcriptional activity. Therefore, it is likely that Dex may increase other SMADs rather than just SMAD3 alone. However, SMAD3 has been shown to be required for TGF␤2-induced ocular hypertension (36). Consistent with this understanding, we observed that inhibition of SMAD3 was able to prevent Dex-induced ocular hypertension. These studies indicate an essential role for activated TGF␤2 signaling in Dex-induced ocular hypertension. Because GCs are known to influence a variety of pathways, the influence of these other pathways on Dex-induced fibrosis cannot be eliminated.
Although it is known that ER stress plays a role in fibrosis (48,49), it is not understood how ER stress can lead to fibrosis. Based on the findings of our study, it is plausible that chronic ER stress can modify TGF␤ signaling, which may worsen already existing fibrosis. Chronic ER stress-induced transcriptional factors such as ATF4 and CHOP are known to be involved in cell dysfunction and death (45,50,51). However, it is not understood whether these chronic ER stress factors play any role in fibrosis. In our previous (25) and current studies, we observed that Dex-induced ATF4 and CHOP is associated with Dexinduced abnormal ECM accumulation and TGF␤2 activation. Genetic knockdown of ATF4 or CHOP prevents Dex-induced TGF␤2 activation and ECM accumulation in TM cells. Consistent with this, the overexpression of ATF4 alone is sufficient to induce TGF␤2 activation without Dex treatment. Interestingly, we have demonstrated previously that Chop knockout mice are protected from Dex-induced ocular hypertension. These findings suggest that chronic ER stress leads to TGF␤2 activation, which in turn leads to fibrosis in the TM. It is interesting to note that both ATF4 and CHOP are significantly elevated in glaucomatous TM tissues, and chronic ER stress is associated with abnormal ECM accumulation in glaucomatous TM tissues (44,52). Considering the known role of TGF␤2 in glaucoma, it is possible that chronic ER stress in the TM may activate TGF␤2, which can further exacerbate ER stress via induction of abnormal ECM accumulation. Alternatively, elevated TGF␤2 levels in glaucoma may induce ER stress via abnormal ECM accumulation in the TM. In addition, TGF␤2 may directly induce chronic ER stress factors including ATF4 and CHOP, which can further potentiate TGF␤ signaling, worsening abnormal ECM accumulation in TM. Future studies will be aimed at understanding whether TGF␤2 can directly induce ER stress in the TM.
We have shown previously that the knockout of Chop in mice protects against Dex-induced ocular hypertension. However, the exact mechanism was not clear. Based on our current findings, it is likely that the knockout of Chop in mice protects against Dex-induced ocular hypertension via inhibiting Dexinduced TGF␤2 signaling, thus preventing abnormal ECM accumulation in the TM. Consistent with this understanding, we demonstrated previously that Chop knockout mice prevent Dex-induced ECM accumulation (25). Although it is not clear from this study how ATF4 and CHOP activate TGF␤2, it is possible that ATF4 and CHOP may interact with GC-induced response elements to modulate TGF␤2 activation and signaling. A previous study demonstrates that a major ER chaperone, calreticulin, is required for TGF␤-induced fibrosis (53). Because TGF␤2 and ECM proteins are synthesized and processed in the ER, it is possible that ER stress plays an important role in TGF␤-mediated ECM remodeling.
Although it is clear that ATF4 and CHOP are involved in Dex-induced TGF␤2 activation, there are several other factors including integrins, proteases, and TSP1 that may be involved in Dex-induced TGF␤2 activation in the TM (46,54). It is possible that Dex increases these extracellular factors involved in the activation of TGF␤2. For example, TSP1 has been shown to induce the activation of TGF␤2 (55). Interestingly, TSP1 levels were found to be increased in one-third of POAG patients, and TSP1 is induced by Dex in vitro (56). Furthermore, TSP1-null mice have a lower IOP when compared with their WT littermates (57). Because TGF␤2 lacks an RDG motif (46), it is unlikely that integrins are involved in Dex-induced TGF␤2 activation. Both MMP2 and MMP9 have been shown to activate TGF␤ (54). It is likely that both MMP and TSP1 play a role in modulating Dex-induced TGF␤ activation in the TM.
In non-ocular cell types, the cross-talk between GC and TGF␤2 pathways has different effects depending on the cell type. Some studies have demonstrated synergistic effects when cells are treated with both Dex and TGF␤2. Wickert et al. (58) report a strong induction of TGF␤-mediated PAI-1 and con-

TGF␤ signaling promotes GC-induced ocular hypertension
nective tissue growth factor (CTGF) expression in rat hepatocytes upon simultaneous stimulation with TGF␤ and Dex compared with that of TGF␤ alone. Similarly, Kimura et al. (59) report that GC enhance TGF␤-induced PAI-1 expression in human renal proximal tubular cells. In contrast, other studies have described opposing interactions between GC and TGF␤ in lung fibroblasts (60) and hepatic stellate cells (40). It appears that these interactions are cell type-dependent. Differential expression of GC receptors may play a major role in determining these contrasting effects of GC on TGF␤ signaling. GR␣ is responsible for the physiological and pharmacological effects of GC, whereas GR␤ acts as a negative regulator of GR activity (61). The lower expression of GR␤ in TM cells (62) may contribute to these differential interactions.
In summary, this study demonstrates that cross-talk between GC and TGF␤2 signaling regulates GC-induced ECM remodeling in the TM. Similar cross-talk between these pathways may also play a major role in the pathophysiology of POAG. Furthermore, our studies demonstrate that the manipulation of these interactions can provide novel interventions for steroid-induced glaucoma.

Antibodies
Antibodies were purchased from the following sources:  The conditioned medium was examined for TGF␤2 by Western blot analysis. n ϭ 3 replicates. C, GTM3 cells, transfected with SMAD-luciferase and CRISPR-Cas9 targeting SMAD3 or CHOP or ATF4 constructs, were treated with Veh or Dex for 24 h. Knockdown of ATF4 or CHOP in Dex-treated TM cells prevented SMAD activation in a SMAD-luciferase assay (n ϭ 3, one-way ANOVA). ***, p Ͻ 0.001). D, GTM3 cells were transduced with adenovirus 5-expressing ATF4 for 48 h. The conditioned medium was examined for TGF␤2 by Western blot analysis (n ϭ 3 replicates). E ϭ control adenovirus. E, GTM3 cells were transfected with SMAD-luciferase constructs and treated with Veh, Dex, or adenovirus 5-expressing ATF4 (Ad5-ATF4). Treatment with recombinant TGF␤2 was utilized as a positive control for activation of SMAD signaling. Both Dex and overexpression of ATF4 alone were able to increase SMAD-luciferase activity significantly. A combination of Dex with ATF4 overexpression further increased SMAD-luciferase activity compared with ATF4 alone. n ϭ 3, one-way ANOVA; *, p Ͻ 0.05; ***, p Ͻ 0.001.

TGF␤ signaling promotes GC-induced ocular hypertension
Experimental animals C57BL/6J (BL6) mice and Smad3 knockout (Smad3 Ϫ/Ϫ ) mice on pure 129 background were obtained from The Jackson Laboratory. Smad3 Ϫ/Ϫ mice were bred and genotyped to ensure that each mouse had the genotype as reported previously (36,63,64). Animals were fed standard chow ad libitum and housed in 12-h light/12-h dark conditions. All experimental procedures were approved by the University of North Texas Health Science Center Institutional Animal Care and Use Committee review board.

Dex treatment of mice
Topical 0.1% Dex phosphate (Bausch & Lomb Inc.) or vehicle eye drops containing sterile phosphate-buffered saline (PBS) were applied to 3-month-old C57BL/6J mice three times daily for 8 weeks (Figs. 5 and 6) as described previously (25). At the end of the treatment, aqueous humor or anterior segment tissues were collected for Western blot analysis, whereas whole eyes were fixed and sectioned for immunostaining. We also utilized our recently developed mouse model of Dex-induced ocular hypertension (9). BL6 mice were obtained from The Jackson Laboratory, and baseline IOP were measured to ensure that each mouse had normal IOP before treatment. Bilateral periocular vehicle or Dex acetate injections were performed every week as described previously (9). To evaluate the effects of genetic knockdown of SMAD3, WT or Smad3 Ϫ/Ϫ littermates were given weekly periocular injections of vehicle or Dex acetate for 3 weeks. Conscious IOPs (IOPs measured without keeping mice under any anesthesia) were measured weekly. To determine the effects of LY364947 on Dex-induced ocular hypertension, BL6 mice were given weekly bilateral periocular injections of Veh or Dex acetate suspension. One week after periocular injections, the right eyes were treated with LY364947 eye drops (1%), whereas the contralateral left eyes were treated with vehicle control (1% DMSO in water) for 2 weeks. IOP was monitored weekly.

Dex treatment of human TM cells
Four different primary human TM cell strains and a transformed GTM3 cell line (65) were cultured in DMEM (Sigma) supplemented with 10% fetal bovine serum (Atlas Biologicals, Fort Collins, CO), L-glutamine (Gibco, Life Technologies), and penicillin-streptomycin (Gibco, Life Technologies). For the characterization of primary human TM cells, cells were examined for the expression of fibronectin, collagen, and laminin as well as for Dex induction of cross-linked actin networks and myocilin as described previously (15,66,67). The human primary TM cells (n ϭ 4) or GTM3 cells were treated with either vehicle (0.1% ethanol) or Dex (100 nM) (Sigma-Aldrich) in serum-free conditions at various time points, and the conditioned medium and lysates were collected for Western blot analysis. The proteins in the media were concentrated using previously described StrataClean resin (Agilent Technologies) (68). For immunostaining analysis, the cells were fixed in 4% paraformaldehyde and stained with the appropriate primary and secondary antibodies.

IOP measurements
Conscious IOP measurements were carried out using rebound tonometry as described previously (69). Mice were restrained in plastic cones and secured in a custom-made restrainer. After a few minutes of acclimatization, IOPs were recorded in a masked manner. All IOP measurements were recorded between 10 a.m. and 1 p.m.

Immunostaining
Eyes from vehicle-or Dex-treated mice were enucleated, fixed in 4% paraformaldehyde, dehydrated, and embedded in paraffin. 10-m thick sections were made using rotatory microtome. The deparaffinized and rehydrated sections were subjected to antigen retrieval in a sodium citrate buffer (pH 6). The slides were blocked in 10% goat serum containing 0.5% Triton X-100 for 2 h and incubated overnight with the appropriate primary antibody (1:50) in 10% goat serum followed by a 1.5-h incubation with the appropriate Alexa Fluor secondary antibodies (1:200, Life Technologies). The slides were mounted in DAPI mounting solution, and images were taken using a Keyence microscope (Itasca, IL). Slides incubated without primary antibody served as a negative control (data not shown). The primary human TM or transformed GTM3 cells cultured in 8-chamber slides were fixed in 4% paraformaldehyde and permeabilized with 0.5% Triton X-100 for 10 min followed by 1 h of blocking with 10% goat serum. The slides were incubated overnight with the appropriate primary antibodies (1:500 for fibronectin and type I collagen). Next day, the slides were washed three times with 1ϫ PBS and incubated with the appropriate Alexa Fluor secondary antibodies for 1.5 h. The slides were washed three times with 1ϫ PBS before being mounted in DAPI mounting solution. To examine extracellular fibronectin in primary human TM cells, TM cells were fixed and stained with fibronectin antibody without Triton permeabilization. To examine intracellular fibronectin and its co-localization with KDEL, TM cells were fixed, permeabilized with 0.5% Triton for 15 min, and stained with fibronectin and KDEL.

Western blot analysis
TM cells or mouse anterior segments dissected from enucleated eyes were lysed in lysis buffer as described previously (68). Equal protein concentrations of lysates were loaded and run in 4 -12% bis-Tris gels (NuPAGE bis-Tris gels, Life Technologies) and transferred onto polyvinylidene difluoride membrane. The blots were blocked in 10% nonfat dried milk prepared in 1ϫ PBST (PBS with Tween 20) for 2 h and then incubated with the appropriate primary antibodies (1:1000) overnight at 4°C. The blots were washed three times with 1ϫ PBST followed by a secondary antibody (horseradish peroxidase-conjugated) incubation for 1.5 h. The blots were developed using ECL detection reagents (SuperSignal West Femto maximum sensitivity substrate, Life Technologies). For phosphorylated Smad3, SuperBlock (PBS) blocking buffer (Life Technologies) was used instead of 10% nonfat dried milk for blocking and antibody incubation.

TGF␤ signaling promotes GC-induced ocular hypertension SMAD reporter assay
TGF␤2-induced signal transduction was assessed using the Cignal SMAD reporter assay kit (Qiagen, Germantown, MD) in GTM3 cells. The SMAD reporter is a mixture of an inducible SMAD-luciferase construct (encodes firefly luciferase reporter gene under the control of a minimal cytomegalovirus promoter and tandem repeats of SMAD-binding element) and a constitutively expressing Renilla luciferase construct (40:1). GTM3 cells were plated into 96-well plates and transfected with negative, positive, and SMAD reporter constructs using Lipofectamine 3000 (Life Technologies). To determine the effects of inhibition of TGF␤ signaling on Dex-mediated SMAD activity, GTM3 cells were treated with or without chemical inhibitors of TGF␤ signaling (SIS3 (10 M) and LY364947 (5 M)) along with Dex in serum-free medium for 16 h. To determine the effects of ATF4 or CHOP on Dex-induced activation of TGF␤ signaling, GTM3 cells were treated with CR-Cas9 control, CR-SMAD3, CR-CHOP, and CR-ATF4 plasmid constructs with SMAD reporter constructs. After 24 h of transfection, cells were treated with vehicle, Dex (100 ng/ml), or TGF␤2 (5 ng/ml) for 16 h. Luciferase assays were carried out using the Dual-luciferase reporter assay system (Promega, Madison, WI) according to the manufacturer's protocol.

ELISA for mouse TGF␤2
A mouse TGF␤2 DuoSet ELISA kit (catalogue No. DY7346-05, R&D Systems) was utilized to quantify total and active TGF␤2 levels in the aqueous humor samples from mice. Ninemonth-old C57 mice were treated weekly with periocular injections of vehicle (left eye) and Dex suspension (right eye). After 2 weeks of injections, ϳ5 l of aqueous humor was obtained from each eye. 3 l of aqueous humor was utilized for the analysis of active TGF␤2, and the remaining 2 l of aqueous humor was used to quantify the total TGF␤2. As per the manufacturer's instructions, samples were processed by acid activation and neutralization (1 N HCl and 1.2 N NaOH/0.5 M HEPES) prior to quantifying the total TGF␤2; the results obtained were multiplied by a dilution factor of 1.4, whereas active TGF␤2 levels were determined in samples without processing. Briefly, a 96-well microplate was coated with 100 l/well of a working concentration of capture antibody and incubated overnight at room temperature. Next day, the plate was washed three times with 1ϫ wash buffer (400 l/well), blocked with 300 l/well reagent diluent, and incubated at room temperature for 1 h. The plate was washed three times with 1ϫ wash buffer and incubated with 100 l/well sample, control, or standards in reagent diluent for 2 h at room temperature. Following the three washes, 100 l of the detection antibody diluted in reagent diluent was added to each well, and the mixture was incubated for 2 h at room temperature. The plate was washed three times with 1ϫ wash buffer, a 100 l of the working dilution of streptavidin-horseradish peroxidase was added to each well, and the mixture was incubated for 20 min at room temperature. Following the washing step, 100 l of substrate solution was added to each well, and the mixture was further incubated for 20 min at room temperature. Finally, 50 l of stop solution was added to each well, and the optical density of the plate was determined immediately using a microplate reader at a 450-nm wavelength with a 570-nm wavelength correction.

ELISA for human TGF␤2
A human TGF␤2 Quantikine ELISA kit (catalogue No. DB250, R&D Systems) was used to quantify total and active TGF␤2 levels in the conditioned medium from GTM3 cells treated with vehicle and Dex for 72 h in serum-free media. As per the manufacturer's instructions, samples were processed by acid activation and neutralization (1 N HCl and 1.2 N NaOH/ 0.5 M HEPES) prior to quantifying the total TGF␤2; the results obtained were multiplied by a dilution factor of 7.8. Active TGF␤2 levels were determined in samples without processing. Briefly, 100 l/well assay diluent RD1-17 was added to an ELISA plate precoated with mAb specific for TGF␤2. This was followed by the addition of 100 l/well standards, control, and both activated and nonactivated test samples (in duplicates). After 2 h of incubation at room temperature, the ELISA plate was washed with 1ϫ wash buffer (400 l/well) at least three times, and 200 l/well TGF␤2 conjugate was added. The ELISA plate was further incubated for 2 h at room temperature and then washed three times. 200 l/well substrate solution was added, and the mixture was incubated for 20 min at room temperature. Finally, 50 l/well stop solution was added to the plate, and the absorbance was read using a microplate reader at 450 nm with a wavelength correction of 570 nm.

Real-time PCR analysis
Quantitative real-time PCR analysis was carried out in primary human TM cells (n ϭ 3 cell strains) treated with vehicle or Dex for 48 h in serum-free conditions. Total RNA was isolated using a Mini Total RNA kit (IBI Scientific) according to the manufacturer's instructions. The purity and concentration of the isolated RNA was examined by NanoDrop 2000 (Thermo Fisher Scientific). An equal amount of RNA was used for cDNA synthesis using a SuperScript VILO cDNA synthesis kit (Thermo Fisher Scientific). Quantitative PCR was performed using 2ϫ SsoAdvanced SYBR Green supermix (Bio-Rad) in a CFX96 thermocycler (Bio-Rad). The PCR thermal profiles consisted of an initial incubation at 95°C for 60 s and 40 cycles at 95°C for 60 s, 60°C for 45 s, and 72°C for 45 s followed by a final dissociation curve step. The primer pairs used for PCR included: fibronectin, 5Ј-AGCGGACCTACCTAGGCAAT-3Ј and 5Ј-GGTTTGCGATGGTACAGCTT-3Ј; TGF␤2, 5Ј-ATCCC-GCCCACTTTCTACAG-3Ј and 5Ј-GCCATTCATGAACA-GCATCA-3Ј; and GAPDH, 5Ј-GGATGATGTTCTGGAG-AGCC-3Ј and 5Ј-CATCACCATCTTCCAGGAGC-3Ј). The PCR cycle threshold (Ct) values were obtained from the CFX96 thermocycler (Bio-Rad). Dex-induced mRNA expression of fibronectin and TGF␤2 compared with vehicle was calculated by the ⌬⌬Ct method using GAPDH as an internal control as reported previously (68,70).

Nuclear and cytoplasmic extraction
GTM3 cells treated with vehicle or Dex at various time points were used to extract nuclear and cytoplasmic fractions with NE-PER nuclear and cytoplasmic extraction reagent kits (catalogue No. 78835, Thermo Fisher Scientific) according to the TGF␤ signaling promotes GC-induced ocular hypertension manufacturer's instructions. Briefly, the treated GTM3 cells were harvested by trypsin-EDTA and centrifuged at 500 ϫ g for 5 min. The cell pellet was suspended in ice-cold cytoplasmic extraction reagent I (CER I) with vigorous vortexing. The suspension was incubated for 10 min on ice followed by the addition of ice-cold cytoplasmic extraction reagent II (CER II), vortexing, and incubating for 1 min on ice. The tube was centrifuged at high speed (16,000 ϫ g) for 5 min. The supernatant, which contained the cytosolic fraction, was collected in a different prechilled tube. The pellet, which contained the nuclei, was suspended in ice-cold nuclear extraction reagent by repeated cycles of vigorous vortexing plus 10 min of incubation on ice for four times. The tube was centrifuged at high speed (16,000 ϫ g) for 10 min, and the supernatant (nuclear fraction) was collected in a different prechilled tube for subsequent Western blot analysis.

ATF4 or CHOP knockdown/overexpression
GTM3 cells were transfected with CRISPR-Cas9-ATF4 or CRISPR-Cas9-CHOP plasmid constructs using Lipofectamine 3000 and then treated with vehicle or Dex for an additional 48 h (Fig. 9A). We also knocked down ATF4 using adenovirus 5-expressing CRISPR-Cas9 -targeting ATF4 or dominantnegative ATF4⌬RK (which inhibits endogenous ATF4 activity) as described previously (45) at multiplicity of infection ϭ 100 for 48 h. Cells were treated with vehicle or Dex for another 48 h (Fig. 9B). For overexpression of ATF4, GTM3 cells were transduced with Ad5-empty or ATF4 at multiplicity of infection ϭ 100 for 24 h (Fig. 9D). The conditioned medium was subjected to Western blot analysis.

Statistical analysis
All data are presented as mean Ϯ S.E. Statistical significance between two groups was analyzed using the unpaired 2-tailed Student's t test. For data between multiple groups, one-way ANOVA or two-way ANOVA was used. p Յ 0.05 was considered statistically significant.