Corneal Resistance to Shear Force After UVA-Riboflavin Cross-Linking

Abstract

Purpose.: We evaluated whether UVA-riboflavin collagen cross-linking (CXL) increases transverse stromal shear moduli ex vivo, whether the shear moduli are greater in the anterior compared to the posterior stroma, and whether the shear moduli are affected by CXL.

Methods.: The resistance to unidirectional transverse shear of human (n = 18) and porcine (n = 42) corneas was measured in a custom engineered biaxial biomechanical setup at different hydrations. The corneas were separated into untreated, riboflavin solution–treated, and CXL–treated groups. The depth dependency of shear moduli within groups was assessed in femtosecond laser cut sheets. Dry weights were obtained for solids correction.

Results.: In porcine full-thickness buttons and 300 μm anterior sheets, a significantly increased unidirectional transverse shear modulus was detected in riboflavin-treated and CXL-treated groups compared to the respective untreated groups. There was no significant difference in shear modulus between riboflavin- and CXL-treated groups. In all groups, the shear moduli were greater in the anterior sheets compared to posterior sheets. Similar results were detected in human corneas.

Conclusions.: A method for unidirectional transverse shear resistance measurement was developed. The shear moduli were greater in the anterior compared to the posterior sheets. Increase in shear moduli was observed in the riboflavin and CXL groups compared to the untreated group, indicating that the immediate effects of the riboflavin or CXL treatment may be due partly to ground substance/dextran-5-phosphate interaction.

Introduction

The corneal stroma is a very dense matrix composed of several hundred sheets of highly organized collagen lamellae, stacked on each other preferably in parallel with respect to the corneal surface. Within a given lamella, the majority of fibrils lie parallel to each other and to the corneal surface, but at large angles with those in adjacent lamella. Overall, they are found at all angles within the corneal plane. Anterior and midstromal lamellae bifurcate and interweave, ¹ whereas posterior lamellae tend to lie noninterwoven in stacked layers. ² The collagen fibrils in the central cornea have preferred directions in the inferior-superior and nasal-temporal meridians in the posterior two-thirds, whereas the anterior third is arranged more isotropically. ^3,4 This microstructure of the corneal stroma suggests that its elasticity will depend on direction during mechanical testing (anisotropy) and that the in-plane properties will be different from the transverse properties. ^5–7 The cornea shears readily in the tangential plane without separating, even at high shear deformations. ^8,9 Thus, the stroma possesses self-cohesion between corneal lamellae, ^9–12 which may be attributed to lamellar interweaving, which varies with depth, and appears maximal at the anterior surface and reduces posteriorly. ^11,13–16 The corneal stroma has an innate tendency to imbibe fluid and swell. When the stroma swells, it loses its transparency ^17–20 because of increased light scattering. ²¹ Therefore, the physiologic control of swelling is crucial to the maintenance of transparency. ²² This swelling tendency and the transparency in vivo are unusual properties for a connective tissue. ^23,24 The leak of aqueous fluid into the stroma is driven by the stromal swelling pressure and is controlled by the ionic permeability of the endothelium. ²⁵ Stromal hydration has an important role in corneal biomechanics, ^26,27 and, therefore, affects the transversal shear properties of the stroma. ^28,29  

Biomechanical properties of the cornea are essential in relation to understanding pathogenesis of corneal disease and predicting outcome of refractive procedures. Previously, measurement techniques to characterize the biomechanical properties of the cornea almost exclusively have comprised ex vivo tensile strip extensometry and inflation tests. ^{6,26,27,30–37} However, these modalities do not introduce shearing deformations and, therefore, do not characterize shear stiffness. Few attempts have been made in the literature to quantify the magnitude of transverse stromal shear. ^7,9,28 In a recent work by Petsche et al., shear throughout the stromal depth of the human cornea was measured by torsion rheometry. ⁷ They found the transverse shear modulus to be depth-dependent and its magnitude to be greatest in the anterior third stroma. 

In our study, resistance to unidirectional tangential shear force is measured in the corneal stroma as a function of axial compression. Further, the stiffening effect of UVA-riboflavin collagen cross-linking (CXL) treatment is evaluated in terms of whether the treatment increases shear modulus in porcine and human corneas ex vivo. 

Materials and Methods

Riboflavin Solution

A solution of 0.1% riboflavin-5-phosphate was prepared as described previously. ³⁸  

Ambient Light Measurement

The riboflavin-only group and the untreated group were prepared under dim light conditions to avoid effects of incident light. ³⁹  

Preparation of Full-Thickness Porcine Corneal Buttons

The research was conducted in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. We divided 24 fresh porcine eyes into three groups: Untreated (n = 8), riboflavin-treated (n = 8), and CXL-treated (n = 8), and prepared the specimens as full-thickness buttons. The eyes were obtained from the local abattoir (Danish Crown, Horsens, Denmark) on the day of experiments and used within two hours. One eye from each pig was used. The central corneal thickness (CCT) was measured in the intact eye ex vivo with ultrasound pachymetry (Sonomed Pacscan 300P; Sonomed, Inc., Manhasset Hills, NY). After a brief application of 96% volume alcohol, the corneal epithelium was removed with a Beaver knife, and the CCT was measured. Two drops of 0.1% riboflavin-5-phosphate solution were applied to the denuded, anterior stromal surface every 5 minutes for 30 minutes. In the riboflavin group (n = 8), the riboflavin application was continued another 30 minutes under dim conditions. In the CXL group (n = 8), the denuded surface was irradiated with an UV-A lamp (UV-X; PESCHKE Meditrade GmbH, Huenenberg, Switzerland) at 370 nm with an intensity of 3 mW/cm² (spot size 11 mm) for 30 minutes, giving a total dose of 5.4 J/cm². The radiation source was calibrated with the UV-meter (Lutron YK-34UV; PESCHKE Meditrade GmbH), supplied by the UV-lamp manufacturer. During the 30-minute radiation period, riboflavin application was continued. After cross-linking or riboflavin application alone, 8 mm buttons were trephined and the endothelium was removed with a Beaver knife. The stromal buttons then were weighed using an Ohaus Discovery DV215CD digital scale (Ohaus Europe GmbH, Nänikon, Switzerland). The weighing was repeated after two hours of preswelling in isotonic saline at room temperature and again immediately after biomechanical testing after gentle blotting. Finally, the dry weights of the specimens were obtained. 

Porcine Corneal Sheets

In 21 fresh porcine eyes, 300 μm thick (ø = 8 mm), anterior and posterior sheets were cut with a ZEISS VisuMax femtosecond laser (FSL; Carl Zeiss Meditech, Jena, Germany), as reported previously. ³⁹ Leaving the sheet in situ, the specimen was reverted and an anterior sheet was cut (300 μm, ø = 8 mm). The laser raster pattern left sufficient bridged tissue across the cutting plane to allow the lamellae sheets to stay attached. Before removing the sheets, the corneas were divided into three groups and either riboflavin-treated (n = 7), CXL-treated (n = 7), or left untreated (n = 7) as in the full-thickness groups. With both lamellae sheets still attached, two drops of 0.1% riboflavin-5-phosphate solution were applied to the denuded, anterior stromal surface every 5 minutes for 30 minutes in the riboflavin and CXL groups, and further every 5 minutes for 30 minutes with UV-A radiation (CXL group) or under dim light conditions (riboflavin group). The sheets (anterior or posterior) to be tested last biomechanically were kept in a moist chamber until preswelled and tested (approximately 4 hours from detachment). Finally, the dry weights of the specimens were determined as above. 

Human Donor Corneas

A total of 18 human corneas from 12 different donors was obtained from the Minnesota Lions Eye Bank (Saint Paul, MN). The tissue was procured and processed in compliance with the Eye Bank Association of America Medical Standards and government regulations, and according to the tenets of the Declaration of Helsinki. The corneas with scleral rings were stored and transported in Optisol GS (Bausch and Lomb, Inc., Rochester, NY). The tissue was used for testing within four weeks from preparation. To dehydrate the corneas before the experiments, the human corneas were placed 24 hours at 5°C in a MEM-based storage medium. ^38,39  

Human Corneal Buttons

Nine human corneal buttons were prepared using the same procedure as for the porcine buttons. The corneas were divided equally into three groups: Full-thickness buttons, full-thickness buttons treated with riboflavin, and full-thickness buttons cross-linked with riboflavin and UV-A exposure. To standardize thicknesses, an initial in vivo stromal thickness of 540 μm was assumed for calculating axial compression levels. Dry weights were obtained for solids correction. 

Human Corneal Sheets

Sheets from the riboflavin and CXL groups were prepared as in the porcine groups above. In the FSL procedure, nine human corneas were divided equally into anterior and posterior parts. The targeted cut depth was calculated from the ultrasound pachymetry reading after epithelial removal divided by two. The FSL cut segment diameter was 8.0 mm. Finally, a 6.5 mm button was punch-trephined through both lamellae sheets of the 8.0 mm button. 

BioTester

A biaxial load cell–based system was custom engineered for swelling pressure and unidirectional shear force measurement (BioTester 5000; CellScale Biomaterials Testing, Waterloo, Canada). This design allowed simultaneous recording of x- and y-axis displacement, force, and time. The semiconductor strain gauge-based load cells were sampled at 100 Hz with hardware 8 sample averaging to reduce noise. A stepper motor under open loop control drove the actuators. Synchronously, a CCD-camera (an integrated part of the BioTester 5000; CellScale Biomaterials Testing) recorded images at 1280 × 960 pixels for documentation. Customized software (LabJoy version 9.05; CellScale Biomaterials Testing) showed real time plots of force, displacement, and time. 

Fixation Unit

The excised corneal button or FSL-cut sheet was fixed vertically by compression between two flat, porous plates to permit bulk diffusion of water to the corneal surfaces and to keep the corneal specimen in place (Fig. 1). The connection arms were made of spring steel plates, so the system had no moving bearings and, thereby, no friction or other sources of hysteresis. The plates were stiff in the vertical direction to carry the weight of the brackets, stiff in the loading direction to carry the load from the actuator to the specimen, and flexible in the transverse direction so as not to contribute significantly to loads in this direction. In calibration this load was subtracted. 

Figure 1

 

Schematic drawing of the customized BioTester fixation unit (top view).

The fixation unit was submerged into a heated isotonic saline bath (9 mg/mL sodium chloride solution; B. Braun Melsungen AG, Melsungen, Germany) at 37°C for 30 minutes before measurements to ensure full hydration and correct temperature of the plates. 

Software and Data Acquisition

Data were recorded in LabJoy version 9.05, provided by the BioTester manufacturer (CellScale Biomaterials Testing). A distance and force calibration routine was performed before measurements of every specimen. Calculations were made in Microsoft Excel 2011 (Microsoft Corp., Redmont, WA) and Prism 5 (GraphPad Software, Inc., La Jolla, CA). 

Mechanical Testing

The shear modulus measures the resistance of the tissue to shearing strains. The layered arrangement of the fibrils and limited number of naturally occurring cross-links suggest an anisotropic biomechanical behavior and a low resistance to shearing loads. Thus, the tangential properties (x-plane, Fig. 2) are expected to be different from the transverse properties (y-plane, Fig. 2). In the full-thickness corneal buttons and in the posterior porcine FSL-cut sheets, the swelling force (y-axis, Fig. 2) and shear force (x-axis, Fig. 2) were recorded at 1 Hz at 5 compression steps: +30%, +20%, +10%, +0%, and −5% of initial stromal thickness as a function of time. In the anterior FSL-cut sheets and posterior human sheets, the compressions were +5%, −5%, −10%, −15%, and −20% of initial sheet thickness, respectively. Each compression step was fixed for 30 minutes allowing the swelling force to reach a plateau phase. For each plateau phase, four cycles of shear deformation were performed. The first three rounds of deformation served as preconditioning and the fourth round was used for analysis. The number of rounds of preconditioning, amount of shear strain, levels of axial compression, and velocity were chosen after pilot studies. 

Figure 2

 

Axis notation within a corneal button (left). Shear force, F from Δ x transverse displacement of a specimen with the area (A) and the thickness (l) (right).

The tangential shear modulus, G was calculated for each compression step (Equation 1). Linear regression parameters were calculated for the linear phase of the loading and unloading curve in the positive displacement intervals (Table 1). 

Table 1

 

Shear Strains and the Intervals Used for Analysis in the Respective Groups

	x-Displacement, μm	Linear Phase for Analysis, μm	Shear Rate, μm/s
Porcine full thickness	±600	0–300	50
Porcine sheets	±300	0–150	25
Human full thickness	±300	0–150	25
Human sheets	±200	0–100	17

In the test series, the force F is measured in the x-axis as a function of x-axis displacement, Δ x at different thicknesses (l; Fig. 2). The shear modulus, denoted G can be estimated as the ratio of shear stress to shear strain:  

 

where τ_xy = (F/A) = shear stress and F is the force that acts on the area (A).  

Statistics

Statistical analysis performed in Prism GraphPad (GraphPad Software, Inc.) and Microsoft Excel (Microsoft Corp.) included unpaired Student's t-test, and one- and two-way ANOVA. Logarithmic transformation of shear moduli was performed to homogenize variance. All results are reported as mean ± SEM or median (range), unless otherwise stated. In the human corneas, statistical analysis was omitted due to the small number of human corneas available. R ² was computed to assess the goodness of fit for the linear relationship between corneal thickness and shear modulus. To assess whether stromal hydration affects the shear modulus, as reported previously by Wollensak et al., ²⁸ the corresponding P values were calculated to evaluate whether the linear regression slopes were significantly nonzero. In biomechanical testing, the reported changes in stromal button or sheet thickness are changes caused by axial compression. To control for solids and hydration differences, including collagen content, shear moduli were divided by dry weight and the above analyses run to test whether any detected differences within or between groups were attributed to collagen content. 

Results

Porcine Corneas

Mean CCT in full-thickness porcine corneas after epithelial removal was 765.3 μm (SEM = 11.61 μm) with no significant difference between groups. Mean dry weight was 8.72 ± 0.51 mg, with no significant difference between groups (P = 0.47, one-way ANOVA). 

The mean CCT of the porcine corneas for subsequent FSL-cutting without epithelium was 751.6 μm (SEM = 13.36 μm) before cutting, with no significant difference between groups (P = 0.41, one-way ANOVA). 

The dry weights (Table 2) did not differ significantly between groups in the anterior (P = 0.27, one-way ANOVA) or posterior (P = 0.22, one-way ANOVA) porcine sheets. The dry weights of all the anterior sheets were significantly lower than that of all posterior sheets (P < 0.001, unpaired t-test). 

Table 2

 

Dry Weights of Porcine 300-μm Sheets

	Dry Weight, 10−3 g
Anterior untreated, n = 7	3.471 ± 0.047
Anterior riboflavin, n = 7	3.257 ± 0.127
Anterior CXL, n = 7	3.343 ± 0.078
Posterior untreated, n = 7	4.343 ± 0.113
Posterior riboflavin, n = 7	3.986 ± 0.211
Posterior CXL, n = 7	4.271 ± 0.084

Corneal hysteresis was observed in the loading and unloading curves for all specimens. Outside the quasilinear ranges, the shear stress increased steeply (Fig. 3). Therefore, the slope of the linear regression is an approximation, when mean shear modulus is calculated within the central assumed linear ranges (Table 1, column 2). 

Figure 3

 

Examples of shear force versus displacement measurements at different compressions (relative compression given in % of initial CCT for each plot) in a porcine full-thickness corneal button.

Untreated Porcine Corneas

In untreated full-thickness porcine corneas, median shear modulus at 5% axial compression was 0.34 kPa (0.040 kPa/mg, Table 3). Shear modulus was significantly correlated inversely to CCT (r ² = 0.11, P < 0.05, Fig. 4) and likewise when corrected for dry weight (r ² = 0.13, P < 0.05, Fig. 4). In anterior sheets shear modulus was independent of sheet thickness (r ² = 0.01, P = 0.82 ) and measured 0.24 kPa (0.062 kPa/mg) at 5% compression (Table 3). In contrast, posterior sheets showed significant increased shear modulus to decreased sheet thickness (r ² = 0.15, P < 0.05) and measured 0.06 kPa (0.013 kPa/mg) at 5% compression (Table 3). Likewise, when corrected for dry weight, the posterior sheets showed significant increased shear modulus to decreased sheet thickness (r ² = 0.15, P < 0.05) and measured 0.013 kPa (0.013 kPa/mg) at 5% compression (Table 3). In summary, all untreated porcine specimens, except anterior sheets, exhibited increased shear moduli with decreased thickness and also when corrected for dry weight. 

Figure 4

 

Full-thickness porcine corneal buttons. Shear modulus as a function of axial compression (given as relative CCT).

Table 3

 

Porcine Full-Thickness and FSL-Cut 300-μm Anterior and Posterior Sheets

	Median Shear Modulus at 5% Axial Compression, kPa	r 2 Shear vs. Thickness	Shear/Dry Weight, kPa/mg	r 2 Shear/Dry Weight vs. Thickness
Full thickness untreated, n = 7	0.34 (0.18–0.57)	0.11*	0.040	0.13*
Full thickness riboflavin, n = 7	0.79 (0.52–1.38)	0.26*	0.095	0.28*
Full thickness CXL, n = 7	1.04 (0.58–1.80)	0.30†	0.128	0.34†
Anterior untreated	0.24 (0.11–0.63)	0.01	0.062	<0.01
Anterior riboflavin	0.48 (0.35–0.71)	0.07	0.151	0.12
Anterior CXL	0.41 (0.05–1.09)	0.14*	0.116	0.09
Posterior untreated	0.06 (0.01–0.18)	0.15*	0.013	0.15*
Posterior riboflavin	0.22 (0.02–1.05)	0.31†	0.055	0.09
Posterior CXL	0.33 (0.06–2.23)	0.24*	0.078	0.10

 Median shear moduli at 5% compression and range. Shear/dry weight is median shear modulus at 5% axial compression divided by dry weight. R ² is calculated for shear versus thickness and shear/dry weight versus thickness, respectively.

*   P < 0.05.

†   P < 0.001.

Riboflavin- and CXL-Treated Porcine Corneas

In full-thickness porcine specimens, the median shear moduli at 5% axial compression were 0.79 kPa (0.095 kPa/mg) and 1.04 kPa (0.128 kPa/mg) in the riboflavin and CXL groups, respectively. Comparing the CXL versus the untreated group, a significant difference was detected (P < 0.05, two-way ANOVA) and likewise in the riboflavin versus untreated group (P < 0.05, two-way ANOVA). No significant difference between CXL and riboflavin groups was detected (P = 0.43, two-way ANOVA; Fig. 4, Table 3). Significantly increased shear moduli with decreased specimen thickness was detected in the untreated, riboflavin, and CXL groups, as determined by correlation coefficients being significantly nonzero (Table 3). When corrected for dry weight, by dividing shear moduli by dry weight, the correlation remained significant in all groups (Table 3). 

When corrected for dry weights to adjust for collagen content in the different depths of the corneal stroma, shear moduli at 5 % compression did not differ significantly between the full-thickness riboflavin and CXL groups (P = 0.30, two-way ANOVA), measuring 0.095 kPa/mg tissue and 0.128 kPa/mg tissue, respectively (Table 3). 

In the porcine anterior sheets, median shear moduli at 5% axial compression were 0.48 kPa (0.151 kPa/mg) and 0.41 kPa (0.116 kPa/mg) in the riboflavin and CXL groups, respectively (Table 3). Shear moduli in the anterior CXL group increased significantly with decreased sheet thickness (r ² = 0.14, P < 0.05). No significant difference in shear modulus was detected in the anterior porcine sheets between the three groups (two-way ANOVA). 

In the posterior porcine sheets, median shear moduli at 5% axial compression were 0.22 kPa (0.055 kPa/mg) and 0.33 kPa (0.078 kPa/mg) in the riboflavin and CXL group, respectively (Table 3). No significant differences in median shear moduli or shear moduli/dry weights were detected in the CXL versus riboflavin group (two-way ANOVA, Table 3). However, the untreated group showed a significant difference compared to the other groups (P < 0.05, two-way ANOVA, Fig. 5, Table 3). Shear moduli in all posterior groups increased significantly with decreased sheet thickness (Table 3). When corrected for collagen content by dividing median shear modulus by dry weight, this parameter was increased significantly at 5% axial compression in the CXL and riboflavin groups compared to the untreated group (P < 0.05, two-way ANOVA). 

Figure 5

 

Porcine 300-μm sheets. Shear moduli as a function of axial compression. Note the difference in compression in the anterior versus posterior sheets.

Human Corneas

Mean human donor age was 53.4 years (SD ± 14.4 years). Mean endothelial cell count at the time of preparation for storage was 2500 cells/mm² (SD ± 273.5 cells /mm²). At inspection at the day of experiments, the corneas had moderate diffuse edema. After 24 hours in the 8% dextran medium, the CCT at the day of experiments ranged from 564 to 850 μm (mean = 700 ± SD 76.4 μm). The targeted stromal thickness for the FSL laser cuts ranged from 240 to 340 μm (mean = 303 ± SD 31.5 μm). The natural stromal thickness in vivo of the full-thickness specimens was presumed to be 540 μm. 

Untreated Human Corneas

In full-thickness human corneal buttons, the median shear modulus at 0% axial compression was 2.00 kPa (0.477 kPa/mg). At 5% axial compression the shear modulus was 2.05 kPa (0.472 kPa/mg, Fig. 6, Table 4). The shear moduli appeared to increase with decreasing CCT (Fig. 7). The median shear moduli of the anterior and posterior sheets (Fig. 6), at 5% axial compression, are listed in Table 4. The moduli appeared increase with decreasing sheet thickness as in full-thickness specimens (Fig. 6). 

Figure 6

 

Human full-thickness buttons (540 μm). Shear modulus as a function of axial compression. Horizontal lines represent median values.

Figure 7

 

Human sheets. Shear modulus as a function of relative axial compression. Horizontal lines represent median values.

Table 4

 

Human Full-Thickness Buttons and Sheets

	Median Thickness, μm (Range)	Dry Weight, 10−3 g (Range)	Median Shear Modulus at 5% Axial Compression, kPa (Range)	Shear Modulus/Dry Weight, kPa/mg
Stromal
Full thickness untreated	691 (653–773)	4.3 (4.1–4.3)	2.05 (1.98–2.87)	0.472
Full thickness riboflavin	625 (537–648)	4.7 (4.7–4.8)	7.18 (4.31–7.35)	1.462
Full thickness CXL	636 (528–748)	4.4 (4.3–4.8)	5.00 (4.13–5.89)	1.171
	Median Targeted FSL-Cut Sheet Thickness, μm (Range)
FSL–cut sheet
Anterior untreated	300 (285–325)	1.9 (1.1–2.3)	2.93 (1.85–3.93)	1.292
Anterior riboflavin	325 (310–340)	2.1 (2.0–2.7)	5.00 (4.97–5.02)	2.000
Anterior CXL	275 (240–325)	2.3 (2.2–3.0)	4.86 (2.73–11.0)	2.750
Posterior untreated	300 (285–325)	1.7 (1.7–2.3)	0.64 (0.30–3.84)	0.337
Posterior riboflavin	325 (310–340)	2.5 (1.9–3.0)	6.59 (3.21–7.34)	2.671
Posterior CXL	275 (240–325)	2.0 (1.7–2.2)	3.35 (3.00–4.34)	1.703

 Median dry weight and range sheet and stromal thickness given as median and range. Sheet values appear twice, as the human corneas were divided equally into anterior and posterior sheets. Shear modulus/dry weight is shear modulus at 5% axial compression divided by dry weight.

Riboflavin- and CXL-Treated Human Corneas

In riboflavin-treated and CXL-treated full-thickness human corneal buttons, the median shear moduli at 5% axial compression were 7.18 kPa (1.462 kPa/mg) and 5.00 kPa (1.171 kPa/mg), respectively (Table 4). Both were increased 2- to 3-fold compared to the untreated group (Table 4). The shear moduli were significantly dependent on CCT (Fig. 7). 

In the human anterior and posterior sheets, the CXL and riboflavin groups had higher shear moduli at all axial compressions compared to the untreated groups (Fig. 6). The posterior riboflavin and CXL sheet groups demonstrated similar shear moduli at 5% axial compression (Table 4). Both groups had higher shear moduli compared to untreated sheets (Table 4). In anterior sheets, the shear modulus/dry weight was highest in the CXL group, and not in the riboflavin group as in the uncorrected shear parameter. In posterior sheets, the shear modulus/dry weight parameter was highest in the riboflavin group, as in the uncorrected shear modulus (Table 4). 

We noted that CXL-treated anterior sheets in human and porcine corneas returned to their original convex shape after 3 hours of compression in the BioTester (CellScale Biomaterials Testing) (Fig. 8). All other specimens remained flat after testing. 

Figure 8

 

A human CXL-treated FSL-cut anterior sheet. Immediately after 3 hours of testing, compressive force from the porous plates (white blocks) is released and the sheet returns immediately to its original convex shape. This phenomenon was observed in human and porcine CXL-treated anterior sheets. Untreated specimens and riboflavin only–treated specimens remained flat after compression.

Discussion

The corneal stroma consists of a three-dimensional network of collagen fibrils embedded in a viscous ground substance matrix. The complex orientations of the fibrils give rise to anisotropic responses. For this reason it is not sufficient to study the material behavior using a single uniaxial test, but rather employ in-plane biaxial stretching test or inflation experiments. This complex stromal structure, however, is not characterized fully biomechanically by in-plane testing. As cells interact mechanically, chemically, and electrically with other cells, they likewise interact with the extracellular matrix. Thus, there is a need to understand the mechanics of particularly adhesion molecules, that govern these interactions. ⁴⁰ The stromal ground substance may thereby have an important role in treatment procedures, such as the photochemical modulation in CXL. The cornea shears readily in the tangential plane, which agrees with the layered, lamellar corneal architecture. Its depth-dependent interweaving suggests further that resistance to shearing strains should be greatest anteriorly. Shear testing reveals transverse biomechanical properties not characterized by in-plane experiments. Further, the corneal shear properties may be more sensitive to the biomechanical effects of therapeutic procedures, such as CXL, in which the ground substance is altered chemically, even though this may be transient. The role of the stromal ground substance is important in tangential shear force loads, where tension is transferred between neighboring fibrils. ^29,36 The dextran of the Optisol GS medium (Bausch and Lomb, Inc.), and presumably from the riboflavin-5-phosphate solution, enter the stroma through its denuded anterior surface. ^41,42 This may alter the composition, and biochemical and biomechanical properties of the stroma. 

The detected shear modulus in untreated human full-thickness specimens (2.0 kPa at 0% axial compression) agreed with previous findings in the range of 2 kPa (Wollensak ²⁸ ) to 9.5 ± 2.9 kPa (Petsche ⁷ ) at 0% axial compression. These cited works used different measurement modalities (dynamic mechanical analysis ⁴³ and torsional rheometry, respectively). The deviations in torsional shear moduli in the work of Petsche et al. are considerable. In comparison, the present observations of deviations in uniaxial shear moduli demonstrated considerable variations as well. Allowing for part of the variation to be caused by the method, this indicates a biological variation in shear moduli, even when corrected for dry weight. The observed hysteresis in the force versus displacement curves (Fig. 3) agrees with findings in preliminary rabbit studies by Maurice, ²⁹ which is expected due to the viscoelastic properties of the stroma. ⁴⁰ Full-thickness porcine specimens exhibited approximately one-fifth less resistance to shear (G = 0.38 kPa) compared to human specimens at 5% axial compression. 

This difference in shear moduli between species might be attributed to structural differences of the stroma ⁴⁴ or the natural occurring cross-linking associated with age. ^45,46 In untreated human and untreated porcine sheets, the shear moduli were greater in the anterior compared to posterior sheets (Table 2). This was expected, as the interweaving of lamellae varies with depth, and appears maximal at the anterior surface and reduces posteriorly. ^1,11,13–15 These findings agreed with the recent study by Petsche et al. ⁷ In porcine corneas, the riboflavin/CXL modulus is approximately twice the physiologic value in the anterior samples, whereas in the posterior samples the ratio is approximately five times the physiologic value. Likewise in the human tests, the ratios are approximately 2 and 8. This is surprising since riboflavin is diffusing from the anterior surface. 

To evaluate whether stromal hydration (thickness) affects the shear modulus, the linear relationship between corneal thickness and shear modulus was assessed (Tables 3, 4). It was proposed by Maurice ⁹ that shear modulus should increase with increasing stromal hydration (thickness), as proposed cohesive structures within the stroma should be more stretched at higher hydrations. However, in the recent experimental study by Petsche et al., ⁷ stromal shear modulus was found to increase with decreasing thickness, which agree with our results in porcine and human samples. In the present study, this correlation appeared more pronounced in full-thickness samples in porcine and human corneas compared to FSL-cut sheets (steeper slope). It should be considered that thin samples with small values of transversal shear resistance possibly are more difficult to measure, which may affect the sensitivity of the method to detect small differences between groups. 

In the test protocol the tissue is swollen and then compressed in different steps. In this way, the risk of sliding of the sample in all tangential directions is minimized. In pilot studies and in a recent publication, ³⁹ swelling pressure measurements were not affected by stepwise compression versus stepwise decompression. This also was observed in the work of Olsen and Sperling. ⁴⁷ In regard to swelling pressure, an unconfined compression does proportionately reverse the swelling deformation. However, the hydration mechanics within the stroma is unknown and, therefore, the water content in different depths of the stroma may be altered by decompression followed by compression. Displacement of fluid then may affect tangential shear measurements. In pilot studies this tangential shear did not appear to be affected when compressing the specimen increasingly or with increasing decompression. Dry weights were measured to control for differences in hydration at different stromal depths to aid interpretation of the results. 

If we compare the present transverse measurements to in-plane experiments, the increased strengthening from CXL is considerably more pronounced in the latter. This indicates that the effect on the fibril properties is greater than on the ground substance to fibril effect or within the ground substance itself. The immediate effects of the riboflavin or CXL treatment may in be due partly to stromal extracellular matrix ground substance/dextran-5-phosphate interaction, as increase in shear moduli interestingly are observed in the riboflavin and CXL groups compared to the untreated group. To control for this, testing of shear moduli could be performed months after CXL or riboflavin treatment in porcine eyes, to ensure absence of short-term riboflavin solution effects. In a study by Zhang et al., the effects of CXL on the interaction of collagen and proteoglycan was assessed. ⁴⁸ In their study, the CXL procedure caused cross-linking of collagen molecules among themselves and proteoglycan core proteins among themselves, together with limited linkages between collagen and leratocan, lumican, mimecan, and decorin in the extracellular matrix. Effects of riboflavin application alone without UV-radiation, however, were not assessed. The interaction of the riboflavin-5-phosphate dextran solution with the stromal ground substance remains to be investigated. The precise mechanism of CXL at the molecular level has not yet been determined to our knowledge. However, CXL is thought to increase the degree of covalent bonding between and within the molecules of the extracellular matrix, such as collagen type I and proteoglycans. ⁴⁹ However, in vitro studies of crosslinking of collagen I hydrogels lacking proteoglycans have shown biomechanical strengthening and reduction of swelling tendency similar to findings in corneal experiments. ⁵⁰  

Fibril density varies with hydration, which implies that if a sheet was cut from an overhydrated sample, its collagen content would be expected to contain less collagen than a sheet of the same thickness cut from a dehydrated sample. As the displacement of fluid within the stroma post mortem to our knowledge is unknown, the dry weights are important in interpretation of the results. In porcine sheets, dry weights were greatest posteriorly, indicating a higher fibril content in the posterior 300 μm stroma compared to the anterior 300 μm. Although greater in dry weight, the posterior sheets were considerably less resistant to shearing strains, confirming the expected low shear resistance of a less interwoven stromal architecture. A similar trend was observed in human sheets, although differences in dry weight complicates interpretation of the shear moduli in relation to depth. To help interpretation, shear moduli at 5% compression were divided by dry weights in the respective groups. In human specimens, this showed increased values in riboflavin- or CXL-treated anterior sheets compared to posterior sheets. Further, comparing the anterior riboflavin group to the CXL group, a pronounced difference was observed in this parameter (Table 4). 

In full-thickness porcine buttons, an increased shear modulus–to-dry tissue ratio also was observed in the riboflavin-treated and CXL groups (Table 3). In all porcine anterior groups, the ratio was higher compared to posterior groups. Unexpectedly, the anterior CXL group ranged second to the riboflavin-only group (Table 3). 

We observed that only the cross-linked anterior sheets returned to their original convex shape immediately after shear testing and compression (Fig. 8). Apparently, the stromal shape is stabilized structurally after CXL as observed previously, ^51,52 although this cannot be measured in mechanical testing. Form stability might require only a slight increase of the internal cohesive structures. 

Methods Considerations

Distances.

The CCT was measured in the center of the cornea by ultrasound pachymetry as an estimate for the global cornea thickness, although the thickness varies over the corneal surface and increases toward the periphery. Ultrasound pachymetry readings depend on the level of stromal hydration due to changes in sound velocity and backscatter ⁵³ ; that is, an increase of 20% in CCT by overhydration will overestimate the CCT reading from ultrasound pachymetry by 7 μm. The CCT measured in the intact porcine eye within 2 hours was used as the normo-hydrated value, as the in vivo value was not measured. 

In the human donor corneas, edema was observed at the day of experiments. To correct for differences in stromal hydration at the day of experiments, the full-thickness human corneas all were assumed to be 540 μm in their normo-hydrated state in vivo. ⁵⁴ The porcine corneas may likewise swell ex vivo or dry during handling and dissection, although dissecting and preparing times were short. To correct for this, dry weights were obtained and used in the analysis. 

Force Measurements.

The corneal stroma is an anisotropic material, and will exhibit different properties when tested in different directions. In our study, linear shear strain was applied. However, the orientation of the specimen and, thereby, the fibril direction was not controlled for in the setup. At different hydrations in vitro, the directions of the stromal lamellae do not change significantly. ² Strain must be as uniform as possible throughout the specimen, so that homogeneity of deformation is maintained. For specimens with an aspect ratio of at least 1:4 (the corneal buttons were approximately 0.8 mm/8.0 mm = 1:10), homogeneity should be maintained for shear stresses up to the order of the shear modulus. ⁵⁵  

Although the fixation unit is made of a rigid material and the compression plates ideally are kept parallel, a thinner specimen increases the importance of plates being parallel. 

The present results obtained with the developed biaxial setup agreed with previously reported results of physiologic shear resistance by other investigators. Although other studies are limited in numbers and use other measuring methods, we considered the present method valid for quantifying the magnitude of unidirectional transverse shear modulus in the corneal stroma, and quantify changes in shear modulus induced by CXL. 

Conclusions

We developed a method for measuring unidirectional transverse shear resistance as a function of axial compression. The shear moduli were quantified in human and porcine stromal buttons, and depth dependency was assessed in femtosecond laser-cut sheets. The anterior stroma exhibited increased shear moduli compared to the posterior stroma in human and porcine corneas. Interestingly, an increase in shear moduli was observed in the riboflavin and CXL groups compared to the untreated group, indicating that the immediate effects of the riboflavin or CXL treatment may, in part, be due to ground substance/dextran-5-phosphate solution interaction. Considering shear moduli in relation to stromal dry mass, a more pronounced CXL effect was observed. From this, caution should be taken to conclude whether the increased resistance to shear force is due to CXL treatment or changes induced by the riboflavin solution alone. 

Acknowledgments

The authors thank Carl Christian Danielsen and Eva Mikkelsen at the Department of Anatomy at Aarhus University for their kind assistance in freeze-drying. 

Supported by The Danish Eye Health Society, Jørgen Bagenkop Nielsens Myopia Foundation, and Maskinfabrikant Jochum Jensen og hustru Mette Marie Jensen f. Poulsens Mindelegat (Fond). 

Disclosure: A.P. Søndergaard, Carl Zeiss Meditec AG (F); A. Ivarsen, None; J. Hjortdal, None