Human limbal tissues were harvested from post-mortem donation corneas not suitable for transplantation. Donor corneas were used for research purposes after a written consent was obtained by the donor’s next of kin following the guidelines of the Italian Transplant Centre (CNT, Rome, Italy), the guidelines of the German Medical Association for the collection of donor corneas and managing an eye bank, as well as the Tenets of the Declaration of Helsinki. LSCs were cultured as previously described (23). Briefly, limbal rims were isolated and treated with 3-4 cycles of trypsin digestion at 37°C for 30 min. Isolated LSCs were plated at a seeding concentration of 20 000/cm2 onto lethally irradiated 3T3 fibroblasts using DMEM (Gibco, Massachusetts, USA) supplemented with F-12 Nutrient Mixture (Gibco), 10% fetal bovine serum (Gibco), and supplemented with 4 mM L-glutamine (C.C pro, Oberdorla, Germany), 0.18 mM adenine (Sigma‐aldrich, Missouri, USA), 0.4 μg/mL hydrocortisone (MERCK, Darmstadt, Germany), 5 μg/mL insulin (Sigma‐aldrich), 2 nM triiodothyronine (Sigma‐aldrich), 8.1 μg/mL cholera toxin (Sigma‐aldrich), 10 ng/mL recombinant human epidermal growth factor (EGF) (Peprotech, Hamburg, Germany), and 2% penicillin/streptomycin (Gibco). The medium was changed every other day and LSCs maintained at 37°C in 5% CO₂ humidified atmosphere.

Lentiviral vectors encoding for the GFP gene sequence as reporter and encoding for β2M- or CIITA-specific shRNAs, or a control non-sense shRNA encoding vector were produced by transfection of HEK-293T cells in HYPERFlask^® vessels (Corning, Darmstadt, Germany). For lentiviral vector particle production, HEK-293T cells were cotransfected with the shRNA-sequence encoding plasmid, lentiviral packaging plasmid (psPAX2) and VSV-G envelope expressing plasmid (pMD2.G) using polyethylenimine. After 48 hours, vector-containing culture supernatant was collected, filtered and centrifuged for 3 hours at 20000 rpm at 16°C. Vector pellets were resuspended in LSCs medium and stored at -80°C.

Approximately 2x10⁶ LSCs were transduced in presence of protamine sulphate (Sigma‐Aldrich, Missouri, USA), as previously described (24, 25). Next day, LSCs were enzymatically detached and seeded onto feeder cells with fresh supplemented LSC medium.

Total RNA was isolated from LSCs (RNeasy Mini Kit, Qiagen, Hilden, Germany) and reverse transcribed to cDNA using the high‐capacity cDNA reverse transcription kit (Applied Biosystems, Darmstadt, Germany). Transcripts levels of β2M, CIITA, HLA-DR, ABCB5, p63α and CK12 were analyzed by qPCR using specific predesigned TaqMan Gene Expression Assays (HS00984230_m1, Hs00932860_m1, Hs00219575_m1, Hs02889060_m1, Hs00978344_m1, Hs00165015_m1, respectively; Thermo Fisher, Massachusetts, USA). Samples were analyzed in triplicates and target gene levels were normalized to glyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) (Hs02758991_g1) (Thermo Fisher).

LSC were carefully detached from culture plates using TrypLE (Thermo Fisher, Massachusetts, USA) and incubated for surface marker detection with anti-HLA-ABC PE-conjugated antibodies (clone W6/32; Biorad, California, USA) and anti-HLA-DR APC/Cy7-conjugated antibodies (clone L243; Biolegend, California, USA). ABCB5 protein detection was performed with an anti-ABCB5 unconjugated antibody (polyclonal; Thermo Fisher Scientific) and a PE-conjugated secondary antibody. For intracellular markers, LSCs were permeabilized (IntraPrep Permabilization Kit, Beckman Coulter, Krefeld, Germany) and stained with primary antibodies specific for p63α (clone I504; Abbexa, Cambridge, UK) and CK12 (clone EPR17882; Abcam, Cambridge, UK) followed by PE- and APC/Cy7- conjugated secondary antibodies, respectively. Transduction efficiency was evaluated by detecting GFP expression. Data acquisition was performed using a FACSCanto II Flow Cytometer (Becton, Dickinson & Company, New Jersy, USA) and the results were analyzed using FlowJo software (Becton, Dickinson & Company). Evaluation of anti-human antibodies cross-reactivity was performed after staining 3T3 feeders cells ( Supplementary Figure 1 ).

ADCC Reporter Bioassay core Kit (Promega, Wisconsin, USA) was used following manufacturer’s instructions. Briefly, silenced and non- silenced LSCs were stimulated with interferon (IFN)ɣ (100ng/mL) for 48 hours. Twenty four hours before the assay, 1.25x10⁴ LSCs were seeded in each well of a 96-well plate. At the day of the assay, antibodies specific for human HLA-ABC and HLA-DR were added. Afterwards, LSCs were incubated in the presence of effector cells (Ratio 1:6, T: E) provided in the kit for 6 hours at 37°C. Cell activation rates were evaluated by adding Bio-Glo™ Luciferase assay substrate reagent to the cultures and luminescence was measured using a Synergy 2 Multi-Detection Microplate Reader (Biotek, Winooski, USA).

Human T cells were isolated from healthy donors. First, peripheral blood mononuclear cells were isolated from whole blood by density gradient centrifugation using Lymphosep (C.C pro, Oberdorla, Germany). Negative T cell isolation was performed with the Pan T cell kit (Miltenyi Biotech, Bergisch Gladbach, Germany) following the manufacturer’s instructions. T cells were stained with the cell proliferation dye efluor 670 (Thermo Fischer Scientific, Massachusetts, USA) and exposed to HLA-silenced and non-silenced LSCs which had been previously stimulated for 48 hours with IFNɣ (100ng/mL). Proliferation assay was performed using a target: effector ratio of 1:3 in RPMI 1640 supplemented with 5% AB serum and interleukin (IL)-2 (100U/ml) (Prepotech, New Jersey, USA) for 7 days. At day 5 of the proliferation assay, T cells were re-stimulated with the same native or genetically engineered LSC targets also previously treated with IFNɣ (100ng/mL). After 7 days of experiment, T cell proliferation rates were analyzed by flow cytometry.

All experiments were performed with T cells isolated from healthy donors as described above. Isolated T cells were primed with fully HLA-expressing LSCs which were previously stimulated with IFNɣ (100ng/mL) for 48 hours. T cell priming was performed for 7 days in RPMI 1640 Medium (Lonza, Basel, Switzerland) supplemented with 5% AB serum and IL-2 (100U/ml) (Prepotech, New Jersey, USA), IL-7 (100ng/mL) (Prepotech, New Jersey, USA) and IL-12 (50ng/mL) (Prepotech, New Jersey, USA). Real time cytotoxicity was measured using XCelligence RTCA DP analyzer (Agilent Technologies, California, USA). First, background signal was measured using E-plate 16 (Agilent Technologies, California, USA) with 50µl medium. Silenced and non-silenced LSCs (target cells) were seeded and let adhere for 24 hours. Subsequently, pre-stimulated allogeneic T cells (effector cells) were added in a 1:2 ratio (target: effector). Changes in electrical impedance were expressed as cell index values, which correlates with cellular coverage of electrode sensors at the bottom of each well, and normalized to baseline impedance values before adding effector cells.

NK cells were isolated from healthy donors by negative selection using magnetic‐activated cell sorting (Miltenyi Biotec Bergisch). Isolated NK cells were maintained overnight in RPMI-1640 (Lonza, Basel, Switzerland) supplemented with 5% human serum (C.C pro, Oberdorla, Germany) and IL‐2 (100U/mL) (Prepotech, New Jersey, USA). HLA-silenced and non-silenced LSCs unstimulated or IFNɣ-stimulated (100ng/mL) for 48 hours were used to evaluate NK cell degranulation. At the day of the experiment, LSCs were cultured with NK cells in a target: effector ratio of 1:2 in 96-well plates (Falcon, Corning Brand) for 5 hours at 37°C. After incubation time, NK cells were collected, stained with anti- CD3-APC/Cy7 (clone: HIT3a; Biolegend, California, USA), anti-CD56-AF700 (clone: B159; BD Biosciences, New Jersey, USA) and CD107a-PE (clone: H4A3; Biolegend, California, USA) and analyzed by flow cytometry. NK cell activation was evaluated by measuring CD107a expression as a marker of degranulation.

T cells isolated from peripheral blood of healthy donors were pre-stimulated (primed) by incubating them with fully HLA-expressing LSCs which were previously treated with IFNɣ (100ng/mL) for 48 hours. Priming was performed in RPMI 1640 Medium (Lonza, Basel, Switzerland) supplemented with 5% AB serum and 100U/mL of IL-2 (Prepotech, New Jersey, USA), 100ng/mL of IL-7 (Prepotech) and 50ng/mL of IL-12 (Prepotech) for 7 days. To perform the assay, β2M-, CIITA-silenced LSCs or non-silenced LSCs previously treated with IFNɣ (100ng/mL) for 48 hours were incubated with primed T-cells in a ratio target: effector 1:5 for 6 hours in RPMI 1640 Medium (Lonza, Basel, Switzerland) supplemented with 5% AB serum and IL-2 (100U/mL). Cell culture supernatants were collected and analyzed for cytokines secretion of IL-5, IL-6, IL-8, IL-10, IL-17A, granulocytes-macrophage colony-stimulating factor (GM-CSF), IFNγ, IFNɣ-induced protein 10 (IP-10), and tumor necrosis factor (TNF)α using MILLIPLEX^® Map Human Cytokine/Chemokine Magnetic Bead Panel (Merck Millipore, Darmstadt, Germany) and a Luminex^® 100/200™ analyzer (Luminex Corp., Texas, USA) according to manufacturer’s instructions. Cytokine concentrations were calculated using the Xponent software version 3.1 (Thermo Fischer, Massachusetts, USA).

Data were analyzed using GraphPad Prim 8 software (GraphPad Software, San Diego, USA). Results are shown as mean ± standard deviation (SD). One-way ANOVA or student t test were used to compare differences between groups. Level of significance was set at p<0.05 (*p<0.05, **p<0.01, ***p<0.001, ****p<0.0001).

Immediately after transplantation, LSCs are subjected to an inflammatory microenvironment due to transplantation surgery itself. Under such circumstances, allogeneic LSCs and their mature cell progeny are expected to respond to this inflammatory microenvironment. One of the cytokines playing a relevant role in inflammation is IFNɣ, which is known to induce upregulation of MHC expression (26). Here, we used 48 hours stimulation with IFNɣ (100ng/mL) to mimic an inflammatory microenvironment.

LSCs were stimulated with IFNγ to evaluate their capacity to upregulate HLA class I and class II levels. β2M, CIITA and HLA-DR transcripts as well as surface expression of HLA class I and II proteins were analyzed on stimulated and non-stimulated LSCs ( Figures 1A – C ). After IFNγ stimulation, transcript level analyses showed an upregulation of 9-, 200-, 500 folds of β2m, CIITA and HLA-DR, respectively. These data correlated with an increase of HLA class I and II surface expression compared to unstimulated LSCs. Expression of cell surface HLA class I molecules on LSCs was 2-fold higher (unstimulated: 28.4% ± 14.3%, stimulated: 74.1% ± 21.5%, p<0.01) after exposure to IFNγ. In the case of HLA class II molecules, their surface expression was 10-fold higher than unstimulated LSCs (unstimulated: 7.6% ± 6.3%; stimulated: 71.0% ± 17.6%, p<0.001) ( Figures 1D, E ).

An efficient transduction is the key to ensure LSCs genetic engineering. Lentiviral vectors harboring GFP reporter gene allowed for the assessment of LSC transduction efficiency. All used lentiviral vectors showed transduction efficiencies with means of 68.0% ± 13.1%. Transduction efficiencies for each of the vectors used were 74.9% ± 10.9%, 61.7% ± 10.9% and 67.3% ± 15.6% for shNS (non-sense, control LSCs), shβ2M (β2M-silenced LSCs) and shCIITA (CIITA-silenced LSCs), respectively ( Figures 2A, B ).

Colony formation is an in vitro characteristic of stem cells associated with their capacity of single cell clonal expansion and their stemness (27, 28). In our study, we observed that transduction with lentiviral vectors did not alter LSCs ability to form colonies. Likewise, LSC growth rates and dynamics were similar in either non-transduced (non-TD) LSCs, or transduced with shNS-, shβ2M- and shCIITA-encoding vectors ( Figure 2C ). Already by day 2 to 3 after seeding on lethally irradiated 3T3 feeder cells, LSCs colonies were already observed; and by day 10 to 11 LSCs were confluent.

p63α is associated with the proliferative potential of LSCs as well as self-renewal (29). Evaluation of this marker allowed for the quality assessment of the LSCs during cultivation. After genetic engineering, no difference in p63α transcripts levels between non-transduced (non-TD) LSCs, non-silenced LSCs or silenced LSCs (shβ2M or shCIITA) was observed ( Figure 3A ). Furthermore, detection of p63α protein levels showed no differences between the different conditions ( Figure 3D ). In addition, ATP-binding cassette, sub-family B, member 5 (ABCB5) protein was demonstrated to play a role in corneal development and repair (30). In our settings, we observed that frequencies of ABCB5+ LSCs were not significantly altered by silencing HLA expression ( Figures 3B, E ). Furthermore, evaluation of CK12 transcripts and expression levels on cultivated and genetically engineered LSCs remained comparable to those measured in non-modified LSCs ( Figures 3C, F ).

HLA class I molecules maturation and loading depend on β2M and a failure of its expression affects HLA class I presentation on the cell surface. In addition, a disruption or interruption over CIITA expression, which is the master regulator of all HLA class II genes transcription, affects directly their expression (31, 32). LSCs transduced with lentiviral vector encoding for shβ2M and shCIITA resulted in an HLA silencing effect ( Supplementary Figure 2 ). Importantly, genetically engineered LSCs after the treatment with IFNɣ (100ng/mL) for 48 hours mimicking a pro-inflammatory microenvironment maintained the downregulation of HLA expression. β2M transcript levels analysis showed a decrease of 80.2% ± 24.1% (p<0.001) compared to the levels detected on non-TD cells ( Figure 4A ). Remarkably, β2M transcript silencing caused a 50.4% ± 20.6% (p<0.01) downregulation on HLA class I surface expression ( Figure 4C ). Comparable results were detected on LSCs transduced with shCIITA-encoding vector. CIITA and consequently HLA-DR transcripts levels were reduced by 68.1% ± 21.9% (p<0.01) and 91.6% ± 9.1% (p<<0.0001), respectively ( Figure 4B ). CIITA downregulation led to a reduction on the expression of HLA class II molecules to 52.6% ± 10.7% (p<0.01) in comparison to LSCs non-TD ( Figure 4D ).

Graft rejection might be triggered by pre-formed or de novo donor specific antibodies (33). In order to evaluate the allogeneic humoral response potentially targeting LSCs and the effect of silencing HLA expression, antibody-dependent cellular-mediated cytotoxicity (ADCC) assays were performed. Anti-HLA specific antibodies mediated a significant reduced cytotoxic effect over HLA class I silenced LSCs (4370 ± 336.5 RLU, p<0.0001) compared to non-silenced (14938 ± 3220 RLU) or non-TD (20370 ± 4666 RLU) ( Figure 5A ). Similarly, HLA class II silenced LSCs showed decreased cell lysis rates due to ADCC in comparison to non-silenced (2579 ± 223.2 RLU vs 4128 ± 127.4 RLU, p<0.001) or non-TD (2579 ± 223.2 RLU vs 4019 ± 409.6 RLU, p<0.001) LSCs ( Figure 5B ).

Notably, proliferation of non-primed T cells was significant reduced by the downregulation of HLA class I (11.9% ± 5.7%; p<0.05) or HLA class II (12.2% ± 6.7%; p<0.05) expression on LSCs in comparison to non-TD (20.6% ± 11.5%) or non- silenced cells (16.6% ± 7.2%) LSCs ( Figures 5C, D ).

Furthermore, the strength of primed T-cell cytotoxic responses targeting HLA class I or II-silenced, non-silenced or non-TD LSCs was evaluated. β2M- and CIITA-silenced LSCs showed higher cell survival rates (CI, cell Index) than HLA-expressing control LSCs (Non-silenced and non-TD) ( Figure 5E ). For instance, at 15 hours after initial contact with primed T cells, shβ2M- and shCIITA-expressing LSCs showed a higher CI (3.3 ± 1.3, p<0.001 and 2.9 ± 0.8, p<0.01) in contrast to non-TD or non-silenced LSCs (0.8 ± 0.2 or 1.0 ± 0.1). In contrast to HLA-expressing cells, higher cell survival indexes were maintained over time (24 h) for HLA-silenced LSCs ( Figure 5F ).

Cytokines and chemokines secreted by T cells are crucial regulators of the immune homeostasis. Increases of pro-inflammatory cytokine levels support graft rejection and failure after transplantation (34, 35). To evaluate the impact of HLA silencing on LSCs in T cell cytokine secretion, allogeneic T cells were co-cultured with silenced and non-silenced LSCs. Higher levels of T-cell cytokine secretion were observed in non-TD and shNS LSC groups. In comparison with non-TD or non-silenced, HLA class I silenced LSCs induced a significantly reduced pro-inflammatory cytokine release [Interleukin (IL)-6 (p<0.01), IL-8 (p<0.01), IFNγ (p<0.0001), TNFα (p<0.0001) and GM-CSF (p<0.0001)]. Moreover, downregulation of HLA class II molecules on LSCs using shRNA-encoding vectors triggered a reduction of IL-6 (p<0.01), IL-8 (p<0.05), IFNγ (p<0.001), TNFα (p<0.0001), GM-CSF (p<0.05) release. Furthermore, IL-17a, which has been described to play a relevant role in cornea transplantation (36) was also reduced [shβ2M-silenced LSCs (p<0.001) and CIITA-silenced LSCs (p<0.0001)] compared to non-TD or non-silenced LSCs ( Figure 6 ).

The activation of NK cells leads to a strong cytolytic effect against their target cells. NK cell cytotoxicity involves the release of granules containing perforins and granzymes and associated with the transient surface expression of lysosomal-associated membrane protein-1 (CD107a) (37). The expression of this molecule on the surface of NK cells has been used as an indirect marker of NK cell cytotoxic function (38). In order to evaluate the effect of HLA class I downregulation on LSCs towards NK degranulation, non-silenced and HLA class I-silenced LSCs were co-cultured with NK cells isolated from healthy donors. The expression of degranulation marker (CD107a) was observed to be similar between Non-TD, shNS and shβ2M LSCs, suggesting that silencing of HLA class I expression on the surface of LSCs does not induce NK cell activation ( Figure 7 ). Hence, the residual HLA class I expression on LSCs protect them from being a target for NK cells.

These observations suggest that silenced LSCs are protected against humoral and cellular allogeneic immune responses.