Biomechanical Modification of the Sclera: An Ex Vivo Study on Porcine Eyes

Abstract

Purpose

Scleral biomechanics are altered in various ocular pathologies. Modification of sclera biomechanics has been proposed as a potential treatment strategy. This study characterizes the biomechanical effect of collagenase (COL), glyceraldehyde (GAD), microbial transglutaminase (mTG), transglutaminase 1 (TG1), transglutaminase 2 (TG2), and lysyl oxidase (LOX) on ex vivo porcine sclera.

Methods

Tissue biomechanics were assessed by uniaxial tensile testing. Tangent modulus (E_T) was calculated from the stress–strain curve. The biomechanical response was analyzed based on dose–response curves. Locally applied treatments to the intraocular, extraocular, and combined intra- and extraocular scleral surface were compared with incubated tissue strips. Treatment safety was investigated on human adult retinal pigment epithelium cells (ARPE19) and mouse retinal ganglion cells.

Results

GAD 0.1 M and mTG 1 U/mL led to a 10.53-fold (P = 0.0011) and 4.71-fold (P = 0.0210) increase in E_T, respectively. COL 0.05 mg/mL decreased tissue stiffness by 2.33-fold (P < 0.0001). Incubation with TG1, TG2, and LOX did not lead to significant changes in tissue E_T. The effect of strip incubation was significantly higher for GAD (P < 0.0001) and mTG (P < 0.01) compared with local applications, with no quantitative difference for COL. Viability assays showed a relatively safe application of mTG and COL on retinal ganglion cells and ARPE19, but increased cytotoxicity at higher GAD concentrations.

Conclusions

COL, GAD, and mTG induced dose-responsive biomechanical changes in ex vivo scleral biomechanics with acceptable safety. Locally applied treatments showed reduced biomechanical impact compared with strip incubation. Further experiments need to confirm these findings in vivo and determine its role in diseased eyes.

Tissue biomechanics are a fundamental aspect of pathological tissue remodeling and new approaches for therapeutic interventions. Crosslinking (CXL) enhances the biomechanical properties of collagenous tissues by inducing covalent bonds between polymer chains.¹ The novel intracollagenous and intercollagenous crosslinks lead to subsequent stiffening of the tissue.

Scleral CXL has been suggested as a potential treatment approach to biomechanically strengthen the sclera in myopic and glaucomatous eyes and counteract disease progression.² Myopia affects more than 28% of the global population, with an increasing prevalence estimated to reach one-half of the global population by 2050.^3,4 Myopia-associated scleral remodeling may lead to posterior staphyloma⁵ and a decrease in scleral thickness with increasing axial length and disease severity.⁶ Glaucoma is worldwide the primary cause of irreversible blindness and will, according to current projections, globally affect 111.8 million people by 2040.⁷ Elevated IOP, the primary risk factor attributed to glaucoma, contributes to remodeling and stiffening of the peripapillary sclera (PPS).⁸ It has been hypothesized that the stiffer PPS provides neuroprotection for IOP-induced damage⁸ by reducing the biomechanical strain within the lamina cribrosa and optic nerve head (ONH).⁹

CXL may be performed light induced, chemically, or enzymatically. However, scleral CXL with UVA light has been reported to damage the retina and compromise its function.^10,11 A further challenge of light source–dependent scleral treatments lies in the limited accessibility of its anatomical location within the orbital cavity. Therefore, neuroprotective, light-independent treatment alternatives need to be considered regarding this application.

Glyceraldehyde (GAD) has been shown in previous studies to effectively stiffen scleral tissue,^12,13 although its retinal safety and translational applicability remain insufficiently characterized. Generally, the profile of endogenous enzymes is potentially attractive owing to their natural biocompatibility. Given that microbial transglutaminase (mTG) has shown a scleral stiffening effect in prior studies,^14,15 we now aim to evaluate whether two endogenous members of the transglutaminase enzyme family, transglutaminase 1 (TG1), and transglutaminase 2 (TG2), as well as the endogenous enzyme lysyl oxidase (LOX), may serve as similarly effective and potentially safer alternatives for clinical application.

Furthermore, extracellular matrix (ECM) digesting agents like collagenase (COL) have been studied in the scope of improving transscleral drug application.¹⁶ Therefore, we decided to investigate the effect of COL on biomechanical effects and safety applications to our study protocol.

The purpose of this study was to determine the biomechanical potential of ECM enzymes and reagents on ex vivo porcine sclera. COL, GAD, mTG, TG1, TG2, and LOX were tested and biomechanical scleral properties were assessed with uniaxial tensile testing. Locally applied treatments to the intraocular, extraocular, and combined intraocular/extraocular scleral surfaces were compared with tissue strip incubation to quantify and compare different treatment modalities. Safety and cytotoxicity were assessed in vitro on human adult retinal pigment epithelial (ARPE19) cells and mouse retinal ganglion cells (RGC) cell lines.

Methods

Preparation and Treatment of Ex Vivo Porcine Scleral Strips

Porcine globes of male and female pigs ranging in age from 1 to 2 years with a weight between 100 and 200 lbs were obtained from Vision Technologies (Sunnyvale, TX, USA), washed in 1% penicillin/streptomycin solution (Sigma-Aldrich, St Louis, MO, USA), and frozen at −80°C. The design of animal experiments adhered to the ARVO guidelines for the use of animals in ophthalmic and vision research. After thawing the globes immediately before experiment, the extraocular muscles and fascia were dissected from the globe (Figs. 1a, 1b) and laterality was determined ¹⁷ (Fig. 1c). A total of 57 globes were included in this study (strip study, n = 30 globes [Table; Fig. 4]; application approach study, up to n = 27 globes [Figs. 2, 7]). A total of 57 globes were included in this study (strip study, n = 30 globes [Table; Fig. 4]; application approach study, up to n = 27 globes [Figs. 2, 7]). The globes were cut in the axial plane along the long posterior ciliary arteries¹⁷ into two hemispheres. Lens, vitreous, retina, and choroid were carefully removed leaving just the bare sclera (Fig. 1d). The two halves of the scleral shell were further cut in half to yield four quarters (Fig. 1e). Tissue strips from each quarter were precisely punched with a custom-made cutting die with a width of 3.2 mm (Fig. 1f) circumferentially from the PPS adjacent to the ONH (Fig. 1g). All scleral strips were cut to a uniform width of 3.2 mm based on the distance of the cutting die's parallel blades (Figs. 1f, 1h) and further trimmed to a length of 9 mm with a scalpel (Fig. 1i). Four samples were dissected from the PPS of each globe of each quarter superior-temporal, superior-nasal, inferior-temporal, and inferior-nasal (Fig. 1c).

Figure 1.

Preparation of scleral strips. (a) anterior view of porcine globe. (b) Extraocular muscles dissected. (c) Posterior view of porcine globe, the ONH inserts in the inferior temporal quarter. IN, inferior-nasal; IT, inferior-temporal; SN, superior-nasal; ST, superior-temporal. (d) Globe cut along long posterior ciliary arteries in superior and inferior hemispheres. (e) Superior hemisphere cut in half into quarters. (f) Custom-made cutting blade with 3.2 mm of distance between razor blades. (g) Punch of scleral strip. (h) Scleral strip with width of 3.2 mm. (i) Scleral strip trimmed with scalpel to final length of 9.00 mm. (j) Thickness of scleral strip measured at 25%, 50%, and 75% of total length with an EVOS light microscope and Image J. (k) Scleral strip final dimensions of length: 9 mm, gauge length: 5 mm, width: 3.2 mm, thickness: 1.08 mm with tissue ink marks for clamping orientation shown. (l) grip support (outermost to innermost layers: grip tape, cardstock, double sided tape, sandpaper). (m) scleral strip glued at tissue ink marks with cyanoacrylate to grip support. (n) Dynamic mechanical analysis of scleral biomechanical properties and humidifier for tissue hydration. (o) Bone shape scleral strip just before breaking point.

Table.

Test Solutions and Concentrations for Each Reagent

	Test Concentrations				Test Solutions
	Unit	Clow	Cmed	Chigh	Buffer	Cofactors
GAD	M	0.001	0.01	0.1	PBS
COL	mg/mL	0.02	0.05	0.1	PBS	2 mM CaCl2
						1 mM MgCl26(H2O)
mTG	U/mL	0.01	0.1	1	PBS	10 mM CaCl2
TG1 from E. coli	U/mL	0.1	1	10	PBS	10 mM CaCl2
TG2 from E. coli	U/mL	0.1	1	10	PBS	10 mM CaCl2
LOX from Trichoderma viride	U/mL	0.001	0.01	0.1	25 mM Tris base pH 7.4	5 µg/mL copper
					100 mM NaCl, 2 mM CaCl2	5 µg/mL ascorbic acid
					1 mM MgCl26(H2O)

Concentrations: C_low = lowest; C_med = medium; C_high = highest.

Each treatment condition was tested on five scleral strips from five different porcine globes.

Figure 4.

Scleral thickness for control strips and intervention groups compared by unpaired two-tailed t-test with 30 control strips (one matching each intervention globe) and 5 for each intervention group. Error bars show mean and standard deviation. *P < 0.01; **P < 0.001; ***P < 0.0001.

Figure 2.

Diagram of three treatment approaches with blue dots indicating treatment solution. (a) Scleral strips cut and then exposed to treatment solution (n = 5). (b) inner and outer scleral treatment to mimic combined “intra- and extraocular” (intra+extra) treatment (up to n = 12). (c) Inner sclera exposed to treatment solution to mimic “intraocular” (intra) treatment (e.g., intravitreal or suprachoroidal injection) sealing with parafilm and cyanoacrylate glue (yellow) (up to n = 12). (d) Outer sclera exposed to treatment solution to mimic “extraocular” (extra) treatment (e.g., sub-Tenon's or retrobulbar injection) (up to n = 12).

Figure 7.

Biomechanical response with respect to treatment application with 5 for the strip group and up to 12 (up to 4 samples from one globe; i.e., 3 eyeballs in total) for the local applications. The standardized biomechanical effect of fully immersed strip incubation compared with ordinary one-way ANOVA to local drug applications to the intraocular (intra), extraocular (extra), or combined application to the intraocular and extraocular (intra+extra) scleral surface for COL 0.05 mg/mL, GAD 0.1 M and mTg 1 U/mL. Error bars show mean and standard error of the mean. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Test Solutions and Concentrations of ECM-Modulating Reagents

TG1 and TG2 were purchased from Zedira (Darmstadt, Germany). Unless other specified, all other reagents were purchased from Sigma-Aldrich. Test solutions were freshly prepared from frozen enzyme aliquots, with the exception of GAD, which was stored at room temperature, and titrated to the test concentrations summarized in Table. Tissue strips were incubated in 2 mL of the test solution at 37°C for 24 hours protected from light in an incubator with CO₂ levels equal to ambient air. After the completed incubation period, all tissue strips were washed three times with the same media used during incubation for 10 minutes each in a tube rotator. Each test concentration consisted of five strips obtained from five different porcine globes. For each globe, four adjacent strips were prepared from the PPS and assigned to the different treatment concentrations to minimize intertissue variability.

Locally Applied Treatments to the Intraocular and Extraocular Scleral Surface

In these experiments, the extraocular muscles and fascia were removed from the globe, and then the cornea was excised along with the lens, vitreous, choroid, and retina leaving a scleral shell. The treatment followed by placing the test solution into the scleral shell and sealing it with parafilm (Fig. 2c, yellow) and cyanoacrylate glue (Fig. 2c, termed “intraocular” [intra]) or by placing the whole globe into the test solution (Fig. 2d, termed “extraocular” [extra]), or by combined application, placing the whole scleral shell in the test solution (Fig. 2b, termed “intra- and extraocular” [intra+extra]) for 24 hours at 37°C. The test solutions were either COL 0.05 mg/mL, GAD 0.1 mol/L, or mTG 1 U/mL prepared the same as in the previous experiment. After treatment incubation, the globe was then processed as described into scleral strips and subjected to the same mechanical testing protocol. The strip-incubated group consisted of the same five strips used in the control group of the strip comparison study in Table and Figure 4, whereas the intra-, extra-, and combined intra+extra application groups each comprised up to four strips derived from the same porcine globe, yielding a total of 12 strips originating from three globes (Fig. 2).

Tissue Dimensions and Measurement of Tissue Thickness

It was necessary to determine the thickness of the scleral strips before mechanical testing to measure the cross area of the sample. A length of 9 mm and width of 3.2 mm were standardized since they were cut to these lengths. The scleral strip was laid on the surface of a glass slide and a photo taken (Invitrogen EVOS XL Core Configured Cell Imager, Fisher Scientific, Hampton, NH, USA). Sample thickness was analyzed from standardized images in ImageJ¹⁸ at 25%, 50%, and 75% of the tissue length (Fig. 1j). The mean value of the three thickness measurements was multiplied by the standardized tissue width of 3.2 mm to calculate the sample cross area.

Uniaxial Tensile Testing

After preparation and treatment, the scleral strips were tested with uniaxial tensile testing with a dynamic mechanical analysis machine (TA Instruments, New Castle, DE, USA). A five-layer approach was used to optimize clamping, prevent slippage, and minimize tear at the grip (Figs. 1l, 1m). The scleral strip was sandwiched between sandpaper P120 (Dura Gold Limited, Downers Grove, IL, USA), cardstock (Sanzix, Ellicott City, MD, USA), and grip tape (CatTongue Grips, Park City, UT, USA), which were all glued together with cyanoacrylate glue (Gorilla Glue Company, Cincinnati, OH, USA). We used 2 mm of each strip end for clamping, resulting in a gauge distance of 5 mm between the clamp accessories. The rectangular tensile accessory of an ARES-G2-Rheometer (TA Instruments) was used to clamp the layered grip supports. The force was tared after clamping, torque removed, and a preload of 0.05 Newton applied. Tissue preconditioning included up to 10 cycles of loading and unloading with a displacement rate of 0.03 mm/seconds for 13.3 seconds (Fig. 3a) and a maximum loading gap of 13 mm to match prior studies.^19,20 Subsequent to preconditioning, the force rate of 0.03 mm/seconds was further applied until tissue failure (Fig. 3b). All samples were measured at ambient temperature and a humidifier prevented tissue dehydration during tensile testing (Fig. 1n). Tensile testing associated data was collected with the TRIOS software version 5.1.1.46572 (TA Instruments).

Figure 3.

Biomechanical protocol for uniaxial tensile testing. (a) Ten preconditioning cycles. (b) Preconditioning cycles (gray) with stress–strain curve (black) and E_T calculation (red) as the slope of the most linear region.

Biomechanical Evaluation and Statistical Analysis

Data were collected in Microsoft Excel version 16.81 (Microsoft, Redmond, WA, USA), and statistical analysis done in R (R 4.3.2 GUI 1.80 Big Sur Intel build) and R Studio version 2023.09.0+463 (Posit PBC, Boston, MA, USA).

Tangent modulus (E_T; unit [MPa]) was experimentally approximated as the ratio of stress Δσ to strain Δε. Stress (σ) was computed as the applied force (F; unit [N]) divided by the initial cross section area of the sample (A₀; unit [mm²]; measured as described elsewhere in this article):

$$\begin{array}{c}\hfill \sigma =\frac{F}{A_0}.\end{array}$$

Strain (ε) was defined as the change in sample length (ΔL; unit [mm]) during uniaxial testing normalized to the initial gauge distance length (L₀; fixed loading gap of 5 mm between the clamp accessories):

$$\begin{array}{c}\hfill \varepsilon =\frac{\Delta L}{L_0}.\end{array}$$

The true E_T is mathematically defined as the ratio of the infinitesimal stress dσ to infinitesimal strain dε, describing the exact derivative (slope) of the stress–strain curve at a single point. In our study, the E_T is experimentally approximated by calculating the finite differences of stress Δσ and strain Δε between nearby data points. Practically, our R code first identifies the most linear region of each stress–strain curve by iteratively scanning the dataset using a sliding strain window of 5%. The interval exhibiting the highest linearity (i.e., best linear fit) is selected as the linear region. A final linear model is then fitted to this interval, and the regression slope coefficient is extracted and reported as the E_T, providing a numerical estimate of the local derivative of the stress–strain relationship.

The true E_T (exact mathematical derivative) is defined as the instantaneous slope of the stress–strain curve at a single strain value:

$$\begin{array}{c}\hfill E_T=\frac{d\sigma }{d\varepsilon }.\end{array}$$

The approximated E_T (experimental finite difference) calculates the average slope of the stress–strain curve over a small interval of data points:

$$\begin{array}{c}\hfill E_T\approx \frac{\Delta \sigma }{\Delta \varepsilon }.\end{array}$$

The mean and standard deviation or standard error of the mean were calculated. For dose–response curves, the extra sum-of-squares F-test was applied to statistically distinguish the best-fit Hill slope from the hypothetical slope of 0, representing no significant dose response. The Hill slope, also known as the slope factor, quantifies the steepness of a dose–response curve by comparing it with a standardized slope.²¹ The sample thickness of control scleral strips was compared with treated samples by unpaired two-tailed t test. Local drug application to the intraocular, extraocular, or intraocular and extraocular scleral surface was evaluated by ordinary one-way ANOVA. The significance level was set to an α of ≤0.05.

Safety and Cell Toxicity

Owing to the close proximity of the sclera to the retina, protection of the retinal cells and structures is imperative for clinical translation. We studied the safety of COL, GAD, and mTG application in vitro on ARPE19 and RGC cell lines. The Stanford University Institutional Animal Care and Use Committee reviewed and granted approval for the RGC culture experiments (protocol number: APLAC-34543).

ARPE19 Culturing and MTT Viability Assay

ARPE19 cell lines purchased from the American Type Culture Collection (Manassas, VA, USA) were seeded in 96-well plate with cell densities of 10 × 10³ cells per well in culture media (DMEM-F12, Ca²⁺ and Mg²⁺ free, 10% fetal bovine serum, 1% penicillin-streptomycin). ARPE19 cultured cells were incubated overnight at 37°C and 5% CO₂ with titrated test solutions at four different concentrations henceforward referred to CTL (i.e., control), C_low (i.e., lowest concentration; COL, 0.02 mg/mL; GAD, 0.001 M; mTG, 0.01 U/mL); C_med (i.e., medium concentration; COL, 0.05 mg/mL; GAD, 0.01 M; mTG, 0.1 U/mL) and C_high (i.e., highest concentration; COL 0.1 mg/mL; GAD, 0.1 M; mTG, 1 U/mL). The colorimetric assay Cell Proliferation Kit I (Roche, Basel, Switzerland) was used to quantify viable cells based on the cleavage of the yellow tetrazolium salt MTT to soluble purple formazan crystals. The intensity of the resulting purple colored solution, which is directly proportional to cell viability, was measured by the BioTek Cytation 5 Cell Imaging Multimode Reader (Agilent, Santa Clara, CA, USA).

RGC Panning, Culturing, and Survival Assay

Postnatal day 2 mouse RGC cells were purified by immunopanning as previously described by Barres et al.²² In short, postnatal retinas were dissected, digested with papain (Worthington Biochemicals, Lakewood, NJ, USA), titrated with trypsin inhibitor (Roche, Basel, Switzerland) inhibiting the enzymatic digestion, and incubated with rabbit anti-mouse macrophage antibody serum (Accurate Chemical, Carle Place, NY, USA). In the first two steps of the panning procedure, the retinal cell suspension was panned on goat anti-rabbit IgG-coated petri dishes to negatively select the RGCs. Nonadherent cells containing RGCs were subsequently positively selected by final panning on goat anti-mouse IgM bound CD90 (anti-Thy-1) (BioRad, Hercules, CA, USA) coated petri dishes. CD90-bound RGC cells were released by trypsin and resuspended in pre-equilibrated Neurobasal Sato growth media including BDNF (50 µg/mL), CNTF (10 µg/mL), Forskolin (5 µM), FDG (20 µg/mL) and GDNF (20 µg/mL), and cultured in a poly D-lysine–bound laminin-coated 96-well plate with cell densities of 5 × 10³ cells per well. Purified RGC cells were incubated overnight at 37°C and 10% CO₂ with the same titrated test solutions C_low to C_high used in the ARPE19 study. Hoechst-33342 was used for nuclear staining and RGC survival was quantified by imaging Calcein-AM⁺ (Life Technologies, Carlsbad, CA, USA) live cells and Sytox⁺ (Life Technologies) dead cells on a CellInsight CX5 HCS Platform (Thermo Fischer Scientific, Waltham, MA, USA) with the Thermo Scientific HCS Studio Cellomics Scan Version 6.6.0 Build 8153 (Thermo Fischer Scientific) using a 10× objective.

Results

Physical Appearance of Scleral Tissue Strips Treated With ECM Modifying Reagents

Sclera is composed mostly of collagen,²³ and collagenous crosslinks can alter the tissue's ability to imbibe water and thus its thickness after treatment. The sample thickness was measured after incubation and before biomechanical testing. Compared (unpaired two-tailed t test) with the control strip average tissue thickness of 1.88 ± 0.21 mm, the sample thickness was significantly lower for strips incubated with 0.1 M GAD (P = 0.0024) or 1 U/mL mTG (P = 0.0085), and significantly higher for strips incubated with 0.1 mg/mL COL (P = 0.0026). Incubation with LOX 0.1 U/mL (P = 0.4060), TG1 10 U/mL (P = 0.4153), or TG2 10 U/mL (P = 0.4400) did not affect the sample thickness compared with controls (Fig. 4). Hemisphere dependent sample thickness of non-treated control scleral strips for the differentiated scleral quarters is compared in Supplementary Fig. S1

Figure 5 shows representative pictures for COL, GAD- and mTG incubated scleral strips from the PPS region. The curvature of the tissue indicates its inherent stiffness and ability to resist the gravitation force of its own weight. Increasing COL concentration causes the tissue to sag under its own weight, suggesting weaker tissue. In contrast, GAD and mTG were able to resist sagging. LOX, TG1, and TG2 did not change appearance relative to control (data not shown in Fig. 5). Maillard browning led to the characteristic yellow–brownish change in scleral color characteristic for melanonids resulting from the chemical reaction between amino acids and reducing sugars.¹³

Figure 5.

Representative pictures for COL, GAD, and mTG incubated scleral strips from the PPS region. Enhanced acute angle for COL incubated tissue strips signaling tissue weakening and sagging. Relatively mild increase in obtuse angle for GAD and mTG with increasing concentrations as tissue stiffening owing to CXL persists the natural convex curvature of the scleral shell. C_high, highest concentration (for COL, 0.1 mg/mL, GAD, 0.1 M, mTG, 1 U/mL); C_low, lowest concentration (for COL, 0.02 mg/mL; GAD, 0.001 M, mTG, 0.01 U/mL); C_med, medium concentration (for COL, 0.05 mg/mL, GAD, 0.01 M, mTG, 0.1 U/mL); CTL, control; mTG, microbial transglutaminase.

Changes in Biomechanical Properties With ECM Modulating Reagents

Dose–Response Analysis

Dose–response curves were generated for scleral strips (refer to Fig. 2: treatment approach of group a) relating treatment concentration and tissue stiffness. Enzyme activities of TG1, TG2, and LOX were analyzed and confirmed before tissue incubation with a fluorescent transglutaminase assay kit (Zedira) or fluorescent LOX assay kit (Sigma-Aldrich), as appropriate. Dose–response curves for COL, GAD, mTG, TG1, TG2 and LOX incubated scleral strips were created with five for each treatment group and concentration (Fig. 6). The E_T of the treatment groups were normalized to the E_T value of the control strip within the individual eye. The P values derived from the F-test statistic evaluate whether the Hill slope of each best-fit curve differs significantly from the hypothetical value of zero, thereby indicating whether tissue stiffness changes meaningfully across the tested range. The F-test for the Hill slope were significant for COL (P < 0.0001), GAD (P = 0.0011), and mTG (P = 0.0210), suggesting a dose-dependent change in tissue stiffness. Treatment with COL at the highest concentration of 0.1 mg/mL resulted in a 12.5-fold decrease in E_T. Treatment with GAD at 0.1 M and mTG at 1 U/mL led to increases in E_T of 10.53-fold and 4.71-fold, respectively. No significant biomechanical effect on porcine sclera was seen with TG1 (P = 0.2962), TG2 (P = 0.4691), or LOX (P = 0.6361).

Figure 6.

Dose–response curves of scleral strip approach for test solutions with five for each treatment group and concentration. Error bars show mean and standard deviation. The E_T of the treatment groups were normalized to the E_T value of the control strip within the individual eye.

Effect of Locally Applied Treatments to Sample Surface on Tissue Stiffness

Scleral strips (nominally 9.00 mm × 3.20 mm × 1.88 ± 0.21 mm) cut before treatment exposed its six surfaces to the treatment solution, including four artificially created surfaces. To better understand the effects of surfaces treated, strips (with six treated surfaces) were compared with local drug application to the intraocular, extraocular, or intraocular and extraocular scleral surface. For COL, the surfaces treated did not statistically affect tissue stiffness between the different groups (P = 0.0985). For GAD and mTG, however, tissue stiffness was significantly greater in the strip-incubated group compared with locally applied treatments to either the intraocular or extraocular scleral surface or combined application, with P < 0.0001 for GAD and P = 0.0044 for mTG (Fig. 7). The change in stiffness did not scale with the exposed surface because intraocular, extraocular, and combined intraocular and extraocular represent 28%, 28%, and 56% of the nominal surface area of the scleral strip (3.20 mm × 9.00 mm × 1.88 ± 0.21 mm rectangular prism).

Safety and Cell Toxicity

We tested treatment in vitro safety of COL, GAD, and mTG with RGC and ARPE19 assays (Fig. 8). The RGC viability assay showed concentration-dependent cell survival for COL and GAD with RGCs remaining at 66.91 ± 17.87% for COL at 0.1 mg/mL and 0 ± 0% for GAD at 0.1 M (the greatest concentration tested for both reagents). mTG did not affect either RGC or ARPE19 cell viability at the tested concentrations between 0.001 and 1.0 U/mL. COL; however, it also showed no diminishing effect on cell survival in the ARPE19 assay.

Figure 8.

Cell viability assays for GAD, COL, and mTG on RGC (red) and ARPE19 (blue) cell lines with three for each treatment group and concentration. Error bars show mean and standard error of the mean. COL and mTG showed acceptable toxicity for increasing concentrations. GAD led to increased death of RGC cells with higher concentrations. The amplified cell viability of GAD on ARPE19 cells may be due to an artefact from interference with the assay.

Discussion

This work showed significant biomechanical modification for GAD, COL, and mTG on ex vivo porcine sclera with an acceptable cell toxicity profile in vitro on RGCs and ARPE19 cells. Locally applied treatments for GAD, COL, and mTG to the intraocular, extraocular, or combined intraocular and extraocular application quantitatively differed from whole strip incubation. No change in scleral tissue stiffness was found after treatment with TG1, TG2, or LOX.

Tissue Selection and Sample Preparation

Porcine sclera serves as a suitable animal model owing to its comparable histology and collagen fiber organization to human sclera, considering that porcine sclera is twice as thick as and two to three times less permeable than human sclera.²⁴ The advantage of porcine tissue is its greater availability compared with equivalent tissue of human origin. Previous studies have shown that consecutive freeze–thaw cycles down to −80°C do not affect either scleral uniaxial tensile properties (i.e., biomechanics)²⁵ or scleral permeability.^24,26 Freezing the samples ensured that all globes originated from a similar cohort, that is, animals from the food industry slaughtered on the same day and processed under comparable post mortem handling and cooled transport conditions. As enucleation, packaging, and delivery procedures at the abattoir are not under direct experimental control, acquiring and freezing a large batch decreases the variability introduced by uncontrolled differences in storage times and transport.

Scleral strip preparation from the PPS was based on previous studies.^27–29 The inhomogeneity and anisotropy of biological tissues are based on their collagen fiber orientation and ECM component distribution, leading to high intraindividual and interindividual sample variability.³⁰ Some studies exclusively use scleral strips nasal to the ONH considering the smaller spatial tissue variation in this region,¹⁹ because the ONH of pigs is located within the inferior temporal quadrant.^17,31 However, our work provides four scleral strips from the same porcine donor eye limiting interocular tissue variability. Further, the PPS has the most consistent scleral thickness in the porcine eye.³²

The sample aspect ratio of 1.0:1.5 in this work (sample width to gauge length, 3.2:5.0 mm) was adjusted to physiological constraints. Lin et al.³⁰ generally suggest an aspect ratio of at least 1:3 to 1:5 for tampered and 1:10 for nontampered soft materials. Nevertheless, the aspect ratio in this study is comparable with previous biomechanical studies on the sclera, for example, 1.00:1.25 (4:5 mm) by Nagase et al.,¹⁹ 1.0:1.5 (4:6 mm) by Wollensak et al.,¹² or 1.0:2.5 (2:5 mm) by Gawargious et al.²⁷ The reinforced (by cyanoacrylate glue, cardstock, and sandpaper) clamping area at both ends of the sample minimized clamping artefacts and slippage during testing. Even though the aspect ratio of the sample was artificially enhanced, the overall aspect ratio was lower than desired owing to the anatomical limitations of porcine tissue. Absolute stiffness numbers need to be interpreted with caution, because a short gauge length may provide lower strain measurements, leading to an overestimation of the E_T and tissue stiffness.

Uniaxial Tensile Testing and Biomechanical Assessment

Previous studies have shown that axial tensile testing on scleral tissue strips is a suitable method to study ex vivo tissue stiffness^12,14 and to quantitatively compare changes in tissue stiffness. The goal of this study was to quantify the global biomechanical tissue response of different reagents. The clamping area at the dynamic mechanical analysis accessories was enhanced by the five-layer grip support (Figs. 1l, 1m), distributing locally induced strains and preventing sample slippage during tensile testing. Preconditioning affects the matrix organization by realignment of the collagen fibers,^33,34 leading to increased peak stress for preconditioned tissue.³⁵ Nagase et al. applied ten preconditioning cycles to porcine sclera, based on preliminary experiments demonstrating that preconditioning effectively reduces the variability in hysteresis between successive stress–strain cycles.¹⁹ Park et al.²⁸ described a stabilized preconditioned state after two to three cycles on human sclera, discussing a possibly negligible preconditioning effect for the primarily anisotropic PPS. The hysteresis loop area (i.e. the area under the curve; AUC) during preconditioning is further analyzed in Supplementary Fig. S2a and compared between the different treatment groups (controls, COL, mTG and GAD) in Supplementary Fig. S2b.

Young's modulus (i.e., elastic modulus) is defined as the ratio of axial stress to axial strain within the true linear elastic region of a material. It describes the intrinsic elastic stiffness of homogeneous and isotropic materials that exhibit a well-defined elastic linear region. Soft biological tissues such as the sclera, however, are anisotropic, heterogeneous and nonlinear and do not exhibit a clearly defined linear elastic region. Therefore, E_T was experimentally assessed as a local stiffness parameter to characterize scleral stiffness in this study. By identifying the most linear region of each stress–strain curve and calculating the slope within this interval, E_T provides a physiologically meaningful estimate of local tissue stiffness. The experimentally extracted E_T resembles the Young's modulus but takes into account the nonlinear and anisotropic features exhibited by biological tissues.

Scleral CXL: Biomechanical Characterization and Treatment Safety

Various crosslinkers have been tested on scleral tissue such as riboflavin and UVA light,^12,36–39 Rose Bengal and 532 nm green light,^40,41 methylene blue and 660 nm light,^10,11 genipin,^10,42,43 GAD,^2,12,44 glutaraldehyde,^9,12 and mTG.^14,15 Light-dependent scleral CXL is challenging for clinical translation owing to the light sensitivity of the adjacent retina and the challenging accessibility of locally applied treatments of the equatorial and posterior sclera within the orbit. Further, UVA CXL releases reactive oxygen species,⁴⁵ which are toxic to the neuroretina.⁴⁶ Therefore, light-independent crosslinkers are favorable for scleral stiffness enhancement. The goal is to minimize the cytotoxic effects on the surrounding tissues while preserving treatment efficacy.⁸ Endogenous enzymes suggest a safe profile owing to their natural origin. Although treatment safety has been considered histologically in multiple studies,^{2,10,11,15,44} our work provides further insights with respect to cell toxicity on RGC and ARPE19 in vitro cell lines.

COL

Our study found a 2.33-fold decrease in the E_T for strip-incubated sclera with concentrations of 0.05 mg/mL compared with controls. The softening effect for locally applied treatment with 0.05 mg/mL COL was approximately reduced by 58.19% (1.36-fold) for intraocular, 66.43% (1.55-fold) for extraocular, and 46.17% (1.08-fold) for combined intraocular and extraocular application compared with the strip-incubated group. Spoerl et al.²⁰ tested the biomechanical response of the human and porcine lamina cribrosa and PPS to treatment of 0.1% COL and found a decrease in Young's modulus within the porcine lamina cribrosa (COL, 7.6 MPa; control, 17.1 MPa) and porcine PPS (COL, 3.4 MPa; control, 29.3 MPa). Interestingly, the relative biomechanical effect was lower for the human lamina cribrosa (COL, 9.8 MPa; control, 15.6 MPa) and human PPS (COL, 16.5 MPa; control, 28.5 MPa),²⁰ even though the human sclera has been described as more permeable compared with the porcine sclera.²⁴ The specimen-specific biomechanical tissue response independent from the tissue's permeability properties underlines the complexity of quantifying biomechanical effects.

The dose range for COL was constrained by practical and tissue-related limitations. For COL, increasing concentrations progressively weaken the scleral tissue. Scleral strips incubated with very high COL concentrations become too soft to be reliably clamped for uniaxial testing. Based on preliminary experiments, we therefore selected the highest COL concentration that still permitted secure clamping and reproducible mechanical measurements, which may have prevented us from fully capturing the upper plateau of a potential S-shaped dose–response curve.

The COL enzyme has wide application in ophthalmology. Its potential for enhancing scleral drug delivery and its application within enzymatic sclerostomy have been previously explored. Jiang et al.⁴⁷ showed that injection of COL improved scleral microparticle delivery on ex vivo human sclera. Jung et al.¹⁶ described how COL improved suprachoroidal space microparticle delivery to the posterior pole ex vivo and in vivo in rabbits, without macroscopic or microscopic adverse effects on the sclera, choroid, or retina. Further, Dan et al.⁴⁸ investigated the potential of COL for enzymatic sclerostomy in vivo in rabbits and ex vivo on calf eyes. A case study from the same study group with 15 blind patients demonstrated effective IOP control from 43.5 ± 9.8 mm Hg at baseline to 24.8 ± 10.6 mm Hg (43% decrease) at 1 day and to 34.8 ± 10.5 mm Hg (20% decrease) at 1 year after the intervention.⁴⁹

GAD

We found a 10.53-fold increase in the E_T for the strip group incubated with 0.1 M GAD compared with the control group. The stiffening effect for local CXL with GAD of the same concentration was approximately reduced by half down to a third and 36.98% (3.89-fold) for intraocular, 50.97% (5.37-fold) for extraocular, and 46.44% (4.89-fold) for combined intraocular and extraocular application compared with the strip-incubated group. Our results show a vastly stronger effect compared with a study by Wollensak et al.,¹² who described a 487% increase in Young's modulus in porcine sclera and 34% increase in Young's modulus in human sclera for ex vivo CXL with 0.2 mol GAD incubated for 5 days.

A further in vivo study by Wollensak et al.² in chinchilla rabbits described a 1027% increase in Young's modulus after 5 replicated sub-Tenon's injections of 0.15 mL 0.5 M GAD within 2 weeks, but no statistically significant change in scleral thickness, in contrast with our results, and inflammatory infiltrates and moderate keratocytes loss in the adjacent peripheral cornea. A long-term in vivo study in chinchilla rabbits by the same study group revealed a persisted increase in Young's modulus of 989.6% after 4 months and 554.17% after 8 months compared with controls without any adverse histological tissue effects.⁵⁰

Danilov et al.⁵¹ investigated the time-dependent change in Young's modulus for ex vivo GAD CXL on rabbit sclera, describing a maximized plateau after approximately 16 hours of incubation at 350% to 550%. Campbell et al.⁵² analyzed the dose response for ex vivo GAD crosslinked rat sclera with roughly 16 hours’ incubation by inflation testing (tested concentrations of 10, 30, 62.5, and 125 mM) for an IOP equivalent to 13 mm Hg (normotension) and 28 mm Hg (hypertension) and received a plateaued stiffening effect for GAD at a concentrations of approximately 30 mM. However, porcine sclera is vastly different from the rabbit and rat sclera. We chose porcine sclera because its biomechanical properties are similar to the human sclera. Dose–response analysis varies between different tissue species and needs to be considered individually. With reference to our results, a plateau has not been reached for GAD-incubated porcine sclera for concentrations of up to 0.1 M and within 24 hours of incubation. The ex vivo study mentioned by Spoerl et al.²⁰ showed an increase in Young's modulus within the porcine (GAD, 22.4 MPa; control, 17.1 MPa) and human (GAD, 24.2 MPa; control, 14.7 MPa) lamina cribrosa and the porcine (GAD, 54.7 MPa; control, 29.3 MPa) and human (GAD, 65.5 MPa; control, 28.9 MPa) PPS.

mTG

Our study found a 4.71-fold increase in the E_T for strip-incubated sclera with 1 U/mL mTG compared with controls. The stiffening effect for local CXL was approximately reduced by one-half and 47.54% (2.24-fold) for intraocular, 49.66% (2.34-fold) for extraocular and 49.20% (2.32-fold) for combined intraocular and extraocular application compared with the strip-incubated group. Our results are quantitatively comparable with the findings of Sun et al.,¹⁵ who injected 1 U/mL mTG in vivo in the sub-Tenon's space (comparable with our extraocular application) of New Zealand white albino rabbits with 15.79 ± 2.93 MPa for mTG-treated sclera and 6.91 ± 2.23 MPa for controls (corresponds with a 2.29-fold increase in Young's modulus).

An earlier study also by Sun et al.¹⁴ quantitively compared the ex vivo CXL effect on porcine sclera of "double-sided" mTG strip incubation (equal to our strip group) with "single-sided" mTG incubation to the extraocular surface (equal to our extraocular group). It is worth emphasizing that the double-sided label in the study is somewhat misleading, because the group used precut scleral strips, allowing not only enzyme incubation on the intraocular and extraocular surface, but also increased diffusion of the enzyme across the artificially cut strip edges, akin to the scleral strips tested here. In the double-sided group, mTG increased the tissue E_T by 2.21-fold, and in the single-sided group the E_T increased by 1.84-fold. The enhanced stiffening effect in our strip group (4.71-fold) and extraocular group (2.34-fold) may be explained by the longer incubation period of 24 hours compared with the 30 minutes applied in the study by Sun et al. In addition, the overall ratio of single-sided to double-sided incubation was higher with approximately 83.26% compared with 49.68% (extraocular group to strip group) in our study. In contrast with our findings (Fig. 4), CXL with mTG did not reduce scleral thickness, possibly also a result of the shorter incubation time.¹⁴

LOX, TG1, and TG2

The potential of LOX, TG1, and TG2 as alternative scleral crosslinkers were investigated in this study. Buffers in our study were optimized for enzymatic temperature and pH requirements. Calcium⁵³ as well as ascorbic acid and copper^54,55 were added as previously recommended cofactors for TG1/TG2 and LOX, respectively. Although the three enzymes, based on their endogenous nature,^53,56 suggest a low risk profile for potential human application, no measurable biomechanical stiffening was detected in our study. This finding must be interpreted in the context of the incubation environment, tissue permeability, and associated substrate accessibility, as well as the inherent substrate selectivity of each enzyme.

TG1 (also known as keratinocyte transglutaminase, expressed in skin keratinocytes, encoded by the TGM1 gene) and TG2 (also known as tissue or endothelial transglutaminase, an ubiquitously expressed multifunctional enzyme, encoded by the TGM2 gene) are calcium-dependent protein crosslinkers that form novel peptide bonds by transaminating glutamine and lysine residues.⁵³ mTG is a calcium-independent smaller enzyme inducing crosslinks between the glutamine and various primary amines, especially of lysine. Enzymes of the LOX family induce covalent intramolecular and intermolecular bonds between collagen and elastin by oxidizing lysine residues to aldehydes.⁵⁶

Nicoli et al.²⁴ have demonstrated that porcine sclera is technically permeable for high molecular weights up to 120 kDa. Thus, LOX, TG1, and TG2 with a molecular weight of 112 kDa, 90 kDa, and 78 kDa, respectively, should be intrinsically permeable, particularly for long incubation periods of 24 hours under optimized conditions. The treatment duration of 24 hours was chosen in accordance with a previous study that showed similar clearance times for macromolecules of similar size.⁵⁷ However, their scleral tissue penetration may not be comparable with that of mTG, which is considerably smaller, with a molecular weight of only 38 kDa. Furthermore, in addition to substrate accessibility, the inherent substrate selectivity of each enzyme may also contribute to the absence of a detectable stiffening effect. Although mTG may consist of a broader substrate permissiveness and fewer regulatory constraints, endogenous enzymes might more likely exhibit more selective substrate profiles and greater regulation within their physiological biological environments. Notably, even substantially longer LOX incubation times of up to 120 hours with concentrations as high as 10 U/mL did not lead to measurable biomechanical stiffening in our preliminary experiments (considering that such extended exposure times are unlikely to be clinically realistic given expected clearance kinetics).

Although LOX preferentially crosslinks elastin, CXL formation within collagen-rich tissues like the sclera is theoretically feasible through oxidation of telopeptide lysine and hydroxylysine residues, albeit with lower efficiency. Importantly, the PPS contains a meaningful amount of elastin: Terai et al.⁵⁸ demonstrated in an ex vivo study on porcine eyes that the PPS region contains a substantially higher elastin content than the remaining sclera, with elastin occupying approximately 15.5% of the tissue on average. Supporting the functional relevance of LOX in scleral biomechanics, an in vivo study by Yuan et al.⁵⁹ showed that LOX expression and Young's modulus of the sclera were both decreased in form-deprived myopic (FDM) guinea pigs. In the same study, scleral Young's modulus was decreased by irreversible LOX inhibition with β-aminopropionitrile and increased by TGF-β–induced amplified LOX expression. Furthermore, Nguyen et al.⁶⁰ found that recombinant LOX treatment significantly stiffened the mechanical properties of calcaneus tendons of in ovo and ex ovo chick embryos, highlighting the enzyme's capacity to enhance ECM rigidity in vivo.

Strip Incubation vs. Local Drug Application

Ex vivo studies of different CXL reagents typically fully immerse tissue strips during reagent incubation. Full strip incubation, however, is only a limited representation of a clinical administration. Within in vivo scleral CXL applications, drug administration is either possible to the extraocular scleral surface by sub-Tenon's or retrobulbar injection, or to the intraocular scleral surface by intravitreal, suprachoroidal, or suprachoroidal-to-optic-nerve injection.⁶¹ Full tissue strip immersion allows an adequate quantification of the maximum biomechanical effect an agent may have on the scleral tissue. To account for clinically feasible in vivo administration, locally applied treatments to either the extraocular or intraocular scleral surface are necessary.

Although strip-incubated and locally applied treatments did not significantly affect tissue stiffness for COL, both GAD and mTG produced markedly greater stiffness in the strip-incubated group compared with locally applied intraocular, extraocular, or combined surface treatments (Fig. 4). This finding suggests that GAD- and mTG-treated tissues are more resistant to tissue hydration owing to increased crosslink formation, whereas COL-treated samples are more susceptible to tissue hydration owing to fewer crosslinks. The latter is likely attributable to the COL-mediated cleavage of collagen fibers, increasing tissue permeability during incubation. The impact of drug administration was also demonstrated by Sun et al.,¹⁴ who showed in an ex vivo study with mTG on the porcine sclera that the quantitative stiffening effect varied between full tissue strip immersion and local incubation to the extraocular scleral surface, with an increase in Young's modulus compared with controls for the strip group of 121% and for the local enzyme application of 84%, respectively.

ARPE19 and RGC Culturing and Assay

COL and mTG showed acceptable toxicity in the ARPE19 and RGC assay, whereas GAD led to increased death of RGC cells with higher concentrations. The amplified ARPE19 survival in the GAD group is most likely an artefact, potentially owing to interference with the MTT assay. Because the MTT test relies on the reduction of MTT to formazan,⁶² GAD could theoretically interfere with this reaction by acting as a reducing agent with its aldehyde group.⁶³ Furthermore, GAD may also influence cellular metabolism, particularly glycolysis, by altering the availability of the reducing equivalents NADH and NADPH on which the MTT test primarily depends.^62,64 Such chemical interference is expected to be more pronounced for GAD, a small, highly reactive aldehyde, than for COL or mTG, which are large protein enzymes with high molecular weights and complex tertiary structures, which are therefore far less likely to interact directly with the MTT reaction. Nonetheless, we acknowledge this as a methodological limitation and interpret the viability results with appropriate caution.

Clinical Implications

Changes in scleral biomechanical properties, driven by remodeling processes, are increasingly recognized as important factors in myopia and glaucoma. In myopia, axial elongation is associated with scleral remodeling and decreased tissue thickness,⁶ down to approximately one-third compared with normal controls.⁶⁵ In experimentally induced myopia, an increase in scleral creep rate⁶⁶ and transition strain,⁶⁷ as well as a cyclic softening effect,⁶⁸ have been described. In glaucoma, an elevated IOP, the most identifiable and only modifiable risk factor,⁶⁹ leads to remodeling and stiffening of the PPS.⁸ This result has been confirmed in inflation^70–72 and uniaxial strip⁷³ tests. It has been hypothesized that the stiffer PPS provides neuroprotection for axonal RGC damage⁸ by decreasing the biomechanical strain within the lamina cribrosa and ONH.⁹ Scleral CXL has been suggested to counteract myopia-associated scleral thinning^6,65 and potentially protect the ONH from IOP damage by biomechanical enhancement of the PPS.

Lin et al.⁴⁴ reported successful myopia prevention with sub-Tenon's injection of GAD in New Zealand rabbits with no histological damage to the retina or choroid. In contrast, Chu et al.⁷⁴ described that in vivo scleral CXL with sub-Tenon's injection of GAD in FDM guinea pigs did not prevent myopia progression. However, Guo et al.⁷⁵ described a decrease in scleral stiffness in FDM guinea pigs, whereas scleral stiffness was increased in the scleral CXL group and FDM plus scleral CXL group compared with controls. Although scleral CXL decreased myopic changes in this study, the IOP was significantly elevated in all three intervention groups and glaucomatous changes in the ONH described.⁷⁵ This finding emphasizes the ongoing uncertainty as to whether PPS stiffening in glaucoma constitutes a protective adaptation or whether it may, in fact, contribute to adverse biomechanical remodeling. A mouse model study by Kimball et al.⁷⁶ showed that GAD-treated sclera had no adverse effects on retinal structure and function, but whole globe CXL increased glaucoma damage and led to greater loss of RGC axons. These conflicting findings underline the need for further investigations regarding clinical translation. Further, the relevance of local CXL to targeted scleral regions instead of whole globe applications require further considerations.

Limitations

The ex vivo model used in this study provides only a limited representation of real-world conditions. Although human and porcine sclera share similar histological features, differences in permeability and thickness need to be considered carefully. Although freezing a large batch of porcine globes was intended to minimize interglobe variability related to abattoir handling and transport conditions, freeze–thaw cycle–induced biomechanical alterations cannot be eliminated completely. Nevertheless, the use of ex vivo tissue enabled testing a greater number of treatment agents, more concentrations, and more treatment locations than possible with in vivo models and will surely guide in vivo work.

A further limitation of our mechanical testing protocol is the poor aspect ratio of the PPS strips imposed by anatomical constraints, which may introduce nonuniform strain fields and clamping artefacts and thereby affect the absolute magnitude of the calculated E_T. Further, the E_T was determined measured up to tissue failure, which is beyond the physiological forces typically encountered in vivo. In vivo studies demonstrate that the sclera experiences a 3D strain magnitude on the order of 5% in normal conditions. Uniaxial tensile testing does not fully represent the multidirectional strain that scleral tissue is subjected to under in vivo conditions. Although E_T provides a useful approximation of tissue stiffness in soft biological materials, its interpretation must consider the inherently anisotropic, heterogeneous, and nonlinear behavior of these tissues.

Although the incubation environment for LOX, TG1, and TG2 was optimized based on supplier information and previous studies, and substrate availability and specificity were carefully considered, not all relevant factors may have been fully accounted for, and further exploration is required to elucidate the absent effect observed in this study. Furthermore, it is possible that the full S-shaped dose–response curve was not entirely captured for these reagents, particularly for COL, where pronounced tissue weakening prevented reliable clamping for uniaxial tensile testing beyond a certain concentration.

In addition to the earlier discussed potential metabolic interference caused by GAD, the inherent limitations of the MTT assay must be acknowledged. Because MTT quantifies metabolic activity rather than a true cell number or survival, changes in mitochondrial function or cellular redox state may influence the readout independent of actual viability. Limitations of the RGC assay include its reliance on metabolic and membrane integrity markers, limited sensitivity for early apoptosis, and a lack of functional readout. Further, potential dye interference by reagents as well as quantification inaccuracies regarding segmentation or threshold settings need careful consideration.

Conclusions

Biomechanical modification of the ex vivo porcine scleral strips was successfully performed for GAD, mTG, and COL with an acceptable safety profile. We found no biomechanical changes with TG1, TG2, and LOX. Understanding the quantitative effect of local effects compared with ex vivo strip incubation is essential regarding clinical translation. Its potential for various applications in ophthalmology underline the need for further investigation.