The effect of prior long-term recellularization with keratocytes of decellularized porcine corneas implanted in a rabbit anterior lamellar keratoplasty model

Decellularized porcine corneal scaffolds are a potential alternative to human cornea for keratoplasty. Although clinical trials have reported promising results, there can be corneal haze or scar tissue. Here, we examined if recellularizing the scaffolds with human keratocytes would result in a better outcome. Scaffolds were prepared that retained little DNA (14.89 ± 5.56 ng/mg) and demonstrated a lack of cytotoxicity by in vitro. The scaffolds were recellularized using human corneal stromal cells and cultured for between 14 in serum-supplemented media followed by a further 14 days in either serum free or serum-supplemented media. All groups showed full-depth cell penetration after 14 days. When serum was present, staining for ALDH3A1 remained weak but after serum-free culture, staining was brighter and the keratocytes adopted a native dendritic morphology with an increase (p < 0.05) of keratocan, decorin, lumican and CD34 gene expression. A rabbit anterior lamellar keratoplasty model was used to compare implanting a 250 μm thick decellularized lenticule against one that had been recellularized with human stromal cells after serum-free culture. In both groups, host rabbit epithelium covered the implants, but transparency was not restored after 3 months. Post-mortem histology showed under the epithelium, a less-compact collagen layer, which appeared to be a regenerating zone with some α-SMA staining, indicating fibrotic cells. In the posterior scaffold, ALDH1A1 staining was present in all the acellular scaffold, but in only one of the recellularized lenticules. Since there was little difference between acellular and cell-seeded scaffolds in our in vivo study, future scaffold development should use acellular controls to determine if cells are necessary.


130
In the normal cornea, keratocytes are in general quiescent, that upon injury to the cornea are 131 stimulated into repair phenotypes, but a fibrotic response can lead to scarring and loss of corneal clarity [24]. To 132 investigate the phenotype of stromal cells introduced into the scaffold, three different culture conditions were 133 examined. In the first group (the short-expansion group), culture was for 14 days in the serum-supplemented 134 media already described. In the second group, culture was for 28 days in serum-supplemented medium (the 135 long-expansion group). For the third group, the cell seeded scaffolds were cultured for 14 days in the serum-136 supplemented media and then switched for a further 14 days, (the keratocytic-expansion group), to a serum-137 free keratocyte medium composed of DMEM/F12 medium (HyClone, ThermoFisher) supplemented with 50 138 µg/mL ascorbic acid (Sigma) and 1x insulin-transferrin-sodium selenite (Gibco), giving a final concentration of 139 insulin 10 µg/mL, transferrin 5.5 µg/mL and sodium selenite 6.7 ng/mL. This medium has been shown to 140 stabilise the cells into a more native corneal phenotype [25]involving self-assembly of cell distribution on the 141 scaffold [23]. Scaffolds that were to be implanted decellularized, were treated for the same time and 142 keratocytic-expansion media, without cells.

143
Limbal epithelial cells were isolated by cutting the remaining donor limbal tissue into quarters and 144 incubating with 2.5 mg/mL dispase (Life Technologies, ThermoFisher) for 1 hour. The cells were removed by 145 scraping the limbal surface with a scalpel blade, then pooled, and centrifuged at 170 G for 5 min. They were 146 then resuspended in epithelial medium and 15 µl placed directly onto the scaffold to give a density of 10 5 147 cells/cm 2 . The epithelium medium was composed of ThermoFisher). After overnight freezing to -20 °C, centrifugation was repeated, and the supernatant discarded.

218
The pellet was washed in 70% ethanol in RNAse-free water, centrifuged, the supernatant removed, air-dried, 219 then dissolved in RNAse-free water. RNA yield and purity were measured using a NanoDrop-1000 (ThermoFisher  (Fig 1A) was removed by decellularization and no nuclei could be seen ( Fig 1B).

288
Supporting this, Picogreen quantification of DNA showed a reduction from the 728.9 + 110.2 ng/mg of the 289 native cornea to 14.89 ± 5.56 ng/mg following decellularization ( Fig 1C). The display of fraction size of DNA > 290 1500 base pairs, taken to be supercoiled DNA, was a bright band for the native cornea, while only faint smearing 291 for the decellularized scaffold lanes ( Fig 1F). GAGs were also reduced, visualised by the reduction of alcian blue 292 staining (Fig 1A & B) and quantified by the biochemical assay ( Fig 1D); in the native cornea GAG was 40.13 ± 2.15 293 µg/mg, while in the decellularized this was 10.22 ± 5.44 µg/mg.

294
Collagen was not noticeably reduced by decellularization, as seen by picrosirius red staining (Fig 1A & B) 295 and quantified by the hydroxyproline assay ( Fig 1E). The collagen content rose from 0.59 ± 0.07 mg/mg to 0.63 ± 296 0.07 mg/mg following decellularization, as its proportion increased as other constituents were reduced. From 297 the histological images, the collagen structure appeared similar between the groups, other than the collagen in 298 the anterior third of the decellularized scaffold appeared more separated, as if there had been more osmotic 299 swelling resulting in lacunae.

305
To assess chemical contamination of the scaffold, the concentration of SDS after decellularization was 306 measured to be 52 ± 23 µM.

307
In PBS the scaffold showed haze, but when dehydrated in glycerol there was good optical clarity (Fig 2).

308
Light transmittance was poor in water (0.95% at 300 and 43.72% at 700 nm), but when transferred to glycerol it 309 improved comparable with that of the native tissue. between the three culture conditions (Fig 3D). Deeper cells had a more native organisation with a dendritic 320 morphology and interconnecting pseudopodia that appeared to form a 3D network.

340
Freshly isolated human limbal epithelial cells colonized the surface when seeded onto the scaffold.

341
There was confirmation of native morphology, as the cells adopted the typical cobblestone appearance with a 342 high nucleus-to-cell size ratio (Fig 6A).

346
Axon regrowth into the scaffold for 14 days was demonstrated by βIII-tubulin-positive neurites from the 347 rat DRG extending through to the centre of the scaffold (Fig 6B). The scaffolds were successfully implanted using conventional surgical technique (Fig 7A&B). Some 351 scaffolds appeared contracted prior to surgery, but readily relaxed into place. Implants were easily transferred 353 characteristics. OCT showed that the defect was repaired with close apposition of the scaffold with the native 354 tissue (Fig 8B). Fluorescein staining demonstrated regeneration of the epithelium over most of the scaffold surface 3 364 weeks post-implantation (Fig 9A, B). By 2 months there was complete cover (Fig 9C, D). Over the 365 experimentation period, the scaffolds appeared to decrease in diameter. The sutures became looser, to the 366 point that a few were unburied and were removed for animal welfare. Although the original whiteness of the 367 scaffold did subside, with some areas appearing to be more translucent. Neovascularization developed, but 368 intraocular pressure did not elevate.

371
Three months after implantation, decellularized and recellularized showed comparable epithelial layer 372 re-establishment transparency had not returned to the implanted scaffolds (Fig 7C,D). The location of the 373 opacification appeared to be in the anterior layer.

375
The animals were euthanized, and OCT showed that the scaffold remained in place, but it appeared 376 swollen (Fig 8C). Three regions were discerned: a posterior native tissue, a central scaffold and an anterior area.

377
The eyes were removed, and the corneas used for histological analysis, with the unoperated eye as control. 3.6.1 Post-mortem corneal ECM appearance 379 The stroma of both groups gave the appearance of three layers: a posterior native area, a central 380 scaffold area and an anterior regenerating area (Fig 10). The central scaffold area appeared denser than the 381 other two areas when viewed with picrosirius red staining. The scaffold area was strongly positive for collagen I,

382
indicating retention of the implanted collagen in both groups. Collagen III was localised to the apical area and 383 adjacent to the scaffold in the posterior native area; it was not present in the scaffold itself. Similarly, 384 fibronectin was also located in this apical area, particularly with the recellularized group (Fig 10).  (Fig 11A). There was a sparse, apparently normal, distribution of cells in the central scaffold and 393 posterior native area of both groups (Fig 11A,B). The anterior area appeared to be regenerating, with an 394 increased density of cells. The corneal endothelium was intact in all cases.

395
There was positive cell staining for ALDH1A1 in the epithelium, the native stroma and endothelium (Fig   396  11B). ALDH1A1 positive cells were also sparsely distributed in the central scaffold area of all of the 397 decellularized and one of the recellularized implants. With α-SMA staining (Fig 11C), there was extensive 398 staining in both groups in the anterior regenerating area, indicating fibrotic cells. Stained cells were also present 399 at blood vessels.

431
The human cornea accepts allogenic tissue more readily than other tissues, with normal corneal 432 avascularity conferring some immune privilege, such that HLA matching is not usually required if combined with 433 topical immunosuppression. However, humans have pre-formed antibodies to some epitopes, and these can 434 result in the rejection of xenotransplants. For example, native porcine tissue contains galactose-alpha-1,3-435 galactose (α-Gal) [40] and also N-glycolylneuraminic acid (NeuGc), a non-Gal red blood antigen. These 436 xenoantigens can trigger rejection in pig-to-human implants. In the current study we did not quantify such 437 epitopes, but note that the expression of both Gal and NeuGc have previously been shown to be greatly 438 reduced after decellularization [41]. Additionally, genetically-engineered pigs which express reduced 439 xenoantigens could be used for tissue supply to reduce the human immune response [42].

440
The detergents used for decellularization can also invoke an immunological problem in their own right.