Determining the Relationship Between Corneal Stiffening and Tissue Dehydration After Corneal Cross-Linking

Abstract

Purpose: To quantify corneal cross-linking (CXL)–induced stiffening via mechanical testing to estimate the impact of changes in hydration levels (H) and evaluate depth-dependent tissue hydration after CXL.

Methods: Eighty-three porcine corneal buttons were divided into three groups: Standard protocol CXL (S-CXL), accelerated CXL (A-CXL), and untreated (nonirradiated riboflavin-only) controls. Samples were hydrated or dehydrated to modulate H and dynamic mechanical analyzer compression tests were performed to measure Young's modulus (E). To extract the solid tissue network modulus, the cornea was modeled as a biphasic material after measuring E at different H. Corneal hydration was correlated with depth-dependent tissue thickness characterized by confocal reflection microscopy (CRM).

Results: Young's modulus increased fourfold after S-CXL (0.72 ± 0.1 MPa) and threefold after A-CXL E (0.53 ± 0.12 MPa) versus controls (0.17 ± 0.045 MPa). However, H decreased from 4.07 ± 0.35 in controls to 2.06 ± 0.2 after S-CXL and 2.79 ± 0.12 after A-CXL. After H modulation and biphasic mechanical modeling, Young's modulus for corneal solid tissue network showed only a 1.8-fold increase after S-CXL (2.25 MPa) and 1.5-fold increase after A-CXL (1.85 MPa) versus controls (1.22 MPa). With CRM, the overall thickness of the corneal tissue was found to linearly correlate to hydration H as expected. No appreciable depth dependence of hydration-induced thickness changes throughout the corneal buttons were observed.

Conclusions: Corneal tissue hydration changes significantly impact measured corneal stiffness after CXL using mechanical testing. Not considering H leads to major overestimation of the stiffening effect of the CXL procedure. Depth-dependence of corneal thickness because of changing hydration is strongly dependent on the integrity of the tissue.

Keratoconus is a progressive corneal disorder resulting in focal thinning and increased focal curvature with vision loss caused by underlying focal corneal weakening.^1,2 Corneal cross-linking (CXL) is standard-of-care treatment to halt this process and stabilize corneal morphology,^3,4 achieved presumably by increasing the biomechanical strength.^5,6 The Standard, or Dresden, CXL procedure involves the corneal epithelial removal with subsequent application of a riboflavin-dextran solution for 30 minutes followed by UV-A light exposure to induce the photochemical reaction. Riboflavin serves as a photosensitizer that absorbs UV-A light energy to induce covalent bonds between collagen molecules among themselves and with proteoglycans^7,8 and is typically in a dextran solution used to prevent corneal swelling. Tissue hydration control is an important step in the procedure because it has been proven that mechanical properties are affected by tissue hydration, with increasing hydration resulting in decreased tissue stiffness.^9–11 Overall, the CXL process results in loss of water content, as observed by the thickness-hydration relationship^12,13 and the thinning of the corneal tissue^3,14 after the procedure. 

Consequently, there are two separate factors contributing to the observed stiffening of the tissue ex vivo after CXL: the covalent bonds that are created between the collagen-collagen molecules and collagen-proteoglycans by the photochemical reaction with riboflavin; and the dehydration experienced by the tissue after the soaking in the riboflavin-dextran solution and the UV-A light exposure. However, the hydration of the tissue in vivo changes over time, with initial tissue dehydration followed by recovery over time to baseline preoperative levels, while maintaining the visual improvement of the patient.^3,4 Therefore the observed stiffening effect in corneas measured ex vivo immediately after treatment should not be presumed to represent the total effect persisting in vivo over an extended period after the CXL procedure. This work aims to understand these factors separately and calculate the stiffening effect of only the photochemical crosslinking between molecules (i.e., the effect on the corneal solid tissue network). 

Previous studies have investigated the correlation between the biomechanical properties of the cornea and its hydration level, demonstrating that as tissue hydration increases, the overall biomechanical strength decreases. These studies include tensile^9–11 and compression testing¹⁵ on both regular and cross-linked corneas.^16–18 Noncontact method Brillouin light spectroscopy has also shown the hydration-dependent mechanical properties of the cornea.^19–21 The results of these studies show that failing to account for tissue hydration when measuring biomechanical properties leads to inaccurate results. The goal of this study was to separate and quantify the individual contributions of the photochemical effect and the hydration effect on the measurement of the biomechanical properties after CXL. 

Methods

Freshly enucleated porcine eyes were obtained from a slaughterhouse (Wagner Meats, Mount Airy, MD, USA) and maintained in ice until the start of experiments, which were completed within a few hours from enucleation. Eyes were inspected, and damaged or cloudy corneas were discarded before starting experiments. Eyes that were selected for experimentation were taken out of the ice and all the following procedure were at room temperature. For the selected eyes, the epithelium tissue was carefully removed with a razor blade, and the cornea was dissected to obtain a 5 mm punched button (Integra Miltex Disposable Biopsy Punch; Integra LifeSciences, Princeton, NJ, USA) from the central cornea. Corneal buttons were then randomly assigned to one of three groups (Table): 

• 1. Control (n = 27)
• 2. Accelerated CXL (A-CXL, n = 27)
• 3. Standard protocol (S-CXL, n = 29)

Table.

 

CXL Methods

 

CXL Procedures

Riboflavin solution at 0.1% with dextran 10% to limit the intrinsic swelling of the procedure²² was prepared and topically applied to the corneal buttons every three minutes for 30 minutes. Afterward, either S-CXL or A-CXL was applied. In S-CXL,⁶ the button was exposed to a constant 365 nm UV-A light (UV Curing LED System; ThorLabs, Newton, NJ, USA) with a surface irradiance of 3 mW/cm² (measured with a UV Digital Radiometer) for 30 minutes, with riboflavin solution application every five minutes. In A-CXL,^23–27 the surface irradiance was 9 mW/cm² for 10 minutes, also with riboflavin solution application every five minutes. Control group corneal buttons were not exposed to UV-A light, but the riboflavin solution was also topically applied every three minutes for 30 minutes. Full details are included in the Table CXL Methods. 

Corneal Hydration Testing

After CXL, the cornea's hydration level (H) was controlled by immersing the corneal button in distilled water or left out in air (over a porous tape to let air flow underneath) for either five, 10, 15, 20, 25, 30, 35 or 40 minutes to induce different levels of hydration or dehydration, respectively, similar to other studies.^19,21 Before mechanical testing, the corneal button was weighed, and afterward they were left out to dry in air for at least 72 hours and weighed again. With this information, the hydration level (ratio of the amount of water and the amount of dry mass of the cornea) was measured:  

\begin{eqnarray*} H = \frac{(\textit{total weight of the button}\,\hbox{--}\,\textit{dry weight})}{{\textit{dry weight}}} \end{eqnarray*}

 

Mechanical Testing

The thickness of the corneal buttons was measured with a digital caliper before mechanical testing. Dynamic mechanical analyzer (DMA Q800; TA Instruments, New Castle, DE, USA) was used to perform compression tests. Sandpaper was attached to the top plate of the compression clamp to avoid slippage. The corneal buttons were placed anterior side up on the bottom plate of the clamp and the top one was brought down to contact, and a 0.001 N force was applied as a preload force to avoid floating of the top plate. A 1N/min compression was applied to the corneal button and stopped at 3N. For the analysis, the points for −1% to −5% compressive strain were taken, and the compressive stress was calculated by dividing the measured force by the area of the corneal button. It would be valuable to comprehensively report the metrics of the stress-strain test (e.g., Poisson's ratio); however, here we focused on the Young's modulus (E) because it is the most used to characterize CXL procedures. Young's modulus (E) was calculated and reported here for this compressive strain as the tangent of the stress-strain curve at a given strain value. This also means that we were not considering the viscoelastic behavior of the cornea in this study. 

Theoretical Model To Calculate Corneal Solid Tissue Network Modulus

The corneal tissue was modeled as a hydrated porous material, a biphasic material made up of a solid \( { ( \alpha =s)} \) and a fluid \( { ( \alpha =f)} \) component. In the case of corneal tissue, the solid component is the solid tissue network, and the fluid component is the water that hydrates the tissue. Each one of these occupies a volume fraction \( { (\phi _{a})} \) of the mixture, and the mixture saturated \( (\phi _{s} + \phi _{f} = 1) \). The Young's modulus (E) relates to the shear modulus \( { (G)} \) via the Poisson ratio \( { (v)} \):  

\begin{eqnarray} G = \frac{E}{{2(1 + v)}},\quad E = 2G(1 + v) \end{eqnarray}

(1)

 

Previous studies have demonstrated that corneal biomechanical properties change as \( { \phi _f} \) of the mixture changes.^19,21 The variation of the shear modulus can be calculated from the variation of the \( { \phi _f} \)²⁸:  

\begin{eqnarray} G = {{G}_s}\left( {1 - {{\phi }_f}} \right) \end{eqnarray}

(2)

Where \( G_s \) is the shear modulus of the mixture when \( \phi _f \) = 0, this is only the solid network of the mixture, without pores and no fluid component, and is the fitting variable to adjust to find the photochemical stiffening effect of the CXL procedure. With Equations 1 and 2, E can be calculated from \( G_s \) and the \( \phi _f \):  

\begin{eqnarray} E = 2{{G}_s}\left( {1 - {{\phi }_f}} \right)\left( {1 + v} \right) \end{eqnarray}

(3)

 

As the hydration level changes, Poisson's ratio also varies and is calculated in terms of the \( \phi _f \)²⁹:  

\begin{eqnarray}v = \frac{{{{H}_A} - 2{{G}_S}\left( {1 - {{\phi }_f}} \right)}}{{2\left( {{{H}_A} - {{G}_S}} \right)\left( {1 - {{\phi }_f}} \right)}}\end{eqnarray}

(4)

Where \( H _A \) is the “aggregate” elastic modulus of the solid network³⁰ and is defined as  

\begin{eqnarray}{{H}_A}\, =\, \frac{{{{D}_0}{{{\left( {1 - {{\phi }_f}} \right)}}^n}}}{{\alpha {{{\left( {2 - {{\phi }_f}} \right)}}^2}}}*{{\left({{{{\phi }_f}}} \right)}^{2 - n}}\end{eqnarray}

(5)

\( {D _0} \) is the value of water self-diffusion coefficient (\({2.3\ \times \ {{10}^{ - 9}}\frac{{{{m}^2}}}{S}}\) at \( {25^{\circ}} {\rm{C}} \)³¹; \( {n} \) and \( {\alpha} \) are fit parameters used for hydrated porous materials³² such as the corneal tissue (\({n = 3.236,\alpha = 3.39\ \times {{10}^{ - 18}}\frac{{{{m}^2}}}{{{{N}_S}}}}\)). Combining Equations 4 and 5 with Equation 3, there is an equation to calculate a theoretical relationship of E in terms of the variation of the fluid component \( {\phi_f} \) (i.e., hydration level of the corneal tissue):  

\begin{eqnarray}E = 2{{G}_S}\left( {1 - {{\phi }_f}} \right)*\left( {1 + \frac{{\frac{{{{D}_0}{{{\left( {1 - {{\phi }_f}} \right)}}^n}}}{{\alpha \left( {2 - {{\phi }_f}} \right)}^{2}}{{{\left( {{{\phi }_f}} \right)}}^{2 - n}} - 2{{G}_S}\left( {1 - {{\phi }_f}} \right)}}{{2\left( {\frac{{{{D}_0}{{{\left( {1 - {{\phi }_f}} \right)}}^n}}}{{\alpha {{{\left( {2 - {{\phi }_f}} \right)}}^2}}}{{{\left( {{{\phi }_f}} \right)}}^{2 - n}} - {{G}_S}\left( {1 - {{\phi }_f}} \right)} \right)}}} \right) \quad \end{eqnarray}

(6)

 

A curve of the Young's modulus E in terms of hydration level is obtained and fitted with Equation 6, obtaining the value of G_s, the shear modulus of only the solid tissue network when hydration level is 0. With the correlation between shear and Young's modulus, E_s is calculated as the Young's modulus of only the solid tissue network of each condition: control, A-CXL and S-CXL. The disparity in the E_s observed across the different conditions represents the real stiffening effect on the biomechanical properties of the corneal tissue and not a hydration effect. 

Confocal Reflection Microscopy

Cross-sectional images of the corneal tissue were captured using a confocal microscope (Olympus FLUOVIEW 3000; Olympus, Tokyo, Japan) with a 405 nm light, and a ×20 lens. Different levels of hydration were induced into the corneal buttons as before. Afterward, the button was cut in half to obtain a cross-section image without the need of further tissue preparation. The ability of the collagen fibrils to reflect light³³ was used to generate high-resolution images without any chemical processing or staining, and only the cutting process required for the cross-sectional imaging. The images were then processed, and the different corneal regions (anterior, posterior and total thickness) were measured with the image processing program ImageJ. 

Results

The S-CXL group, showed a ∼fourfold increase of the Young's modulus (0.72 ± 0.1 MPa) when compared to the control group (0.17 ± 0.045 MPa), statistically significant (P < 0.0001). However, dehydration was also evident on the corneal buttons after the procedure, decreasing the hydration level H from 4.07 ± 0.35 in the control condition to 2.06 ± 0.2 after the CXL, also statistically significant (P < 0.0001). This effect was seen in the protocol even though the drops of riboflavin to reduce dehydration were periodically applied to the corneal buttons. A-CXL showed a similar trend, but with reduced effect both in stiffening (∼threefold increase of Young's modulus (0.53 ± 0.12 MPa, P < 0.01). and less change in hydration level (H decreased to 2.79 ± 0.12) compared to controls (Fig. 1) (P < 0.0001). 

Figure 1.

 

(A) Hydration level of the different conditions of the cornea. Hydration level decreases when the CXL procedure is applied. (B) Young's modulus (E) of the different conditions of the CXL procedure. E increases with both S-CXL and A-CXL, but S-CXL shows a higher increase in stiffness.

To evaluate the correlation between H and E, H was modulated after the CXL protocols, immersing the corneal button in water or leaving them out to dry, reaching H levels from 1 to 12. E was measured afterward following the same procedure as before, and data points obtained represent the behavior of the Young's modulus versus hydration level obtained for the different conditions. With Equation 6, modeling the corneal tissue as a biphasic material, we fitted the relationship E-H for the different conditions, leaving the modulus of only the solid tissue network as the fitting variable (Fig. 2). The fitting variable for Equation 6 is the shear modulus of the solid network G_s; but E_s is easily calculated with Equation 1. Because this is a value where the hydration level is 0, with no water in the tissue, the values of the mechanical modulus for this solid network is higher than values of Figure 1, where the groups have specific levels of hydration according to their CXL protocol. 

Figure 2.

 

Young's modulus versus hydration curves for the different conditions of the CXL protocols. Each point in hydration level represents increments of 1. The asterisk represents the condition where hydration is the result of the procedure, it is not manipulated after the procedure, same data points as in Figure 1.

The photochemical stiffening effect due to the collagen fiber cross-links was calculated (Fig. 3), with a ∼1.8-fold increase (2.25 ± 0.19 MPa) for S-CXL and a ∼1.5-fold increase (1.85 ± 0.18 MPa) for A-CXL compared to controls (1.22 ± 0.11 MPa). Visualizing these values in Figure 2, these would be found where the fitting curves intercept the Y axis, meaning H = 0, removing any hydration effect on the Young's modulus of the tissue. 

Figure 3.

 

Young's modulus exclusively attributable to the photochemical effect. The result comes as a fitting parameter of the generated curves of Figure 2, modeling the corneal tissue as a hydrated porous material and a biphasic material made up of a solid and a fluid component. Error bars: 95% confidence.

The fitting variable for Equation 6 is the shear modulus of the solid network G_s; but E_s is easily calculated with Equation 1. Because this is a value where the hydration level is 0, with no water in the tissue, the values of the mechanical modulus for this solid network is higher than values of Figure 1, where the groups have certain level of hydration. 

An important relevant topic to the hydration effects on CXL is the spatial behavior of hydration in vivo versus ex vivo and in different experimental settings. To analyze how the hydration level affects the tissue in corneal buttons, virgin corneal buttons were immersed in DI water, or left to dry on open air to modulate H. The images obtained with confocal reflection microscopy were analyzed (Fig. 4) and the overall thickness of the corneal tissue was found to linearly correlate to hydration H: Thickness [µm] = 368.3 + 188.7 * H; R² = 0.91 (Fig. 4B). We defined the anterior part of the stroma, as the section with interweaving lamellae, and the posterior part as the section with parallel lamellae (Fig. 4C). No depth dependence of hydration induced changes in thickness throughout the cornea was observed. The ratio of the anterior/posterior part showed a small increase with hydration level (Fig. 4D), showing that at some point at highly hydrated conditions, the anterior cornea swelled more in percentage than the posterior cornea. 

Figure 4.

 

(A) Representative image of the corneal tissue with confocal reflection microscopy. Anterior part of the stroma is defined as the part with interweaving fibrils. Posterior part is defined as when fibrils are parallel to each other. (B) Total thickness, of the cornea at different Hydration levels. Thickness [µm] = 368.32 + 188.67 * H; R² = 0.9064. (C) Anterior and posterior sections of the cornea at different hydration levels. At low hydration levels H < 3, the sections definition is not clear. (D) In these corneal buttons, the anterior/posterior ratio of the cornea increases at high levels of hydration. Linear regression y = 0.0279x + 0.2892; R² = 0.7909; P < 0.001.

Discussion

The results of this study show that the measured impact of CXL on corneal stiffening in ex vivo samples is significantly impacted by corneal hydration and that tissue dehydration leads to an overestimation of the impact of CXL on the cornea solid tissue network when using mechanical testing. This works presents a quantitative model for the biomechanical properties of the cornea, accounting for both mechanical stiffening of the solid tissue network and hydration level of the porcine corneae. 

Although previous studies have reported the importance of hydration impacting directly the corneal biomechanical properties,^{9–11,15–21} our model and the resulting modulus-hydration curves (Fig. 2) allowed us to isolate and calculate only the photochemical effect of CXL (Fig. 3). Our model also demonstrates, consistent with previous studies,¹⁸ that the photochemical effect of CXL persists at different hydration levels, showing a higher modulus of the CXL corneas when compared to controls at all hydration levels. In our model, the fitting variable is the modulus when hydration is zero, not affected by the hydration level, and can be understood as the stiffening of only the solid tissue network of the tissue (Fig. 3). We theorize that this value can be understood as a more accurate long-term effect of the CXL procedure on the cornea. 

The cornea is mainly composed by the stroma (about 90%), a layer that consists of highly organized collagen fibrils that form sheets of lamellae that are oriented orthogonally to each other and parallel to the corneal surface.³⁴ The lamellae in the anterior part of the cornea interweave, while the posterior lamellae are parallel to each other.^35,36 This specific architecture results in a significant increase of stiffness on the anterior part of the cornea and the maintenance of corneal curvature.³⁷ Additionally, the anterior part of the cornea also exhibits strong resistance to swelling when immersed in water, whereas the posterior part of the cornea absorbs water and swells, as demonstrated by electron microscopy on fixed corneoscleral tissue immersed in water for different days.³⁷ We corroborated these results with porcine corneas and using confocal reflectance microscopy, which allowed for the generation of high-resolution images without any chemical processing or staining because of the ability of the collagen fibrils to reflect light.³³ The images were then processed to measure the different parts of the cornea (anterior, posterior, and total thickness, Fig. 4A) based on the assumption of the tissue hydration and thickness relationship.^12,13 In accordance with a previous study in human corneas,³⁷ when using corneoscleral tissue, the anterior part of the cornea presents resistance to swelling, whereas the rest of the cornea swells significantly (Supplementary Fig. S1). However, the object of study in this work was the corneal tissue presented as corneal buttons, so the architecture of the cornea under hydration was studied in this condition. 

In intact corneas in vivo, the primary regulator of the tissue hydration is the continuous pumping action of endothelial cells.³⁸ In contrast, in ex vivo corneal samples, hydration is only controlled by the medium in which the cornea is immersed. The result was that at highly hydrated conditions (>6), the anterior part of the cornea also presents swelling (Fig. 4B), even increasing the ratio of interweaving/parallel lamellae, meaning that the anterior part with interweaving lamellae presents greater swelling than the posterior part at some point (Fig. 4C). The difference between the corneoscleral tissue and the corneal button is the ease of water penetration into the tissue of the latter. In the corneal button, the stroma is exposed to the water on the sides of the button, plus the corneal fibers in the anterior may be disrupted when cutting the tissue for the cross-section imaging, whereas in the corneoscleral tissue, the sclera presents a barrier of the water penetrating the tissue. We theorize that this is why corneal buttons achieve comparable high levels of hydration, calculated by corneal thickness, in a time within hours compared to days in corneoscleral tissue. 

As limitations of the study, we note that the swelling protocol we used could cause loss of soluble protein and proteoglycans, affecting mechanical properties, possibly in a different manner among the different categories, and this effect was not considered. In addition, the drying protocol only relied on air/time and did not include other means (e.g., oven), thus it is conceivable that only the free water was removed from the tissue, leaving some intrafibrillar/interfibrillar water. Finally, many of the modeling studies that we used and compared were designed and performed with human corneas, and although some mechanical behavior trends and architecture are similar, they might have a different swelling behavior than our porcine cornea samples. 

In conclusion, in this work we present a quantitative model to calculate the stiffening effect of the CXL procedure by separating the photochemical cross-linking of the solid tissue network and tissue dehydration. Not considering tissue dehydration when measuring the effects of the different CXL procedures ex vivo leads to an overestimation of the stiffening effect. Calculating only the stiffening due to the photochemical cross-linking may better estimate how corneal tissue is affected in vivo, where hydration level changes initially after the CXL but eventually reach a basal steady state. Our results suggest that the hydration of corneal buttons is different to hydration of corneoscleral buttons, possibly due to different boundary conditions in water transport. This is an important behavior to consider when designing ex vivo experiments that involve highly hydrated conditions. 

Acknowledgments

Supported in part by National Institutes of Health (R01 EY028666, R01 EY032537), and an unrestricted departmental grant to the Cole Eye Institute, Cleveland Clinic, from Research to Prevent Blindness (New York, NY, USA). 

Author Contributions: GS and JNW conceived and designed the project. RRL performed the experiments and data analysis. ME and PK helped with experiments and discussion of results. WF, HZ, JBR, helped in the discussion of results. RRL and GS wrote the manuscript with input from all the other authors. 

Disclosure: R. Rodríguez-López, None; J.N. Webb, None; M. Erdi, None; P. Kofinas, None; W. Franco, None; H. Zhang, None; J.B. Randleman, None; G. Scarcelli, None