Lysyl oxidase-like 4 involvement in retinoic acid epithelial wound healing

Vitamin A and its active forms (retinoic acids/RAs) are known to have pro-healing properties, but their mechanisms of action are still poorly understood. This work aimed to identify the cellular and molecular processes by which atRA (all-trans RA) improves wound healing, using an in vivo model of mouse corneal alkali burns and an in vitro cellular human corneal epithelial injury model. Regulation by atRA has been studied on most of the cellular events that occur in wound healing. We investigated the direct influence of atRA on a specific target gene known to be involved in the extracellular matrix (ECM) dynamics, one of the pathways contributing to epithelial repair. Our results demonstrate that atRA promotes corneal epithelial wound healing by acting preferentially on migration. The induction of lysyl oxidase-like 4 (LOXL4) expression by atRA in the corneal epithelium environment was established as essential in the mechanism of atRA-dependent wound healing. Our study describes for the first time a direct link between a retinoic-induced gene and protein, LOXL4, and its general clinical pro-healing properties in ECM dynamics.

transcriptionally regulated by vitamin A and involved in various healing process, such as increase in hyaluronic acid or mucin (MUC16) production or modulation of pro-inflammatory cytokine production 17,29 . Moreover, recent studies describe one component of the plasminogen activator system, tissue type plasminogen activator (tPA) (retinoid-regulated gene involved in ECM dynamics) 30 , which must be precisely controlled for ideal wound healing of rat cornea 31 . Focusing on the ECM dynamics, the lysyl oxidase (LOX) family appears to be of particular interest for a full explanation of this process. The LOX family, a group of copper-dependent amine oxidase enzymes, is involved not only in ECM remodeling, but also in many biological functions in tumors or healthy tissues (migration, cell adhesion, metastasis [32][33][34], where it initiates covalent crosslinking between component chains of collagen and elastin. The first identified LOX plays a critical role in ECM formation, stabilization and repair by contributing to the biogenesis of collagen and elastin 35,36 . Four other LOX-like proteins (LOXL 37 , LOXL2 38 , LOXL3 39 and LOXL4 40 ) have been described, sharing 95% homology between their C-terminal domains, containing the catalytic domain 41 . The members of the LOX family are essential, and must be precisely controlled: a deregulation of LOX expression has already been described in connective tissue diseases such as Menkes syndrome 42 and in fibrotic diseases (arteriosclerosis) 43 . Concerning ocular disorders, the deregulation of LOX expression and activities is also the cause of keratoconus 44 , proliferative diabetic retinopathy and rhegmatogenous retinal detachment 45 . Among these LOX-linked disorders, previous results have also shown that in the case of keratoconus the healing process is remarkably inefficient 46 .
In this work, we characterize and demonstrate for the first time a cellular link between the pro-healing properties of vitamin A and a directly induced gene: LOXL4. These results, obtained for cornea wound healing treatment, open the way to a fully documented understanding of the positive effects of the vitamin A pathway, not only in the ocular sphere, but more generally in the clinical management of epithelial wound treatment.

Results
atRA is able to induce corneal wound healing on in vivo and in vitro models. In vivo study models of wound healing showed that treatment of mouse corneal alkali burns with atRA (1 μ M) for 7 days improved ulcer resorption, as seen by fluorescein staining (Fig. 1a). The percentage of remaining ulcer was only 26 ± 7% with atRA compared with the DMSO control group (100%) (Fig. 1b). In vitro scratch assay on HCE cells (Fig. 1c) confirmed that the wound area of cells treated with atRA shrank faster than that of DMSO-treated cells. From 12 h to 48 h (Fig. 1d), the percentage wound area between atRA-and DMSO-treated cells was significantly different at all four test points (12, 24, 36 and 48 h). atRA improves wound healing by acting principally on epithelial cell migration. By what cellular processes does atRA improve wound healing ? Migration and proliferation are the most classical events already described as important in this process, and experiments on HCE cells show that treatment with atRA increases the number of migrating cells by 72 ± 5.6% (Fig. 1g) compared with the DMSO condition. No significant difference was found for cell proliferation (Fig. 1e,f) after atRA treatment. To be sure that atRA did not affect the epithelial characteristics of HCE cells during the healing process, staining with keratin 12 (K12) (specific marker of the intermediate filaments in corneal epithelial cells 47,48 was performed. Results (Fig. 1h, left panel) revealed no significant change in the distribution or intensity of the signal between the two treatment conditions (DMSO or atRA). No vimentin staining was detected after atRA treatment, ruling out epithelial-mesenchymal transition (Fig. 1h, right panel).

LOXL4 is induced by atRA in HCE cells at transcript and protein levels.
To clarify the atRA wound healing action, we first identified the presence of all the isoforms (α , β , γ ) of the retinoic nuclear receptors RAR and RXR in the cellular model of HCE (Fig. 2a, upper panel). Based on targeted transcriptomic results obtained to design retinoic acid target genes potentially involved in wound healing, we classified them according to their physiological pathways using the genomatix program. We identified the LOX family as strongly involved with its presence for 3 out of 6 identified pathways (Fig. 2b). PCR analysis demonstrated that all LOX-like genes were expressed in HCE cells (Fig. 2a, lower panel). Nevertheless, RT q-PCR experiments on HCE cells treated with atRA showed that only LOXL4 was induced from 12 h (1.28 ± 0.1) to 60 h (2.71 ± 0.35) by atRA treatment (Fig. 2c and Supplementary Data 1 for LOX, LOXL1, LOXL2 and LOXL3). Transcript results were confirmed at the protein level by immunocytochemistry experiments (confirmed by ELISA assays, Supplementary Data 2), where protein accumulation was increased as illustrated by a factor of 1.9 ± 0.2 after 48 h of treatment (Fig. 2d,e). atRA induction of LOXL4 gene transcription involves RARα fixation on one DR5 site. To identify the nuclear retinoid actors involved in this LOXL4 regulation, triple mutant (RARα , RARβ , and RARγ ) and RARα rescue MEF cells were used. The expression of the different LOX family members was first demonstrated (Fig. 3a, top panel) in this cellular model. To determine which isoforms of RAR were involved in the activation of LOXL4, transient transfections of RARα , β or γ triple mutant cells were carried out (Fig. 3a, bottom panel), which showed that only the RARα transfection could lead to an atRA induction of LOXL4 expression (1.7 ± 0.1). This result was confirmed by the use of TM rescue RARα MEF cells, where the induction of LOXL4 by atRA was present and stronger (8.5 ± 0.7) (Fig. 3b). Bio-informatics studies using Genomatix ® software highlighted the presence of two DR5 49 in the promoter of LOXL4: DR5-1 (− 1247 to − 1271 pb) and DR5-2 (− 4122 to − 4146 pb) (Fig. 3c, left panel). Transient transfection in HCE cells of both constructs (DR5-1 and DR5-2) demonstrated after atRA treatment that DR5-2 (induction factor of 2.3 ± 0.3) seemed to be involved in the atRA-dependent regulation of LOXL4 transcription. This result was confirmed by mutagenesis studies, showing a total loss of LOXL4 induction when the DR5-2 was mutated, whereas no change was detected for DR5-1 (Fig. 3c, right panel). LOXL4, induced by atRA is essential for in vitro wound healing of HCE cells. Involvement of LOXL4 in the wound-healing process promoted by atRA was investigated using the scratch assay technique and a well-known enzymatic inhibitor of the LOX family, β APN (β-aminopropionitrile). In agreement with the Scientific RepoRts | 6:32688 | DOI: 10.1038/srep32688 literature 50,51 , we first found that a concentration of 500 μ M was not toxic for the corneal cell model by obtaining stability of LDH (lactate dehydrogenase) concentrations in culture media (Supplementary Data 3). Treatment of HCE cells by atRA or a combination of atRA + β APN clearly showed that the significant difference in the percentage of wound healing between DMSO-and atRA-treated cells disappeared when β APN was used in combination with atRA. This was illustrated at 48 h of scratch assay, with the percentage of wound area equal to 100% for vehicle, 40 ± 7% for atRA treatment and 111 ± 9% for cells treated with atRA + β APN (Fig. 4a). The influence of LOX enzymatic inhibition was then tested on cell proliferation and migration as depicted in Fig. 1f,g. No difference in the percentage of proliferating cells could be observed between cells treated with vehicle and cells treated with atRA, or with cells treated with β APN (Fig. 4b). On the other hand, β APN treatment clearly prevented cell migration, as shown in Fig. 4c (108 ± 5% against 172 ± 6% for cells treated with atRA + β APN and atRA respectively).
Because β APN is not specific to any one member of the LOX family, the specific involvement of LOXL4 in wound healing was further confirmed by a siRNA technique. First, the efficiency of siRNA transfection was tested on LOXL4 expression at transcript (affecting induced and endogenous) and protein levels. In detail, RT-qPCR experiments showed that in cells transfected by siRNA scramble, LOXL4 was induced by atRA by a factor of 1.6 ± 0.1 at 48 h of scratch assay, but in cells transfected by siRNA against LOXL4, the expression of LOXL4 was strongly reduced (Fig. 4d). At protein level, immunochemistry experiments revealed that in cells transfected by Nuclei were counterstained with Hoechst (blue). For all graphs, each bar represents mean ± SD. Mann-Whitney U-test; **p < 0.01; ***p < 0.001; ns: not significant.
Scientific RepoRts | 6:32688 | DOI: 10.1038/srep32688 siRNA scramble, LOXL4 staining was stronger for cells treated with atRA and strongly reduced for cells transfected with siRNA against LOXL4 (Fig. 4e). Negative control of gene extinction by siRNA against LOXL4 was also done at the mRNA levels for atRA induced genes (RARβ and STRA6) and for a non-aTRA induced gene member of the lysyl oxidase family (Supplementary Data 4). Quantification of LOXL4 staining (Fig. 4f) confirmed an induction of LOXL4 (1.9 ± 0.2) by atRA in cells transfected by siRNA scramble, and a reduction of LOXL4 accumulation protein (1.0 ± 0.1) when cells were transfected by a siRNA against LOXL4. A scratch assay was carried out in the presence of this siRNA (Fig. 4g), which showed that wound closure for cells transfected with siRNA against LOXL4 reverted to the level found for DMSO treatment (between 94 ± 1% for 12 h and 118 ± 14% for 36 h). LOXL4, induced by atRA, is essential for in vivo wound healing of mouse cornea. Going back to our in vivo mouse model of wound healing, the major role that LOXL4 plays in this process was investigated. Using PCR, we demonstrated the expression of all LOX-like genes (Fig. 5a, top panel), and all the RAR isoforms (α , β , γ ) (Fig. 5a, bottom panel) in mouse cornea. Thus to find out whether LOXL4 was induced by atRA at the protein level in mouse cornea, unburned or burned corneas treated with atRA (or vehicle DMSO) were analyzed by immunohistochemistry. When corneas, burned or unburned, were treated with atRA, a strong accumulation of LOXL4 protein appeared specifically in the epithelium zone (Fig. 5b,c). Lastly, the same experiments on mice were conducted as previously described, but in the presence or absence of β APN. They confirmed that LOXL4 was essential for the wound healing process promoted by atRA. As illustrated in Fig. 5d, fluorescein staining of the ulcer surface demonstrated that the percentage of ulcer remaining after 7 days of treatment was still near 90% for eyes treated with atRA + β APN, against around 26% for eyes treated with atRA (Fig. 5e).

Discussion
Corneal wound healing is an essential clinical step in obtaining the best possible recovery of visual acuity after, for example, eye trauma caused by ocular chemical burns. Our study focused on the effect of atRA on this mechanism using in vivo and in vitro approaches. First, we showed on a mouse model of alkali burns that atRA was able to promote wound healing as previously demonstrated on other animal models and by using vitamin A and/ or its active derivatives [17][18][19][20] . Switching to an in vitro model of wounded HCE cells enabled us to investigate the still-unknown cellular and molecular processes by which atRA acts. Migration, proliferation and differentiation are the most important and best-described events in epithelial wound repair. For the first time, atRA was shown to act preferentially on migration rather than epithelial proliferation and differentiation to stimulate wound healing. These results agree with those obtained by Hattori et al., but extend our knowledge of atRA influence by underlining the importance of the migration process. These authors have already demonstrated in a model of HCE-T cells that atRA combined with nanoparticle vehicles, used at high concentration, has a positive effect on wound healing, with no effect on cell proliferation or differentiation 18 .
Tissue remodeling and especially cell migration are dependent on the balance between the destruction and formation of ECM and its ability to mediate cell adhesion 52,53 . The link between atRA regulated genes and ECM dynamics and remodeling has already been described in several tissues, e.g. the skin 54,55 . For the eye, it has been shown that a treatment with atRA on corneal stromal keratinocytes cultured in vitro increases the production of ECM components such as collagen type 1, and reduces the expression of matrix metalloproteinase (MMP)-1, MMP-3 and MMP-9 (responsible for the disintegration of the ECM) 56 . Interestingly, it has been demonstrated that MMP9 is essential in the wound healing process for regulating re-epithelialization and migration of cells 17,57 , and that an abnormal elevation of its amount is correlated with abnormal re-epithelialization in other tissues 58 . In addition, and again in tissues other than ocular, it has been demonstrated that atRA influences overall balance rather than only ECM degradation, as seen for example in the MMP-2/TIMP-2 ratio in a model of cultured human fetal palate mesenchymal cells 59 . In our study, to describe the direct relation between a vitamin A regulated gene and wound healing, we opted to focus on another component of ECM remodeling, the lysyl oxidase family, which plays a crucial role by acting on collagen biosynthesis 60 . This choice was made because the deregulation of these oxidases leads to severe connective tissue disorders and ocular diseases such as keratoconus [42][43][44] . We demonstrate that one of the LOX-like proteins, LOXL4, is directly induced by atRA at transcript and protein . The residual wound area obtained after DMSO treatment is fixed as 100% (***p < 0.001). Each bar shows mean ± SD. ns: not significant. levels in HCE cells and mouse corneas. In this paper, this positive transcription of LOXL4 by retinoic acid is described for the first time. In the LOX family, only LOX and LOXL2 were already established as retinoic acid target genes, but in tissue environments other than corneal 38 , demonstrating the great plasticity of retinoic acid as a gene regulator.
The LOXL4 regulation by atRA involved nuclear retinoic acid receptor RARα fixation on the second DR5 in LOXL4 promoter (− 4122, − 4146 pb). Furthermore, we found that this gene was essential for corneal wound healing promoted by atRA on HCE cells and on a mouse model of corneal injury by promoting cell migration. LOXL4 regulation by atRA presents the twofold advantage of positively regulating one molecular step (ECM reinforcement with its action on collagen) and one cellular step (epithelial migration). This combined effect of atRA on the corneal environment, added to others previously reported, could synergize in wound healing, reversing corneal keratinization, preserving epithelial barrier function or increasing stromal keratocyte number in vitro 9 .
Ocular injury is found in approximately 15-20% of patients with facial burns 61 . Among the ocular injuries, ocular chemical burns represent some 10% of eye trauma 62,63 . Understanding the pathophysiology of these injuries is essential for improving the management of corneal burns. Several studies on animal models have already demonstrated the positive effect of retinoic acid on corneal wound healing by preventing harmful events, such as neovascularization and opacity of the cornea 64,65 . By describing a gene by which atRA and more generally vitamin A permit wound healing, this work opens the way to the study and direct targeting of other genes involved in ECM formation and remodeling. Our findings could lead on to future clinical therapies using vitamin A alone or in combination with other medications and already long-used surgical treatment, such as amniotic membrane grafts 66,67 . were performed with the approval of the Auvergne Regional Ethics Committee (CEMEA Auvergne) in the animal facility of the School of Medicine -University of Clermont-Ferrand (approval No. 63.113.15). All the experiments were conducted in accordance with the ARVO Resolution for the Use of Animals in Ophthalmic and Vision Research. Twenty-eight white male CD1 mice aged 4-6 weeks were used. At Day 0, under general anesthesia with intraperitoneal injection of pentobarbital (0.82 mg), a standardized chemical corneal burn was performed by placing a filter paper (circular, diameter 3.0 mm) saturated with NaOH (1 N) for 15 s on the sclero-corneal limbus. The wound surface was then washed with Balance Salt Solution ® , and antibiotic (norfloxacin: Chibroxine ® , 0.3% eye drops) was applied on it three times per day. Finally, each injured cornea was treated with eye drops six times daily for 7 days. Mice were divided into four groups of seven mice receiving a treatment with atRA (1 μ M) (Sigma) dissolved in dimethyl sulfoxide (DMSO/atRA vehicle) (Sigma) or with DMSO (0.1%) (control group) or with DMSO and β APN (ß-aminopropionitrile) (500 μ M) (Sigma) or with atRA (1 μ M) and β APN (500 μ M). Same dilution ratio (0.1%) for DMSO was always realized in physiological serum. Wound size was determined by staining with fluorescein (0.5%) and photographing at Days 0 and 7. Wound area was quantified from photographs using imageJ V.1.31. Unburned or burned corneas treated with DMSO or atRA were collected and stored frozen at − 80 °C for PCR experiments, or cryopreserved in OCT for immunohistochemistry.
Cell cultures. Human corneal epithelial (HCE) cells transformed with SV40 (ATCC/CRL11135) were cultured under standard conditions (5% CO 2 , 95% humidified air, 37 °C) as previously described 24  Transcriptomics study. RNA was extracted and quality-controlled using the Agilent RNA 6000 Nano Kit. cDNA was then obtained using the Superscript III First-Strand-Synthesis System for RT-PCR. The cDNA was hybridized on a Human genome 8 × 60 K transcriptomics chip (Agilent, Santa Clara, CA) by a local lab (Hybrigenics, Saint-Beauzire, France). Genes were classified according to fold change and biological pathway using the Genomatix ® pathway system (GePS). The results were furnished by the company after this bioinformatics analysis as a genes list containing gene name, gene abbreviation, log fold change and accession number. Enzyme-linked immunosorbent assay. ELISA assay was assessed using a commercial LOXL4 ELISA kit (DL-LOXL4-Hu; Wuxi Donglin Sci and Tech Development Co. Ltd) on HCE cell lysates according to the manufacturer's instructions. Absorbance of each sample was measured at 450 nm. The concentration of LOXL4 in the samples was then determined by comparing the absorbance of the samples with the standard curve. Results were normalized by measuring the total protein concentration in each sample using a BSA protein assay kit (Pierce). Each sample was loaded on the microplate in duplicate.

Immunohistochemistry and immunocytochemistry experiments. Immunohistochemistry and
immunocytochemistry experiments were performed on cryosections (10 μ m) of mouse corneas (four mice per group) and on HCE cells after 48 h of treatment with atRA or DMSO respectively. Slides were fixed with paraformaldehyde (4% in PBS) and incubated overnight with a rabbit polyclonal anti-LOXL4 antibody (1/200) (ab88186; Abcam) (slides of tissues and cells) or with a mouse monoclonal anti-vimentin antibody (1/40) (090M4817; Sigma-Aldrich) or a goat polyclonal anti-keratin 12 (1/50) (L-20 sc-17099) for slides of cells. Slides of tissues and cells were then respectively incubated with a goat anti-rabbit antibody (Jackson) or anti-mouse (Life Technology) antibodies conjugated with Alexa488 or donkey anti-goat conjugated with cyanine 3 (Jackson) for 2 h. Nuclei were counterstained with Hoechst (bisBenzimide H 33258) (Sigma) (1/10000) for 10 min at RT. Finally, slides were examined under fluorescent microscopes (Zeiss Axiophot or Leica SPE5). Control samples were obtained without the primary antibody, and after incubation of primary antibody with blocking peptide. For induction of LOXL4 protein, quantification of LOXL4 staining was performed using imageJ and, in cells, was confirmed by ELISA test (Supplementary Data 2).
Cell transfections. LOXL4 expression was depleted in HCE cells using a small interfering RNA (siRNA) technique. First, HCE cells placed in six-well plates were treated with atRA or DMSO for 24 h. For transfection, 1 μ L of a commercial siRNA (10 μ M) (SI00142450; QIAGEN) against LOXL4, or 1 μ L of a nonspecific siRNA (10 μ M) (siRNA scramble; SI03650318; QIAGEN) in 125 μ L of optiMEM ® I (1X) medium (Gibco), was mixed with 2.5 μ L of lipofectamine ® RNAiMAX reagent (Invitrogen) in 125 μ L of optiMEM ® I (1X) and incubated for 20 min before transfection in cells. Cells were treated with atRA or vehicle DMSO during the transfection, and for 48 h afterwards. For LOXL4 promoter analysis, HCE cells placed in six-well plates were transiently transfected with 0.5 μ g of each human RARα and RXRα expression plasmids, 1 μ g of different LOXL4 promoter constructs, and 0.05 μ g of pRL-tk vector with 6.4 μ L of nanofectin reagent (PAA). After incubation overnight, the cells were treated for 48 h with atRA or DMSO. Finally, luciferase activity was measured using the Dual-Luciferase ® reporter assay system (Promega) according to the manufacturer's instructions. Triple mutant (RARα , RARβ , and RARγ ) and rescue RARα MEF cells were transiently transfected in six-well plates with 0.5 μ g of each human RARα , β or γ and RXRα expression plasmids in the presence of 3.2 μ L of nanofectin reagent, and simultaneously treated with atRA or DMSO for 48 h. These transfections were repeated three times in duplicate.
Quantitative RT-PCR and PCR experiments. Total RNA was extracted from mouse corneas and cell cultures using RNeasyMini Kit (Qiagen). For cDNA synthesis, 1 μ g of RNA was reverse-transcribed using a Superscript III First-Strand Synthesis System for RT-PCR (Invitrogen). Quantitative RT-PCR reactions were performed using SensiFAST ™ SYBR (Bioline) and Light Cycler ® 480 SYBR Green I Master (Roche). Transcripts were quantified independently three times in three independent experiments, and normalized by the geometric mean of three housekeeping genes (RPLP0, 18S rRNA and β-actin). All the steps followed the MIQE guidelines. Primers used for PCR and qPCR are detailed in Table 1.
Statistics. The data expressed as means ± SDs are the average of duplicates or triplicates of three or four independent experiments. Comparison of means was performed by a nonparametric U-test (Mann-Whitney) or Kruskal-Wallis test using GraphPad PRISM software (GraphPad Software Inc.). Throughout, values were considered significantly different at p < 0.05.