Restoration of Corneal Stiffness in Rabbits Following Withdrawal of Travoprost

Abstract

Purpose

To evaluate if the effect of travoprost on corneal material stiffness could be restored after drug withdrawal.

Methods

Seventy-two rabbits were randomly allocated into three groups: medicine (M), medicine withdrawal (MW), and blank (B). Within the M and MW groups, treatment with travoprost was administered to the right eyes (MT and MWT) over a period of 12 weeks. Subsequently, the M group was killed, but the MW group underwent an additional 12-week period for treatment withdrawal. No treatment was given to the contralateral eyes (MC and MWC) in the M and MW groups. A separate blank control (BC) group remained untreated for the entire 24-week duration. In each group, corneas from 18 rabbits were tested mechanically under inflation conditions to estimate their tangent modulus (E_t). The corneas of the remaining six rabbits underwent electron microscopy analysis, which focused on fibril diameter and interfibrillar spacing.

Results

Central corneal thickness (CCT) of the treated eyes (MT and MWT groups) decreased with 12 weeks of travoprost treatment (P < 0.05). The CCT in the MWT group increased after 12 weeks of withdrawal but was still lower than that in the BC group (P < 0.05). The E_t of the MT group was significantly lower than that of the MC group at mean tissue stresses of 2, 4, and 6 kPa (P < 0.05). Conversely, no significant difference in E_t values was observed between the MWT, MWC, and BC groups, indicating recovery after treatment cessation. Furthermore, the stromal interfibrillar spacing of the treated MT group was significantly larger (P < 0.05) than that of the control MC group, but no disparity was noted among the MWT, MWC, and BC groups following treatment withdrawal. Additionally, there were no significant differences in the mean diameter of collagen fibrils among all groups (all P > 0.05).

Conclusions

Travoprost treatment appears to soften corneal tissue, decrease tissue thickness, and reduce the density of stromal collagen fibers by increasing the interfibrillar spacing. These changes were partially reversed after treatment cessation. Travoprost could further inhibit corneal growth, so its use in childhood and adolescence should be carefully considered. Additionally, the effect of travoprost in reducing corneal stiffness may lead to underestimations of intraocular pressure (IOP) measurement and hence overestimations in the effect of treatment in lowering IOP.

Glaucoma is a prevalent eye disease worldwide, often resulting in irreversible blindness.¹ Lowering intraocular pressure (IOP) can mitigate the risk of glaucoma progression,² making it an effective strategy to halt or slow advancement of the disease.³ Prostaglandin analogs (PGAs) are hypotensive agents that are frequently prescribed for the treatment of primary open-angle glaucoma (POAG) or ocular hypertension.⁴ Following the use of PGAs, increased activity of matrix metalloproteinases (MMPs), reduced presence of tissue inhibitors of MMP (TIMPs),^5,6 decreased collagen fibril density, and remodeling of the extracellular matrix have been observed in the trabecular meshwork,⁷ sclera,⁸ and cornea.⁹ Moreover, PGAs are associated with corneal stromal degradation, reductions in central corneal thickness (CCT),^10,11 and alterations in biomechanical properties of the cornea.¹² Collectively, these changes may lead to decreased stiffness and potentially contribute to an elevated risk of keratoconus progression, myopic regression, and ectasia development following refractive surgery.^13,14 In addition, the corneal stiffness reduction may result in underestimations in IOP measurement¹² and inaccuracies in procedures (such as refractive surgeries) whose outcomes depend on the stiffness of corneal tissue.¹⁵

Although the implications of applying PGAs have been studied, the effects of PGA withdrawal on corneal tissue biomechanics remain unclear. This point was assessed in an earlier study by Meda et al.,¹⁶ who reported partial reversible effects of PGA withdrawal on corneal hysteresis (CH). However, given the ongoing debate surrounding the true meaning of CH, uncertainty about stiffness recovery following drug withdrawal remains. This study aimed to address this gap by employing direct mechanical inflation testing to estimate the tangent modulus (E_t) of the tissue, the most common measure of material stiffness. Additionally, the study assessed the microstructure of the tissue, specifically collagen fibril spacing and diameter, due to their direct link to the biomechanical behavior of the tissue. Nevertheless, as the rabbits tested in this study were subjected to an experimental period that coincided with their growth, this study sought to distinguish between stiffness increases due to natural maturation and those due to PGA withdrawal.

Materials and Methods

Experimental Animals

Seventy-two Japanese white rabbits, 2 to 3 months old and weighing between 1.5 and 2.5 kg, were given a standard diet and water. They were individually housed in cages with a 12-hour light/dark cycle. The temperature and humidity of their environment were maintained at 16° to 26°C and 40% to 70%, respectively. These rabbits were sourced from the Animal Breeding Unit of Wenzhou Medical University and were acclimated for 1 week before the study commenced. All animal procedures adhered to the guidelines outlined in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Additionally, the Animal Care and Ethics Committee of Wenzhou Medical University approved the study prior to its initiation.

The rabbits were randomly divided into three groups of 24 each (Fig. 1; Table 1): medicine (M), medicine withdrawal (MW), and blank (B). In the M and MW groups, travoprost eyedrops (TRAVATAN, 0.04 mg/mL; Novartis, Basel, Switzerland) were administered once daily between 4:30 PM and 7:30 PM to the right eyes for 12 weeks, thus forming the MT and MWT subgroups. The left eyes of these groups remained untreated, thus constituting the MC and MWC control subgroups, respectively. Following the initial 12-week period, the MWT and MWC subgroups underwent no further treatment for an additional 12 weeks, facilitating travoprost withdrawal in the former subgroup. Additionally, the right eyes of the B group were designated as the blank control (BC) subgroup. Subgroups MT and MC were killed after 12 weeks, and subgroups MWT, MWC, and BC were killed after 24 weeks.

Figure 1.

Rabbit groups and administration methods in the study. Rabbits were randomly assigned to three groups of 24 rabbits each—namely, medicine (M), medicine withdrawal (MW), and blank (B). In the M and MW groups, the right eyes were treated with travoprost for 12 weeks and formed the MT and MWT subgroups, whereas the left eyes remained untreated and formed the MC and MWC subgroups. Additionally, after the first 12 weeks, the MWT and MWC subgroups had no treatment for a further 12 weeks. The right eyes of the B group were renamed the BC subgroup. The MT and MC subgroups were killed after 12 weeks, but the MWT, MWC, and BC subgroups were killed after 24 weeks.

Table 1.

Specimen Groups of Study

Group	Specimens, n	Description
BC	24	A separate blank control group that remained untreated for the entire 24-week duration of study
MWT	24	Right eyes of the medicine withdrawal group, in which travoprost was administered over a period of 12 weeks followed by another 12-week period when treatment was stopped
MWC	24	Left eyes of the medicine withdrawal group, in which no treatment was introduced
MT	24	Right eyes of the medicine group, in which travoprost was administered over a period of 12 weeks.
MC	24	Left eyes of the medicine withdrawal group, in which no treatment introduced

Central corneal thickness (CCT) was measured at baseline (pretreatment) and weekly thereafter, from 1 week (pos1w) to either 12 weeks (pos12w) or 24 weeks (pos24w). After applying topical anesthesia (a single drop of 0.5% proparacaine), CCT was determined using a portable pachymeter (PachPen; Accutome, Malvern, PA, USA). Three measurements were taken for each eye by the same operator (CW) during the same morning time window (between 8 AM and 11 AM), and the average results were recorded for analysis.

Experimental Design

Similar to our previous studies,^12,17 all rabbits were euthanized via intravenous injection of a high concentration (100 mg per kg of body weight) of pentobarbital sodium (Merck, Darmstadt, Germany) after the experimental observation period. The experimental observation period was different for each group. For the MT and MC subgroups, the experimental observation period was 13 weeks, whereas for MWT, MWC and BC subgroups the experimental observation period was 25 weeks (all groups included 1 week of adaptation before commencing experiments). Upon euthanization, both eyes were promptly enucleated, the corneas of 18 eyes in each subgroup were carefully separated along with a 3-mm-wide ring of scleral tissue and mounted onto a pressure chamber. Mechanical clamps were used to securely attach the entire scleral ring to the chamber, allowing only the corneal cap to deform freely. The chamber was filled with phosphate-buffered saline (PBS; Maixin, Fujian, China), inside which the internal pressure (simulating IOP) was controlled using a syringe pump.^12,17,18 Movement of the pump was controlled by custom-built LabVIEW software (National Instruments, Austin, TX, USA). Corneal geometric features, such as central and peripheral tissue thickness (the latter taken approximately 1.5 mm away from the limbus) at locations across the surface and diameter in horizontal, vertical, and two 45° diagonal directions, were determined using an ultrasonic pachymeter (SP-3000; Tomey Corporation, Aichi, Japan), and a vernier caliper, respectively. Side images were also taken by three digital cameras positioned around the specimen with 120° between their longitudinal axes. These images were analyzed to determine the anterior topography of the cornea.

Biomechanical Inflation Testing

After they were mounted on the pressure chamber, cornea specimens were inflated with an initial pressure of 2.0 mmHg to maintain a wrinkle-free surface. Corneal apex displacement and the pressure inside the chamber were continuously monitored by a CCD laser sensor (LK series; Keyence, Osaka, Japan) and A pressure sensor (DMP-HS, Hangzhou Yunm Instrument Co., Ltd., Hangzhou, China), which were connected to a computer to record pressure and displacement data. Three loading and unloading cycles were then conducted with a pressure change rate of 0.40 mmHg/s and a maximum pressure of 32.0 mmHg to condition the behavior of the tissue.¹⁹ To ensure that the behavior was not influenced by the strain history of previous loading cycles, a 90-second recovery period was allowed between each two cycles.²⁰ Subsequently, a final pressure loading up to 32.0 mmHg was applied, and the corresponding pressure–deformation behavior was recorded and utilized in an inverse analysis exercise to assess the biomechanical properties of the tissue. All testing procedures were completed within 3 hours postmortem.

Inverse Analysis

Specimen-specific numerical models, constructed using experimentally recorded corneal geometry data, were employed in an inverse analysis process based on the finite element (FE) method to derive the stress–strain behavior of the tissue. The iterative inverse analysis process used the FE solver Abaqus (Dassault Systèmes, Waltham, MA, USA) and the optimization software package LS-OPT (Livermore Software Technology, Livermore, CA, USA).^20,21 The analysis was based on trialing different corneal stress–strain relationships until the mismatch between the model predictions of corneal apical displacement and the experimentally measured values reached the lowest value. With the stress–strain behavior determined, the tangent modulus of the material at any stress could be estimated.

Each model was comprised of 1728 15-noded continuum elements (C3D15H) and organized into 12 rings and two layers, similar to our previous studies.¹² To mimic the connection to the mechanical clamps, the models were provided with fully fixed nodes along their edge. A first-order hyperelastic Ogden constitutive model²² was employed to characterize the behavior of corneal material, using a strain energy density function in the following form:

$$\begin{array}{c}\hfill W=\frac{2\mu }{{\alpha }^2}\left({\overline{{\lambda }_1}}^{\alpha }+{\overline{{\lambda }_2}}^{\alpha }+{\overline{{\lambda }_3}}^{\alpha }-3\right)+\frac{1}{D}{\left(J-1\right)}^2\end{array}$$ (1)

In Equation 1, µ and α are material parameters that represent the shear modulus and the strain hardening exponent, respectively; the other parameters were described and set up the same as in our previous study.^12,22 W is the strain energy per unit volume; λ_k represents the deviatoric principal stretches, where J^−1/3 × λ_k (k = 1, 2, 3); and λ₁, λ₂, and λ₃ are the principal stretches, J = λ₁λ₂λ₃. D is a compressibility parameter, the value of which is dependent on Poisson's ratio, ν: D$=\frac{3\left(1-2\nu \right)}{\mu \left(1+\nu \right)}$. In this study, ν = 0.49 was assumed to represent the near incompressibility of corneal tissue.^23,24 In Equation 1, µ and α were the only unknowns, and their values were estimated using the inverse analysis process.

The inverse analysis process was performed by setting µ and α to vary within certain ranges while monitoring corneal deformations under the applied inflation pressure. This process continued until the numerically predicted experimental displacement of the corneal apex matched the experimentally recorded displacement. The optimization to estimate µ and α employed the following objective function to minimize the root mean squared error (RMSE) of the mismatch between the experimental and numerical displacements at the corneal apex:

$$\begin{array}{c}\hfill RMSE=\sqrt{\frac{1}{P}.\sum _{p=1}^P{({\delta }_p^{exp}-{\delta }_p^{num})}^2}\end{array}$$ (2)

In Equation 2, P represents the total number of pressure levels between 2 and 32 mm Hg, and ${\delta }_p^{exp}$ and ${\delta }_p^{num}$ represent the experimental and numerical displacements of corneal apex at each pressure level p. To execute the optimization process, we employed LS-OPT software. We initially defined wide search ranges for µ and α and then analyzed 30 specimens, including five randomly drawn from six subgroups. The analysis showed that the optimal values of µ and α were always in the ranges of 0.0003 to 0.6 for µ and 10 to 300 for α; therefore, these ranges were adopted in subsequent analyses. These settings have been validated in previous published studies,¹² demonstrating the reliability and robustness of the method. By setting an appropriate material parameter range, the match between the corneal apical displacement predicted by the model and the experimental measurement was optimized. Based on this, a stress–strain relationship and tangent modulus closely representing corneal behavior and stiffness could be obtained. The inverse analysis process typically required 150 to 200 iterations. A uniqueness test was conducted and reported earlier,¹² and, although it was repeated in this study, its results are not reported herein.

Histological Analysis

The remaining six corneas from each specimen group underwent histological analysis to quantify corneal collagen fibril diameter, interfibrillar spacing, number of fibrils in a given unit area (µm²), and number of fibrils per unit area adjusted based on the thickness of the corneal stromal layer (CST). The analysis leading to these values followed the process detailed in a previous study.¹² An expert pathologist fixed the corneas, embedded them in paraffin, and sectioned them along the sagittal plane. These sections were stained with Masson's trichrome. Three sections, each about 80 nm thick, were obtained from each cornea at depths of approximately 80, 180, and 280 µm beneath the epithelium, representing the stromal anterior, intermediate, and posterior layers, respectively. Transmission electron microscope (TEM) microstructure images were captured by analyzing the sections with an H-7500 transmission electron microscope (Hitachi, Tokyo, Japan) at 40,000× magnification (refer to Fig. 2). The TEM images were analyzed for fibril diameter and interfibrillar spacing using ImageJ software (National Institutes of Health, Bethesda, MD, USA), as described in our previous study.¹⁷ Only fibrils with clearly defined circular boundaries and high contrast were included. Fibril cross-sections displaying an elliptical shape were excluded from the determination of fibril diameter.²⁵

Figure 2.

Five representative electron microscopy images for all groups: (A) MT, (B) MC, (C) MWT, (D) MWC, and (E) BC. (F) A schematic map showing the fibril diameter and the interfibrillar spacing.

Statistical Analysis

All analyses were conducted using SPSS Statistics 26.0 (IBM, Chicago, IL, USA). To assess differences between the treated and lateral control specimen groups, we employed the paired t-test or the Wilcoxon test. Similarly, one-way analysis of variance (ANOVA) or the Kruskal–Wallis H test was performed to compare biometric, biomechanical, and fibrillar parameters between the BC subgroup and the other four subgroups, with consideration of the results of a normal distribution test. P < 0.05 was considered statistically significant.

Results

CCT Changes

The baseline CCT was similar in all groups (P > 0.05) (Fig. 3A). During travoprost treatment, the CCT of the treated eyes (MT and MWT groups) and the corresponding control eyes (MC and MWC groups) increased slightly over the first 4 to 6 weeks. Then, the CCT in the MC and MWC groups became stable, but the CCT in the MT and MWT groups underwent a gradual decrease. Compared with the corresponding control eyes (MC and MWC groups), the CCT of the treated eyes in the MT and MWT groups decreased by 20.78 ± 17.93 µm and 17.72 ± 13.13 µm, respectively, after the 12th week (pos12w) (P < 0.01). After the treatment was stopped, the CCT in the MWT group gradually increased to be non-significantly different from that of the MWC group at the 24th week (pos24w). In contrast, the CCT of the BC group exhibited a gradual upward trend, and the differences between the BC group and the other groups gradually increased over time (all P < 0.05).

Figure 3.

(A) Changes in CCT during the 24-week follow-up period. (B) The ratio of each group to the BC group for CCT during the 24-week follow-up period. (C) Trends in CCT differences between each group and the BC group during the 24-week follow-up period. In A and B, right eyes treated with travoprost for 12 weeks formed the MT group. Right eyes treated with travoprost for 12 weeks then underwent treatment withdrawal over another 12 weeks and formed the MWT group. Fellow eyes were untreated and formed the MC and MWC subgroups. The right eyes in the blank control group were assigned to the BC subgroup. In C, the right eyes of the M and MW groups were combined to form the T group. The left eyes of the M and MW groups were combined to form the C group. W indicates 4 weeks after drug withdrawal in the T group.

After 4 to 5 weeks of travoprost treatment, the CCTs in the T group (right eyes of the M and MW groups) and the W group were slightly and non-significantly (P > 0.05) affected relative to the BC group (Figs. 3B, 3C). This effect grew gradually and became more pronounced with longer treatment time. However, after treatment withdrawal, the CCT ratio in the W group recovered to the levels observed in the C group (left eyes of the M and MW groups) (Fig. 3C) but remained lower than in the BC group (P < 0.01).

Results of Inflation Tests and Inverse Analyses

There were noticeable differences in the mean nonlinear pressure–apical displacement behavior between different subgroups, although there were also similarities in the main trends (Fig. 4). In the low stiffness stage, all subgroups displayed nonlinear pressure–displacement behavior. When a pressure threshold of 10 to 15 mmHg was exceeded, the behavior became almost linear. A significant difference in pressure–apical displacement behavior was observed between the MT group and the BC group (P < 0.05). However, the pressure–apical displacement behavior of the MC, MWT, and MWC groups was not significantly different from that of the BC group (all P > 0.05). Meanwhile, the difference in apical displacement between the treated MT group and the control MC group was significant (P < 0.05), but there were no significant differences in pressure–apical displacement behavior between the MWT and MWC groups (P > 0.05). For each cornea, an inverse analysis optimization process was used to estimate the µ and α coefficients of the material, enabling the closest possible match between the predicted and measured pressure–displacement results. This optimization aimed to find the parameter values that achieved the lowest possible RMSE described in Equation 2.

Figure 4.

Mean pressure–displacement behavior at the corneal apex in all subgroups. Error bars represent standard deviations of the displacement values.

Similar to the pressure–displacement results, the mean stress–strain (σ–ε) behavior for each subgroup, as depicted in Figure 5 and based on the optimized values of µ and α in Table 2, showed no significant differences between any of the subgroups and the BC group (all P > 0.05); the only exception was α in the MT group. The mean stress–strain behavior patterns were then employed to determine the tangent modulus (E_t = dσ/dε) for each group at a number of specific stress levels.

Figure 5.

Mean stress–strain behavior of corneas in all subgroups. Error bars represent standard deviations of the stress values.

Table 2.

Mean and SD of Constitutive Parameters µ and α in All Subgroups

	Mean ± SD
Group	µ (MPa)	α	RMSE (µm)
BC	0.1319 ± 0.1351	163.44 ± 33.06	38.02 ± 17.92
MWT	0.0799 ± 0.1010	137.15 ± 24.77	20.96 ± 18.39
MWC	0.0811 ± 0.0571	153.49 ± 36.04	26.58 ± 17.37
MT	0.0914 ± 0.1282	100.23 ± 36.15	29.96 ± 29.43
MC	0.0944 ± 0.0804	139.87 ± 33.17	28.86 ± 21.11
P	0.782	<0.001	—

Group comparisons were made at three stress levels of 2, 4, and 6 kPa, all falling within the initial nonlinear range of the specimens’ behavior. The E_t values are presented in Table 3. As shown in Table 4, compared with the BC group, E_t in the MT group was lower (P < 0.05), unlike E_t of the MC, MWT, and MWC groups, for which no significant differences with the BC group were observed (all P > 0.05). The E_t of the treated MT group was also lower than that of the control MC group (all P < 0.05 at 2, 4, and 6 kPa stress stages). Furthermore, the E_t of the treated MT group was not significantly different from the treatment withdrawal MWT group (MT vs. MWT, all P > 0.05 at three stress stages). Additionally, there were no significant differences in E_t between the MWT and MWC groups at the stress levels considered (all P > 0.05).

Table 3.

Mean and SD of Tangent Modulus in All Subgroups at Various Stress Levels

	Et, Mean ± SD (MPa)
	Stress Level
Group	2 kPa	4 kPa	6 kPa
BC	0.60 ± 0.36	0.89 ± 0.36	1.19 ± 0.37
MWT	0.44 ± 0.29	0.68 ± 0.29	0.93 ± 0.31
MWC	0.46 ± 0.17	0.73 ± 0.20	1.01 ± 0.25
MT	0.33 ± 0.14	0.50 ± 0.19	0.68 ± 0.25
MC	0.48 ± 0.21	0.72 ± 0.20	0.97 ± 0.22
P	0.037	0.001	<0.001

Table 4.

Significance (P) of Differences in E_t Values Calculated at Various Stress Levels

	Stress	MT	MC	MWT	MWC	BC
MT	2 kPa	—	0.016	1.000	1.000	0.017
	4 kPa		0.003	0.394	0.093	<0.001
	6 kPa		0.001	1.000	1.000	<0.001
MC	2 kPa	—	—	1.000	1.000	1.000
	4 kPa			1.000	1.000	0.552
	6 kPa			1.000	1.000	0.264
MWT	2 kPa	—	—	—	0.754	0.571
	4 kPa				0.504	0.183
	6 kPa				0.350	0.083
MWC	2 kPa	—	—	—	—	0.957
	4 kPa					0.698
	6 kPa					0.681

Histological Analysis

As shown in Tables 5 and 6, there was no significant difference in the mean diameter of collagen fibers among all of the groups in the three stromal layers (anterior, intermediate, and posterior) tested (all P > 0.05). The mean interfibrillar spacings in all three stromal layers in the MT and MC groups were significantly larger than the corresponding values in the BC subgroup (P < 0.05), except for the intermediate layer in MC and BC group specimens (P = 0.089). Notably, the comparison between the MWT and BC groups, as well as between the MWC and BC groups, did not reveal any significant differences (P > 0.05), indicating similarities in the mean interfibrillar spacings in these groups.

Table 5.

Mean and SD of Interfibrillar Spacing and Fibril Diameter in Corneal Stroma in All Subgroups

			Mean ± SD
Stroma	Group	n	Fibril Diameter (nm)	Interfibrillar Spacing (nm)	Fibrils per Unit Area, n	Fibrils Normalized by Stromal Thickness × 1000, n
Anterior	BC	6	31.7 ± 2.6	15.3 ± 1.4	579.0 ± 45.0	207.2 ± 20.9
	MWT	6	32.7 ± 3.2	17.3 ± 3.3	527.8 ± 120.7	173.2 ± 41.1
	MWC	6	32.0 ± 2.3	16.0 ± 2.8	564.9 ± 95.7	188.5 ± 30.9
	MT	6	33.4 ± 2.4	23.1 ± 2.0*,†	402.3 ± 51.6*,†	118.5 ± 14.6
	MC	6	32.1 ± 2.3	20.1 ± 2.0*	469.9 ± 41.0	154.7 ± 13.6
	P		0.303	<0.001	<0.001	<0.001
Intermediate	BC	6	31.7 ± 2.1	15.3 ± 1.1	580.5 ± 60.0	207.8 ± 25.8
	MWT	6	33.4 ± 3.5	17.7 ± 4.9	509.7 ± 126.3	167.1 ± 41.8
	MWC	6	32.6 ± 4.2	16.2 ± 3.9	557.4 ± 128.4	185.7 ± 40.8
	MT	6	33.4 ± 2.7	25.1 ± 2.3*,†	376.1 ± 46.8*,†	110.7 ± 13.0
	MC	6	33.4 ± 2.0	20.6 ± 1.9	440.9 ± 49.1	145.2 ± 16.4
	P		0.329	<0.001	<0.001	<0.001
Posterior	BC	6	32.3 ± 2.5	15.0 ± 1.7	576.8 ± 80.5	206.6 ± 33.0
	MWT	6	34.0 ± 4.2	17.4 ± 3.1	503.7 ± 133.1	165.8 ± 43.7
	MWC	6	32.8 ± 2.4	15.9 ± 2.9	548.6 ± 87.4	183.0 ± 27.6
	MT	6	33.5 ± 2.2	24.1 ± 1.5*,†	385.9 ± 28.0*,†	113.7 ± 83.8
	MC	6	33.3 ± 1.8	20.1 ± 1.6*	448.9 ± 42.1	147.8 ± 14.5
	P		0.330	<0.001	<0.001	<0.001

P indicates the significance of differences in fibril diameter in the different subgroups and different stromal layers.

^* P < 0.05 difference against BC subgroup.

^† P < 0.05 difference against MC subgroup.

Table 6.

Significance (P) of Differences in Number of Fibrils Normalized by Stromal Thickness in Anterior Stroma, Intermediate Stroma, and Posterior Stroma

Stroma		MT	MC	MWT	MWC	BC
MT	Anterior	—	<0.001	0.015	0.001	<0.001
	Intermediate		<0.001	0.044	0.003	<0.001
	Posterior		0.004	0.052	0.004	<0.001
MC	Anterior	—	—	1.000	0.370	0.021
	Intermediate			1.000	0.333	0.019
	Posterior			1.000	0.493	0.020
MWT	Anterior	—	—	—	0.303	0.362
	Intermediate				0.044	0.329
	Posterior				0.347	0.243
MWC	Anterior	—	—	—	—	1.000
	Intermediate					1.000
	Posterior					1.000

The number of fibrils per unit area (µm²) in the MT group was significantly lower than in the BC group (all P < 0.05), whereas the MC, MWT, and MWC groups had fibril density values similar to those of the BC group (all P > 0.05) (Table 4). The number of fibrils normalized by stromal thickness in the MT and MC groups was also significantly lower than that in the BC group (all P < 0.05). Nevertheless, there were no significant differences among the MWT, MWC, and BC groups (all P > 0.05).

Bilaterally, as indicated in Table 4, the interfibrillar spacing in the MT group was larger than in the MC group in all three stromal layers (all P < 0.05). This resulted in the number of fibrils per unit area and the number of fibrils normalized by stromal thickness being smaller in the MT group than in the MC group. Additionally, there were no significant differences in interfibrillar spacing, number of fibrils per unit area, and number of fibrils normalized by stromal thickness between the MWT and MWC groups in all three stromal layers (all P > 0.05), except for the number of fibrils normalized by stromal thickness in the intermediate stromal layer (P = 0.044).

Discussion

Due to their high antihypertensive efficiency, stable effect, and low incidence of adverse reactions, PGAs, including travoprost, have become the first-line drugs in the management of primary open-angle glaucoma and normal-tension glaucoma.^26,27 Travoprost has been reported to have superior IOP lowering characteristics, including a shorter time to peak activity^28,29 and a longer sustained effect in a 24-hour dosing regimen even with a single instillation.^30–32 However, PGAs also have side effects, including deepening of upper eyelid sulcus,³³ conjunctival congestion,³⁴ eye irritation,³⁵ iris pigmentation,³⁶ and eyelid skin darkening.³⁷ But, some studies have shown that, among prostaglandins, travoprost has fewer side effects, including the appearance of changes in eyelashes,³⁸ hyperemia in eyes,³⁹ and iris hyperpigmentation.²⁸ It is worth noting that travoprost was also shown to lead to unintended significant reductions in the stiffness of corneal tissue. Several studies have shown that the extent to which travoprost reduces CCT^40,41 and corneal stiffness¹² is greatest in the PGA family. Because IOP measurements are affected by CCT and corneal stiffness, Travoprost may cause the greatest error in IOP measurement in the PGA family. Few studies have further investigated the effect of PGAs on corneal biomechanics and how this, in turn, affects the measurement of IOP, which the treatment is intended to lower.¹⁶ This study aimed to address a specific related question as to whether travoprost withdrawal allows restoration of the mechanical stiffness of the tissue. The study employed mechanical testing and histological analysis, in which rabbit eyes were treated with travoprost, and, in some eye groups, treatment withdrawal was allowed. The results provided clear evidence of changes in corneal biomechanics, thickness, and stromal microstructure following treatment, as well as their partial reversal after withdrawal.

Most studies in the field to date have relied on in vivo biomechanics measurement methods, primarily utilizing an Ocular Response Analyzer (ORA; Reichert Technologies, Buffalo, NY, USA) and the Corvis ST (OCULUS, Wetzlar, Germany). However, the parameters provided by these instruments mainly describe the amplitude and speed of corneal response under mechanical action rather than directly estimating the material stiffness. Additionally, a connection between the parameters of most of these devices and standard biomechanical measures such as the tangent modulus has not been established. An earlier study observed no change in the ORA CH parameter but noted a decrease in the corneal resistance factor (CRF) in POAG patients treated with prostaglandins compared to untreated patients.⁴² Other studies have reported various outcomes, including increases in CH after prostaglandin treatment, increases in CH with no change in the CRF,⁴³ and increases in both CH and CRF.⁴⁴ These disparate findings indicate that the effect of prostaglandins on corneal biomechanical properties remains controversial and requires further investigation using alternative methods.

In this study, inflation testing, which is considered a more representative testing protocol of physiological conditions compared to uniaxial testing,^45,46 was utilized to quantify the tangent modulus of corneal tissue and its alterations with prostaglandin treatment and withdrawal. This test maintained the integrity of the cornea and physiological hydration levels on both the front and back surfaces while subjecting the tissue to repeated pressure application to simulate intraocular pressure, allowing us to observe the resulting deformations on the front surface. Subsequently, an inverse finite element analysis was employed to determine the stress–strain characteristics of the tissue and tangent modulus. The results revealed that the E_t of the cornea in the treated (MT) group at stress levels of 2, 4, and 6 kPa was significantly lower than in the two corresponding untreated control groups (MC and BC) (all P < 0.05). At the other extreme, the other treatment group (MWT), which underwent withdrawal, did not exhibit significantly different E_t values compared to the relevant control groups, BC and MWC (all P > 0.05). Furthermore, in comparison to the BC group, the E_t values of the MT, MC, MWT, and MWC groups were lower by 42.91% to 44.87%, 18.20% to 20.14%, 21.76% to 26.77%, and 14.89% to 23.38%, respectively. These results indicate consistent trends wherein travoprost treatment led to a decline in corneal stiffness, which can be partially restored with withdrawal.

However, it is worth noting that the CCT of rabbits in the BC subgroup increased, indicating that the rabbits were still growing in the 12-week experimental period. As a result, it is anticipated that the stiffness recovery noted in the MWT group was partly due to natural growth, and this study sought to separate the effects of drug withdrawal and natural growth. Because the recovery of CCT after drug withdrawal mainly occurred in the WMT group from pos12w to pos16w, the degree of increase in CCT during this period was therefore compared between the MWT group and the BC group. It was found that the CCT increase in the MWT group was much larger than that in the BC group during the period from pos12w to pos16w. In addition, the slope of CCT changes in the MWT group during this period was much steeper than in the BC group. The above two points indicate that the recovery effect of corneal and the natural growth effect of the cornea during this period each played a role.

Using Brillouin microscopy, other studies have reported that the relative decrease of these conventional moduli in the area of pathology in keratoconic corneas compared with the healthy corneas was approximately 50%.⁴⁷ This reduction is large and exceeded the stiffness reduction we observed with PGAs using a different measurement method. However, this reduction can potentially have an effect on IOP measurement and lead to clinically relevant overestimations of the effect of PGAs in lowering IOP.

Prostaglandin analog monotherapy significantly reduces corneal thickness in patients with glaucoma or ocular hypertension.^10,11,48,49 This effect stems from the alteration in the balance of MMPs and TIMPs triggered by inflammatory cytokines, which in turn are induced by prostaglandin eye drops, leading to accelerated degradation of corneal stroma and a substantial reduction in stromal thickness.¹¹ The literature indicates that all prostaglandin analogs induce decreases in corneal thickness; an example study demonstrated a maximal reduction in thickness of 11.57 µm caused by travoprost use in glaucoma patients.¹⁰ In another study, travoprost was linked to a significant reduction in CCT, with a mean value of 10.10 µm in patients with POAG.⁵⁰ In our investigation, CCT in the BC group exceeded that in the other groups at the 24th week (P < 0.05). Compared to their corresponding control eyes, the CCTs of the experimental eyes in the MT and MWT groups decreased by 20.8 ± 17.9 µm and 17.7 ± 13.1 µm, respectively, after 12 weeks of exposure to travoprost (P < 0.01).

Although many studies have shown that the use of prostaglandin eye drops can cause corneal thinning, most of these studies have been conducted over large time intervals of months⁴⁸ or years,^11,49 which may lead to the neglect of some potential impacts of the effect of prostaglandins on the cornea. In this study, CCT was continuously monitored during the use and withdrawal of travoprost every week. Therefore, we were better able to determine drug effects on the cornea at different stages.

The CCTs of the MT and MWT groups increased slowly at the beginning of treatment, more slowly than those of their corresponding control subgroups (MC and WMC), a finding that differs from a continuous decrease in CCT after the use of prostaglandins in other studies.^11,48,49 This may be due to the fact that the rabbit cornea in this study was still in the growth stage; the cornea would continue to thicken, and the corneal development was inhibited by the travoprost. As a result, the degree of corneal thickening was less than that of the corresponding control subgroups. In other studies, corneas had grown completely, so this phenomenon did not occur. In addition, CCT in the BC group continued to increase from pos16w to pos24w, whereas CCT in the MWT group remained stable from pos16w to pos24w after drug withdrawal, possibly suggesting a delayed inhibition of corneal growth by travoprost. This phenomenon suggests that there is corneal growth inhibition when using travoprost in children with ocular hypertension or glaucoma. Prostaglandins should be used with caution in children and adolescents, which was confirmed by other study.⁵¹

In addition, it was found that the effect of travoprost on corneal thinning gradually weakened, and CCT decreases changed from being rapid to slow, suggesting that there is a threshold value of the effect of the drug on corneal thinning. When this threshold is reached, the effect of the drug on cornea thinning might gradually diminish. One study¹⁰ showed that the reduction in CCT with travoprost occurred mainly in the first 6 months, suggesting a similar maximum threshold, but that study did not obtain detailed weekly data for the first 6 months.

The results indicate that travoprost gradually reduced CCT in both treated and contralateral eyes during the medication period. This effect becomes more pronounced over time, with the CCT reduction due to exposure to travoprost becoming apparent after an initial period. Additionally, the results suggest CCT recovery following drug withdrawal, reaching levels observed in the contralateral eye. However, some residual effects of travoprost on CCT persisted, resulting in both the experimental and contralateral eyes remaining thinner than the blank control eyes.

The decreases in corneal material stiffness and tissue thickness associated with travoprost suggest significant overall stiffness reductions, which persist even after drug withdrawal. Whether this effect completely disappears with longer withdrawal periods remains an unanswered question. More importantly, the role of prostaglandin F2α (PGF2α) in reducing the stiffness of corneal material may lead to an underestimation of IOP readings, which could negatively affect the management of ocular diseases, including glaucoma.⁵² Research shows that, compared with the blank control group, travoprost results in IOP underestimations of 14.7% to 16.5%,¹² indicating that a significant part of the reductions in IOP readings may be an artifact caused by the effect of PGF2α in lowering the stiffness of the cornea. It is for this reason that this study has implications for setting IOP targets in patients treated with travoprost, as it is even more important to use techniques that provide biomechanically corrected IOP measurements.

The corneal stroma consists of regularly arranged collagen fibers with uniform diameter and interfibrillar spacing, forming a highly organized collagen layer composed primarily of type I and type V collagen.⁵³ In our rabbit experiments, the reduction in collagen in the corneal stroma, observed through immunohistochemical staining after PGA use, resulted in reduced stiffness, stretching, and thinning,⁹ which would have influenced the structure and distribution of stromal collagen fibers. Our experiment revealed that the spacing between collagen fibers in the stromal layer of the treated (MT) eyes was significantly wider than that in the control (MC) group (P < 0.01), whereas there was no significant difference in collagen fiber diameter (P > 0.05). These findings suggest that travoprost increases collagen interfibrillar spacing, leading to a decrease in corneal stiffness.

There was also no significant difference in collagen fiber diameter and collagen interfibrillar spacing between the MWT and MWC groups following travoprost withdrawal (P > 0.05). The mean collagen interfibrillar spacings in the MT and MC groups were also larger than those of the BC subgroup in all three stromal layers (P < 0.05), except in the intermediate layer of the MC and BC groups (P = 0.089), whereas there were no significant differences among the MWT, WMC, and BC groups (all P > 0.05). Furthermore, when the number of fibers was normalized by the stromal thickness, its lowest value was observed in the MT group (P < 0.05), whereas there was no significant difference between the MWT and MWC groups following withdrawal. These findings suggest that the decrease in collagen fibril density in the stroma contributes to reductions in corneal stiffness, and withdrawal of the drug may facilitate partial restoration of the stromal microstructure characteristics prevalent prior to treatment.

The study also observed the effect of PGAs on the lateral eyes, which were not directly treated, although the effect was small compared to the treated eyes.^12,54 One possible mechanism responsible for the observed tissue softening in the lateral eye could be the absorption of the drug through the nasolacrimal mucosa, followed by blood transport to the other eye. This implies that the drug may have systemic effects beyond the directly treated area. Furthermore, at the 24th week, the CCT of the MWC group was significantly lower (P < 0.01) than that of the BC group, providing further support for this conclusion.

Rabbit eye globes have been frequently utilized in research due to their biomechanical resemblance to human eyes.^46,55 However, it is crucial to exercise caution when extrapolating the findings of this study, given the distinct behavior between ex vivo rabbit corneas and in vivo human corneas.⁴⁶ Additionally, a limitation of this study lies in the finite-element models, which solely account for the corneal segment, with fixed edges that do not adequately replicate the natural connection to the sclera.

Conclusions

This study assessed the impact of travoprost application and subsequent withdrawal on corneal biomechanical behavior. The findings indicate a notable reduction in corneal stiffness following the use of travoprost eye drops. Upon cessation of the drug application, stiffness levels increased, but they did not return to their initial pretreatment levels in 12 weeks. These results are of particular importance because the tissue softening caused by travoprost may lead to underestimations in IOP measurements, which can exaggerate the IOP-lowering effect of the treatment. The further finding that tissue stiffness recovers gradually with travoprost withdrawal indicates that the tissue softening is not permanent and could be reversed, if required. In addition, travoprost has been shown to inhibit corneal growth during the childhood development period, so it should be used with caution to treat ocular hypertension or glaucoma in children and adolescents. Travoprost has the potential to heal corneal scars and restore clarity, which could be significant for tailoring travoprost treatment plans and guiding its clinical use and withdrawal of prostaglandins.