Mechanical outcome of accelerated corneal collagen cross-linking evaluated by Brillouin microscopy

Abstract

Purpose

To use Brillouin microscopy to quantify corneal mechanical changes induced by CXL procedures of different UV-A intensity and exposure time.

Settings

University of Maryland, College Park, MD

Design

Laboratory Study

Methods

Porcine cornea samples were debrided of epithelia and soaked with 0.1% riboflavin solution. Next, the samples were exposed to a clinically-standard 5.4 J/cm² of UV-A radiation with varying intensity and exposure time: 3 mW/cm² for 30 minutes, 9 mW/cm² for 10 minutes, 34 mW/cm² for 2.65 minutes, and 50 mW/cm² for 1.80 minutes. Using Brillouin microscopy, the Brillouin modulus for each cornea sample was computed as a function of radiation intensity/exposure time. For validation, using a compressive stress-strain test, the Young’s Modulus was found and compared for each irradiation condition.

Results

The standard 3 mW/cm² irradiance condition produced a significantly larger increase in corneal Brillouin modulus compared to the 9 (p ≤ 0.05), 34 (p ≤ 0.01), and 50 mW/cm² conditions (p ≤ 0.01). Depth-analysis revealed similar anterior sections of the standard and 9 mW/cm² conditions, but significantly less stiffening in the central and posterior of the 9 mW/cm² sample than the standard. The stiffening of the standard protocol was significantly larger (p ≤ 0.01) in all sections of the 34 and 50 mW/cm² conditions. For all samples, the overall change in Brillouin-derived Brillouin modulus correlated to the increase in Young’s Modulus (R² = 0.98).

Conclusions

Although the UV-A light dose was kept constant, accelerating the irradiation process decreased the effectiveness of CXL stiffening. Specifically, Brillouin analysis reveals that accelerated protocols are especially ineffective in the deeper portions of the cornea.

Introduction

Corneal ectasia, due to progressive keratoconus or refractive surgery, can result in progressive loss of vision and the need for corneal transplantation^1,2. Clinically observed alterations in corneal morphology, a distinguishing characteristic of ectasia, are believed to be the consequence of a non-uniform decrease in stiffness of the corneal stroma^3–6. To combat ectasia, the US Food and Drug Administration has recently approved corneal collagen cross-linking (CXL). The accepted CXL procedure involves the debridement of the corneal epithelium followed by a 30-minute application of riboflavin solution (0.1% riboflavin, 20% dextran) and an additional 30-minutes of UV-A exposure (3 mW/cm²; 5.4 J/cm²). The photochemical reaction between the riboflavin photosensitizer and UV light has been shown to increase the covalent bonding within the stroma, thereby, increasing the overall stiffness of the cornea and halting ectasia^7,8.

Recently, a great deal of effort has focused on decreasing the overall treatment time of the CXL procedure. It was proposed that treatment time could be shortened by increasing the radiation intensity, as biological effects should only depend on the total energy dose^8,9. While protocols featuring intensities up to 50 mW/cm² for under 2 minutes of exposure time were suggested⁹, evidence is not conclusive in supporting accelerated cross-linking for clinical settings. Wernli et al. reported a dramatic reduction in CXL-stiffening when the procedure was performed above a 45 mW/cm² intensity¹⁰. Hammer et al. showed a decrease in CXL effectiveness when comparing the standard protocol to that of just a 9 mW/cm² radiation intensity¹¹. Additionally, all previous mechanical studies observed the stiffness changes as a function of irradiance regime without spatial resolution^10–12. To assess the spatially-varying effects of accelerated CXL, indirect techniques have been used such as fluorescence imaging to assess riboflavin penetration¹¹ or Optical Coherence Tomography to quantify the depth location of refractive index changes within the stroma (i.e. the demarcation line)^9,13. Here, we used recently-developed Brillouin microscopy, which can directly assess corneal mechanics with three-dimensional resolution, to observe the depth-dependence of stiffening following accelerated CXL protocols. As expected, we found the standard 3 mW/cm² radiation intensity for 30 minutes of exposure time produced the largest stiffness increase. Depth-analysis revealed that the 9 mW/cm² intensity condition could match the stiffening of the standard protocol in the anterior stroma, but produced lower stiffening in the central and posterior sections.

Methods

Corneal Cross-Linking (CXL)

Fresh porcine eyes were obtained from a local slaughterhouse (Frederick, Maryland). For all eyes, the epithelium was carefully removed by scraping with a razor blade. The eyes were then dissected in order to punch two 5 mm corneal disc samples (Integra Miltex ® Disposable Biopsy Punch). Both disc samples were treated with one drop of 0.1% riboflavin/20% Dextran solution every 3 minutes for 30 minutes. After undergoing identical procedures, one of the two samples was set aside as the control while the other sample was exposed to UV radiation. In all experimental settings, a constant UV-A energy of 5.4 J/cm² was provided by a high-power UV Curing LED System (Thorlabs, Newton, NJ). We performed four treatment regimens, featuring 3 [n = 4], 9 [n = 6], 34 [n = 6], and 50 [n = 6] mW/cm² radiation intensity at 30, 10, 2.65, and 1.80 minutes of illumination time respectively, as summarized in Table 1.

Table 1.

Corneal Cross-Linking Treatment Regime

Treatment Group	Number of Eyes	Irradiation Energy (J/cm2)	Irradiation Intensity (mW/cm2)	Irradiation Time (min)
1	4	5.4	3	30
2	6	5.4	9	10
3	6	5.4	34	2.65
4	6	5.4	50	1.80

Brillouin Microscopy

Following the CXL procedure, both the control and cross-linked corneal discs were imaged via Brillouin microscopy with setup and procedures described previously^16–19. Briefly, the confocal Brillouin microscope utilized a 532 nm laser with an optical power of 10 mW. Light was focused into the sample by a 20× objective lens with numerical aperture of 0.4 (Olympus) with transverse resolution of ~1 μm and depth resolution of ~4 μm. The scattered light, collected through the same objective, was coupled into a single mode fiber and delivered to a two-stage VIPA spectrometer featuring an EMCCD camera (Andor, IXon Du-897). Each Brillouin spectrum was acquired in 0.2 seconds. To quantify the Brillouin shift at each sample location, raw spectra from the camera were fitted using a Lorentzian function and calibrated using the known frequency shifts of water and glass.

From the Brillouin frequency shift, the local mechanical properties of the cornea can be estimated using the following relationship:

$$M^{\prime }=\frac{\rho {{\nu }_B}^2{{\lambda }_i}^2}{4n^2}$$

where M’ is the longitudinal elastic modulus (we will refer to it as Brillouin modulus), $v_B$ is the measured Brillouin frequency shift, $n$ is the refractive index of the material, ${\lambda }_i$ is the wavelength of the incident photons, and $\rho$ is the density of the material. The spatially varying ratio of $\rho /n^2$ was approximated to the constant value of 0.57 g/cm³ based on literature values^18,22; we estimate this to introduce a 0.3% uncertainty throughout the cornea^24,25.

Brillouin Image Analysis

The corneal samples were set next to each other on the Brillouin microscope and imaged within the same acquisition run. Therefore, each scan imaged the frequency shift of both the control and cross-linked sample as a function of depth. For each scan, a depth cross-section (XZ) was collected, producing a 1000 μm (lateral) × 1400 μm (axial) image of Brillouin shift. A central sliver of each corneal cross-section was chosen for consistent post-processing analysis. In addition to the entire cornea sample, the cross-sections were divided into three equal segments (anterior, central, and posterior) for depth analysis. For such analysis, the depth of cornea (d) was normalized for each set of corneal samples in order to compare the Brillouin modulus of each sample in the anterior (0 < d ≤ 0.33), central (0.33 < d ≤ 0.66), and posterior (0.66 < d ≤ 1).

Compressive Biomechanical Testing

All samples were measured with compressive mechanical testing immediately after Brillouin imaging using a home-built compressive stress-strain instrument. The instrument consisted of a metallic base-plate, topped with finely gritted sandpaper to prevent unwanted slippage and a downward-moving plunger containing a force-measuring loading cell (Futek, Irving, CA) which is controlled via a motorized translational stage (Zaber, Vancouver, BC). Prior to every sample, the plunger was systematically moved downward until a reaction force from the base-plate was detected. The recorded plunger position was then recorded and later used to calculated the total thickness of the sample.

The corneal sample was placed on the bottom plate of the instrument and the plunger compressed the sample at a constant downward rate of 10 μm/s. Using a custom designed LabView software program (National Instruments, Austin, TX), the increasing compression force from the plunger and the corresponding material displacement were measured to produce the stress-strain curve of the material. The plunger position at which a reaction force was first sensed was noted in conjunction with the previously recorded base-plate position to accurately quantify the thickness of each sample. To obtain the Young’s Modulus of a sample, we plotted respective Stress (Force/Area) vs Strain (Displacement/Thickness) graphs and quantified the slope of the linear segment of the curve following the sleek strain. Using the Stress vs Strain curve for each sample, the Young’s Modulus was reported by fitting the tangent line at 15% strain. A consistent 15% strain was chosen to observe the linear elastic behavior of the biological tissue as demonstrated by Wernli et al.¹⁰

Statistical Analysis

First, to characterize individual CXL protocols, a paired, two tailed t-test was performed by comparing CXL samples with their respective controls cut from the same eye. Then, to compare different irradiation conditions, the stiffening effects of CXL protocols were analyzed by comparing the percent difference of both Brillouin and Young’s Moduli. The percent difference for both the Brillouin modulus and Young’s Modulus were found per eye using the following equation: Modulus (X) Percent Difference = 100(X_CXL − X_Control)/(X_Control). The respective moduli percent differences for the samples in each irradiation group were averaged ± standard error of the mean. A Wilcoxon Rank-Sum Test was used to assess the significance in comparisons between samples and/or corneal sections. P values less than or equal to 0.05 were considered statistically significant.

Results

First, to assess the stiffening effects of each CXL procedure, the Young’s modulus of the CXL samples were compared to their respective controls punched from the same porcine eye. This comparison showed that all CXL protocols produced a statistically significant (p ≤ 0.05) increase in stiffness compared to non-irradiated control conditions. Then, we compared the mechanical outcome of the varying CXL procedures by quantifying the percent change in Young’s Modulus with respect to each control. Figure 1 depicts the average percent change in Young’s Modulus ± standard error of the mean for the four irradiation conditions. The standard protocol, 3 mW/cm² of UV-A power for 30 minutes, was significantly more effective than the 34 (p ≤ 0.05) and 50 mW/cm² (p ≤ 0.01), the 9 mW/cm² condition was statistically significantly higher than the 50 mW/cm² (p ≤ 0.01), and there was no significant difference between the 34 and 50 mW/cm² conditions.

Figure 1.

The averaged percent change in Young’s Modulus for each irradiation condition from 3 to 50 mW/cm². As the intensity increases, the change in Young’s Modulus decreases. The error bars represent the standard error of the mean for each condition.

To compare with the compressive mechanical test, we calculated the Brillouin modulus from the Brillouin frequency shift and averaged the value over the whole sample. First, as for the traditional modulus analysis, the Brillouin modulus of CXL samples were compared to their respective non-irradiated controls. All CXL protocols produced a statistically significant (p ≤ 0.05) increase in corneal stiffness compared to their control. Figure 2 shows the stiffness outcome of the CXL protocols as measured by Brillouin microscopy. Figure 2a depicts representative cross-sectional Brillouin images of corneas from each irradiation condition with the color encoding the Brillouin frequency shift at each location. We then computed the percent change of Brillouin modulus for each irradiation condition. Figure 2b summarizes the stiffening as a function of radiation intensity. The 3 mW/cm² CXL condition resulted in higher corneal stiffening than the 9 mW/cm² (p ≤ 0.05), 34mW/cm² (p ≤ 0.01), and 50 mW/cm² (p ≤ 0.01), the 9mW/cm² had significantly higher stiffening than the 34mW/cm² (p ≤ 0.01) and 50 mW/cm² (p ≤ 0.01), and the 34 and 50 mW/cm² irradiation conditions did not significantly differ in CXL-induced stiffening.

Figure 2.

a- A representative image (700 μm × 100 μm), produced via MATLAB software (color map: jet), of the Brillouin shifts for each condition: (left to right) 3, 9, 34, and 50 mW/cm². Due to the relationship between shift and Brillouin modulus, a higher Brillouin shift correlates to a higher Brillouin modulus. The corneal slices are positioned top down from the anterior to the posterior.

b- The averaged percent change in Brillouin modulus for each irradiation condition from 3 to 50 mW/cm². As the intensity increases, the overall change in Brillouin modulus decreases. The error bars represent the standard error of the mean for each condition.

Figure 3 shows the relationship between Young’s Modulus and the average Brillouin modulus of the cornea sample. Each of the four points in Figure 3 represents the average of all the samples at a given irradiation condition. As radiation intensity decreased, both the Young’s Modulus percent change and the Brillouin modulus percent change increased at a similar rate. The similar rate of increase resulted in a highly statistically significant linear correlation between the percent change of Young’s Modulus and the percent change of Brillouin modulus with R²=0.985.

Figure 3.

The correlation between percent difference of mechanically yielded Young’s Modulus and Brillouin microscopy derived Brillouin modulus. The corneas were exposed to a constant energy dose of 5.4 J/cm² at a variety of light intensities and exposure times: (from left to right) 50mW/cm² for 1.80 minutes, 34mW/cm² for 2.65 minutes, 9mW/cm2 for 10 minutes, 3mW/cm² for 30 minutes. The correlation yielded a line of best fit of y = 14.82x + 2.605, with a correlation coefficient of R² = 0.9849. The error bars represent the standard error of the mean for each condition.

Finally, we used Brillouin microscopy to perform a depth-dependent analysis on the corneas. We calculated the percent change in Brillouin modulus in anterior, central and posterior sections of the cornea. Figure 4 shows the percent change in Brillouin modulus for each irradiance condition in the anterior, central, and posterior of the cornea. Comparing the stiffening of the 3 mW/cm² and 9 mW/cm² conditions, there was no statistical significance in the anterior section. However, the difference between the two conditions was statistically significant (p ≤ 0.05) in the central and posterior sections of the corneas. In all sections, the 3 mW/cm² produced significantly more stiffening (p ≤ 0.01) than the 34 and 50 mW/cm² conditions. The stiffening from the 9 mW/cm² sample significantly differed from that of the 34 mW/cm² sample in the anterior (p ≤ 0.01) and the 50 mW/cm² sample in the anterior (p ≤ 0.01) and central (p ≤ 0.05). There were no significant differences in stiffening at any section between the two most accelerated conditions.

Figure 4.

The Brillouin modulus at a constant energy dose of 5.4 J/cm² as a function of radiation intensities and exposure times ((green) 3 mW/cm2 for 30 minutes, (blue) 9 mW/cm2 for 10 minutes, (purple) 34 mW/cm² for 2.65 minutes, and (yellow) 50 mW/cm² for 1.80 minutes) as well as corneal section (anterior, central, and posterior). The error bars represent the standard error of the mean for each condition.

Discussion

In this study, we evaluated the stiffening effects of accelerated corneal cross-linking protocols. To minimize the control-to-sample variability, two discs were punched from each cornea to act as the non-irradiated control and cross-linked sample respectively. Keeping a constant energy dose of 5.4 J/cm², we performed four corneal cross-linking regimens with varying power and exposure time. To first validate our cross-linking procedures, we compared the resulting stiffness of our cross-linked samples to the respective control samples via stress-strain testing and Brillouin modulus derivation. The two techniques consistently showed a significant stiffening effect of CXL for all conditions.

Next, we compared the standard protocol, 3 mW/cm² of UV-A radiation intensity for 30 minutes of exposure time, to different accelerated protocols. We first used a commonly accepted compression test to compare the stiffening effects of each regime. The standard CXL protocol, 3mW/cm² of UV-A radiation for 30 minutes of exposure time, produced a significantly higher percent difference in Young’s Modulus than the 34 and 50 mW/cm² conditions. The 9 mW/cm² condition yielded a significantly higher stiffening effect than the most accelerated protocol. Thus, while the Bunsen-Roscoe law predicts that the crosslinking effects should be similar at a constant energy, our mechanical analysis yielded a radiation intensity/exposure time dependence of stiffening. This is in agreement with previous papers that analyzed accelerated CXL stiffness with compressive mechanical testing. The observed increases in Young’s Modulus at each irradiance condition, varying from 83% to 18% as a function of radiation power, all fall between the reported values of Hammer et al. and Wernli et al. The consistency between our values and that found in similarly conducted literature furthered validated our results. However, the studies presented notable differences in the trend between irradiance condition and stiffening effect. Hammer et al. reported a significant decrease in stiffening when the standard 3 mW/cm² irradiation condition was compared to 9 and 18 mW/cm² accelerated protocols¹¹. Wernli et al. reported a statistically significant difference in comparing cross-linked samples to their respective controls up until a maximum radiation intensity of 45 mW/cm² for 2 minutes of exposure time. However, unlike our findings and that of Hammer et al., the study demonstrated a relatively constant effect of cross-linking from 3 mW/cm² through roughly 45 mW/cm² before experiencing a significant decrease in stiffening magnitude¹⁰. The differences in results could depend on the variability in experimental procedures. For example, Wernli et al. kept the corneas immersed in a pool of riboflavin solution for 30 minutes prior to UV-A exposure. Our protocol more closely resembled that of Hammer et al. in that riboflavin drops were incrementally applied to the cornea for a total of 30 minutes prior to radiation exposure. Differing from Hammer et al, our protocol used corneal punches rather than the entire globe. This distinction, by minimizing the solution run-off due to a decreased sample curvature, may be responsible for our greater stiffening effects.

The results of the stress-strain testing were also used to validate Brillouin microscopy. We have previously determined a log-log linear correlation between Young’s Modulus and Brillouin-derived Brillouin modulus¹⁴. Therefore, our linear correlation between the percent change in Young’s Modulus and Brillouin modulus agrees with the previously found relationship.

Using Brillouin microscopy, we were able to uniquely expand on our findings as well as those of relevant studies by observing the effects of CXL as a function of depth. This analysis yielded particularly interesting results when comparing the standard protocol (3 mW/cm² for 30 minutes) to the slightly accelerated protocol (9mW/cm² for 10 minutes). When the entire cornea was analyzed, the standard condition displayed a significantly higher Brillouin modulus than the 9 mW/cm² protocol. From depth-dependent analysis, the two treatments did not show statistically significant differences in anterior stroma, but they showed statistically significant differences in the central and posterior portions of the cornea. This suggests the difference between the two conditions was primarily due to the deeper sections. Our findings on depth-dependent stiffening via CXL are in agreement with those of Aldahlawi et al.²⁶, who used the resistance to enzymatic degradation properties of cross-linked corneas to test the effective depth of the procedure. The study similarly showed, up until a maximum radiation power of 18 mW/cm² for 5 minutes of exposure time, that accelerating the cross-linking procedure has little effect on the anterior of the corneal stroma but reduces the effective depth of CXL.

It has been suggested that oxygen is a key limiting factor in the CXL process, as CXL had no significant stiffening results when conducted in a low-oxygen environment⁷. Moreover, oxygen is expected to be consumed in the CXL process due to its transformation into the reactive species that catalyze the covalent bonding of collagen and other matrix proteins in the stroma¹⁵. Thus, at high radiation intensity, the rate of oxygen depletion could exceed the rate of oxygen replenishment via diffusion. In particular, the available oxygen concentration is a depth-dependent quantity as it is increasingly difficult for the oxygen to diffuse deeper into the cornea. Therefore, the difference between oxygen depletion and replenishment correlates to both depth and radiation intensity. As a result, when comparing 3 mW/cm² and 9 mW/cm² conditions for example, it is expected that the anterior portions of the corneas would have similar results while larger stiffening differences should be observed deeper into the corneas. This phenomenon would explain the lower stiffening we observed at the deeper sections of the accelerated cross-linked corneas. However, this effect is primarily seen in the 9 mW/cm² condition as the 34 and 50 mW/cm² conditions significantly lack stiffening at all three sections of the stroma when compared to the standard protocol. The reported anterior lack in stiffening of the most accelerated protocols may be an effect of the measurement technique used. Brittingham et al., when comparing the standard 3 mW/cm² and 9 mW/cm² conditions, demonstrated a significant decrease in depth of the demarcation line following accelerated protocols²⁷. Therefore, it is possible that for our highest accelerated conditions, the most effected depth of the CXL is too shallow to significantly detect using our Brillouin microscopy parameters. However, it is worth noting that a similar threshold of stiffening efficiency was also observed from Wernli et al. It is possible that when the procedure is performed at such a rapid rate, oxygen is unable to diffuse back into the cornea to a significant depth. Past a threshold of a sufficiently oxygenated stroma, we would expect to observe very little stiffening effects of CXL due to the low oxygen environment as reported by Richoz et al.⁷

In conclusion, our study has confirmed the sub-optimal effects of accelerated cross-linking compared to the standard 3 mW/cm² condition. Furthermore, by employing the depth-dependent analysis of Brillouin microscopy, we contribute the lack of CXL stiffening in deeper sections of the cornea as an important difference when comparing the standard to accelerated regimes.

What was known

Accelerated corneal collagen cross-linking, when compared to the standard 3 mW/cm² intensity for 30 minutes regime, provides suboptimal results regarding the stiffening effects on the entire cornea.

What the paper adds

Brillouin microscopy allowed for a depth-dependent analysis following corneal collagen cross-linking. The confirmed decrease in effectiveness of accelerated cross-linking is primarily due to the lack of stiffening deeper in the cornea when compared to the standard protocol.