Development of ex vivo organ culture models to mimic human corneal scarring

Cornea collection from the cornea bank

Corneas from human donors (ages 40–80) were collected in agreement with the tenets of the Declaration of Helsinki, including the consent of families for clinical and scientific use, and then stored in the cornea bank of the Hôpital E. Herriot (Lyon, France), at 31 °C in organ culture medium, for a maximum of 21 days before use. Since corneas were collected for grafting purposes, patient selection followed a rigorous set of criteria, as defined by the French Agence de la Biomedicine in conjunction with the Hospices Civils de Lyon. These include absence of viral, bacterial, or parasite infections, low risk of Creutzfeldt-Jacob disease, absence of tumors, lack of treatment with immunosuppressants or blood products, unknown cause of death, lack of exposure to toxic substances, and absence of corneal pathology or corneal surgery. The corneas used for the experiments described here were rejected for grafting because of low endothelial density.

Epithelium removal and excimer laser treatment

Depending on the conditions studied, the corneal epithelium and limbus (denoted collectively by E in Figure 1) were sometimes removed before culture. In these cases (the W and WT conditions), the epithelium was peeled off using a 6 mm lancet shaped Desmarres-Paufique keratoplasty dissector knife (Moria, Antony, France) in such a way as to leave the basement membrane intact, and the limbus was removed using scissors or a trepan (8 mm). Wounding (denoted by W in Figure 1) was performed using an excimer laser (diameter 4 mm, depth 250 µm).

Ex vivo cultures

Corneas were cultured, for 30 days at 37 °C in 5% CO₂ using 10 ml of medium per six-well plate, either fully immersed (IM; epithelium uppermost; Figure 1) or maintained at the air-liquid interface (AL; Figure 1). Medium composition was as follows: Dulbecco's Modified Eagle Medium (DMEM)/Glutamax and Ham’s F-12 (both from GIBCO/Life Technologies, Saint Aubin, France) mixed in a 1:1 ratio with 10% fetal bovine serum (Hyclone/Thermo Fisher Scientific, Brebières, France), 50 µg/ml vitamin C (Bayer, Berlin, Germany), 100 UI/ml penicillin, 20 µg/ml streptomycin (Panpharma, Fougères, France), and 1 µg/ml Fungizone (Bristol Myers Squibb, Rueil-Malmaison, France). The medium was changed three times a week. In some conditions (denoted by T in Figure 1), the medium was supplemented with 10 ng/ml TGF-β1 (R & D Systems Europe, Lille, France). When cultured at the air-liquid interface, the posterior (“endothelial”) face of the cornea was in contact with sterile blotting paper, soaked with culture medium, and placed on a conformer to maintain the curved shape of the cornea, while the anterior (“epithelial”) face was in contact with air (Figure 1). A drop of medium was added twice a day either to the top of the cornea or directly into the wound (when present) to mimic the effect of tears.

Histological analyses and immunofluorescence

After 30 days in culture, corneas were cut into two halves: the first was fixed with formalin before paraffin inclusion while the second was frozen with optimal cutting temperature embedding medium (Tissue Tek). Histology was performed after hematoxylin phloxine saffron (HPS) staining. For immunofluorescence, frozen sections (5 µm) were first blocked in phosphate buffered saline containing 1% (w/v) bovine serum albumin and 5% normal goat serum. Mouse monoclonal primary antibodies were used as follows: anti-α-SMA (Novocastra, Newcastle-upon-Tyne, UK; dilution 1:100), antitenascin C (Dako, Trappes, France; dilution 1:200), antifibronectin (Santa Cruz Biotechnology, Heidelberg, Germany; dilution 1:50), antithrombospondin 1 (Thermo Fisher; dilution 1:100). The secondary antibody was either fluorescein isothiocyanate–conjugated sheep antimouse immunoglobulin G (IgG) (TEBU Bio, Le Perray en Yvelines, France; dilution 1:100) or Alexa Fluor 488-conjugated goat antimouse IgG (Invitrogen/Life Technologies, Saint Aubin, France; dilution 1:100). Control IgG was from Abcam (Paris, France). Nuclei were stained with 4',6-diamidino-2-phenylindole (DAPI) using Vectashield mounting media for fluorescence (from Eurobio/Abcys, Les Ulis, France). Photos were taken with an Eclipse 50i optical microscope using NIS-Elements software (Nikon, Champigny sur Marne, France).

Immunostaining for collagen III was attempted using the following antibodies: goat anticollagen type III (Southern Biotech, Birmingham, AL) and mouse anticollagen type III (Novotec, Lyon, France). However, although positive results on control human skin samples were obtained, the signal observed with human corneas was much less satisfactory. Thus, collagen III was assessed indirectly by measuring RNA levels with real-time PCR (see below).

Semiquantitative evaluation of immunofluorescence was performed by three observers independently using a scale from 0 to 3, where 0=no staining in the wound area (or central area when no wound was present), 1=presence of stained dots in the wound area, 2=marked staining of localized regions of the wound, and 3=intense/uniform staining all over the wound area. Similarly, the scale used for histology was 0=no stromal disorganization, 1=weak stromal disorganization/few activated keratocytes, 2=moderate stromal disorganization/activated keratocytes, and 3=strong stromal disorganization/numerous activated keratocytes. Statistical significance was determined using the Mann–Whitney test, using a p value of <0.05.

Real-time polymerase chain reaction detection of collagen III

Total RNA was extracted from corneal stromas (after the epithelium was removed when present) following culture in the IM/E condition (controls; six samples) and the IM/WT condition (wounding and TGF-β1 in the absence of the epithelium and limbus; six samples). Corneas were then frozen and transferred from liquid nitrogen to 500 µl RLT buffer (RNeasy Mini kit, Qiagen, Courtaboeuf, France)/1% β-mercaptoethanol mix and kept on ice. This mix was then subjected to eight cycles of 20 s shaking followed by 5 min of cooling at 4 °C on a FastPrep-24 System (MP Biomedicals, Illkirch, France). The supernatant was then retrieved, and total RNA was extracted and further purified using an RNeasy Mini kit and RNase-Free DNase set (Qiagen, Courtaboeuf, France). RNA integrity was assessed with a 2000 Nano Kit (Agilent, Massy, France); each sample used had an RNA integrity number of at least 7. Reverse transcriptase PCR was performed using the Invitrogen Superscript III VILO kit according to the manufacturer’s recommendations. Real-time PCR was performed on 2.5 ng cDNA in a Roche (Meylan, France) LightCycler 480 using Roche SYBR Green I Master mix. Primers were from Eurogentec (Angers, France; COL3A1 sense: 5’-GGT GCT CGG GGT AAT GAC G-3’; COL3A1 antisense: 5’-TCC AGG GAA TCC GGC AGT T-3’). Three genes, already validated for human corneas [37], were used as standard baseline genes: TATA box binding protein, fibrosin, and PIH1 domain-containing protein 1. PCR efficiency was determined for each primer set to calculate the expression ratio. The fold-change in gene expression was calculated using the 2-ΔΔCT ratio. PCR products were checked with sequencing (Millegen, Labege, France).

Evaluation of corneal transparency

Corneal transparency was evaluated, after 15 and 30 days, by reading text written in Arial characters (8-point font size) through the cornea. Corneas were classified as “opaque” when the text could not be seen through the cornea.

Air-liquid interface model

Figure 2 (the AL/E column) shows the characterization of human control corneas cultured for 1 month at the air-liquid interface. With HPS staining, the stroma showed the lamellar organization typical of the cornea, and the epithelium remained differentiated and pluristratified. No immunofluorescence for α-SMA or fibronectin was detected. Immunofluorescence for tenascin C was sparse. Immunofluorescence for thrombospondin-1 occurred in a punctate manner throughout the cornea with a slightly more intense signal beneath the basement membrane. None of these control corneas were opaque.

To determine if the wound by itself was sufficient to trigger a change in the histological appearance of the cornea or in the expression of the usual markers of wound healing, we compared the control corneas (AL/E) with the wounded corneas (AL/WE) cultured in the absence of added TGF-β1. Fluorescein staining indicated that complete reepithelialization was obtained after 7 days in wounded corneas, resulting in a pluristratified (five to six layers) epithelium. Bowman’s membrane was not reconstituted, as found in in vivo wounds [38,39]. Using HPS staining, we observed an increase in the number of activated keratocytes, as well as some stromal disorganization (Figure 2, the AL/WE column). Semiquantitative evaluation of several samples (Figure 3A) showed this increase was significant. Immunofluorescence for fibronectin and tenascin C also began to appear (Figure 2, the AL/WE column), but these changes were not significant (Figure 3A). In contrast, expression of thrombospondin-1 was unchanged (compared to the AL/E control) and α-SMA remained undetectable (Figure 2). Wounding alone did not result in corneal opacity.

Since the response to injury alone did not recapitulate all the features of wound healing and scarring, especially the lack of opacity and the failure of the keratocytes to differentiate into α-SMA-expressing myofibroblasts, we next determined the effects of adding TGF-β1 to the ex vivo cultured corneas. TGF-β1 was added to the medium at a concentration of 10 ng/ml. Using a single non-wounded control cornea, addition of TGF-β1 promoted limited expression of α-SMA, only in regions in direct contact with the culture medium (data not shown). In contrast, when the wound and TGF-β1 were present, α-SMA began to be expressed in the stroma (Figure 2, the AL/WET column), but this increase was not significant (Figure 3A). In addition, keratocyte activation was again observed (as for the AL/WE condition) as well as marked stromal disorganization, while the number of cell layers in the epithelium was slightly decreased in the presence of TGF-β1, in agreement with previous reports [24]. Adding TGF-β1 also led to marked increases in the expression of fibronectin, tenascin C, and thrombospondin-1 (compared to the AL/WE condition, Figure 2). In contrast, none of the seven corneas studied in this condition showed any sign of opacity, as found in the absence of TGF-β1 (the AL/E and AL/WE conditions).

We then tested the effect of removing the epithelium and limbus before wounding on the repair process (the AL/WT condition; Figure 1). Interestingly, in this condition, following wounding and the addition of TGF-β1, α-SMA expression was now significantly upregulated (Figure 2, the AL/WT column; Figure 3A) compared to the control condition (AL/E) and the AL/WE and AL/WET conditions. As also found with the AL/WET condition, all other parameters (keratocyte activation and stromal disorganization, expression of fibronectin, tenascin C, and thrombospondin-1) remained highly elevated (Figure 2; Figure 3A). Most strikingly, for the first time, five of the six corneas used for this condition were opaque, as illustrated in Figure 4. In contrast, in the absence of added TGF-β1, these corneas remained transparent (the AL/W condition; data not shown). In addition, even in the presence of added TGF-β1, no opacity was seen after 15 days in culture (data not shown), indicating that haze first appeared between 15 and 30 days.

Immersed model

When corneas were cultured in IM conditions, in all cases the epithelium was reduced in thickness, as previously reported [33,40], and generally consisted of only one cell layer (Figure 5). In control conditions (Figure 5, column IM/E), the stroma showed a typical lamellar organization interspersed with mostly quiescent keratocytes. Furthermore, as with the air-liquid equivalent (Figure 2), no immunofluorescence staining for α-SMA, fibronectin, or tenascin C was detected. Weak staining was seen for thrombospondin-1, in the region of the epithelial basement membrane, as with the air-liquid equivalent. None of the three controls studied were opaque.

Following wounding, but permitting subsequent reepithelialization (the IM/WE condition), with HPS staining activated keratocytes and stromal disorganization/turgescence were seen (Figure 5; the IM/WE column). Semiquantitative analysis (Figure 3B) showed this was statistically significant. In addition, reepithelialization was complete after 7 days, as for the air-liquid equivalent, even though it consisted of a single cell layer, unlike the air-liquid equivalent. As with the air-liquid equivalent, α-SMA remained undetectable. Immunofluorescence staining for fibronectin and tenascin C was however greater than in the air-liquid equivalent, reaching statistical significance, compared to the controls (Figure 3B). Small increases in thrombospondin-1 staining were also observed (Figure 5), but these were not significant (Figure 3B).

Following wounding and with added TGF-β1 (the IM/WET condition), HPS staining continued to show keratocyte activation and stromal disorganization (Figure 5; the IM/WET column), though with less stromal disorganization than in the A/L conditions. As with the air-liquid equivalent, α-SMA began to be expressed in this immersed model, but in this case the increase was statistically significant (Figure 3B), compared to controls. Expression of thrombospondin-1 also increased (Figure 5) but not to a significant extent (Figure 3B). As with the air-liquid equivalent, none of the corneas were opaque.

Finally, as with the air-liquid model, when the epithelium and limbus were removed before wound healing, and in the presence of added TGF-β1, α-SMA expression in the immersed IM/WT condition was significantly upregulated (Figure 3B, Figure 5) compared to controls, as found in the air-liquid culture. Keratocyte activation and stromal disorganization were also seen (though with less stromal disorganization than in the A/L conditions), and fibronectin and tenascin C were strongly expressed, comparable to the IM/WET condition. Thrombospondin-1 was also highly expressed, reaching statistical significance. Most strikingly, however, as with the air-liquid equivalent, most of the corneas were now opaque, as illustrated in Figure 4.

We also investigated changes in the expression of collagen III. Since we were unable to obtain a satisfactorily specific signal with immunofluorescence, despite using different antibodies, we measured collagen III expression with real-time PCR. Collagen III mRNA was 175.7±61.8-fold more abundant in the stroma of corneas cultured in the IM/WT condition than in corneas cultured in the IM/E control condition.

Comparison of the air-liquid interface and immersed models

Using air-liquid and immersed culture, the models for which wound healing occurred in the presence of added TGF-β1 and in the absence of the epithelium/limbus led to the most severe scars, characterized by corneal opacity, keratocyte activation, and stromal disorganization, with marked expression of α-SMA, fibronectin, tenascin C, and collagen III (immersed model). When the results of the semiquantitative analyses were compared, between control conditions (E) and those without epithelium plus TGF-β1 (WT), the same trends were apparent in the air-liquid and immersed cultures (Figure 3C). There were no significant differences between AL/E and IM/E, or between AL/WT and IM/WT, except in the case of α-SMA for which more intense staining was observed in the IM/WT condition compared to AL/WT. In conclusion, air-liquid and immersed cultures appeared suitable for reproducing most of the events of corneal scar formation after injury. A summary of the main findings of this study is shown in Table 1.