Prostaglandin E₂ and F_2α alter expression of select cholesteryl esters and triacylglycerols produced by human meibomian gland epithelial cells

Abstract

Purpose:

PGF_2α analogs are commonly used to treat glaucoma and are associated with higher rates of meibomian gland dysfunction (MGD). The purpose of this study was to evaluate the physiological effects of PGF_2α and PGE₂ on immortalized human meibomian gland epithelial cells (HMGECs).

Methods:

HMGECs were immunostained for the four PGE₂ receptors (EP1, EP2, EP3, EP4) and the one PGF_2α receptor (FP) and imaged. Rosiglitazone-differentiated HMGECs were exposed to PGF_2α and PGE₂ (10⁻⁹ to 10⁻⁶ M) for three hours. Cell viability was assessed by an ATP-based luminescent assay, and lipid extracts were analyzed for cholesteryl esters (CEs), wax esters (WEs), and triacylglycerols (TAGs) by ESI-MSMS^ALL in positive ion mode by a Triple TOF 5600 Mass Spectrometer using SCIEX LipidView 1.3.

Results:

HMGECs express three PGE₂ receptors (EP1, EP2, EP4) and the one PGF_2α receptor (FP). Neither PGE₂ nor PGF_2α showed signs of cytotoxicity at any of the concentrations tested. WEs were not detected from any of the samples, but CEs and TAGs both exhibited a diverse and dynamic profile. PGE₂ suppressed select CEs (CE 22:1, CE 26:0, CE 28:1, CE 30:1). PGF_2α dose dependently increased several CEs (CE 20:2, CE 20:1, CE 22:1, CE 24:0) yet decreased others. Both prostaglandins led to nonspecific TAG remodeling.

Conclusion:

PGE₂ and PGF_2α have minimal effect on HMGEC viability. PGF_2α influences lipid expression greater than PGE₂ and may do so by interfering with meibocyte differentiation. This work may provide insight into the mechanism of MGD development in glaucoma patients treated with PGF_2α analogs.

Introduction

There are greater than 64.3 million people between the ages of 40 and 80 who suffer from glaucoma worldwide.¹ Up to 80 percent, more than 50 million of these patients, may also have concurrent meibomian gland dysfunction (MGD).² This number is even more alarming when considering that ocular surface diseases, such as MGD, may exacerbate glaucoma by interfering with intraocular pressure (IOP) reduction and/or surgical outcomes.^3,4 Aggressive management of glaucoma-associated MGD with both oral and topical medications has shown to improve not only the signs and symptoms of MGD, but also IOP management.³ The mechanism that underlies the onset of MGD in glaucoma patients is not fully understood, but there is strong evidence that the medical management of glaucoma is culpable.^2,3,5–13

Prostaglandin analogs (PGAs), which are primarily PGF_2α analogs, are a common first-line treatment in glaucoma patients, owing to their efficacy, affordability, convenient dosing schedule, and favorable systemic side effect profile.¹⁴ Their therapeutic benefits are thought to be mediated through the PGF_2α FP receptor, a seven transmembrane G-protein coupled receptor.^15,16 Upon ligand binding, the FP receptor induces a calcium-dependent chloride current, as well as accumulation of inositol triphosphate.^17,18 Signaling through FP receptors promotes upregulation of several matrix metalloproteinases capable of degrading the extracellular matrix.¹⁹ This mechanism is largely credited for the FP receptor’s favorable effect on IOP: a degraded extracellular matrix reduces aqueous humor outflow resistance, permits increased outflow, and ultimately lowers the IOP.^15,19

Unlike the minimal adverse effects observed systemically, the ocular side effects of PGAs are numerous and may be partly mediated through crossover binding to other prostaglandin receptors, such as those for PGE₂ (EP1, EP2, EP3, and EP4).¹⁴ The PGA side effects consist of redness, burning, itching, pigmentation changes, and eyelash growth, among several others.¹⁴ More recently, development and/or progression of MGD has also been associated with chronic PGA use.¹² The mechanism, though currently unknown, is likely related to pathologic changes induced by the topical application of both the PGA and its preservative system. One hypothesis is that the increase in MGD may be due to an alteration in the secretions from the meibomian gland secondary to stimulation by PGAs.²⁰

The meibomian glands produce a lipid-rich fluid, termed meibum, that spreads across the ocular surface to form the tear film lipid layer.²¹ Meibum is comprised mostly of nonpolar lipids with significant contributions provided by the hydrophobic wax esters (WEs, 48 percent) and cholesteryl esters (CEs, 40 percent).^22,23 Another nonpolar lipid class, triacylglycerols (TAGs), is of less abundance, ranging between 0.05 to 6 percent,^24–28 but preliminary work has suggested that several TAGs are upregulated in a preclinical model of MGD²⁹ that uses the same human meibomian gland epithelial cell line (HMGEC) as in this study. We have previously described the CE and TAG profiles from differentiated HMGECs;^30,31 WE expression from these cells, however, appears to be minimal or undetectable.^32–35 As the only human meibocyte cell line available, this HMGEC model was used to better understand the outcomes provoked by PGE₂ and PGF_2α on the differential expression of CEs, WEs, and TAGs. We hypothesize that HMGECs undergo a shift in their lipid expression following supplementation with PGE₂ and PGF_2α and that these lipidomic alterations are likely mediated through prostaglandin receptors found on HMGECs.

Materials Methods

Reagents Materials

Rosiglitazone, PGE₂, and PGF_2α were purchased from Cayman Chemical (Ann Arbor, MI). Stock solutions were made by dissolving each compound in sterile-filtered dimethyl sulfoxide (DMSO, Hybri-Max™, Sigma-Aldrich, St. Louis, MO) and stored under nitrogen at −20°C. Rosiglitazone, PGE₂, and PGF_2α were added fresh to media preparations just prior to each experiment. DMSO concentration was maintained at 0.5% in all samples. Falcon 8-well chambered cell culture slides were used for immunocytochemistry and purchased from ThermoFisher (Waltham, MA). Clear-bottom, white-walled 96-well plates were used for luminescent cell viability assays and were also purchased from ThermoFisher (Waltham, MA). Glass petri dishes were used for lipidomics experiments and were purchased from Sigma Aldrich (St. Louis, MO).

Immunocytochemistry

HMGECs were cultured on 8-well chambered slides and maintained in proliferating conditions consisting of KSFM with 5 ng/ml epidermal growth factor (EGF) and 50 μg/ml bovine pituitary extract. Media changes were performed every other day until 80 to 90% confluence when the culture media was changed to DMEM/F12 with 10 ng/ml EGF, 2% fetal bovine serum (FBS), and 50 μM rosiglitazone. After 24 hours, HMGECs were fixed in freshly prepared PBS containing 4% paraformaldehyde for 10 mins at room temperature, followed by four consecutive wash cycles. Fixed HMGECs were blocked with PBS containing 10% goat serum (ThermoFisher, Waltham, MA), 1% bovine serum albumin (BSA), and 0.3% Triton-X 100 for 30 mins on ice, then washed four times. Blocked HMGECs were incubated at room temperature for one hour in a humidified chamber with a rabbit polyclonal antibody against one of the following receptors: EP1, EP2, EP3, EP4, or FP. All primary antibodies were purchased from ThermoFisher and diluted 1:50 in PBS with 1% BSA and 0.3% Triton-X 100. Following four wash cycles, HMGECs were incubated with goat anti-rabbit IgG (ThermoFisher, Waltham, MA) conjugated to Alexa Fluor Plus 555 for one hour in the dark (1:500 dilution in PBS with 1% BSA and 0.3% Triton-X 100). Each slide was mounted with mounting media consisting of DAPI counterstain, 70% glycerol, 10% n-propyl gallate, and 20% PBS. Slides were imaged with a Zeiss Axioplan 2 Fluorescent Microscope (Jena, Germany). Images were captured with a 40x objective at a z-step of 0.4 μm. To optimize signal-to-noise ratios, laser intensities were set at 25% and 75% and exposure times at 250 ms and 1500 ms for the channels exciting DAPI and Alexa Fluor Plus 555, respectively. Image stacks were corrected for chromatic aberration and crosstalk and deconvolved with Huygens Professional Software v19.10 (Scientific Volume Imaging, Hilversum, North Holland, Netherlands). All wash cycles consisted of a five-minute period on a shaker at 50 rpm. All experiments consisted of two experimental replicates and two technical replicates.

Cell viability

HMGECs were grown to 80 to 90% confluence and then split at a density of 30,000 cells per well in 96-well plates. All cells were incubated in DMEM/F12 with 10 ng/ml EGF, 2% FBS, and 50 μM rosiglitazone for two days to induce differentiation. PGE₂- or PGF_2α-containing media (10⁻⁹, 10⁻⁸, 10⁻⁷, 10⁻⁶ M) were added to HMGECs and allowed to incubate for three hours. Following incubation, cell viability was assessed using the Cell Titer-Glo Luminescent Cell Viability Assay (Promega, Madison, WI) according to the manufacturer’s instructions. Luminescence was quantified using the Wallac Perkin-Elmer 1420–041 Victor2 Multiplate Multifluorescence Reader (Mt. Waverly, Victoria, Australia) over an acquisition time of one second. All experiments were performed with two experimental replicates and three technical replicates.

Lipid extraction and analysis by mass spectrometry

HMGECs were seeded at one million cells per 6-cm glass petri dish in DMEM/F12 with 10 ng/ml EGF, 2% FBS, and 50 μM rosiglitazone for two days to induce differentiation. Following incubation, HMGECs were exposed to PGE₂ or PGF_2α (10⁻⁸, 10⁻⁷, or 10⁻⁶ M) for three hours. Lipids were extracted using a recently described modification of the Folch technique optimized to efficiently extract lipids from cultured HMGECs.^32,36 Briefly, 3 ml of pre-chilled chloroform-methanol (2:1 v/v, −20°C) were added directly to the HMGECs, which were then scraped with a sterile stainless steel scraper. The suspension was then transferred to a glass vial where 0.75 ml of molecular biology-grade water containing ammonium acetate (50 mM) was added. The emulsion was shaken at 350 rpm for 20 mins on ice and centrifuged at 1600 xg for five mins to promote stratification. The lower nonpolar phase was withdrawn and stored at −80°C until analysis. All steps involving organic solvents were performed with glass, stainless steel, or polytetrafluoroethylene (PTFE). All experiments were performed with two experimental replicates and two technical replicates.

Untargeted lipidomics was performed on dried lipids using a SCIEX Triple TOF 5600 Mass Spectrometer (SCIEX, Framingham, MA) in positive ion mode via direct infusion. The direct infusion solvent was methanol-chloroform (2:1) with 5 mM ammonium acetate. Each sample was delivered to the source by isocratic flow at 7 μl/min using a 500 μl Hamilton Gas Tight Syringe (Reno, NV). Prior to and after each sample, the syringe was cleaned with two flushes each of 100% methanol, 100% acetonitrile, 10% isopropyl alcohol, and 100% direct infusion solvent. Calibration runs were performed in positive mode with the APCI Positive Calibration Solution (SCIEX, Framingham, MA). The analyte mass evaluation range was 200 to 1200 m/z. A high-resolution time-of-flight (TOF) scan was acquired initially for 250 ms, then a series of high-sensitivity product ion scans were acquired per one Dalton (1 m/z) mass starting at 200 m/z through 1200 m/z. The collision energy was fixed at 35 eV, curtain gas to 20.00, GS1 to 20.00, GS2 to 15.00, spray voltage to 5000 volts, and interface temperature to 400°C. The acquisition time per sample was six minutes.

The acquired mass spectrometry data were processed with SCIEX LipidView 1.3 software (Framingham, MA). Lipid identities were assigned by LipidView, which utilizes a database of known ion fragmentations. To confirm selected lipid identities, SCIEX PeakView 2.2 was used to further investigate fragments. The mass tolerance window was set to 5 mDa, and the peaks greater than a signal-to-noise ratio of three were considered for analysis. Identification of individual lipid species from LipidView assignments was based on mass accuracy (<5 ppm) and MS/MS spectra obtained from PeakView.

Data analysis

For mass spectrometry data, CEs, WEs, and TAGs were the predominant focus of this study. To be included in the analysis, a given lipid species had to be present in all replicates of all samples. Each lipid was normalized to the sum intensity per class and reported as percent of the overall class. For TAGs, only those that were present at a threshold concentration of 0.1% were included. All data for both cell viability and lipidomics were analyzed by one-way ANOVA with Tukey post-hoc testing (SPSS v26, Armonk, NY) when tests of normality (Kolmogorov-Smirnov) and homogeneity of variance (Levene’s Test) were satisfied. If the assumption of equal variance was violated, then Games-Howell post-hoc testing was used. If the assumption of normality was violated, then the non-parametric Kruskal Wallis test was used. A p-value of 0.05 was considered significant.

The labeling convention for CEs consists of two numbers, separated by a colon. The two numbers represent the numbers of carbons and double bonds, respectively, in the fatty acyl chain. For some species, a third number is present, which denotes that the CE has been oxidized with the specified number of oxygenations. Each parent TAG is identified similarly by the total numbers of carbons and double bonds in the three fatty acyl chains. Following each TAG, however, is a product ion, either a fatty acid (FA) or another TAG, which is labeled similarly. This notation represents that the LipidView software identified a neutral loss corresponding to that specific fatty acid or TAG. As an example, TAG 54:3 (FA 18:1) is a TAG that has 54 total carbons and three total double bonds among its fatty acyl chains, where one of those chains consists of 18 carbons and one double bond.

Results

Expression of FP- and EP-type receptors

HMGECs were immunostained against the four PGE₂ receptors (EP1, EP2, EP3, and EP4) and the one PGF_2α receptor (FP) and imaged by fluorescent microscopy. Positive signal (pseudocolored red) was observed for EP1, EP2, EP4, and FP receptors (Figure 1). There was no appreciable EP3 signal above background.

Figure 1:

Fluorescent microscopy images (40x) of HMGECs stained with primary antibodies against PGE₂ (EP1, EP2, EP3, or EP4) or PGF_2α (FP) receptors and counterstained with DAPI (blue) after 24 hours of culture in differentiation media containing DMEM/F12, 10 ng/ml EGF, 2% FBS, and 50 μM rosiglitazone. Positive signal for each primary antibody (pseudocolored red) was detected for EP1, EP2, EP4, and FP receptors.

PGE₂: prostaglandin E₂

PGF_2α: prostaglandin F_2α

Influence of prostaglandins on cell viability

To determine whether PGE₂ or PGF_2α affects cell viability of differentiated HMGECs, ATP was quantitated from HMGECs exposed to 10⁻⁹, 10⁻⁸, 10⁻⁷, 10⁻⁶ M PGE₂ or PGF_2α for three hours. As shown in Figure 2, none of the PGE₂ concentrations (p = 0.20) or PGF_2α concentrations (p = 0.82) reduced viability compared to the vehicle control. Further, there were no differences in viability between any of the PGE₂ and PGF_2α concentrations (p = 0.32). The cytotoxic detergent Triton-X 100 (1%) served as a positive control and strongly reduced viability.

Figure 2:

HMGECs were differentiated for two days (see Methods) prior to exposure to PGE₂ or PGF_2α for 3 hours. After incubation, cell viability was assessed with a luminescent ATP-based assay. There were no significant differences between the vehicle control and any of the PGE₂ or PGF_2α concentrations, suggesting that these prostaglandins at physiologic concentrations do not alter cell viability. Further there were no differences between PGE₂ or PGF_2α. Triton-X 100 (1%) was used as a positive control, which differed significantly from all other concentrations (p < 0.001). n = 6 per condition

HMGEC: human meibomian gland epithelial cell

PGE₂: prostaglandin E₂

PGF_2α: prostaglandin F_2α

Description of the CE and TAG profiles across all samples

This study focused on the expression of CEs, WEs, and TAGs. Although WEs were not detected among any of the samples in this study, CEs and TAGs were. There were 107 CEs detected across all samples; however, only 39 met the criteria for inclusion in the analysis (Figure 3A). The chain length varied from 11 carbons to 32 carbons, and the double-bond count varied from zero to five. Six CEs were found to be oxidized. The most abundant CE was CE 18:1. Very long-chain (20 ≤ carbon number [n_c] ≤ 25) and ultra long-chain CEs (n_c ≥ 26) were present from all conditions and comprised 26.2% and 6.9% of the overall CE pool, respectively. Of the 39 CEs, nine were saturated, 15 were monounsaturated, and 15 were polyunsaturated. Monounsaturated CEs were the most abundant (51.1%), followed by polyunsaturated (28.4%) and saturated (20.5%).

Figure 3:

(A) HMGECs expressed 39 unique CEs, including 6 oxidized CEs (oxCEs). CE 18:1 was the most abundant, followed by CE 16:0. The chain length varied from 11 carbons to 32 carbons. The double-bond count ranged from 0 to 5. CEs are labeled by carbon number and double-bond count, respectively. When a third number is present, it denotes an oxCE with the corresponding number of oxygenations. n = 28

(B-D) HMGECs expressed 145 TAGs that met the criteria for analysis (see Methods). The carbon count varied from 44 to 72 (B), and the double-bond count varied from 0 to 12 (C). LipidView 1.3 detected the neutral loss of sixteen unique fatty acyl chains (D) from the parent TAGs. FA 18:1 was the most frequently observed. n = 28

HMGEC: human meibomian gland epithelial cell

CE: cholesteryl ester

TAG: triacylglycerol

FA: fatty acid

There were 3,706 TAGs detected across all samples; however, only 145 met the criteria for inclusion in the analysis. The total carbon count from the three acyl chains, excluding the glycerol backbone, ranged from 44 to 72 with the majority (137/145, 94.5%) falling within the range of 46 to 56 (Figure 3B). All expressed TAGs consisted of an even carbon count (145/145, 100%). The number of double bonds in the acyl chains of the TAGs varied from zero to 12. Very few TAGs were fully saturated (7/145, 4.8%). The degree of unsaturation followed a bimodal distribution (Figure 3C) that was heavily weighted toward a lower degree of unsaturation. TAGs were primarily of lower unsaturation (106/145, 73.1%, one to five double bonds) or of higher unsaturation (22/145, 15.2%, eight to ten double bonds). Only five of 145 (3.4%) had a moderate degree of unsaturation (six or seven double bonds) or a very high degree of unsaturation (11 or 12 double bonds). The LipidView 1.3 software identified the neutral loss of 16 unique fatty acyl chains from the 145 TAGs (Figure 3D). Their individual carbon counts varied from 12 to 26 with double bonds ranging from zero to 4. Similar to the parent TAG molecules, all fatty acyl chains consisted exclusively of even numbers of carbons. The most frequently observed fatty acyl group was oleic acid (FA 18:1), which was present in 23 of 145 TAGs (15.9%). The second most common was palmitic acid (FA 16:0) and palmitoleic acid (FA 16:1), which were each present in 15 (10.3%) TAGs. Several TAGs (18/145, 12.4%) consisted of very long-chain fatty acids (at least 20 carbons).

Influence of prostaglandins on CE and TAG expression

To determine the effects of PGF_2α on CE expression, lipid extracts from differentiated HMGECs were exposed to 10⁻⁸, 10⁻⁷, 10⁻⁶ M PGF_2α for three hours. Of the 39 CEs analyzed, expression of seven CEs was significantly different among the concentrations (Figure 4A). CE 14:1 and CE 26:0 were suppressed relative to control at all tested concentrations of PGF_2α, though the 10⁻⁶ M concentration failed to reach significance for both CEs. One oxidized CE (CE 26:0;2) reached significance, showing a decrease in expression at higher PGF_2α concentrations. The remaining 4 CEs (CE 20:2, CE 20:1, CE 22:1, and CE 24:0) demonstrated a dose-dependent increase in expression among the three PGF_2α concentrations; however, the low-dose PGF_2α suppressed expression relative to the vehicle control.

Figure 4:

HMGECs were differentiated for two days (see Methods) prior to exposure to PGF_2α for three hours. Lipid extracts were analyzed by mass spectrometry. (A) Thirty-nine CEs met the criteria for analysis. Seven of the 39 varied significantly with PGF_2α supplementation (inset). All significant CEs were reduced relative to control with low-dose PGF_2α; however, four of these CEs demonstrated a dose-dependent relationship with PGF_2α. This low-dose suppression following by a dose-dependent upregulation suggests that multiple competing pathways may be activated (see Discussion). CEs are labeled by carbon number and double-bond count, respectively. When a third number is present, it denotes an oxCE with the corresponding number of oxygenations. (B) One hundred forty-five TAGs met the criteria for analysis. To aid legibility, only the 23 that varied significantly with PGF_2α are displayed. Four of the 23 failed to reach significance in pairwise comparisons. PGF_2α promoted generalized TAG remodeling, affecting 15.9% of all analyzed TAGs. TAGs are labeled by two numbers corresponding to the total number of carbons and the total number of double bonds, respectively. The fatty acid in parentheses represents one of the three fatty acids of the parent TAG molecule.

n = 4 per condition

* denotes significance

gray bar p ≤ 0.05, black bar p ≤ 0.01, dashed bar p ≤ 0.001

HMGEC: human meibomian gland epithelial cell

CE: cholesteryl ester

oxCE: oxidized cholesteryl ester

TAG: triacylglycerol

FA: fatty acid

Twenty-three of 145 TAGs showed statistically significant differences in expression among the different PGF_2α concentrations (Figure 4B). Four of the 23, however, failed to reach significance in pairwise comparisons: TAG 48:1 (FA 16:1), TAG 50:2 (FA 16:1), TAG 52:9 (TAG 35:2), and TAG 56:2 (FA 18:1). Six of the 23 were upregulated across all concentrations: TAG 48:2 (FA 18:1), TAG 54:2 (FA 18:2), TAG 54:3 (FA 20:2), TAG 56:4 (FA 20:2), TAG 48:2 (FA 14:0), and TAG 52:3 (FA 20:2). Two were decreased across all concentrations: TAG 50:9 (TAG 33:1) and TAG 54:10 (TAG 35:2). Seven were upregulated at low-dose PGF_2α but reduced at higher doses, including all of the FA 12:1-containing TAGs: TAG 46:2 (FA 12:1), TAG 46:3 (FA 12:1), TAG 48:2 (FA 12:1), TAG 48:3 (FA 12:1), TAG 48:6 (TAG 36:1), TAG 48:7 (TAG 36:1), and TAG 48:11 (TAG 36:2). The remaining four TAGs were reduced at low-dose PGF_2α but upregulated at higher doses: TAG 48:1 (FA 14:0), TAG 48:1 (FA 16:0), TAG 48:1 (FA 18:1), and TAG 50:0 (FA 18:0).

To determine the effects of PGE₂ on CE expression, lipid extracts from differentiated HMGECs exposed to 10⁻⁸, 10⁻⁷, or 10⁻⁶ M PGE₂ for three hours were analyzed by mass spectrometry. Of the 39 CEs that met the inclusion criteria, four showed a statistically significant change in expression (Figure 5A). A dose-dependent decrease was observed in CE 22:1 and CE 26:0. For CE 22:1, significance was achieved between the media control lacking PGE₂ and the 10⁻⁶ M concentration (p = 0.01). For CE 26:0, significance was achieved between the media control and both the 10⁻⁷ M 10⁻⁶ M concentrations (p = 0.01 for both). CE 28:1 and CE 30:1 also reached significance, where the 10⁻⁶ M (p = 0.02) and 10⁻⁷ M (p = 0.04) concentrations, respectively, were decreased relative to 10⁻⁸ M PGE₂.

Figure 5:

HMGECs were differentiated for two days (see Methods) prior to exposure to PGE₂ for three hours. Lipid extracts were analyzed by mass spectrometry. (A) Thirty-nine CEs met the criteria for analysis. Four of the 39 varied significantly with PGE₂ supplementation (inset), and all were reduced with increased PGE₂ concentrations. CEs are labeled by carbon number and double-bond count, respectively. When a third number is present, it denotes an oxCE with the corresponding number of oxygenations. (B) One hundred forty-five TAGs met the criteria for analysis. To aid legibility, only the nine that varied significantly with PGE₂ are displayed. Two of the 9 failed to reach significance in pairwise comparisons. PGE₂ promoted generalized TAG remodeling. TAGs are labeled by two numbers corresponding to the total number of carbons and the total number of double bonds, respectively. The fatty acid in parentheses represents one of the three fatty acids of the parent TAG molecule.

n = 4 per condition

* denotes significance

gray bar p ≤ 0.05, black bar p ≤ 0.01, dashed bar p ≤ 0.001

HMGEC: human meibomian gland epithelial cell

CE: cholesteryl ester

oxCE: oxidized cholesteryl ester

TAG: triacylglycerol

FA: fatty acid

Only nine of 145 analyzed TAGs (6.2%) showed statistically significant differences in expression among the different PGE₂ concentrations (Figure 5B). The remaining 136 (93.8%) did not vary significantly with respect to PGE₂. Two of the nine that were statistically significant, however, failed to reach significance in pairwise comparisons: TAG 48:3 (FA 12:1) and TAG 60:3 (FA 24:1). Two of the remaining seven TAGs showed a dose-dependent increase in response to PGE₂: TAG 52:12 (TAG 34:0) and TAG 54:10 (TAG 35:2). TAG 54:10 (FA 16:1) and TAG 50:3 (FA 18:1) were elevated across all PGE₂ concentrations. TAG 54:5 (FA 18:1) appeared to have reduced expression across all PGE₂ concentrations. The remaining two TAGs that reached significance had varying degrees of expression among the different concentrations of PGE₂: TAG 48:0 (FA 16:0) and TAG 54:10 (TAG 38:3).

Discussion

In light of growing awareness of the association between chronic PGA use in glaucoma and development of MGD,^2,12,20 we sought to identify the physiologic effects of PGE₂ and PGF_2α on HMGECs in culture. We report, for the first time, that HMGECs express three of the four PGE₂ receptors (EP1, EP2, and EP4) and the one PGF_2α receptor (FP). Across a broad range of physiologic concentrations (10⁻⁹ to 10⁻⁶ M), we found that neither PGE₂ nor PGF_2α altered cell viability. Though their effects were not cytotoxic, the observed changes in the HMGEC lipidome suggest that prostaglandins, particularly PGF_2α, affect lipid expression and/or metabolism. Much of this variation is consistent with general lipidomic remodeling, though specific patterns following PGF_2α exposure support the presence of multiple, competing pathways that appear to be concentration dependent. Taken together, these findings suggest that the meibomian glands express FP- and EP-type receptors, making them vulnerable to unwanted side effects associated with commercially available PGAs. Just one “dose” administered in this study, which was orders of magnitude less than commercially available formulations, modulated the lipid profile after only three hours. In vivo, lipidomic remodeling may translate into altered tear film lipid layer quality, accelerated tear evaporation, and/or poor tear film stability.

Expression of FP- and EP-type receptors

A notable advancement from this work is the discovery that HMGECs express FP receptors. This finding suggests that the meibomian gland could be added to the long list of other ocular and adnexal tissues known to express FP receptors: conjunctiva, cornea, sclera, iris, ciliary body, lens, retina, and optic nerve.^37–39 The FP receptor’s role in meibomian gland physiology is largely unknown, but in adipocytes and sebocytes, other cells optimized for high lipid turnover, signal transduction cascades that are initiated by FP binding ultimately inactivate PPARγ,⁴⁰ a known inducer of differentiation.^41,42 Recently, Kim et al have shown that PPARγ regulates differentiation in HMGECs,^43,44 and we have found that most of the CEs and TAGs produced by HMGECs are highly modulated in response to PPARγ agonism.^30,31 Therefore, FP expression in the meibomian gland may have a direct yet adverse influence on cellular differentiation and nonpolar lipid production. Because PGF_2α is capable of binding EP receptors at 10⁻⁸ to 10⁻⁶ M affinities,¹⁷ we were also interested in PGE₂ and the EP-type receptors. We found that HMGECs express EP1, EP2, and EP4 but not EP3 receptors, similar to the expression pattern reported in human sebocytes.⁴⁵ The effect of EP receptor engagement depends upon its exact EP subtype, the specific G protein it is coupled to, and its downstream signal transduction pathways,⁴⁶ a topic further discussed below.

Influence of prostaglandins on cell viability

Across all the concentrations tested in this study, we found that neither PGE₂ nor PGF_2α affected HMGEC viability. Although some reports have linked PGE₂ to apoptosis by promoting calcium influx, these responses are typically at much higher PGE₂ concentrations—approximately 50 times greater than the highest concentration used in this study.^47,48 At physiologic concentrations, PGE₂ and PGF_2α have been shown to promote cell survival and/or inhibit apoptosis,^49–51 supporting our observation of sustained viability during the three-hour incubation with either prostaglandin. Although the exact concentrations of PGE₂ and PGF_2α are unknown in or near the meibomian glands, their concentrations have been measured in the tear film. PGE₂ varies between 0.4 to 2.5 nM in normal patients and 0.5 to 7.7 nM in those with dry eye or MGD.^52–54 PGF_2α shows less disparity, averaging 1.0 nM in normals and 1.4 nM in MGD.⁵² To account for potential differences in fluid concentrations versus tissue concentrations,⁵⁵ we chose to investigate PGE₂ and PGF_2α concentrations in 10x increments from 1 nM to 1 μM. Even after incubating with the upper limit of physiologic concentrations of either prostaglandin, cell viability was not affected.

Influence of prostaglandins on lipid expression

Our results support that PGF_2α influences lipid expression, an outcome strongly influenced in HMGECs by cellular differentiation.^56,57 Nearly 18% of all expressed CEs and 16% of all expressed TAGs varied in response to PGF_2α supplementation. PGF_2α was associated with nonspecific changes in several TAG species, suggesting that it induces generalized TAG remodeling. Regarding CEs, however, more consistent patterns emerged. Low-dose PGF_2α (10⁻⁸ M) decreased expression of six of the seven statistically significant CEs compared to control. Despite this low-dose suppression, four of these six CEs (CE 20:2, CE 20:1, CE 22:1, and CE 24:0) increased thereafter in a dose-dependent fashion with PGF_2α (Figure 4A). For this pattern to consistently emerge, we hypothesize that multiple, competing pathways that influence cellular differentiation and/or lipid synthesis are concurrently activated.

First, we believe that PGF_2α may be inducing a relative suppression of differentiation and thus reducing meibum-relevant lipid expression. Previous reports have not only shown that PGF_2α suppresses adipocyte differentiation but that it does so via the FP receptor, ultimately leading to phosphorylation and inactivation of PPARγ.^40,58 In our experimental model, we differentiated our cells with rosiglitazone, a PPARγ agonist, for two days prior to introducing PGF_2α for three hours. During that three-hour incubation, we believe that PGF_2α led to a heterogeneous mixture of both phosphorylated (inactive) and unphosphorylated (active) forms of PPARγ, manifesting as a relative suppression of lipid expression compared to control. If this were the only pathway at play, however, a dose-dependent decrease would have been observed across the remaining PGF_2α concentrations.

Instead, we observed an increase in four specific CEs and believe this effect may be related to the known opposing effects that different prostaglandins have on adipogenesis, all converging on PPARγ.⁴⁰ PGF_2α-induced upregulation of PGD₂ and PGJ₂ have been previously described in a variety of cell types.^59–61 and PGD₂ and PGJ₂ have both been shown to potently activate PPARγ at a rate of approximately 80-fold and 20-fold, respectively.⁶²

If these two pathways are occurring simultaneously, as hypothesized, a portion of PPARγ protein would be inactivated by PGF_2α and another portion activated by PGD₂ and PGJ₂. For this latter mechanism to explain the dose-dependent increase observed in our study, then PPARγ activation by PGD₂ and PGJ₂ must dominate over the relative PPARγ inhibition by PGF_2α. To better interrogate these pathways, additional work is needed to isolate these mechanisms using receptor antagonists, COX inhibitors, and quantitative methods for other prostaglandins, PPARγ, and associated gene products.

We also assessed the effects of PGE₂ on lipid expression and found that PGE₂ led to a significant decrease in 10% of meibum-relevant CEs (CE 22:1, CE 26:0, CE 28:1, and CE 30:1) and significant remodeling of just 6% of TAGs. The role of PGE₂ on lipid synthesis is complex, yet its seemingly dichotomous effects may be explained by its diverse array of receptors. Some sources cite PGE₂ as an inhibitor of lipogenesis, which is mediated through the EP3 receptor.^63,64 Other sources credit PGE₂ for promoting lipogenesis and fat accumulation, an outcome attributed to EP2 and EP4 signaling.⁶⁵ Here, we did not detect EP3 expression on HMGECs, so the inhibitory effect of PGE₂ on lipogenesis may not exist in these cells. In the absence of this mechanism, it is currently unknown how PGE₂ induced the suppression of several meibum-relevant CEs in this study. Additional work is needed with receptor antagonists to determine which EPs are mediating this observation.

We detected the presence of six unique oxidized CEs (oxCEs) from our samples, only one of which (CE 26:0;2) varied significantly between low-dose and higher-dose PGF_2α. None of the six oxCEs varied with respect to PGE₂. Lipid oxidation of the tears has been previously investigated by Borchman et al, who found that the degree of lipid oxidation is greater in normal subjects compared to those with MGD.⁶⁶ They further described these oxide moieties to be hydrophilic groups among hydrophobic regions, interfering with lipid-lipid interactions and conferring increased fluidity to the overall lipid compartment. In simpler terms, a greater degree of oxidation decreases lipid order and promotes a more fluid secretion. In MGD, a lesser degree of oxidation increases lipid order and, thus, increases viscosity. The origin, mechanism of production, and significance of the oxCEs that we discovered from HMGECs remains to be explored.

Potential for PGF_2α’s effects to translate to the ocular surface with dosing of anti-glaucoma PGAs

Our experiments were designed to identify changes in the HMGEC lipidome and cell viability due to PGE₂ and PGF_2α after one exposure for three hours. Alteration in lipid production is just one mechanism that could lead to obstructive MGD as a result of PGA use in glaucoma.⁵⁷ Other mechanisms almost definitely exist, such as prostaglandin-mediated inflammation and preservative-induced toxicity. It is currently unknown whether the changes we observed in lipid expression are also produced by anti-glaucoma PGAs, but other researchers have begun investigating similar outcomes. Kam et al first reported that bimatoprost adversely influences a marker of HMGEC survivability but that it has no effect on lipid production assessed by LipidTOX, a nonpolar lipid stain.⁶⁷ Unfortunately, the use of bimatoprost in this study complicates generalizability to other PGAs, as bimatoprost is technically a prostamide analog whose functions may or may not be mediated through the FP receptor.^14,68,69 More recently, Rath and colleagues evaluated several preserved and non-preserved formulations of PGAs on HMGECs.⁷⁰ They found that only latanoprost affected cell viability and that bimatoprost upregulated cornulin and involucrin mRNA, two keratins that may be upregulated in MGD.⁷⁰ This study reportedly used commercially available formulations of PGAs, though it was unclear how these eye drops were added to the cell culture media without affecting the original concentration of the PGAs and without diluting the components of the culture media itself. Regardless, it is important to note that antiglaucoma PGAs are significantly more concentrated than the physiologic concentrations of PGF_2α that we used in this study. Specifically, commercially available latanoprost is approximately 100 times more concentrated than our highest dose of PGF_2α; bimatoprost is about 240 to 720 times more concentrated. Ultimately, it remains unknown whether anti-glaucoma PGAs mimic the effects of PGF_2α observed here, what concentrations of PGAs reach the meibomian glands, or whether there are cumulative effects with daily instillation.

In vivo, iatrogenic MGD that is associated with anti-glaucoma therapeutics is a problem compounded by long-term, daily instillation of preserved eye drops.¹² One perceived limitation of our experiments, therefore, could be the short duration of exposure (three hours) to the prostaglandin challenge. This cell culture model, however, is supported by the pharmacokinetics of prostaglandins and prostaglandin analogs. Ishihara et al reported that the half-lives of PGE₂ and PGF_2α in cell culture experimentation are approximately 9 hours and 15 hours, respectively, and further asserted that incubations greater than a few hours would be subject to significant metabolism.⁷¹ Repeated daily introductions of PGE₂ and PGF_2α were considered; however, Sjöquist et al previously reported that the pharmacokinetics of latanoprost, a representative PGF_2α analog used to treat glaucoma, were similar between single doses and repeated doses.⁷² Latanoprost showed no accumulation in tissues, suggesting that repeated dosing of prostaglandins specifically is an unlikely contributor to disease development. Therefore, we chose to perform a single challenge to PGE₂ and PGF_2α for three hours, a duration that is under the reported half-lives of PGE₂ and PGF_2α and equivalent to the known pharmacokinetics of latanoprost (half-life between two to four hours^73,74). Guided by these literature sources, we found that even a short duration of exposure to prostaglandins is sufficient to alter the lipidomic expression from HMGECs.

Another perceived weakness of our study may be our decision to differentiate with rosiglitazone, a PPARγ agonist, considering that PGF_2α is known to influence PPARγ function. Previous research has shown that PGF_2α can block differentiation of adipocytes by inhibiting PPARγ if introduced in the first two days of treatment.⁵⁸ We, however, did not introduce the prostaglandins until two days after the initial exposure to differentiating conditions, allowing the cells to reach a more mature, differentiated state. Further, the decision to utilize rosiglitazone-induced differentiation is well-supported by literature that has more comprehensively defined and characterized its mechanism^{30,31,43,44,75} compared to other differentiating agents,^33,76,77 reinforcing its use as a preferred method for HMGEC differentiation. Lastly, terminally differentiated meibocytes in vivo are believed to be under the influence of PPARγ regulation, so any interference by prostaglandins would likely translate into a similar interference that would occur at the ocular surface.⁷⁸

In conclusion, we have reported that HMGECs express FP receptors and three EP receptors (EP1, EP2, and EP4), potentially making them vulnerable to undesirable side effects caused by the PGF_2α analogs used in clinical practice to treat glaucoma. Just one exposure to PGF_2α, to an extent greater than PGE₂, led to lipidomic remodeling of HMGECs with significant changes observed in the expression of both CEs and TAGs. Alterations to the lipid chemistry of the meibomian gland secretions could affect the biochemical and biophysical interactions of the tear film lipid layer, potentially altering tear film viscosity and tear film stability. Further work is needed to determine how these observations translate to commercially available PGAs on the ocular surface.