Pro-Homeostatic Effects of Estrogen on Intraocular Pressure and the Trabecular Meshwork

Abstract

Purpose

Elevated intraocular pressure (IOP) results from the dysregulation of aqueous humor outflow through the mechanoresponsive trabecular meshwork (TM). Evidence suggests that low estrogen (E2) and high TGFβ2 levels are risk factors for elevated IOP and primary open-angle glaucoma. We sought to examine whether E2 signaling is involved in TM regulation of IOP homeostasis.

Methods

IOP and central cornea thickness were measured in male and female Esr1^−/− mice and wild-type littermate controls from 3 to 12 months of age. Primary human TM cells (n = 10, 6 females and 4 males) were treated with TGFβ2 and/or E2 in the presence/absence of cyclic mechanical stretch (CMS) for 24 hours. Expression differences of 17 TGFβ2-responsive genes were assayed by quantitative RT-PCR.

Results

Higher IOP was observed in Esr1^−/− females compared to wild-type females (P < 0.05). Treatment with TGFβ2 and/or E2 significantly affected the expression of 10 genes (BMP1, CCN2, FST, GREM1, LAMC1, NFATC1, SERPINE1, SMAD2, SMAD3, and VCAN), with CMS exacerbating the effects of TGFβ2. The addition of E2 ameliorated the effects of TGFβ2 on two genes (NFATC1 and SMAD2) under static conditions and five genes (BMP1, FST, LAMC1, NFATC1, and SMAD2) under CMS.

Conclusions

These findings suggest that estrogen signaling promotes homeostatic TM regulation of IOP.

Glaucoma is characterized by the loss of retinal ganglion cells (RGCs) and optic nerve cupping.¹ The only treatable risk factor for the main subtype, primary open-angle glaucoma (POAG), is elevated intraocular pressure (IOP).^2,3 IOP is maintained by the production and outflow of aqueous humor (AH).^4,5 The predominant AH outflow pathway—the conventional pathway—comprises the trabecular meshwork (TM) and Schlemm's canal.^6–8 The TM is the predominant regulator of AH outflow and IOP.⁹ IOP alterations are thought to be sensed by the TM cells as changes in mechanical stretch, allowing the TM cells to act as a homeostat for IOP.^10–12 In POAG, TM fibrosis, extracellular matrix (ECM) accumulation, and reduced outflow facility may lead to IOP dysregulation.^9,10,12–15

Among many pathways regulating IOP homeostasis, estrogen signaling may affect IOP regulation and glaucoma risk. First, we previously identified that many genes involved in hormone signaling and metabolism were responsive to cyclical mechanical stretch (CMS) of TM cells¹⁶ and that β-estradiol (E2) regulates several IOP-associated genes through ESR1.¹⁷ Second, the onset of menopause, a low estrogen state, increases IOP and glaucoma risk.^18–21 Third, an overall shorter fertility duration, hallmarked by late menarche or early menopause/ovariectomy, increases glaucoma risk.^19,22,23 Fourth, hormone replacement therapy (HRT) among menopausal women decreases their IOP and glaucoma risk as compared to those without HRT.^21,24–30 Fifth, IOP is lower during hyperestrogenic phases of the menstrual cycle and pregnancy.^18,31–36 Sixth, genetic studies have identified associations between sequence variants in the estrogen signaling pathway and elevated IOP and glaucoma risk.^37–40 Seventh, aromatase knockout mice without the ability to convert testosterone to E2 had elevated IOP and RGC cell loss in female homozygous mice by 12 weeks of age.⁴¹ Eighth, estrogen receptors and metabolizing enzymes are expressed in multiple ocular tissues with neuroprotective effects.^{21,27,42–46} These lines of evidence strongly support the role of estrogen receptor signaling in IOP regulation and in protection against POAG risk. However, the ability of E2 to counteract glaucomatous TM changes has not been explored.

We hypothesized that estrogen signaling regulates POAG- and IOP-related genes, promoting proper TM cell function, increasing AH outflow, and reducing IOP and POAG risk. Therefore, this project sought (1) to ascertain the effects of estrogen signaling loss in murine eyes and (2) to characterize the transcriptional effects of E2 on human TM (HTM) cells. Humans and mice have three estrogen receptors: estrogen receptor 1 (Esr1), estrogen receptor 2 (Esr2), and G protein-coupled estrogen receptor 1 (Gper1).^47,48 Previously, we found that Esr1 regulates several IOP-associated genes.¹⁷ For this reason, we sought to determine the effects of Esr1 loss on ocular parameters related to glaucoma using Esr1 knockout mice.

We also explored whether estrogen treatment would restore normal TM processes in a cell culture model treated with TGFβ2. Exogenous overexpression of TGFβ2 may induce ocular hypertension (OHT) in mice and decrease outflow facility in perfused human, primate, and porcine anterior segments.^49–60 The observed IOP changes result from TGFβ2-induced TM changes, including altered ECM production and turnover, increased actin stress fiber formation and subsequent increased contractility, increased phagocytosis, reduced TM cell survivability, and an overall fibrotic stiffening of TM tissue.^14,61–66 TGFβ2 induces the expression of several ECM-related genes and proteins, including but not limited to those in Table 1.^14,59,60,67 Therefore, this study assessed the ability of E2 to ameliorate glaucomatous changes induced by TGFβ2 treatment of HTM cells. Furthermore, HTM cell treatments were conducted with and without CMS to replicate what TM cells regularly experience as AH exits through the tissue.^{10–12,14,68–70}

Table 1.

Genes and Proteins Responding to TGFβ2 in the Trabecular Meshwork

Name(s)	Observed Changes	Supporting Literature
α−Smooth muscle actin (αSMA, ACTA2)	Increased gene and protein expression	65 , 108 , 115 – 123
β-Actin (F-actin, ACTB)	Increased protein expression; formation of stress fibers and cross-linked actin networks (CLANs)	65 , 101 , 116 – 120 , 124 , 125
Bone morphogenetic protein 1 (BMP1)	Increased gene expression; increased protein secretion and activity	126
Connective tissue growth factor (CTGF, CCN2)	Increased gene and protein expression	120 , 127 – 134
CD44	Increased protein expression	135
Collagen type I α 1 chain (COL1A1)	Increased gene and protein expression	65 , 101 , 108 , 118 – 121 , 136
Follistatin (FST)	Increased gene and protein expression; increased protein secretion	137
Gremlin (GREM1)	Increased gene and protein expression	138
Laminin	Increased protein expression	115 , 139 , 140
Lysyl oxidase (LOX)	Increased gene and protein expression; increased protein activity	134 , 141
Nuclear factor of activated T cells 1 (NFATC1/NFAT2/NFATc)	Hypothetical activation	(Personal correspondence with Donna Peters)
Plasminogen activator inhibitor 1 (PAI-1, SERPINE1)	Increased gene and protein expression	60 , 101 , 108 , 127 , 128 , 131 , 142 – 144
SMAD2	Decreased gene expression; phosphorylation and nuclear localization of protein; increased protein activity	108 , 116 , 117 , 120 , 125 , 133 , 136 , 145
SMAD3	Decreased gene expression; phosphorylation and nuclear localization of protein; increased protein activity	108 , 116 , 117 , 120 , 130 , 133 , 136 , 145 , 146
Transglutaminase (TGM2)	Increased gene and protein expression; increased protein expression	98 , 130
Thrombospondin 1 (TSP1, THBS1)	Increased gene and protein expression	60 , 127 , 129
Tenascin C (TNC)	Increased protein expression in response to constitutively active RhoA, a downstream target of TGFβ2	115
Versican (VCAN)	Increased mRNA expression; differential expression of mRNA splice variants	147 , 148

Methods

Animals

All animal husbandry and experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Vision and Ophthalmic Research, and the protocol was approved by the Augusta University Institutional Animal Care and Use Committee (IACUC).

Heterozygote breeding pairs of B6N(Cg)-Esr1^tm4.2Ksk/J obtained from the Jackson Laboratory (Bar Harbor, ME, USA) were used to establish an Esr1^−/− colony. Exon 3 of the Esr1 gene has been deleted in these mice, leading to a truncated transcript with a premature stop codon.⁷¹ Mice were genotyped by PCR using primers (Supplementary Table S1) and the protocol established by the Jackson Laboratory. Measurements were conducted on both homozygous and heterozygous knockout mice. Wild-type littermates were used as controls.

Intraocular Pressure

IOP was assessed by applanation tonometry with an iCare TONOLAB tonometer (Colonial Medical Supply, Franconia, NH, USA). Mice were anesthetized before IOP measurements by placing them in an isoflurane chamber for 1 to 3 minutes until their breathing had slowed. Three measurements per eye were taken between 8:30 AM and 1 PM. The measurements for both eyes were averaged, so the data reported show the average for each individual mouse, where n = number of mice/group. IOP was assessed weekly or biweekly in Esr1^−/− mice and wild-type littermate control mice from 13 to 52 weeks of age (Supplementary Fig. S1). IOP measurements over 4 weeks were averaged to obtain a monthly mean (Supplementary Table S2). The experimenter was not masked to the genotype.

Tissue Histology

Following euthanasia by CO₂ or isoflurane, eyes were immediately placed in Davidson's fixative or JB-4 fixatives. After >24 hours, the eyes in Davidson's fixative were moved to 70% EtOH. The eyes were then either embedded in JB-4 resin or taken to the Augusta University Electron Microscopy and Histology Core (RRID:SCR_026810) for paraffin embedding, sectioning, and hematoxylin and eosin staining. Sections were imaged with an ECHO Revolve 4K Hybrid Fluorescence Microscope (Echo, San Diego, CA, USA).

Optical Coherence Tomography

Iridocorneal angle and central corneal thickness (CCT) were assessed bilaterally by spectral domain optical coherence tomography (OCT) using a Bioptigen Envisu-R2200 System (Leica Microsystems, Wetzlar, Germany) in the Visual Function Assessment core at Augusta University (RRID:SCR_027277). Assessments were performed at 3, 6, 9, and 12 months of age. Mice were anesthetized with 100 mg/mL ketamine (NDC code 11695-0703-1; Covetrus, Portland, ME, USA) and 30 mg/cc xylazine (NDC code 59399-110-20) 10 µL/g. A rectangular scan of each eye was used to assess the openness of the iridocorneal angle. Iridocorneal angles were considered open if the angle was open for the majority of its circumference (i.e., angles with one or two small, isolated synechiae were not considered closed). Mice with consistently closed angles were excluded from IOP analyses. Corneal damage was assessed from a radial scan. Eyes with consistent damage (i.e., unresolved pits, ulcers, etc., within the pupillary axis that affected CCT measurements) were excluded (Supplementary Table S3). CCT was measured manually from the radial scan using the caliper function available on the Bioptigen software. When available, CCT values from the left and right eyes were averaged. The value from the undamaged cornea was used for mice with one damaged cornea. The experimenter was not masked to the genotype.

Cell Culture

Independent strains (n = 10) of primary HTM cells were derived, cultured, and validated from postmortem human cornea rims (n = 10 donors) acquired from the Eye Clinic at Augusta University Medical Center and the Eye Physicians and Surgeons of Augusta–The Eye Guys after corneal transplantation according to the established protocol in the consensus recommendations for TM cell isolation, characterization, and culture.⁷² The study was approved by the Augusta University Institutional Review Board, and the human tissue experiments complied with the guidelines of the ARVO Best Practices for Using Human Eye Tissue in Research (November 2021). Cells were obtained from male and female donors without glaucoma from different ages (Supplementary Table S4). When sex information was not available, cells were genotyped for SRY, a gene present only on the Y chromosome, using primers F-TCGAACTCTGGCACCTTTCA and R-TTCATGGGTCGCTTCACTCT (Supplementary Table S5). The presence of a 243-bp product indicated a male donor, and the absence indicated a female donor. After growing in regular TM media (i.e., low glucose [1 g/L] Dulbecco's modified Eagle's medium [DMEM], 10%–20% fetal bovine serum [FBS], 1% penicillin/streptomycin, or 1% antibiotic/antimycotic), HTM cells (second to fourth passages) were cultured for >24 hours in estrogen-free media (i.e., phenol red–free DMEM) with 10% charcoal-stripped FBS before the treatment with 50 µM E2 (Cayman Chemical Company, Ann Arbor, MI) and/or 10 ng/mL TGFβ2 (Abcam, Cambridge, UK) for 24 hours. The following groups were included: media-only control, DMSO vehicle control, 50 µM E2 only, 10 ng/mL TGFβ2 only, and 50 µM E2 with 10 ng/mL TGFβ2. For each cell strain (n = 9–10), treatments were performed simultaneously on cells cultured on standard plastic 6-well culture plates (stiffness: ∼1000 kPa) or on collagen I–coated BioFlex 6-well culture plates (stiffness: 930 kPa) subjected to CMS (15%, 1 cycle/s, 24 hours) on a Flexcell FX-6000T Tension System (Flexcell International Corporation, Burlington, NC, USA).

Quantitative Real-Time PCR

After 24 hours of treatment, total RNA was extracted using a miRNeasy Mini kit (Qiagen, Hilden, Germany) for a quantitative real-time PCR (qRT-PCR). An Agilent 2100 Bioanalyzer and a High Sensitivity RNA 2000 Pico kit (Agilent Technologies, Santa Clara, CA, USA) were used to assess RNA quality and concentration, requiring a RNA Integrity Number (RIN) ≥5.8 for inclusion. RNA (100 ng) was converted to cDNA using a High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Waltham, MA, USA). Transcriptional differences of a panel of 17 TGFβ2-responsive genes (Table 2) were assessed by qRT-PCR using a StepOnePlus System (Applied Biosystems, Waltham, MA, USA) with a 10-µL reaction using iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories, Hercules, CA, USA) and 10 µM primers (Supplementary Table S5).

Table 2.

TGFβ2/Estradiol Treatment and Alteration of the Expression of Known TGFβ2-Responsive Genes

	P Value
Gene	Full Model	Static	Stretch	Male	Female
ACTA2	—	1.0E-03	4.5E-04	—	3.4E-02
BMP1	2.2E-07	4.1E-09	1.4E-10	1.7E-02	9.2E-05
CCN2	2.2E-02	1.9E-07	7.6E-06	—	—
CD44	—	1.2E-04	3.6E-03	—	—
COL1A1	—	1.9E-03	—	—	—
FST	4.9E-02	1.0E-02	5.2E-03	4.6E-02	—
GREM1	1.2E-10	9.4E-06	2.5E-09	1.6E-04	1.9E-06
LAMC1	1.4E-02	4.5E-02	4.4E-07	—	—
LOX	—	—	—	—	—
NFATC1	3.7E-04	8.6E-06	2.1E-06	2.0E-02	4.4E-02
SERPINE1	2.1E-06	7.5E-11	7.9E-06	3.0E-04	1.3E-03
SMAD2	2.2E-03	5.9E-04	7.5E-05	3.8E-02	—
SMAD3	1.0E-11	2.9E-11	3.1E-09	1.1E-02	2.4E-12
TGM2	—	2.0E-06	—	—	—
THBS1	—	3.6E-02	4.3E-05	—	1.8E-05
TNC	—	4.4E-02	—	—	—
VCAN	6.4E-03	—	8.3E-05	—	3.1E-05

A panel of TGFβ2-responsive genes was assayed for the significant impacts of treatment. Data were analyzed by a linear mixed-effects model using treatment, stretch, and/or sex as cofactors. The full model included all samples and all factors. Data were also stratified to examine the impacts of treatment under static and stretch conditions separately, as well as in female and male cells. P ≤ 0.05 was considered significant.

Statistical Analysis

For all IOP data, a mixed-effects analysis, an alternative to two-way repeated-measures ANOVA, was performed using either genotype × age or sex × age as factors with Tukey multiple comparison correction. A mixed-effects analysis, an alternative to two-way repeated-measures ANOVA, was used for all OCT data using genotype × age as factors with Tukey multiple comparison correction. For qRT-PCR, the transcriptional effects of treatment, stretch, and/or sex of the cell donor were tested using a linear mixed-effects model based on gene C_t relative to ACTB C_t, with treatment, stretch, and sex as covariates as specified in R.4.4.1 with the lme4 package. For post hoc testing, the ΔΔC_t method was used to calculate fold change (2^–ΔΔCt), normalized to ACTB. A repeated-measures ANOVA was employed on 2^ΔΔCt values using Tukey multiple comparison correction. All qRT-PCR comparisons were made to media-only controls. In all cases, differences were considered significant if P ≤ 0.05. Statistical analysis was performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA), version 10.1.1.

Results

Intraocular Pressure Elevation in Female Esr1^−/− Mice

IOP was measured in female and male Esr1^−/− mice from 3 to 12 months of age (Supplementary Table S2). Female Esr1^−/− IOP trended higher than that of their wild-type littermates throughout their lifetime, with a larger separation between the two becoming more evident with advancing age (Fig. 1A). This difference was significant at 6 and 11 months of age (6-month-old Esr1^+/+ 10.4 mm Hg, n = 21 vs. Esr1^−/− 11.2 mm Hg, n = 5, P = 0.028, 95% confidence interval [CI], −1.55 to −0.09; 11-month-old Esr1^+/+ 9.9 mm Hg, n = 14 vs. Esr1^−/− 11.7 mm Hg, n = 3, P = 0.017, 95% CI, −3.19 to −0.43). Assessment of the iridocorneal angle did not show genotype-biased angle closure to obstruct the outflow (Fig. 1C). The IOP of female heterozygous Esr1^+/− mice was similar to that of their wild-type littermates (Fig. 1A). Age was a significant factor (P < 0.0001), with female wild-type and Esr1^+/− IOP decreasing with age. These time-dependent decreases were not observed for female Esr1^−/−. Age was also a significant factor (P = 0.00020) for male mice, with wild-type males having significantly different IOP at 5 versus 12 months of age (11.4 mm Hg, n = 11 vs. 10.0 mm Hg, n = 11 P = 0.046, 95% CI, 0.03–2.70). However, no significant genotype-based difference (P > 0.05) was noted between male mice at any time point (n = 3–28) (Fig. 1B).

Figure 1.

Intraocular pressure was influenced by Esr1 genotype and sex. (A, B) IOP was assessed in female and male mice from 3 to 12 months of age. IOP was significantly different between female wild-type and Esr1^−/− mice at 6 and 11 months of age. (A) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, n = 3–28 group size/month; 6 months, *P ≤ 0.05, n = 21 and 5, 95% CI, −1.55 to −0.09; 11 months, *P ≤ 0.05, n = 14 and 3, 95% CI, −3.19 to −0.43. (B) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, P > 0.05, n = 2–28 group size/month (sample size was insufficient for statistical analysis for 12-month-old Esr1^−/−). (C) The openness of the angle of female Esr1^+/+ and Esr1^−/− was assessed by hematoxylin and eosin (H&E) histology and OCT. Whole-globe H&E images were from paraffin-embedded eyes (n = 5); angle H&E images were from JB-4–embedded eyes (n = 8), and all eyes (n = 30–38) were examined by OCT. AS, anterior segment; CB, ciliary body; CC, collector channel; ONH, optic nerve head; SC, Schlemm's canal. (D–F) There were significant sex differences in IOP in wild-type mice at 6 months and in Esr1^+/− mice at 7 and 8 months but not in Esr1^−/− mice. (D) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, n = 12–28 group size/month; 6 months, *P ≤ 0.05, n = 21 and 27, 95% CI, 0.03–1.28. (E) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, n = 3–20 group size/month; 7 months, *P ≤ 0.05, n = 11 and 18, 95% CI, 0.17–1.65; 8 months, *P ≤ 0.05, n = 11, 95% CI, 0.069–1.57. (F) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, P > 0.05, n = 2–17 group size/month (sample size was insufficient for statistical analysis for 12-month-old Esr1^−/−). (G, H) CCT was also assessed in female and male mice. (G) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, P > 0.05, n = 3–25 group size/month. (H) Mixed-effects analysis as an alternative to two-way repeated-measures ANOVA, n = 5–26 group size/month; 6 months, *P ≤ 0.05, n = 24 and 12, 95% CI, −20.94 to −1.02. All data in Figure 1 are mean ± SEM. Only significant comparisons to wild-type or between sexes are shown.

For wild-type mice, the IOP of male mice trended higher than that of females, with male IOP becoming significantly higher than female IOP at 6 months of age (11.0 mm Hg, n = 27 vs. 10.4 mm Hg, n = 21, P = 0.041, 95% CI, 0.03–1.28) (Fig. 1D). Age was again a significant factor (P < 0.0001). Similarly, Esr1^+/− male IOP trended higher than that of female mice with significant differences at 7 and 8 months of age (7-month-old male 11.0 mm Hg, n = 18 vs. female 10.1 mm Hg, n = 11, P = 0.018, 95% CI, 0.17–1.65; 8-month-old male 11.2 mm Hg, n = 11 vs. female 10.4 mm Hg, n = 11, P = 0.034, 95% CI, 0.07–1.57) (Fig. 1E). Age was a significant factor (P < 0.0001), with the interaction between sex and age also being significant (P = 0.0040). Unlike wild-type and Esr1^+/− mice, female Esr1^−/− IOP trended higher than Esr1^−/− males after 8 months of age, although this difference never reached significance (Fig. 1F).

Because IOP measurements may be affected by CCT,^{3,4,6,73–76} CCT was examined before and after the ages at which significant IOP differences were observed between female wild-type and Esr1^−/− mice. Multiple comparison testing did not show significant differences (P > 0.05) between different genotypes of female mice (Fig. 1G, Supplementary Table S3). Interestingly, male wild-type mice had a thinner CCT than male Esr1^−/− mice at 6 months of age (Esr1^+/+ 87.6 ± 2.9 µm, n = 24; Esr1^−/− 98.5 ± 2.9 µm, n = 12; P = 0.028, 95% CI, −20.94 to −1.02) (Fig. 1H, Supplementary Table S3). However, this difference was not observed at other ages. Six-month-old wild-type males also had significantly thinner CCT than 9-month-old wild-type males (6 months 87.6 ± 2.9 µm, n = 24; 9 months 94.9 ± 2.6 µm, n = 19; P = 0.005, 95% CI, −18.33 to −3.67), suggesting that the significant CCT difference between wild-type and Esr1^−/− males may be due to outliers in the 6-month-old wild-type male cohort.

Transcriptional Response of Human Trabecular Meshwork Cells to In Vitro Estradiol and TGFβ2 Treatment

HTM cells (n = 10, 6 females and 4 males) were treated under both static and mechanical stretch conditions for 24 hours in the following groups: media-only control, DMSO vehicle control, 10 ng/mL TGFβ2 only, 50 µM E2 only, and 10 ng/mL TGFβ2 with 50 µM E2. qRT-PCR was used to examine whether E2 could mitigate the effects of TGFβ2 on a panel of 17 TGFβ2-responsive genes (Table 2).

Using a linear mixed-effects model with treatment, stretch conditions, and cell donor sex as covariates (i.e., full model), 10 genes (i.e., BMP1, CCN2, FST, GREM1, LAMC1, NFATC1, SERPINE1, SMAD2, SMAD3, and VCAN) were significantly altered (P ≤ 0.05) by one or more treatments (Table 2). Data were also stratified by static and stretch conditions as well as by donor sex to analyze the effects of treatment within each of those subpopulations separately. Fifteen genes (i.e., ACTA2, BMP1, CCN2, CD44, COL1A1, FST, GREM1, LAMC1, NFATC1, SERPINE1, SMAD2, SMAD3, TGM2, THBS1, and TNC) and 13 genes (i.e., ACTA2, BMP1, CCN2, CD44, FST, GREM1, LAMC1, NFATC1, SERPINE1, SMAD2, SMAD3, THBS1, and VCAN) were significantly altered (P ≤ 0.05) by treatment under static and stretch conditions, respectively (Table 2). Treatment significantly altered (P ≤ 0.05) eight genes (i.e., ACTA2, BMP1, GREM1, NFATC1, SERPINE1, SMAD3, THBS1, and VCAN) in female TM cells and seven genes (i.e., BMP1, FST, GREM1, NFATC1, SERPINE1, SMAD2, and SMAD3) in male cells (Table 2).

To distinguish the effects of the different treatments on the 10 genes identified by the full model, the expression data were analyzed by repeated-measures ANOVA, analyzing static and stretch conditions separately. Under static conditions, E2 significantly upregulated the expression of eight of the genes identified by the full model (i.e., BMP1, P = 0.0091; CCN2, P = 0.0030; FST, P = 0.044; GREM1, P = 0.0013; LAMC1, P = 0.0056; SERPINE1, P < 0.0001; SMAD2, P = 0.013; and SMAD3, P = 0.010) while TGFβ2 significantly downregulated three of the genes (i.e., NFATC1, P = 0.0005; SMAD2, P = 0.028; and SMAD3, P < 0.0001) (Fig. 2). The addition of E2 with TGFβ2 returned NFATC1 and SMAD2 to nonsignificance (P > 0.05) (Fig. 2, Fig. 3). However, SMAD3 and CCN2 remained significantly downregulated (P < 0.0001) and upregulated (P = 0.026) relative to the media-only control, respectively.

Figure 2.

Transcriptional response of human trabecular meshwork cells to in vitro estradiol and TGFβ2 treatment. The expression differences of the 10 genes identified by a linear mixed-effects model as being significantly impacted by treatment relative to media-only control were examined by repeated-measures ANOVA, n = 9−10, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. E2, estradiol; FC, fold-change.

Figure 3.

Estradiol ameliorated TGFβ2 effects on human trabecular meshwork transcription under static conditions. Under static conditions, the addition of E2 with TGFβ2 ameliorated the effects of TGFβ2 on NFATC1 and SMAD2. Repeated-measures ANOVA, n = 9−10, *P < 0.05, ***P < 0.001.

Under stretch conditions, E2 significantly upregulated BMP1 (P = 0.030) (Fig. 2) while TGFβ2 significantly upregulated SERPINE1 (P = 0.037) and downregulated seven genes (i.e., BMP1, P = 0.0019; FST, P = 0.014; GREM1, P = 0.00020; LAMC1, P = 0.030; NFATC1, P = 0.018; SMAD2, P = 0.0067; and SMAD3, P < 0.0001). The addition of E2 with TGFβ2 returned FST, LAMC1, NFATC1, and SMAD2 to nonsignificance (P > 0.05) (Fig. 2, Fig. 4). Furthermore, the direction of expression change for BMP1 was reversed from downregulation in the TGFβ2-only treatment (P = 0.0019) to upregulation with the addition of E2 (P = 0.046) (Fig. 2, Fig. 4). However, GREM1, SMAD3, and SERPINE1 remained significant (P = 0.017, 0.026, and 0.030, respectively) although the addition of E2 with TGFβ2 decreased the degree of significance for GREM1 and SMAD3.

Figure 4.

Estradiol ameliorated TGFβ2 effects on human trabecular meshwork transcription under stretch conditions. Under stretch conditions, the addition of E2 with TGFβ2 reversed the effects of TGFβ2 on BMP1 and ameliorated its effects on FST, LAMC1, NFATC1, and SMAD2. Repeated-measures ANOVA, n = 9−10, *P < 0.05, **P < 0.01.

Interacting Effects of Mechanical Stretch and TGFβ2/Estradiol in Human Trabecular Meshwork Cells

The effects of stretch/substrate on the 17 TGFβ2-responsive genes were analyzed using a linear mixed-effects model with treatment, stretch, and donor sex as covariates. From this analysis, nine genes (i.e., BMP1, CCN2, CD44, FST, GREM1, LAMC1, NFATC1, SMAD2, and SMAD3) were significantly impacted (P ≤ 0.05) by stretch/substrate (Table 3). Data were also stratified by sex to determine the effects of stretch on male and female cells separately. Stretch/substrate significantly impacted (P ≤ 0.05) 12 genes (i.e., BMP1, CCN2, CD44, COL1A1, FST, GREM1, LAMC1, NFATC1, SMAD2, SMAD3, THBS1, and VCAN) in male cells and 11 of the same genes in female cells (i.e., BMP1, CCN2, CD44, COL1A1, FST, GREM1, LAMC1, NFATC1, SMAD2, SMAD3, and THBS1) (Table 3).

Table 3.

Mechanical Stretch and Expression of Known TGFβ2-Responsive Genes

	P Value
Gene	Full Model	Male	Female
BMP1	1.1E-15	1.0E-05	9.6E-11
CCN2	9.3E-07	1.1E-03	2.1E-04
CD44	2.5E-10	7.3E-06	2.1E-05
COL1A1	—	1.1E-02	1.8E-04
FST	9.05E-07	5.5E-08	2.6E-02
GREM1	8.8E-12	3.7E-05	3.4E-08
LAMC1	7.0E-05	7.7E-03	4.5E-03
NFATC1	6.3E-11	4.1E-03	1.5E-08
SMAD2	2.5E-06	7.2E-06	3.6E-02
SMAD3	3.0E-12	3.1E-05	1.1E-09
THBS1	—	2.9E-03	3.5E-13
VCAN	—	3.7E-03	—

A panel of TGFβ2-responsive genes was assayed for the significant impacts of stretch/substrate. Data were analyzed by a linear mixed-effects model using treatment, stretch, and/or sex as cofactors. The full model included all samples and all factors. Data were also stratified to examine the impacts of stretch/substrate in female and male cells separately. P ≤ 0.05 was considered significant.

Furthermore, examining the effects of treatment separately under static and stretch conditions showed that the treatment with more significant effects (P ≤ 0.05) changed in response to the mechanical environment. Among the 17 genes of interest, more genes were significantly (P ≤ 0.05) altered by E2 under static conditions and by TGFβ2 under stretch conditions (Fig. 2, Fig. 5A).

Figure 5.

Mechanical stretch, but not sex, influenced transcriptional effects of TGFβ2 and estradiol. (A) Genes from the 17-gene panel significantly impacted by either TGFβ2 or E2 treatment under static and stretch conditions are shown. (B) Genes from the 17-gene panel significantly impacted by treatment in either male or female HTM cells are shown.

Sex-Biased Transcriptional Response of Known TGFβ2-Responsive Genes

The HTM cells used for treatment had been derived from six female and four male donors (Supplementary Table S4). The effects of sex on the 17 TGFβ2-responsive genes were analyzed using a linear mixed-effects model with treatment, stretch, and donor sex as covariates. Only two genes (i.e., CD44, P = 0.039; SMAD3, P = 0.038) were significantly impacted by sex differences under the full model (Table 4). Data were also stratified to examine the effects of sex under static and stretch conditions separately. Only SMAD2 (P = 0.0029) was significantly affected by sex differences under static conditions, while three genes (i.e., CD44, P = 0.037; SMAD3, P = 0.045; THBS1, P = 0.048) were significantly affected by sex under stretch conditions. Furthermore, stretch impacted almost all the same genes in male and female cells (Table 4). Moreover, approximately the same number of genes were affected by treatment in male and female HTM cells (Table 4; Fig. 5B).

Table 4.

Sex and the Expression of Known TGFβ2-Responsive Genes

	P Value
Gene	Full Model	Static	Stretch
CD44	3.9E-02	—	3.7E-02
SMAD2	—	2.9E-03	—
SMAD3	3.8E-02	—	4.5E-02
THBS1	—	—	4.8E-02

A panel of TGFβ2-responsive genes was assayed for the significant impacts of the sex of the cell donor. Data were analyzed by a linear mixed-effects model using treatment, stretch, and/or sex as cofactors. The full model included all samples and all factors. Data were also stratified to examine the impacts of sex differences under static and stretch conditions separately. P ≤ 0.05 was considered significant.

Transcriptional Response of TGFβ Receptors to Estradiol Treatment

TGFBR1 expression was not significantly changed by E2 or TGFβ2 (Fig. 6). However, E2 significantly upregulated TGFBR2 (P = 0.017) under static conditions. On the other hand, TGFβ2 significantly downregulated (P < 0.0001) TGFBR2 under stretch conditions. The addition of E2 ameliorated this downregulation. Although TGFβ2 did not significantly alter the expression of TGFBR3 under static conditions, TGFβ2 significantly upregulated (P < 0.0001) TGFBR3 under stretch conditions. The addition of E2 did not mitigate this upregulation.

Figure 6.

TGFβ2/estradiol treatment affected the expression of TGFβ receptors. TGFBR1 was not significantly affected by E2, TGFβ2, or combined treatment under static or stretch conditions. TGFBR2 was significantly upregulated by E2 under static conditions and downregulated by TGFβ2 under stretch conditions. TGFBR3 was significantly downregulated by TGFβ2 under stretch conditions, even when combined with E2. Repeated-measures ANOVA, n = 9−10, *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001.

Discussion

Previously, we showed that hormone signaling–related genes are responsive to CMS in TM cells¹⁶ and that ESR1 is a key upstream regulator of IOP-associated genes.¹⁷ In the current study, we show that loss of Esr1 elevates IOP in female mice at later ages. Furthermore, E2 may counteract the transcriptional effects of TGFβ2 in HTM cells in a CMS-influenced manner.

Effects of Estrogen Signaling Loss on Glaucoma-Related Ocular Parameters

Female Esr1^−/− mice had significantly higher IOP by ∼1 mm Hg than their wild-type littermates at 6 and 11 months of age, confirming the role of estrogen signaling in IOP regulation.^20,21,25 While differences were not significant at all ages, Esr1^−/− IOP levels consistently trended higher than those of wild-type. Although 1 mm Hg is not large, POAG risks can increase by 15% with as little as a 1-mm Hg increase in baseline human IOP.^37,77,78 The observed significant but modest IOP elevation due to Esr1 loss is consistent with the IOP changes (1.5–3 mm Hg) from the aromatase knockout mice and human-based studies during pregnancy, menopause, and estrogen supplementation.^41,79–82 Furthermore, POAG is a lifelong disease influenced by genetic and environmental factors, whose effects on the disease process are magnified with advanced aging.^83,84 IOP elevation did not become significant until later ages, supporting the accumulative effect of chronic estrogen deprivation on IOP and POAG.⁸² Furthermore, only one estrogen receptor was examined. Potentially, the signaling pathways of the other estrogen receptors may also affect IOP.⁷⁹

Wild-type and Esr1^−/− male IOP were not significantly different at any age, perhaps due to intrinsically low estrogen signaling levels in males. Because estrogen levels differ between males and females, and several studies have reported that males have a higher risk of glaucoma than females until menopause,^85–88 the IOPs of male and female mice were compared. The IOPs of wild-type and Esr1^+/− males were significantly higher than those of females by ∼1 mm Hg at 6 and 7 to 8 months, respectively, with male IOP consistently trending higher even when not significant. This phenomenon was not observed in Esr1^−/− mice, with female IOP trending higher than male Esr1^−/− mice for most of their lifetime. These data suggest a role for ESR1 signaling in female IOP regulation that may explain previously observed sex-biased glaucoma risk.^85–88

Although CCT may affect IOP measurements, and estrogen-containing hormone replacement therapy affects CCT,^26,76 the IOP increase observed in female Esr1^−/− mice did not appear to be mediated by estrogen-induced CCT changes. Therefore, the observed IOP elevation was not artifactual under the assumptions of rebound tonometry but may be due to the loss of estrogen signaling in AH outflow. Potential impacts on AH outflow require further study.

Mitigation of TGFβ2 Transcriptional Effects in Human Trabecular Meshwork Cells by Estradiol Treatment

Because female Esr1^−/− IOP was higher than that of wild-type females, we hypothesized that E2 plays a role in TM regulation of IOP homeostasis. We found that treating TM cells with E2 ameliorates TGFβ2-induced effects on genes previously shown to be TGFβ2-responsive. Several are involved in POAG-associated processes, such as ECM and cell adhesions,^89–92 suggesting that E2 plays a protective role in regulating these IOP- and POAG-related genes.

Interestingly, the TGFβ2-responsive genes in our qRT-PCR panel showed more pronounced TGFβ2 effects under CMS, suggesting that high IOP-induced increases in CMS may exacerbate TGFβ2’s impact through a positive feedback loop. In previous studies, CMS of TM cells has affected TGFβ signaling,⁹³ TGFβ-relevant microRNAs, and the expression of long noncoding RNAs with sequence similarity to TGFβ receptor 2 and TGFβ receptor-associated protein 1.¹⁶ CMS induces HTM expression of another TGFβ family member, TGFβ1, leading to subsequent changes in ECM gene expression.⁹⁴ Furthermore, CMS upregulated IL-6 expression in a TGFβ1-dependent manner.⁹⁵ Together, this suggests an interaction between TGFβ signaling and mechanical stretch, whereby mechanical stretch could increase the effects of TGFβ signaling by increasing TGFβ1 expression. Moreover, IL-6, which is upregulated in TM cells by both TGFβ1 and CMS, can increase TGFβ1 promoter activity,⁹⁶ suggesting a positive feedback mechanism for TGFβ signaling and CMS. Importantly, those findings pertained to TGFβ1, not TGFβ2, and the crosstalk between TGFβ2 and CMS in TM cells warrants further study. However, TGFβ2 and CMS affect many of the same genes and proteins,^97–102 potentially allowing for a double-dose effect.

While CMS alone could interact with TGFβ2 signaling, it is worth noting that our static treatments were conducted on plates with hard plastic bottoms (∼1000 kPa), whereas stretch treatments were conducted on plates with stretchable membrane bottoms (930 kPa). Substrate composition and stiffness are known to affect TM cell expression.^103–107 Interestingly, substrate stiffness and TGFβ2 signaling interact with softer substrates, reducing some effects of TGFβ2 and amplifying others.^108,109 As a result, it is possible that the differences in culturing substrates could account for a portion of TGFβ2 effects being exacerbated under stretch conditions.

Surprisingly, response to treatment varied little between sexes, although we did not probe for individual treatments; rather, we examined the effects of treatment overall. The effects of mechanical stretch also had little difference between the sexes, and very few genes were significantly impacted by sex differences. This was surprising because previous studies have shown that higher titers of TGFβ2-loaded adenovirus 5 are required to induce sustained OHT in males than in females, suggesting that TM responsiveness to TGFβ2 differs between the sexes.^110,111 However, Sugali et al.¹¹¹ made this TGFβ2-induced OHT comparison between 3- and 5-month-old male mice and previously published data on 5- to 8-month-old mice.¹¹⁰ This is especially important because Sugali et al.¹¹¹ observed that males had a larger IOP response to Ad5-TGFβ2 at 5 months than at 3 months, suggesting that age also affects the TGFβ2 response. Age was also an important factor for our mouse studies. Therefore, the apparent lack of sex differences in HTM cell response to treatment and mechanical stretch may be influenced by the older age of our female cell donors. Of our six female donors, age information was available for only four with ages of 18, 44, 54, and 56 years (Supplementary Table S4). Two of these were older than the average age at menopause in the United States (i.e., 52 years).¹¹² Hormone changes associated with menopause could cause these cells to respond more similarly to male HTM, thereby confounding analysis of sex differences. However, no information on menopausal status, estrogen levels, or HRT use was available for the donors in this study, and due to the low sample size of cells with donor age information, we could not explore age as a factor.

Limitations

Although our study identifies a role for estrogen signaling in IOP regulation, it also has some limitations. First, the cell study had some limitations: cells were from nonglaucomatous donors, with incomplete demographic data in some cases, and most cells were from Caucasian donors, limiting the applicability of these results to other ethnicities with a higher risk and differing genetic basis of POAG.^4,113 Second, our cell study was limited by a single, relatively short treatment time and higher E2 concentrations than physiological levels in the AH.^17,114 Third, our cell culture experiments focused primarily on transcriptional studies with little information on translational or posttranslational changes. Although TGFβ2 has a range of transcriptional effects, several of the genes we examined in our TGFβ2-responsive gene panel are known to change their protein levels in response to TGFβ2. This could help explain why some of the genes in the panel did not show significant responses to TGFβ2 treatment. Fourth, the stretched TM cells were cultured on plates with stretchable membrane bottoms, whereas the statically treated TM cells were cultured on plates with hard plastic bottoms. Despite their similar stiffnesses (930 vs. 1000 kPa), substrate composition and stiffness could affect TM cell biology.^103–107 Fifth, our Esr1^−/− mouse model is not a tissue-specific knockout. Sixth, AH inflow and outflow dynamics with or without estrogen stimulation could be examined in the future. Despite these limitations, we have presented several lines of evidence suggesting the role of Esr1 signaling in IOP regulation and E2’s anti-TGFβ2 effect on TM cells.

Conclusions

For the first time, we have identified that loss of Esr1 elevates IOP in female mice and that 17β-estradiol ameliorates the transcriptional effects of TGFβ2 in HTM cells. Although the detailed mechanism remains to be elucidated, these results suggest that estrogen signaling plays a role in preserving IOP homeostasis by exerting antifibrotic effects on the TM. Importantly, because of its apparent antifibrotic activity in the TM, estrogen signaling could serve as a therapeutic target for POAG in both males and females.