November 2017
Volume 6, Issue 6
Open Access
Articles  |   December 2017
Differential Gene Transcription of Extracellular Matrix Components in Response to In Vivo Corneal Crosslinking (CXL) in Rabbit Corneas
Author Affiliations & Notes
  • Sabine Kling
    Laboratory of Ocular Cell Biology, Center of Applied Biotechnology and Molecular Medicine, University of Zurich, Switzerland
    Laboratory of Ocular Cell Biology, University of Geneva, Switzerland
  • Arthur Hammer
    Laboratory of Ocular Cell Biology, University of Geneva, Switzerland
    Hoptial ophtalmique Jules-Gonin, Fondation Asile des aveugles, Lausanne, Switzerland
  • Emilio A. Torres Netto
    Laboratory of Ocular Cell Biology, Center of Applied Biotechnology and Molecular Medicine, University of Zurich, Switzerland
    Department of Ophthalmology, Paulista School of Medicine, Federal University of Sao Paulo, Sao Paulo, Brazil
  • Farhad Hafezi
    Laboratory of Ocular Cell Biology, Center of Applied Biotechnology and Molecular Medicine, University of Zurich, Switzerland
    Laboratory of Ocular Cell Biology, University of Geneva, Switzerland
    ELZA Institute AG, Dietikon/Zurich, Switzerland
    University of Southern California, CA, USA
  • Correspondence: Sabine Kling, PhD, University of Zurich, Center for Applied Biotechnology and Molecular Medicine, Winterthurerstrasse 190, 8057 Zurich, Switzerland. e-mail: kling.sabine@gmail.com 
Translational Vision Science & Technology December 2017, Vol.6, 8. doi:10.1167/tvst.6.6.8
  • Views
  • PDF
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Sabine Kling, Arthur Hammer, Emilio A. Torres Netto, Farhad Hafezi; Differential Gene Transcription of Extracellular Matrix Components in Response to In Vivo Corneal Crosslinking (CXL) in Rabbit Corneas. Trans. Vis. Sci. Tech. 2017;6(6):8. doi: 10.1167/tvst.6.6.8.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: We studied changes in gene transcription after corneal crosslinking (CXL) in the rabbit cornea in vivo and identified potential molecular signaling pathways.

Methods: A total of 15 corneas of eight male New-Zealand-White rabbits were de-epithelialized and equally divided into five groups. Group 1 served as an untreated control. Groups 2 to 5 were soaked with 0.1% riboflavin for 20 minutes, which in Groups 3 to 5 was followed by UV-A irradiation at a fluence of 5.4 J/cm2. Ultraviolet A (UVA) irradiation was delivered at 3 mW/cm2 for 30 minutes (Group 3, standard CXL protocol), 9 mW/cm2 for 10 minutes (Group 4, accelerated), and 18 mW/cm2 for 5 minutes (Group 5, accelerated). At 1 week after treatment, corneal buttons were obtained; mRNA was extracted and subjected to cDNA sequencing (RNA-seq).

Results: A total of 297 differentially transcribed genes were identified after CXL treatment. CXL downregulated extracellular matrix components (collagen types 1A1, 1A2, 6A2, 11A1, keratocan, fibromodulin) and upregulated glycan biosynthesis and proteoglycan glycosylation (GALNT 3, 7, and 8, B3GALT2). Also, CXL activated pathways related to protein crosslinking (transglutaminase 2 and 6). In 9.1% of the significantly different genes, CXL at 3 mW/cm2 (Group 1) induced a more distinct change in gene transcription than the accelerated CXL protocols, which induced a lower biomechanical stiffening effect.

Conclusions: Several target genes have been identified that might be related to the biomechanical stability and shape of the cornea. Stiffening-dependent differential gene transcription suggests the activation of mechano-sensitive pathways.

Translational Relevance: A better understanding of the molecular mechanisms behind CXL will permit an optimization and individualization of the clinical treatment protocol.

Introduction
Until recently, corneal ectasia could not be treated and typically required corneal transplantation, involving the risks of infection, protracted wound healing, and rejection. In 1997, Spoerl et al.1 proposed a new technique to increase the biomechanical stiffness of the cornea: corneal crosslinking (CXL). The treatment involves de-epithelialization of the cornea, soaking the corneal stroma with a chromophore (Vitamin B2, riboflavin), and ultraviolet A (UVA) irradiation with 3 mW/cm2 for an additional 30 minutes. Multiple studies have shown that CXL successfully stops keratoconus2 progression and also arrests postsurgical corneal ectasia.3 Since its introduction, a number of modifications of the original treatment protocol have been proposed, aiming at increasing its efficacy, shortening treatment duration, and reducing the risk of postoperative complications. The most widely used modified treatment protocol is accelerated CXL,4 using a higher irradiance in combination with a shorter irradiation time. However, several studies showed a reduced treatment efficacy. In clinical settings, a shallower demarcation line57 and minor corneal flattening7,8 was reported with accelerated CXL compared to standard CXL; in experimental settings, a lower tensile elastic modulus9,10 and a lower dry-weight after enzymatic digestion11 were found. Further modified treatment protocols include iontophoresis-assisted,12 trans-epithelial,13 hypo-osmolar,14 pulsed,15 contact lens–assisted,16 and customized17 CXL. All modified protocols share limited success: the increase in corneal stiffness is lower compared to that of the standard CXL treatment. A reason why it is difficult to optimize CXL is that its working principle is poorly understood. Although most mechanical strengthening would be expected if bonds were formed between collagen lamellae, x-ray scattering experiments indicate that bonds are formed rather at the collagen fibril surface and in the protein network surrounding the collagen.18 Also, the corneal swelling capacity is reduced strongly after CXL,19 suggesting that proteoglycans and glycosaminoglycans are involved.20,21 Clinical trials currently are performed to address the question whether CXL has the potential for primary refractive corrections of myopia22 and hyperopia. A better understanding of the basic mechanisms behind CXL would allow better adaptation of the protocol for different therapies, but also to identify its limitations. 
One might speculate that the arrest of keratoconus progression induced by CXL implies long-term and permanent changes on transcriptional, translational, and/or posttranslational levels. This hypothesis is supported by the fact that the increase in corneal stiffness after CXL lasts23 potentially longer than the actual collagen turnover in the corneal tissue and that significant—sometimes even progressive—corneal flattening is observed after CXL treatment.24 There are different mechanisms of how CXL may change gene transcription: the generation of large amounts of reactive oxygen species (ROS) may activate signaling pathways25,26 with the potential of reintroducing homeostasis. Another mechanism may involve mechanotransduction,27,28 which means the process of converting mechanical signals into biochemical responses. Mechanical signals may result from dynamically acting forces, but also from remodeling of the extracellular matrix (ECM), leading to changes in cell adhesion and cell–cell contact that finally determine the mechanical interaction with the surrounding matrix.29,30 Different mechanisms of action have been identified for mechanotransduction: certain ion-channels open in response to increased tension in the plasma membrane (observed during osmotic changes), proteins can unfold domains upon tension that reveal cryptic-binding, and phosphorylation may increase upon stretching.27 These immediate changes may activate signaling pathways and/or gene transcription within minutes to hours.27,30,31 
The purpose of this study was to analyze the corneal transcriptome before and after CXL treatment to identify differentially transcribed candidate genes that potentially affect corneal stiffness. 
Methods
Eight New Zealand White rabbits (2.5 kg weight) were purchased from Charles River Laboratories (Saint-Germaine-Nuelles, France). All experiments were approved by the local ethical committee and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research. 
CXL Treatment Protocol
Rabbits were anesthetized with a subcutaneous injection of ketamine (Ketalar; Pfizer AG, Zurich, Switzerland) and xylazine (Rompun 2%, 20 mg ml-1; Bayer, Basel, Switzerland). A total of 15 eyes then were assigned to one of five treatment groups (n = 3 per group). The corneas of all groups were de-epithelialized. Group 1 served as untreated control. Groups 2 to 5 corneas additionally received 0.1% riboflavin instillation during 20 minutes, using a suction ring. Group 2 served as riboflavin control. Group 3 corneas subsequently were irradiated with 3 mW/cm2 during 30 minutes, Group 4 corneas with 9 mW/cm2 during 10 minutes, and Group 5 corneas with 18 mW/cm2 during 5 minutes. Riboflavin was not renewed during UV irradiation. Three different irradiances were included to study the effect of different degrees of biomechanical stiffening.10 Directly after treatment, antibiotic ointment (Ofloxacin, Floxal 0.3%; Bausch & Lomb, Zug, Switzerland) was administered prophylactically onto the cornea and repeated twice daily (until epithelial closure on postoperative days 3–4) to avoid infections. In addition, buprenorphin (Temgesic) was administered subcutaneously twice daily at 50 μg/kg until epithelial closure. 
Sample Preparation
One week after CXL treatment the rabbits were sacrificed (intravenous 120 mg/kg, Pentothal; Ospedalia AG, Hünenberg, Switzerland) and the corneas obtained with a trephine (8 mm diameter). The corneal tissue was immersed in RLT lysis buffer + 1% β-mercaptoethanol and homogenized, first with scissors and then with a tissue disruptor (Qiagen GmbH, Hilden, Germany). Afterwards, samples were frozen in liquid nitrogen and stored at −80°C. 
Then, mRNA of the entire cornea, including epithelial, keratocyte, and endothelial cells, was extracted using an RNeasy kit (Qiagen) according to the manufacturer's instructions. mRNA quantity and quality were assessed with a spectrophotometer (Qbit; Life Technologies, Carlsbad, CA, USA) and the Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA), respectively. 
Differential Gene Transcription
Equal amounts of mRNA (300 ng) were reverse transcribed, then cDNA sequencing (RNAseq) was performed with the HighSeq 2500 system (Illumina, San Diego, CA, USA) using the TruSeq stranded mRNA protocol with 100 single-end reads. The sequencing quality was controlled with FastQC v.0.11.2 leading to a Phred quality score of >28 corresponding to a 1/1000 chance of errors. TopHat v2.0.13 software was used for mapping against the reference genome. The alignment percentage was not optimal (∼65%), probably due to low sequencing quality of the rabbit genome. As a consequence, multiple-mapping reads were not considered in the counts. Here, counts corresponded to the total number of reads aligning to a genomic feature. Biological quality control and summarization were done with RSeQC v2.4 and PicardTools v1.92 software, respectively. Only genes with a count above 10 in at least three samples were included for further analysis. The normalization and differential transcription analysis was performed with the R/Bioconductor package edgeR v.3.10.5, for the genes annotated in the reference genome. 
Statistical Analysis
Differentially transcribed genes were determined for each individual treatment group using a General Linear Model (GLM), a negative binomial distribution and a quasi-likelihood test. Ten pairwise comparisons (edgeR, GLM, quasi-likelihood F test) of the experimental groups were analyzed (Table 1). Instead of correcting the P values of the differentially transcribed genes with the Bonferroni method for multiple testing error, a different approach was chosen selecting significant genes according to the response of the whole set of CXL and control conditions. For this purpose, a composite null hypothesis, H0, was created summarizing the five most important comparisons. The condition CXL at 18 mW/cm2 was excluded in this selection process, as its treatment efficacy is smallest, as shown experimentally911 and clinically58 and, hence, its meaningfulness is lower than the other comparisons.  
\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\bf{\alpha}}\)\(\def\bupbeta{\bf{\beta}}\)\(\def\bupgamma{\bf{\gamma}}\)\(\def\bupdelta{\bf{\delta}}\)\(\def\bupvarepsilon{\bf{\varepsilon}}\)\(\def\bupzeta{\bf{\zeta}}\)\(\def\bupeta{\bf{\eta}}\)\(\def\buptheta{\bf{\theta}}\)\(\def\bupiota{\bf{\iota}}\)\(\def\bupkappa{\bf{\kappa}}\)\(\def\buplambda{\bf{\lambda}}\)\(\def\bupmu{\bf{\mu}}\)\(\def\bupnu{\bf{\nu}}\)\(\def\bupxi{\bf{\xi}}\)\(\def\bupomicron{\bf{\micron}}\)\(\def\buppi{\bf{\pi}}\)\(\def\buprho{\bf{\rho}}\)\(\def\bupsigma{\bf{\sigma}}\)\(\def\buptau{\bf{\tau}}\)\(\def\bupupsilon{\bf{\upsilon}}\)\(\def\bupphi{\bf{\phi}}\)\(\def\bupchi{\bf{\chi}}\)\(\def\buppsy{\bf{\psy}}\)\(\def\bupomega{\bf{\omega}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\begin{equation}\tag{1}{H_{\bf{0}}} = {H_{{\rm{virgin}} = {\rm{3mW}}}}\left| {{H_{{\rm{virgin}} = {\rm{9mW}}}}} \right|{H_{{\rm{ribo}} = {\rm{3mW}}}}\left| {{H_{{\rm{ribo}} = {\rm{9mW}}}}} \right|\, \sim\! {H_{{\rm{virgin}} = {\rm{ribo}}}}\end{equation}
and hence:  
\begin{equation}\tag{2}{H_{\bf{1}}} = \, \sim\! {H_0} = \, \sim\! {H_{{\rm{virgin}} = {\rm{3mW}}}}\ \& \, \sim\! {H_{{\rm{virgin}} = {\rm{9mW}}}}\ \& \, \sim\! {H_{{\rm{ribo}} = {\rm{3mW}}}}\ \& \, \sim\! {H_{{\rm{ribo}} = {\rm{9mW}}}}\ \& \, {H_{{\rm{virgin}} = {\rm{ribo}}}}\end{equation}
where H1 is the composite null hypothesis. Hx=y represents an individual null hypothesis, that is there is no difference between x and y. ∼Hx=y represent a rejected null hypothesis, that is there is a difference between x and y. Each comparison between CXL (at 3 or 9 mW/cm2) and control (virgin or riboflavin) is expected to be significant. In contrast, the comparison between the two control conditions is expected not to be significant. A given gene then will be considered significant, if H1 is true. With a confidence interval of 95%, the probability for a false positive in one comparison is:  
\begin{equation}\tag{3}{P_{\bf{i}}} = \left( {{P_{{\rm{virgin}}\, \sim\! = {\rm{3mW}}}}} \right) \cdot \left( {{P_{{\rm{virgin}}\, \sim\! = {\rm{9mW}}}}} \right) \cdot \left( {{P_{{\rm{ribo}}\, \sim\! = {\rm{3mW}}}}} \right) \cdot \left( {{P_{{\rm{ribo}}\, \sim\! = {\rm{9mW}}}}} \right) \cdot \left( {{P_{{\rm{ribo}} = {\rm{virgin}}}}} \right)\end{equation}
The probability of Pribo=virgin cannot be calculated exactly, as it is the power of the test. However, assuming that the power is 1, we have neglected this term resulting in Pi ≤ 0.054. Applied to the entire set of n = 9335 analyzed genes, the probability of having at least one false-positive can be calculated:  
\begin{equation}\tag{4}{P_{{\bf{cumulative}}}} = {\rm{1}} - {\left( {{\rm{1}} - {P_{\rm{i}}}} \right)^{\rm{n}}} \le 0.0{\rm{567}}\end{equation}
This P value, Pcumulative, is comparable to the standard significance level. An alternative correction for multiple testing is the Bonferroni method, which, however, can be applied only to one group at a time. The above-described whole-data-set approach is superior, as it accounts for the reproducibility of the CXL effect before correcting for multiple testing. Figure 1 illustrates that with Bonferroni correction, lower statistical significance (19 significantly different genes) can be reached than with the whole-data-set approach (297 significantly different genes).  
Table 1
 
Differential Gene Transcription was Computed for a Total of 10 Comparisons Between Different Treatment and Control Groups
Table 1
 
Differential Gene Transcription was Computed for a Total of 10 Comparisons Between Different Treatment and Control Groups
Figure 1
 
Comparison of different approaches to correct for multiple statistical testing. (A) Bonferroni method resulting in 19 significantly differently transcribed genes. (B) Whole-data-set method developed in this manuscript resulting in 297 significantly differently transcribed genes.
Figure 1
 
Comparison of different approaches to correct for multiple statistical testing. (A) Bonferroni method resulting in 19 significantly differently transcribed genes. (B) Whole-data-set method developed in this manuscript resulting in 297 significantly differently transcribed genes.
Filter for Stiffening Dependent Gene Transcription
The resulting list of significantly transcribed genes then was subjected to filtering to determine genes that are transcribed differentially in a stiffening-dependent manner. The following criteria Filter(stiffening) was imposed:  
\begin{equation}{C_{\bf{1}}} = \left[ {\left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{3mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{9mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{18mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{3mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{9mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{18mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{3mW}} - {\rm{9mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{3mW}} - {\rm{18mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{9mW}} - {\rm{18mW}}}} \lt 0} \right)} \right]\quad\quad (5\rm a)\end{equation}
 
\begin{equation}{C_{\bf{2}}} = \left[ {\left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{3mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{9mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{virgin}} - {\rm{18mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{3mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{9mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{ribo}} - {\rm{18mW}}}} \lt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{3mW}} - {\rm{9mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{3mW}} - {\rm{18mW}}}} \gt 0} \right)\& \left( {{\rm{logF}}{{\rm{C}}_{{\rm{9mW}} - {\rm{18mW}}}} \gt 0} \right)} \right]\quad\quad (5\rm b)\end{equation}
 
\begin{equation}{\bf{Filte}}{{\bf{r}}_{({\bf{stiffening}})}} = {\rm{\ }}{{\rm{H}}_{\rm{1}}}\& \, \sim\! {{\rm{H}}_{{\rm{3mW}} = {\rm{9mW}}}}\& \, \sim\! {{\rm{H}}_{{\rm{3mW}} = {\rm{18mW}}}}\& \left( {{{\rm{C}}_{\rm{1}}}{\rm{|}}{{\rm{C}}_{\rm{2}}}} \right)\quad\quad (5\rm c)\end{equation}
where logFC is the fold-change in log2 scale between the different tested conditions; & and | represent the logical operators AND and OR, respectively.  
Correlation Analysis
The Pearson's linear correlation coefficient among all treatment conditions was calculated for selected differentially transcribed genes using Matlab software (Mathworks, Bern, Switzerland) to investigate mutual gene interactions. The online tool DAVID32,33 Bioinformatics Resources (Version 6.8) was used to extract related signaling pathways. 
Results
Differential Gene Transcription
From a total of 9335 transcripts, 297 were significantly differentially transcribed between the two clinically efficient CXL conditions (at 3 and 9 mW/cm2) and controls (virgin and riboflavin). Of these differentially transcribed genes, 9.1% (27 genes) were significantly stiffening-dependent, as per the definition above. 
Most of the 297 differently transcribed genes were related to signaling (42), disulfide bonding (34), nucleotide binding (26), ATP binding (21), hydrolase (19), transferase (17), secreted (14), DNA binding (14), extracellular matrix (8), DNA replication (8), immunoglobulin domain (6), helicase (5), tyrosine protein kinase (5), collagen (3), DNA repair (3), and DNA damage (3). Figure 2 presents a subset of pathways and genes that are likely involved in corneal mechanical properties. 
Figure 2
 
Signaling pathways with specific genes that were significantly affected by CXL treatment and are likely to be involved in corneal stiffness.
Figure 2
 
Signaling pathways with specific genes that were significantly affected by CXL treatment and are likely to be involved in corneal stiffness.
Stiffening-dependent and -independent Differentially Transcribed Genes
Table 2 and Supplementary Table S1 present genes that were significantly differentially transcribed in a stiffening-dependent and stiffening-independent manner, respectively. Several genes of either subset have been reported previously to show an altered gene expression in keratoconus (references provided in the Tables). 
Table 2
 
Genes that Were Significantly Differently Transcribed in a Stiffening-Dependent Manner
Table 2
 
Genes that Were Significantly Differently Transcribed in a Stiffening-Dependent Manner
Table 2
 
Extended
Table 2
 
Extended
Table 2
 
Extended
Table 2
 
Extended
Figure 3 shows the change in normalized counts of selected genes for the different treatment and control conditions: Enzymatic crosslinking by transglutaminases 2 and 6 was increased significantly after CXL (Figs. 3A, 3B). Also, the expression of polypeptide N-acetylgalactosaminyltransferase 3 and β-1,3-galactosyltransferase 2, both related to the glycosylation of proteoglycans, was increased in crosslinked corneas (Figs. 3C, 3D). The only collagen type that was significantly upregulated after CXL was type IV, which forms part of the basement membrane. All other collagen types (I, VI, XI) were downregulated (Figs. 3E–H). Downregulation also was observed in noncollagenous ECM components, including thrombosponding 4 and keratocan (Figs. 3I, 3J). At the same time, enzymatic glycolysis by means of enolase 1 and transketolase was reduced in crosslinked corneas (Figs. 3K, 3L). 
Figure 3
 
Changes in the normalized counts of transcription for selected genes: (A, B) related to enzymatic crosslinking, (C, D) related to proteoglycan glycosylation, (E–H) structural ECM components, (I, J) other ECM components, and (K, L) related to ECM degradation.
Figure 3
 
Changes in the normalized counts of transcription for selected genes: (A, B) related to enzymatic crosslinking, (C, D) related to proteoglycan glycosylation, (E–H) structural ECM components, (I, J) other ECM components, and (K, L) related to ECM degradation.
Most Affected Signaling Pathways after CXL Treatment
Table 3 presents the two most affected pathways. Seven genes of the ECM receptor interaction pathway and 19 genes of the glycan biosynthesis and metabolism pathway were significantly differentially transcribed. 
Table 3
 
Significantly Differentially Transcribed Genes of the Two Strongest Affected Pathways 1 Week After CXL Treatment
Table 3
 
Significantly Differentially Transcribed Genes of the Two Strongest Affected Pathways 1 Week After CXL Treatment
Correlation analysis
Figure 4 shows genes that strongly correlated (cpearson>0.8, P > 0.05) with thrombospondin 4, a matricellular protein that is involved in tissue remodeling. Among its highest correlated genes were structural extracellular matrix components, including collagen (types I, II, VI, XI), keratocan, and fibromodulin. 
Figure 4
 
Very strongly correlating (cpearson>0.8) genes with thrombospondin 4.
Figure 4
 
Very strongly correlating (cpearson>0.8) genes with thrombospondin 4.
Discussion
We analyzed differential gene transcription induced by CXL treatment and observed a significant remodeling of the ECM, including changes in collagen synthesis, glycan biosynthesis, and proteoglycan glycosylation. 
Fibrillar collagen types I and XI were downregulated after CXL, while the epithelial basement membrane constituting34 collagen type IV was upregulated. Decreased collagen types I and XI transcription potentially results from a reduced collagen degradation after CXL, while increased collagen type IV may be attributed to the recent re-epithelialization and continuing epithelial remodeling. 
The activity of enzymes related to glycosylation (enolase 1, transketolase) and, hence, to ECM degradation, was decreased after CXL treatment. Previously, enolase 1 and transketolase overexpression had been reported in context with increased ECM degradation and cancer invasion.3537 Interestingly, a reduced expression of enolase, transketolase, and the protease inhibitor α2-macroglobulin-like 1 has been reported in keratoconus,3846 which, however, was not able to prevent corneal ectasia. 
In contrast, other genes were inversely differentially transcribed after CXL treatment when compared to keratoconus: collagen type I, keratocan, and thrombospondin 4 were downregulated after CXL, but upregulated in keratoconus.41,47 These ECM components potentially may be involved in extracellular remodeling resulting from the increased corneal stiffness after CXL. Thrombospondin 4 has been identified previously as a mechano-sensing molecule in the cardiac contractile response to mechanical stress showing upregulation in response to hypertension.48 After CXL-treatment, the mechanical stress resistance increases and, as a consequence, the tissue strain decreases, which may have led to the downregulation of thrombospondin 4. In the same line, in keratoconus, where increased tissue strain in the cone region is observed, an overexpression of thrombospondin 4 has been reported. Further potential mechano-sensitive genes may be involved in the molecular signaling after CXL treatment (see Table 2), which in turn could modify the transcription of nonmechano-sensitive genes (see Supplementary Table S1). 
One of the identified stiffening-independent mechanisms of CXL was the increase in enzymatic proteoglycan glycosylation and glycan biosynthesis (Table 3). β1,3-galactosyltransferase 2 is involved in the N-acetyl-D-glucosamine sugar addition on the keratan sulfate proteoglycan. A deficiency in a similar enzyme, β1–4 galactosyltransferase 7, has been associated with Ehlers-Danlos syndrome,49 which manifests in joint hyperelasticity and previously also has been reported in context with corneal curvature abnormalities, including keratoconus, keratoglobus, and cornea plana.50,51 These pathologies likely arise from an alteration of corneal stiffness. Other conditions that affect corneal stiffness include diabetes and aging, in which nonenzymatic glycation is increased.52,53 In contrast with increased enzymatic glycosylation (as observed after CXL), increased nonenzymatic glycation is a random process that makes it less specific in ECM crosslinking. 
Although the 18 mW/cm2 condition was excluded to identify the significantly differentially transcribed genes between crosslinked and control corneas, its expression levels either were in a similar absolute range as the 3 and 9 mW/cm2 conditions, or did confirm the gradient between the 3 and 9 mW/cm2 conditions. This can be considered as an additional quality control, but at the same time emphasizes the fact that CXL protocols differ on the molecular level in an irradiance/time dependent way. 
In absence of an animal model of keratoconus, we used healthy corneas in the experimental groups. It remains to be investigated, if the identified pathways differ in keratoconic corneas. Also, more studies are needed to fully understand the interaction between gene transcription and phenotypic response after CXL. Although it would have been interesting to validate the significantly transcribed genes on the proteomic level, this aspect was out of scope of this study given the high number of identified genes. A further limitation was that we could not separate the differentially transcribed genes according to their origin (keratocytes, epithelial and endothelial cells). Therefore, the results presented here describe the overall response of ECM relevant differential transcription. Future studies may address the individual contribution of keratocytes and epithelial cells, as well as potential effects on wound healing. 
In summary, several target genes potentially related to the biomechanical stability and shape of the cornea were identified. Our findings suggest that corneal stiffening after CXL likely results from a decreased ECM degradation in combination with an increased enzymatic glycosylation, and hence, an altered proteoglycan interaction with collagen fibrils. A proteoglycan-based stiffening after CXL also would be in line with previous findings from x-ray scattering.18 
Acknowledgements
The authors thank Alain Conti for his skilled technical assistance. 
Supported by the Gelbert Foundation (Geneva, Switzerland). 
Disclosure: S. Kling, None; A. Hammer, None; E. A. Torres Netto, None; F. Hafezi, None 
References
Spörl E, Huhle M, Kasper M, Seiler T. Increased rigidity of the cornea caused by intrastromal cross-linking [in German]. Ophthalmologe. 1997; 94: 902–906.
Wollensak G, Spörl E, Seiler T. Treatment of keratoconus by collagen cross linking [in German]. Ophthalmologe. 2003; 100: 44–49.
Hafezi F, Kanellopoulos J, Wiltfang R, Seiler T. Corneal collagen crosslinking with riboflavin and ultraviolet A to treat induced keratectasia after laser in situ keratomileusis. J Cataract Refr Surg. 2007; 33: 2035–2040.
Tomita M, Mita M, Huseynova T. Accelerated versus conventional corneal collagen crosslinking. J Cataract Refr Surg. 2014; 40: 1013–1020.
Kymionis GD, Tsoulnaras KI, Grentzelos MA, et al. Corneal stroma demarcation line after standard and high-intensity collagen crosslinking determined with anterior segment optical coherence tomography. J Cataract Refr Surg. 2014; 40: 736–740.
Peyman A, Nouralishahi A, Hafezi F, Kling S, Peyman M. Stromal demarcation line in pulsed versus continuous light accelerated corneal cross-linking for keratoconus. J Refract Surg. 2016; 32: 206–208.
Ng ALK, Chan TC, Cheng AC. Conventional versus accelerated corneal collagen cross-linking in the treatment of keratoconus. Clin Exp Opthalmol. 2016; 44: 8–14.
Hashemi H, Miraftab M, Seyedian MA, et al. Long-term results of an accelerated corneal cross-linking protocol (18 mW/cm 2) for the treatment of progressive keratoconus. Am J Ophthalmol. 2015; 160: 1164–1170.
Wernli J, Schumacher S, Spoerl E, Mrochen M. The efficacy of corneal cross-linking shows a sudden decrease with very high intensity UV light and short treatment timecorneal cross-linking efficacy. Invest Opthalmol Vis Sci. 2013; 54: 1176–1180.
Hammer A, Richoz O, Mosquera SA, Tabibian D, Hoogewoud F, Hafezi F. Corneal biomechanical properties at different corneal cross-linking (CXL) irradiances. Invest Opthalmol Vis Sci. 2014; 55: 2881–2884.
Aldahlawi NH, Hayes S, O'Brart DP, Meek KM. Standard versus accelerated riboflavin–ultraviolet corneal collagen crosslinking: resistance against enzymatic digestion. J Cataract Refr Surg. 2015; 41: 1989–1996.
Bikbova G, Bikbov M. Transepithelial corneal collagen cross-linking by iontophoresis of riboflavin. Acta Ophthalmol. 2014; 92: e30–e34.
Leccisotti A, Islam T. Transepithelial corneal collagen crosslinking in keratoconus. J Refract Surg. 2010; 26: 942–948.
Hafezi F, Mrochen M, Iseli HP, Seiler T. Collagen crosslinking with ultraviolet-A and hypoosmolar riboflavin solution in thin corneas. J Cataract Refr Surg. 2009; 35: 621–624.
Mazzotta C, Traversi C, Caragiuli S, Rechichi M. Pulsed vs continuous light accelerated corneal collagen crosslinking: in vivo qualitative investigation by confocal microscopy and corneal OCT. Eye. 2014; 28: 1179–1183.
Jacob S, Kumar DA, Agarwal A, Basu S, Sinha P, Agarwal A. Contact lens-assisted collagen cross-linking (CACXL): a new technique for cross-linking thin corneas. J Refr Surg. 2014; 30: 366.
Seiler TG, Fischinger I, Koller T, Zapp D, Frueh BE, Seiler T. Customized corneal cross-linking: one-year results. Am J Ophthalmol. 2016; 166: 14–21.
Hayes S, Kamma-Lorger CS, Boote C, et al. The effect of riboflavin/UVA collagen cross-linking therapy on the structure and hydrodynamic behaviour of the ungulate and rabbit corneal stroma. PloS One. 2013; 8: e52860.
Kontadakis GA, Ginis H, Karyotakis N, et al. In vitro effect of corneal collagen cross-linking on corneal hydration properties and stiffness. Graef Arch Clin Exp. 2013; 251: 543–547.
Cheng X, Pinsky PM. Mechanisms of self-organization for the collagen fibril lattice in the human cornea. J R Soc Interface. 2013; 10: 20130512.
Pinsky PM, Hatami-Marbini H. Modeling collagen-proteoglycan structural interactions in the corneal stroma. Invest Opthalmol Vis Sci. 2011; 52: 4382–4382.
Kanellopoulos AJ. Novel myopic refractive correction with transepithelial very high-fluence collagen cross-linking applied in a customized pattern: early clinical results of a feasibility study. Clin Ophthalmol. 2014; 8: 697.
Raiskup-Wolf F, Hoyer A, Spoerl E, Pillunat LE. Collagen crosslinking with riboflavin and ultraviolet-A light in keratoconus: long-term results. J Cataract Refr Surg. 2008; 34: 796–801.
Koller T, Pajic B, Vinciguerra P, Seiler T. Flattening of the cornea after collagen crosslinking for keratoconus. J Cataract Refr Surg. 2011; 37: 1488–1492.
D'Autréaux B, Toledano MB. ROS as signalling molecules: mechanisms that generate specificity in ROS homeostasis. Nat Rev Mol Cell Biol. 2007; 8: 813–824.
Hancock J, Desikan R, Neill S. Role of reactive oxygen species in cell signalling pathways. Biochem Soc T. 2001; 29: 345–349.
Hoffman BD, Grashoff C, Schwartz MA. Dynamic molecular processes mediate cellular mechanotransduction. Nature. 2011; 475: 316–323.
Wang N, Tytell JD, Ingber DE. Mechanotransduction at a distance: mechanically coupling the extracellular matrix with the nucleus. Nat Rev Mol Cell Biol. 2009; 10: 75–82.
Liu Z, Tan JL, Cohen DM, et al. Mechanical tugging force regulates the size of cell–cell junctions. Proc Natl Acad Sci. 2010; 107: 9944–9949.
Parsons JT, Horwitz AR, Schwartz MA. Cell adhesion: integrating cytoskeletal dynamics and cellular tension. Nat Rev Mol Cell Biol. 2010; 11: 633–643.
Olson EN, Nordheim A. Linking actin dynamics and gene transcription to drive cellular motile functions. Nat Rev Mol Cell Biol. 2010; 11: 353–365.
Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009; 4: 44–57.
Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2008; 37: 1–13.
Nakayasu K, Tanaka M, Konomi H, Hayashi T. Distribution of types I, II, III, IV and V collagen in normal and keratoconus corneas. Ophthalmic Res. 1986; 18: 1–10.
Gao J, Zhao R, Xue Y, et al. Role of enolase-1 in response to hypoxia in breast cancer: exploring the mechanisms of action. Oncol Rep. 2013; 29: 1322–1332.
Hsiao K-C, Shih N-Y, Fang H-L, et al. Surface α-enolase promotes extracellular matrix degradation and tumor metastasis and represents a new therapeutic target. PLoS One. 2013; 8: e69354.
Xu IM-J, Lai RK-H, Lin S-H, et al. Transketolase counteracts oxidative stress to drive cancer development. Proc Natl Acad Sci U S A. 2016; 113: E725–E734.
Acera A, Vecino E, Rodríguez-Agirretxe I, et al. Changes in tear protein profile in keratoconus disease. Eye. 2011; 25: 1225–1233.
Balasubramanian SA, Wasinger VC, Pye DC, Willcox MD. Preliminary identification of differentially expressed tear proteins in keratoconus. Mol Vis. 2013; 19: 2124–2134.
Chwa M, Kenney MC, Khin H, Brown DJ. Altered Type VI Collagen Synthesis by Keratoconus Keratocytesin Vitro. Biochem Bioph Res Co. 1996; 224: 760–764.
Feng X, Chaerkady R, Kandasamy K, et al. Proteomic Profiling of Corneal Stroma in Keratoconus Patients. Invest Ophthalmol Vis Sci. 2010; 51: 3477–3477.
Joseph R, Srivastava O, Pfister R. Differential epithelial and stromal protein profiles in keratoconus and normal human corneas. Exp Eye Res. 2011; 92: 282–298.
Kopsachilis N, Tsaousis KT, Tsinopoulos IT, Kruse FE, Welge-Luessen U. A novel mechanism of UV-A and riboflavin-mediated corneal cross-linking through induction of tissue transglutaminases. Cornea 2013; 32: 1034–1039.
Nielsen K, Vorum H, Fagerholm P, et al. Proteome profiling of corneal epithelium and identification of marker proteins for keratoconus, a pilot study. Exp Eye Res. 2006; 82: 201–209.
Sawaguchi S, Twining SS, Yue B, et al. Alpha 2-macroglobulin levels in normal human and keratoconus corneas. Invest Ophthalmol Vis Sci. 1994; 35: 4008–4014.
Srivastava OP, Chandrasekaran D, Pfister RR. Molecular changes in selected epithelial proteins in human keratoconus corneas compared to normal corneas. Mol Vis. 2006; 12: 1615–1625.
Wentz-Hunter K, Cheng EL, Ueda J, Sugar J, Yue B. Keratocan expression is increased in the stroma of keratoconus corneas. Mol Med. 2001; 7: 470.
Cingolani OH, Kirk JA, Seo K, et al. Thrombospondin-4 is required for stretch-mediated contractility augmentation in cardiac muscle. Cir Res. 2011; 109: 1410–1414.
Freeze HH, Schachter H. Chapter 42: Genetic Disorders of Glycosylation. In: Varki A, Cummings RD, Esko JD, et al., eds. Essentials of Glycobiology. 2nd ed. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press; 2009.
Robertson I. Keratoconus and the Ehlers-Danlos syndrome: a new aspect of keratoconus. Med J Australia. 1975; 1: 571–573.
Cameron JA. Corneal abnormalities in Ehlers-Danlos syndrome type VI. Cornea. 1993; 12: 54–59.
Malik NS, Moss SJ, Ahmed N, Furth AJ, Wall RS, Meek KM. Ageing of the human corneal stroma: structural and biochemical changes. BBA-Mol Basis Dis. 1992; 1138: 222–228.
Wolff SP, Jiang ZY, Hunt JV. Protein glycation and oxidative stress in diabetes mellitus and ageing. Free Radical Bio Med. 1991; 10: 339–352.
Nielsen K, Birkenkamp-Demtröder K, Ehlers N, Orntoft TF. Identification of differentially expressed genes in keratoconus epithelium analyzed on microarrays. Invest Ophthalmol Vis Sci. 2003; 44: 2466–2476.
Song YH, Shiota M, Kuroiwa K, Naito S, Oda Y. The important role of glycine N-methyltransferase in the carcinogenesis and progression of prostate cancer. Modern Pathol. 2011; 24: 1272.
Figure 1
 
Comparison of different approaches to correct for multiple statistical testing. (A) Bonferroni method resulting in 19 significantly differently transcribed genes. (B) Whole-data-set method developed in this manuscript resulting in 297 significantly differently transcribed genes.
Figure 1
 
Comparison of different approaches to correct for multiple statistical testing. (A) Bonferroni method resulting in 19 significantly differently transcribed genes. (B) Whole-data-set method developed in this manuscript resulting in 297 significantly differently transcribed genes.
Figure 2
 
Signaling pathways with specific genes that were significantly affected by CXL treatment and are likely to be involved in corneal stiffness.
Figure 2
 
Signaling pathways with specific genes that were significantly affected by CXL treatment and are likely to be involved in corneal stiffness.
Figure 3
 
Changes in the normalized counts of transcription for selected genes: (A, B) related to enzymatic crosslinking, (C, D) related to proteoglycan glycosylation, (E–H) structural ECM components, (I, J) other ECM components, and (K, L) related to ECM degradation.
Figure 3
 
Changes in the normalized counts of transcription for selected genes: (A, B) related to enzymatic crosslinking, (C, D) related to proteoglycan glycosylation, (E–H) structural ECM components, (I, J) other ECM components, and (K, L) related to ECM degradation.
Figure 4
 
Very strongly correlating (cpearson>0.8) genes with thrombospondin 4.
Figure 4
 
Very strongly correlating (cpearson>0.8) genes with thrombospondin 4.
Table 1
 
Differential Gene Transcription was Computed for a Total of 10 Comparisons Between Different Treatment and Control Groups
Table 1
 
Differential Gene Transcription was Computed for a Total of 10 Comparisons Between Different Treatment and Control Groups
Table 2
 
Genes that Were Significantly Differently Transcribed in a Stiffening-Dependent Manner
Table 2
 
Genes that Were Significantly Differently Transcribed in a Stiffening-Dependent Manner
Table 2
 
Extended
Table 2
 
Extended
Table 2
 
Extended
Table 2
 
Extended
Table 3
 
Significantly Differentially Transcribed Genes of the Two Strongest Affected Pathways 1 Week After CXL Treatment
Table 3
 
Significantly Differentially Transcribed Genes of the Two Strongest Affected Pathways 1 Week After CXL Treatment
Supplement 1
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×