Engineering of immune checkpoints B7-H3 and CD155 enhances immune compatibility of MHC-I-/- iPSCs for β cell replacement.

Induced pluripotent stem cells (iPSCs) represent a source from which β cells can be derived for diabetes replacement therapy. However, their application may be hindered by immune-mediated responses. Although abrogation of major histocompatibility complex class I (MHC-I) can address this issue, it may trigger natural killer (NK) cells through missing-self recognition mechanisms. By profiling the relevant NK-activating ligands on iPSCs during in vitro differentiation into pancreatic β cells, we find that they express high levels of B7-H3 and CD155. Hypothesizing that such surface ligands could be involved in the amplification of NK-activating signals following missing-self, we generate MHC-I-deprived B7-H3-/-, CD155-/-, and B7-H3-/-/CD155-/- iPSCs. All engineered lines correctly differentiate into insulin-secreting β cells and are protected from cell lysis mediated by CD16dim and CD16+ NK subpopulations both in vitro and in vivo in NSG mice. Our data support targeted disruption of NK-activating ligands to enhance the transplant compatibility of MHC-I-/- iPSC pancreatic derivatives.

Introduction

Over the past decade, induced pluripotent stem cells (iPSCs) have gained considerable attention as a potentially unlimited source of therapeutically relevant cells for regenerative medicine. However, clinical application of iPSCs faces some disease-specific issues. In particular, although patients with type 1 diabetes (T1D) could benefit from therapies based on autologous iPSC-derived insulin-producing cells, the presence of preexisting autoimmune responses against β cell antigens represents a significant hindrance. Likewise, cell therapy with allogeneic iPSCs would result in CD8+ T cell-mediated allograft rejection owing to surface expression of highly polymorphic human leukocyte antigens (HLAs) (Taylor et al., 2011). Several solutions have been proposed to overcome these obstacles, including an HLA-typed stem cell biobank (Taylor et al., 2012), the inactivation of major histocompatibility complex class I (MHC-I) and class II (MHC-II) (Mattapally et al., 2018), and the overexpression of the antiphagocytic CD47 (Deuse et al., 2019) or the CD8+ T cell-inhibitory molecules CTLA4-Ig/PD-L1 (Rong et al., 2014). Among these strategies, the most promising remains the abrogation of MHC-I by genetic inactivation of the β2-microglobulin (B2M) gene (Wang et al., 2015; Norbnop et al., 2020; Petrus-Reurer et al., 2020), the product of which is required for surface expression of classical and non-classical HLA-I molecules. However, MHC-I-deprived iPSCs and their derivatives can trigger natural killer (NK) cytotoxicity via missing-self recognition mechanisms (Bix et al., 1991; Liao et al., 1991). The latest approaches to avoid this drawback have been the ectopic expression of tolerogenic non-classical HLA-I molecules, such as HLA-E (Gornalusse et al., 2017) and HLA-G (Zhao et al., 2014; Shi et al., 2020), since these molecules are high-affinity ligands of the inhibitory CD94:NKG2A and LIR-1 receptors, expressed on most NK cells (Navarro et al., 1999). However, the responsiveness of NK cells depends on a fine balance between cross-antagonizing activating and inhibitory signals, which are differentially integrated by NK cell subsets and may vary according to disease context. For this reason, a single engineering strategy might not be sufficient to fully protect transplanted cells from NK immune attack. To this end, NK-activating signals may represent interesting targets to circumvent missing-self recognition. Moreover, there are no studies showing the expression pattern of activating ligands in iPSCs and their pancreatic β cell derivatives and how these ligands can affect the outcome of future cell replacement strategies in T1D. In this study, we characterize the expression profile of the main NK-activating ligands on human iPSCs before and during in vitro differentiation into functional β cells. Furthermore, we investigate the impact of the most expressed extracellular activating ligands B7-H3 and CD155 on missing-self recognition mechanisms, assessing the activity of different NK cell subsets against MHC-I-deprived iPSCs and their pancreatic derivatives following the abrogation of such molecules through CRISPR-Cas9 technology. Finally, we corroborate the relevance of activating signal exploitation by in vivo experiments on NK-mediated acute allograft rejection, proposing the knockout of B7-H3 and CD155 in iPSC-derived β cells with an MHC-I-null background as an alternative strategy to evade NK cytotoxic activity.

Results

Classical and non-classical HLA class I and HLA class II molecule expression on iPSCs before and during in vitro differentiation into pancreatic β cells

The expression of allogeneic HLA class I molecules can strongly influence the immune response by promoting allograft rejection by CD8+ T lymphocytes and modulating NK cell function via the killer cell immunoglobulin-like receptors (KIRs) and the CD94:NKG2A/C receptors. Analogously, HLA class II recognized by islet-infiltrating CD4+ T cells may play an immunopathogenic role, affecting transplant outcome. We thus evaluated, by flow cytometry, the expression of B2M protein, classical (HLA-A, -B, and -C) and non-classical (HLA-E and HLA-G) HLA-I molecules, and the HLA-DR/-DP/-DQ class II molecules on the CGTRCiB10 iPSC line, both before and during in vitro differentiation to β-like insulin-producing cells, with or without interferon (IFN)-γ, a well-known inducer of MHC expression in many tissues. Undifferentiated cells displayed high levels of B2M (1,652 ± 70 mean fluorescence intensity [MFI]) and HLA-A/B/C (688 ± 183 MFI) (Figures 1A and 1B), whereas we observed significant downregulation of both B2M (10-fold decrease) and HLA-A/B/C (>5-fold decrease) when iPSCs differentiated to posterior foregut (PF), followed by a further reduction in transitioning to pancreatic endoderm (PE; >10-fold decrease). On the other hand, the late stages of differentiation were characterized by reexpression of B2M (431 ± 66 MFI) and HLA-A/B/C (368 ± 91 MFI). No expression of HLA-E/G and HLA-II molecules was detected either on undifferentiated cells or during pancreatic differentiation (Figure 1B). Exposure to IFN-γ increased from 2- to 5-fold the surface levels of B2M and HLA-A/B/C and induced de novo expression of HLA-E/G molecules in undifferentiated iPSCs and during all the steps of differentiation, while no effect was observed on expression levels of HLA class II molecules (Figures 1A and 1B).

Figure 1

Classical and non-classical HLA class I and HLA class II molecule expression on iPSCs before and during in vitro differentiation into pancreatic β cells

(A) Representative histograms of surface staining with monoclonal antibodies (mAbs) for B2M, classical/non-classical HLA-I, and HLA-II molecules on undifferentiated iPSCs, under basal culture conditions (red) and after treatment with IFN-γ (orange; 10 ng/mL for 12 h). Staining with the related isotype control is shown in light blue.

(B) B2M, HLA-II, and classical/non-classical HLA-I expression during pancreatic differentiation of iPSCs in the presence or absence of IFN-γ (10 ng/mL). Data are expressed as mean fluorescence intensity (MFI) and are representative of a total of three experiments (n = 3), intended as independent in vitro differentiations of the same iPSC clone (CGTRCiB10) at different passages. iPSC, undifferentiated iPSCs; PF, posterior foregut; PE, pancreatic endoderm; EN, endocrine cells; iBeta, iPSC-derived β cells. Error bars indicate standard deviation. Statistical significances were obtained by one-way ANOVA following Tukey’s post hoc test; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Profile of NK-activating ligands on iPSCs and during their in vitro differentiation into pancreatic β cells

DNAM-1 (CD226) (Shibuya et al., 1996), NKG2D (Lanier, 2015), and the members of the natural cytotoxic receptors (NCRs) family NKp30 (CD337) and NKp46 (CD338) are pivotal activating NK receptors (Koch et al., 2013; Kruse et al., 2013). We evaluated mRNA and protein expression on an iPSC line of the CD226 ligands PVR (CD155) and Nectin2 (CD112); the NKG2D ligands MICA, MICB, ULBP1, ULBP2, ULBP3, ULBP4 (RAET1E), ULBP5, and ULBP6; the CD337 ligands BAG6/BAT3 and VIM; and the CD338 ligand properdin (CFP). In addition, we assayed the tumor-specific CD337 ligand B7-H6 and its closely related co-stimulatory molecule CD276 (B7-H3) (Brandt et al., 2009).

Both mRNA and protein quantification confirmed the expression of some activating NK receptor ligands in undifferentiated iPSCs (Figures 2A–2C). As shown in Figure 2B, 98.9% ± 0.8% and 85% ± 6.4% of iPSCs were positive for B7-H3 and CD155, respectively. MICA+ and RAET1E+ cells represented 65.5% ± 6% and 36.9% ± 9.3% of the overall iPSC population, respectively. Immunofluorescence assay demonstrated the intracellular expression of the CD337 ligands BAG6/BAT3 and VIM (Figure S2A), while no significant expression of B7-H6, MICB, CFP, or other ULBPs was detected (Figures 2A, 2B, and S2A). CD112 displayed transcriptional levels significantly higher than those of other ligands (p < 0.001; Figure 2B), but protein appeared to be totally retained in the cytoplasm (CD112+ cells: intracellular staining >99%, membrane staining 0%; Figure S2B).

Figure 2

Expression pattern of extracellular NK-activating ligands on undifferentiated iPSCs and during in vitro differentiation into pancreatic β cells

(A) Representative histograms of surface staining with mAbs for NK ligands (red) on undifferentiated iPSCs. Staining with the related isotype control is shown in light blue.

(B) Percentage of positive cells (mean ± SD; n = 6 independent experiments) for surface NK ligands measured by flow cytometry (top) and relative mRNA expression (mean ± SD; n = 3 independent experiments using batches of the same cell line at different passages) of NK ligands assayed by quantitative real-time-PCR on CGTRCiB10 iPSCs. GAPDH was used as a normalizer gene and data are plotted on a logarithmic scale (bottom).

(C) Surface NK ligand expression by flow cytometry in both iPSC (CGTRCiB10, CB4, NeoFiWT, DRI1#11, DRI2#3) and embryonic stem cell (H1-hESC) lines (mean ± SD; n = 6 independent experiments). Statistical significances refer to difference between samples in the expression of related markers and were inferred by one-way ANOVA; ∗∗∗p < 0.01.

(D) Percentage of cells (mean ± SD; n = 3 independent in vitro differentiations) positive for surface NK ligands on iPSCs and during different stages of pancreatic differentiation. Fold change in separation index (mean ± SD; n = 3 independent in vitro differentiations) between the stained population and the negative one (isotype control) of the CGTRCiB10 clone as relative variation in fluorescence intensity during pancreatic differentiation compared with undifferentiated cells.

(E) Relative mRNA expression (mean ± SD; n = 3 independent in vitro differentiation experiments on the CGTRCiB10 clone at different passages) of the NK ligands during pancreatic differentiation (CGTRCiB10 clone). GAPDH was used as the normalizer gene and data are reported as fold change over undifferentiated cells.

(F) Representative histograms of surface staining with mAbs for NK ligands (red) or isotype control (light blue) on terminally differentiated CGTRCiB10 cells (iBeta). iPSC, undifferentiated iPSCs; PF, posterior foregut; PE, pancreatic endoderm; EN, endocrine cells; iBeta, iPSC-derived β cells. Error bars indicate standard deviation. Each experiment has been carried out with the indicated cell line at different passages and states of confluence. Statistical significance obtained with two-tailed t test refers to the comparison with iPSC stage; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

We then wondered if the expression pattern of these ligands was a prerogative of our model or a common feature of stem cell lines, either induced pluripotent or embryonic. Consequently, we evaluated the surface expression of NK ligands in four other iPSC lines (CB4, NeoFiWT, DRI1 clone #11, DRI2 clone #3) and in the H1-hESC human embryonic stem cell line. As shown in Figure 2C, all stem cell lines expressed comparable levels of both CD155 and B7-H3. Interestingly, we found differences in surface expression of CD112, which was exposed at different levels on the plasma membrane of the CB4 (23.54% ± 5.39%), NeoFiWT (12.47% ± 3.38%), DRI1#11 (32.19% ± 3.98%), and H1 (35.54% ± 8.32%) cell lines. Significant variability was observed for MICA and RAET1E levels (ANOVA, p < 0.01; Figure 2C). A small percentage of CB4 and NeoFiWT was also positive for MICB, whereas the embryonic cell line was characterized by 22.42% ± 3.74% of MICB+ cells, according to previously reported findings (Suárez-Alvarez et al., 2010).

Since B7-H3 represented the most expressed surface ligand on iPSCs, we chose to extend our investigations to other B7 family genes. The B7 family consists of structurally related immune-regulatory surface proteins (Collins et al., 2005; Xu et al., 2016). To evaluate in depth the expression profile of B7 family genes, we analyzed a single-cell RNA (scRNA) dataset previously collected by our laboratory (Pellegrini et al., 2020), following validation of the activating (B7-H2 and B7-H6) and the inhibitory (B7-H1 and B7-H4) ligands by quantitative real-time PCR. We confirmed by scRNA sequencing the high expression of B7-H3 in undifferentiated iPSCs, whereas low levels of B7-H2 were detected only by quantitative real-time PCR experiments (Figures S2C and S2D). B7-H6 was detected in few cells of the scRNA dataset, and validation of its transcriptional levels confirmed spurious or null expression of this ligand in our model (Figure S2D).

To evaluate the impact of inflammatory cytokines on NK-activating ligand expression, we exposed iPSCs to IFN-γ and/or tumor necrosis factor (TNF)-α. IFN-γ downregulated the CD314 ligands MICA and RAET1E, whereas TNF-α increased the RAET1E expression only (Figure S3). Expression of NK-activating ligands was also evaluated during in vitro differentiation into insulin-producing cells (Figures 2D–2F, S2F, and S2G). B7-H3 mRNA levels increased over 2-fold during transition to PE, then in the last stages they returned to those observed in undifferentiated cells. Surface expression of B7-H3 does not change significantly, except at the end of differentiation, where we observed a higher variability, likely due to the population heterogeneity of the terminally differentiated cells (Figures 2D and 2E). CD155 displayed a slightly variable, swinging surface expression profile during differentiation, although its mRNA levels grew steadily, increasing over 20-fold in terminally differentiated β cells (iBeta) compared with the undifferentiated ones (Figures 2D and 2E). CD112 was retained in the cytoplasm until the end of differentiation, when surface expression emerged in about 32.17% ± 2.95% of the cells. In contrast, both MICA and RAET1E decreased during differentiation, since only 8.23% ± 1.34% and 2.43% ± 1.01% positive cells were detected, respectively (Figure 2D). A gradual increase during differentiation was observed for the intracellular NCR ligands VIM and CFP (over 10- and 80-fold, respectively), while the CD337 ligand BAG6/BAT3 expression levels remained essentially stable (Figures S2F–S2H). Commitment to endocrine pancreas showed cells expressing variable levels of B7-H1, B7-H2, B7-H4, B7-H5, and B7-H6 as well (Figure S2I). The scRNA-derived expression data were confirmed by quantitative real-time PCR only for B7-H2 and B7-H4 (p < 0.001) (Figure S2L). No expression of MICB or other ULBPs was detected during differentiation (data not shown).

Blocking of NK-activating receptor ligands inhibits in vitro killing of MHC-I−/− iPSC lines by an allogeneic CD16+ NK cell subset

To evaluate the impact of NK-activating ligands on missing-self recognition mechanisms, we generated MHC-I-null iPSCs by CRISPR-Cas9-directed disruption of the B2M gene, preventing HLA-I surface expression. The B2M gene targeting efficiency was about 40%–50%, and MHC-I-null clones were selected by cell sorting based on the expression of B2M and HLA-ABC (Figures S4A and S4B). Moreover, we also engineered B2M−/− iPSC subclones to induce a stable surface expression of HLA-E∗0103 molecules bound to the HLA-G-derived leader peptide VMAPRTLFL (Figures S4A and S4B), as in the previously described immune-escape cell model (Gornalusse et al., 2017). Due to the low homologous recombination efficiency of the pepB2M-HLA-E-IRES-eGFP construct, we subjected the cells to nocodazole treatment to temporarily block them in G2/M phase, when direct homology-directed repair (HDR) is favored over the non-homologous end-joining (NHEJ) repair system (Lin et al., 2014). Thus, we increased HDR efficiency up to 7%, allowing cell sorting of HLA-E-expressing cells by using GFP as a selection marker (Figure S4B).

B2M−/− and B2M−/−/HLA-E+/+ iPSCs maintained pluripotency, as demonstrated by the OCT4, SOX2, and NANOG expression (Figure S4C), and preserved their ability to differentiate into the three germ layers in vitro (data not shown). Edited cells retained a normal karyotype and remained phenotypically stable over 30 passages after gene manipulation (data not shown). Furthermore, the expression pattern of NK-activating ligands on both B2M−/− and B2M−/−/HLA-E+/+ iPSCs was not affected by gene editing (Figure S4D).

We then tested whether allogeneic CD8+ T lymphocytes and donor-derived CD16+ NK cells recognized in vitro unedited B2M−/− and B2M−/−/HLA-E+/+ iPSCs (Figure 3). Allogeneic CD8+ effector T cells were derived by priming donor-derived peripheral blood mononuclear cells (PBMCs) with gamma-irradiated HLA-expressing target cells (MOLT-4, HLA-A∗0101), to boost for T cell recognition of our model of wild-type iPSCs (CGTRCiB10, HLA-A∗0101). CD16+ NKs were magnetically sorted from PBMCs of healthy donors, and each donor-derived CD16+ NK primary line was used individually, as a biological replicate, after preactivation by exposure to IL-2 and IL-12. Unedited iPSCs were killed by CD8+ cytolytic T lymphocytes (52.3% ± 5.82% propidium iodide-positive [PI+] cells at 1:1 effector:target [E:T] ratio; 71.67% ± 1.17% PI+ cells at 5:1 E:T ratio), comparable to the positive control cell line MOLT-4 (54.67% ± 2.32% PI+ cells at 1:1 E:T ratio; 72.12% ± 3.32% PI+ cells at 5:1 E:T ratio). As expected, B2M−/− and B2M−/−/HLA-E+/+ iPSCs escaped CD8+-mediated killing (95% CI 0.41%–1.14% and 0.98%–5.63% PI+ cells, respectively), showing cell death levels equal to those of the negative control cell line Jurkat (95% CI 1.21%–2.29% PI+ cells). On the other hand, B2M−/− iPSCs triggered CD16+ NK cell killing by missing-self recognition mechanisms (46.21% ± 2.82% PI+ cells at 1:1 E:T ratio; 63.24% ± 2.41% PI+ cells at 5:1 E:T ratio). As expected, surface expression of HLA-E molecules protected the B2M−/− iPSCs from NK immune attack (2.95% ± 1.37% PI+ cells at 1:1 E:T ratio; 2.85% ± 0.36% PI+ cells at 5:1 E:T ratio), bringing the levels of cell death even below those of the unedited cells (3.12% ± 0.58% PI+ cells at 1:1 E:T ratio; 11.91% ± 2.88% PI+ cells at 5:1 E:T ratio) (Figure 3A).

Figure 3

Blocking of either NK-activating receptors or their ligands inhibits killing in vitro of MHC-I−/− iPSC lines by allogeneic CD16+ NK cells

(A) CD8+ T cells and NK killing against unedited and edited iPSCs with two target:effector ratios. Jurkat and MOLT-4 were used as positive and negative controls, respectively.

(B) Cytotoxicity test performed on wild-type, B2M−/−, and B2M−/−/HLA-E+/+ iPSCs after preincubation of NK cells with blocking antibodies against NK receptors. Irrelevant IgG antibody was used as mock control. Assays were performed with two target:effector ratios.

(C) Cytotoxicity test performed after preincubation of wild-type, B2M−/−, and B2M−/−/HLA-E+/+ iPSCs with blocking antibodies against NK-activating ligands. Irrelevant IgG antibody was used as mock control. Assays were performed at 1:5 target:effector ratio. In all cytotoxicity tests, percentage of PI-positive target cells was measured by flow cytometry after co-incubation with effectors for 4 h. Effector-mediated lysis was normalized on basal cell death percentage occurring in target cell lines after incubation without effectors. Error bars indicate standard deviation. Statistical significance obtained by two-tailed t test refers to comparison with mock control; n = 3 independent experiments carried out using a different donor each time; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

By using blocking monoclonal antibodies, we investigated which specific activating receptors, if any, mediated NK killing of B2M−/− iPSCs (Figure 3B). Concordant with the expression pattern of activating ligands on target cells, preincubation of CD16+ NK cells with anti-CD226, anti-NKG2D, and anti-CD337 blocking antibodies prevented killing of B2M−/− iPSCs (ANOVA, p < 0.001). Furthermore, by using blocking antibodies against CD94:NKG2A, we observed a decrease in NK inhibition by B2M−/−/HLA-E+/+ iPSCs, while by antagonizing the CD94:NKG2C receptor we highlighted a significant (p < 0.05) reduction in B2M−/−/HLA-E+/+ iPSC lysis levels compared with mock control (Figure 3B). The percentage of NK-mediated lysis was also affected by treatment with antibodies against NK-activating ligands on B2M−/− iPSCs, significantly decreasing killing from 82.1% ± 7.1% (mock control) to 38.2% ± 8.5% (anti-B7-H3), 30.2% ± 5% (anti-CD155), and 37.3% ± 4.4% (anti-MICA) (Figure 3C). Interestingly, we observed that the blockade of B7-H3 on MHC-I-expressing cells resulted in a slight increase in susceptibility to NK killing, suggesting, in our model, a dichotomic role for this molecule depending on the HLA-I surface expression. On the other hand, masking NK-activating ligands on B2M−/−/HLA-E+/+ iPSCs did not potentiate killing by the CD16+ NK population (Figure 3C).

B7-H3 and CD155 abrogation prevents in vitro missing-self recognition of MHC-I−/− iPSCs by an activated CD16+ NK cell subset

On the bases of the above-described results we hypothesized that B7-H3 and/or CD155 may play a key role in NK activation in the absence of MHC-I. We then generated iPSC knockout lines for B7-H3 and CD155 genes on the MHC-I-null background by CRISPR-Cas9 technology (Figure S5A). We derived double-knockout B2M−/−/B7-H3−/− and B2M−/−/CD155−/− iPSC lines, and, to evaluate potential synergic effects, we generated a triple-knockout B2M−/−/B7-H3−/−/CD155−/− iPSC line. Biallelically edited clones were obtained by cell sorting according to the absence of surface expression of B7-H3 and CD155 (Figure S5B). The three lines remained stable in culture over 30 passages, and the expression of OCT4, SOX2, NANOG, and SSEA4 assayed by flow cytometry, quantitative real-time PCR, and immunofluorescence showed a profile comparable to that of the unedited line (Figures S5C–S5E). Moreover, Cas9-mediated B7-H3 knockout did not result in off-target events altering the transcription profile of the other B7 family ligands, as confirmed by the quantitative real-time PCR experiments on both iPSCs and iBeta cells (Figure S5F).

Finally, we also excluded that surface abrogation of B7-H3 and CD155 affected the cytotoxic T cell response. We performed Dunnett’s test using CD8-mediated lysis levels against B2M−/− cells as control, finding no differences among B2M−/− (1.79% ± 0.34%), B2M−/−/B7-H3−/− (2.62% ± 1.32%), B2M−/−/CD155−/− (1.02% ± 0.874%), and B2M−/−/B7-H3−/−/CD155−/− (2.82% ± 1.98%) iPSC lysis percentages (p > 0.99), but only between B2M−/− and wild-type cell lysis percentages (1.79% ± 0.34% vs. 42.10% ± 8.12%; p < 0.001) (Figure S5G). To assay NK-mediated killing, we used freshly isolated CD16+ NK cells as effectors (a different donor for each experiment) and wild-type, B2M−/−, B2M−/−/HLA-E+/+, B2M−/−/B7-H3−/−, B2M−/−/CD155−/−, and B2M−/−/B7-H3−/−/CD155−/− iPSCs as target cells. Jurkat and JY cell lines were used as positive and negative controls, respectively (Figure 4A). Knockout of B7-H3 and CD155 in B2M−/− iPSCs conferred protection from CD16+ NK cell-mediated lysis to the same extent of MHC-I-expressing or B2M−/−/HLA-E+/+ iPSCs. No synergistic effects were displayed by surface abrogation of both B7-H3 and CD155 molecules. Moreover, flow cytometry analysis showed that knockout of these ligands on MHC-I-deprived iPSCs duly resulted in poor degranulation of CD16+ NK cells (less than 25% of CD107a+ cells; p < 0.01) and, generally, in lower expression of both IFN-γ and TNF-α (less than 10% and 5% of positive cells, respectively; p < 0.01), compared with the NK cells engaging the B2M−/− line (Figure 4B).

Figure 4

B7-H3 and CD155 abrogation prevents in vitro missing-self recognition of MHC-I−/− iPSCs by activated CD16+ NK cells

(A) NK cell killing of wild-type and edited iPSC lines at different target:effector ratios after co-incubation with effectors for 4 h. Jurkat and JY cell lines were used as positive and negative controls, respectively. Percentage of PI-positive target cells was measured by flow cytometry and effector-mediated lysis was normalized on basal cell death percentage occurring in target cells after incubation without effectors. Error bars indicate standard deviation; n = 6 independent experiments.

(B) Percentage of NK cells positive for surface CD107a and intracellular TNF-α and IFN-γ by flow cytometry. Number of individual experiments (n) is indicated for each graph. Error bars represent standard deviation. Each experiment was carried out using a different NK donor. One-way ANOVA followed by Tukey’s post hoc test was performed; ∗p < 0.001.

B7-H3 abrogation exerts protective effects against cytotoxic activity of the CD56dim/CD16dim NK subpopulation

The effector CD56dim NK cells can be divided into CD16+ and CD16dim subpopulations, which represent 50%–60% and ∼20%, respectively, of the circulating NK cell repertoire in healthy subjects (Amand et al., 2017; Hofland et al., 2019). Notably, the CD56dim/CD16dim population behaves as a distinct subset, having increased degranulation and a higher cytokine production compared with their CD16+ counterparts (Amand et al., 2017). We sorted CD16+ and CD16dim NK cells from the same donor by magnetic beads (Figure 5A), comparing their receptor repertoire. Specifically, CD16dim NK cells displayed a significantly higher expression of CD337 (86.97% ± 3.76% vs. 77.11% ± 4.69%; ANOVA, p < 0.01), NKG2A (52.07% ± 6.20% vs. 44.08% ± 7.63%; ANOVA, p < 0.05), and NKG2C (27.42% ± 6.29% vs. 13.23% ± 9.20%; ANOVA, p < 0.001) in comparison with their CD16+ counterparts, whereas no substantial differences were found for NKG2D, CD226, CD338, or CD96 (Figure 5B). Finally, CD16dim NK cells showed a significant increase in killing activity against wild-type stem cells (23.88% ± 4.55% vs. 14.93% ± 3.30%, ANOVA, p < 0.05, at 1:1 effector:target ratio, and 49.10% ± 6.22% vs. 37.99% ± 6.62%, ANOVA, p < 0.01, at 5:1 effector:target ratio), B2M−/−/HLA-E+/+ (39.74% ± 4.13% vs. 26.66% ± 4.45%, ANOVA, p < 0.001, at 5:1 effector:target ratio), and B2M−/−/CD155−/− (27.99% ± 5.86% vs. 19.43% ± 4.39%, ANOVA, p < 0.05, at 1:1 effector:target ratio, and 48.70% ± 8.38% vs. 37.21% ± 3.28%, ANOVA, p < 0.01, at 5:1 effector:target ratio) iPSCs compared with the CD16+ subset, whereas B2M−/− cells appeared less susceptible to CD16dim NK-mediated cytotoxicity (31.16% ± 5.99% vs. 48.07% ± 8.36%, ANOVA, p < 0.001, at 1:1 effector:target ratio, and 50.80% ± 7.29% vs. 78.67% ± 4.39%, ANOVA, p < 0.0001, at 5:1 effector:target ratio) compared with their CD16+ counterparts (Figure 5C). Notably, B7-H3 but not CD155 abrogation further reduced lysis mediated by CD16dim NK cells at a 1:5 target:effector ratio (16.39% ± 5.08% vs. 32.57% ± 3.39%, ANOVA, p < 0.0001, for B2M−/−/B7-H3−/−, and 25.04% ± 4.38% vs. 37.26% ± 6.85%, ANOVA, p < 0.001, for B2M−/−/B7-H3−/−/CD155−/−), suggesting a putative key role for the CD337/B7-H3 axes in CD16dim NK activation (Figure 5C). On the other hand, no significant differences were found among wild-type, B2M−/−, and B2M−/−/CD155−/− for CD16dim-mediated killing, suggesting that HLA-I-independent killing mechanisms are probably due to a reduced expression of inhibitory KIRs, as reported in the literature (Amand et al., 2017).

Figure 5

B7-H3 and CD155 disruption exerts protective effects against cytotoxic activity of different NK subpopulations

(A) Representative scatterplots and histograms of sorted CD16+ and CD16dim NK cells after surface staining for either CD16 and CD56 (red) or isotype control (light blue).

(B) Percentage of positive cells following staining with mAbs for NK receptors of CD16+ and CD16dim NK subsets.

(C) Cytotoxicity on iPSC lines after co-incubation with CD16+ and CD16dim NK cell subsets at two target:effector ratios for 4 h. Percentage of PI-positive target cells was measured by flow cytometry, and the data were normalized on basal cell death percentage occurring in target cells after incubation without effectors. Error bars indicate standard deviation; n = 6 independent experiments. All experiments were conducted using six different donors (one for each experiment). Each cytotoxicity experiment was carried out using CD16+ and CD16dim of the same donor. One-way ANOVA followed by Tukey’s post hoc test was performed; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Both a previous study (Gornalusse at al., 2017) and the experiments on the blocking of NKG2C receptor suggested that B2M−/−/HLA-E+/+ cells could still be lysed by NKG2C+ NK cells. Generally, the expression of the NKG2A and NKG2C receptors is mutually exclusive, identifying two different NK subsets: NKG2A+/NKG2C− and NKG2A−/NKG2C+. Therefore, to directly demonstrate the involvement of NKG2C+ cells in the elimination of B2M−/−/HLA-E+/+ cells, we performed cell sorting experiments on total NK cells (Figure S6A) to enrich the two NKG2A+/NKG2C− and NKG2A−/NKG2C+ subsets, which were challenged in cytotoxicity assays with B2M−/−/HLA-E+/+ cells (Figure S6B). As expected, HLA-E-expressing iPSCs failed to properly evade the immune attack of NKG2C+ NK cells compared with their NKG2A+ counterparts (46.12% ± 7.50% vs. 11.30% ± 4.48%, ANOVA, p < 0.001, at 1:1 effector:target ratio, and 65.58% ± 8.55% vs. 17.80% ± 4.32%, ANOVA, p < 0.0001, at 5:1 effector:target ratio). Moreover, these results could explain the increased susceptibility of B2M−/−/HLA-E+/+ to the CD16dim NK cell subset, since CD16dim NKs are characterized by a higher number of NKG2C+ cells. Conversely, downregulation of the activating ligands B7-H3 and CD155 were demonstrated to prevent the cytotoxic response of both NKG2A+ and NKG2C+ subsets (Figure S6B).

Double knockout of B7-H3 and CD155 increases survival of MHC-I−/− iPSC-derived β cells from immune attack mediated by both CD16+ and CD16dim NK cell subsets

We performed in vitro pancreatic differentiation of all gene-edited cell lines. The INS gene expression (as well as other endocrine and β cell markers) of terminally differentiated cells and their ability to secrete insulin in response to glucose in dynamic perfusion were not affected (Figures 6A and 6B and data not shown). We performed NK cytotoxicity experiments on both unedited and gene-engineered iPSC-derived insulin-producing cells (iBeta). As reported in Figure 6C, the B7-H3 and CD155 knockout, as well as HLA-E overexpression, significantly reduced CD16+ NK cell-mediated lysis in an MHC-I-null background (p < 0.001). We observed that double knockout exerted more protective effects than abrogation of CD155 or B7-H3, since B2M−/−/B7-H3−/−/CD155−/− iBeta cells displayed lower CD16+ NK-mediated lysis (34.6% ± 7.9%) compared with B2M−/−/CD155−/− (50.9% ± 6.4%; p = 0.012) and B2M−/−/B7-H3−/− (45.1% ± 7.6%; p = 0.209) cells. Again, B7-H3 knockout showed the strongest effect in reducing killing by the CD16dim NK cells (B2M−/−/B7-H3−/−, 37.23% ± 10.33%; two-tailed t test, p < 0.05) compared with B2M−/−/CD155−/− (55.33% ± 9.19%; two-tailed t test, p = 0.9882) or with the HLA-E-expressing iBeta cells (58.31% ± 6.91%). In addition, B2M−/−/B7-H3−/−/CD155−/− iBeta evaded CD16dim NK-mediated immune attack better than all the other lines (21.45% ± 6.28%, ANOVA, p < 0.001), suggesting a potential synergistic effect of the double knockout of B7-H3 and CD155.

Figure 6

Double knockout of B7-H3 and CD155 increases survival of MHC-I−/− iPSC-derived β cells from immune attack mediated by both activated CD16+ and CD16dim NK cell subsets

(A) Percentage of insulin-positive cells during different stages of pancreatic differentiation. Insulin positivity was assayed by flow cytometry; n = 3 independent in vitro differentiations. Error bars indicate standard deviation. Each experiment used distinct in vitro differentiations of the same iPSC clone at different passages. One-way ANOVA followed by Dunnett’s test was performed using undifferentiated iPSCs as control. iPSC, undifferentiated iPSCs; PE, pancreatic endoderm; iBeta, iPSC-derived β cells.

(B) Dynamic insulin secretion assayed by perifusion on iPSC-derived β cells, upon stimulation with 20 mM glucose + 3-isobutyl-1-methylxanthine (IBMX) and 30 mM KCl. Insulin levels are expressed as fold change over basal secretion after 1 h of acclimatation at 0.5 mM glucose; n = 3 independent in vitro differentiations. Error bars indicate standard deviation. Each experiment used a distinct in vitro differentiation of the same iPSC clone at different passages.

(C) Cytotoxicity test on iPSC-derived β cell clusters after co-incubation with CD16+ or CD16dim NK subsets for 4 h. Percentage of PI-positive target cells was measured by flow cytometry and effector-mediated lysis was normalized to basal cell death percentage occurring in target cells after incubation without effectors. Error bars indicate standard deviation; n = 6 independent experiments. All experiments were conducted using six different donors (one for each experiment). Each cytotoxicity experiment was carried out using CD16+ and CD16dim from the same donor. One-way ANOVA followed by Tukey’s post hoc test was performed; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

B7-H3 and CD155 abrogation enhances cell survival of MHC-I−/− iPSCs after transplantation into NSG mice infused with donor-derived NK cells

Finally, we investigated an in vivo model to assay cell engraftment and survival of firefly luciferase (FLuc)-expressing wild-type, B2M−/−, B2M−/−/HLA-E+/+, and B2M−/−/B7-H3−/−/CD155−/− iPSC-derived pancreatic progenitors (PPs) (Figures S7A–S7C). To properly recapitulate in a preclinical model the in vitro experiments performed with both CD16+ and CD16dim NK subsets and to achieve the number of cells necessary to efficiently repopulate immunodeficient mice, we used allogeneic donor-derived primary NK cells after 12-days of in vitro expansion that allowed us to increase the NK cell number up to 10-fold of that of the starting material (Figures S7D and S7E). The expansion did not alter NK cell subpopulation distribution (Figure S7F) nor affect activation efficiency and killing capability, as assessed by CD107a expression and the cytotoxic activity against K562 cell line in comparison with freshly isolated NK cells (Figures S7G–S7I). For in vivo cytotoxicity experiments we divided NSG mice into two experimental groups: NK cell-injected and control group. Expanded NK cells were activated in vitro and infused via retro-orbital injection of the venous sinus in the NK cell-injected group 24 h before and 24 h after alginate-embedded PP subcutaneous transplantation (Figure 7A). The control group received PP cells only. About 24 h after PP transplantation, we observed a comparable reduction in graft area in all samples of the two experimental groups, mainly due to the transplant technique itself (Figures 7B and 7C). This reduction reflected a marked decrease in the bioluminescence signal, which we observed after 24 h in the NK cell-injected group (Figure 7C). Fascinatingly, after 72 h, B2M−/− PPs displayed a significant decrease in bioluminescence (p < 0.05, one-tailed t test), compared with wild-type, B2M−/−/HLA-E+/+, and B2M−/−/B7-H3−/−/CD155−/− cells (Figures 7B and 7C), whereas mice of the control group still showed a very strong signal, as we observed a slight increase in the graft area, too (Figure 7B). At follow-up (10 days after transplantation), when sample collection was performed, PPs lacking surface expression of both B7-H3 and CD155 resulted in a higher survival rate than their MHC-I-null counterparts, measured as a percentage of retained implants over the total of transplants (85% vs. 57%; p < 0.01). Immunofluorescence on cells transplanted into the NK cell-injected mice finally confirmed the co-expression of PDX-1 and INS, suggesting the beginning of PP maturation into functional β cells in vivo (Figures 7D and 7E). As expected, residual B2M−/− graft cells resulted in less than 20% of PDX1+ cells and around 7% ± 2.8% of PDX1+/INS+ cells, supporting the hypothesis that the degree of susceptibility to the immune system could strongly affect the maturation capability of progenitors (Figure 7E).

Figure 7

B7-H3 and CD155 surface abrogation enhances cell survival of MHC-I−/− iPSC-derived PPs after subcutaneous transplantation into NSG mice injected with allogeneic donor-derived NK cells

(A) Schematic representation of in vivo cytotoxicity protocol using allogeneic donor-derived primary NK cells and alginate-embedded FLuc-expressing iPSC-derived PP cells.

(B) Representative mice of the NK cell-injected group (left) and the control group (right), as acquired by IVIS at different time points (0, 24, and 72 h) after PP cell subcutaneous transplantation. 1, wild type; 2, B2M−/−; 3, B2M−/−/HLA-E+/+; 4, B2M−/−/B7-H3−/−/CD155−/−.

(C) Histogram of bioluminescence signals acquired by IVIS and normalized by area of subcutaneous alginate-embedded implants at different time points (0, 24, and 72 h) after transplantation. Data are plotted on a logarithmic scale and are representative of a total of seven mice (n = 7). Each NGS mouse of the NK cell-injected group was infused with a single different NK donor. Error bars indicate standard deviation. One-way ANOVA followed by Dunnett’s post hoc test was performed using B2M−/− as control; ∗p < 0.05.

(D) Immunofluorescence staining for the endocrine markers PDX-1 (red) and INS (green) on grafts explanted from the NK cell-injected group 10 days after transplantation. Cell nuclei are stained with Hoechst (blue). Scale bar, 100 μm (20× original magnification).

(E) Percentage of PDX1+ (progenitors) and PDX1+/INS+ (β cells) assessed by immunofluorescence in grafts from the NK cell-injected group 10 days after transplantation. Data are expressed as the mean of positive cells in at least three explanted grafts for wild-type, B2M−/−/HLA-E+/+, and B2M−/−/B7-H3−/−/CD155−/− and in two explanted grafts for B2M−/−. Error bars indicate standard deviation. One-way ANOVA followed by Tukey’s post hoc test was performed; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗∗p < 0.0001.

Discussion

Over the past decades, several preclinical experiments using iPSC-derived cells have propelled the field of regenerative medicine forward, increasing the prospects to create improved platforms for the replacement of damaged or dysfunctional tissues in patients. At the same time, human clinical trials involving iPSC-derived cells have been performed, aiming at the evaluation of the long-term safety and efficacy of transplantation of iPSC-derived retinal pigment epithelial (RPE) cells (Fritsche et al., 2014), dopaminergic progenitors (Stoddard-Bennett and Reijo Pera, 2019), NK cells (Hu et al., 2019), and cardiomyocytes (Cyranoski, 2018). However, immune reaction to iPSC derivatives may not be completely avoided by using autologous (Zhao et al., 2011) or fully matched donor (Korula et al., 2018) stem cell replacement, making the engineering of “off-the-shelf” hypoimmunogenic cells a sine qua non condition to overcome current engraftment limitations.

In the scenario of cell therapy for T1D treatment, a comprehensive characterization of the molecular determinants that elicit immune responses against iPSC-derived β cells could be the key point for the generation of a cell product invisible to the immune system that can be safely used in the clinical setting. In the present study, we defined the expression profile of the main intra- and extracellular NK-activating receptor ligands on both iPSCs and iPSC-derived insulin-secreting β cells, drawing special attention to the mechanisms following the loss of inhibition and the amplification of activating signals that leads to NK-mediated immune response. Among others, B7-H3 and CD155 emerged as the most expressed surface ligands in our stem cell model. Moreover, we confirmed that expression of both these molecules represents a common feature of both induced pluripotent and embryonic stem cell lines. Interestingly, CD112 was actively transcribed and translated in iPSC lines as well as an embryonic H1 cell line, although in some of them, such as the CGTRCiB10 clone used for gene engineering experiments, it was retained in the cytoplasm, whereas in others (CB4, NeoFiWT, and DRI1#11), was only partially exposed on the plasma membrane.

It has been reported that the ubiquitin-proteasome system (UPS) is detrimental for surface expression of CD112 (Molfetta et al., 2019). The UPS directly influences the pluripotency and commitment of stem cells (Schröter and Adjaye, 2014; Saez et al., 2018), and therefore different degrees of UPS activation could explain the observed variability in the surface expression of CD112 in both undifferentiated and terminally differentiated cells.

The expression of both CD155 and CD112 in stem cells has been previously reported (Kruse et al., 2015) and, in accordance with Kruse and colleagues, we found that killing of iPSCs may be mediated by the CD226-activating receptor. We also showed, using blocking antibodies, that NK missing-self recognition mechanisms may underlie the signals mediated by CD337 and NKG2D as well. Specifically, NKG2D-mediated cytotoxicity against stem cells was explained by the expression of MICA, and such response could be likely involved in the outcome of transplantation of the iPSC derivatives. As a matter of fact, it has been recently reported that NKG2D ligands are responsible for the engraftment failure of iPSC-derived cardiomyocytes in a syngeneic mouse model (Nakamura et al., 2019). Finally, we observed that in our model the blockade of B7-H3 determines an increase in the NK response in HLA class I-expressing cells, similar to what has been reported in the literature regarding T lymphocytes (Lu et al., 2020). As a matter of fact, it is used as an immune checkpoint in the treatment of tumors (Lee et al., 2017). However, the results we have observed represent only first evidence, which requires further and future investigations, which may be aimed at clarifying the contrasting role of this molecule (Hofmeyer et al., 2008) in association with class I molecules and antigen-presentation mechanisms.

All current engineering strategies to generate immunotolerant iPSCs converge on the abrogation of class I and class II HLA proteins, followed by knockin of tolerance-promoting immunomodulatory molecules, such as HLA-E (Gornalusse et al., 2017), HLA-G (Shi et al., 2020), CTLA4-Ig/PD-L1 (Rong et al., 2014), and CD47 (Deuse et al., 2019). However, single knockins were only partially effective, and further studies proposed to combine the above strategies, knocking in HLA-G, PD-L1, and CD47 in an MHC-I-null stem cell line (Han et al., 2019). However, the susceptibility profile of a graft may reflect the complexity of a wide range of interactions between target and effector cells by a plethora of other molecules acting as stimulatory signals. In particular, the expression on human iPSCs and their derivatives, such as the pancreatic β cells, of molecular ligands binding the activating counterreceptors on NK cell subsets can play a relevant role. The exploitation of their mechanisms of action is needed to finely manipulate NK missing-self recognition and allograft rejection processes, offering a potentially less invasive alternative to that of inserting exogenous molecules.

We demonstrated that gene manipulation of a protein involved in the activating checkpoints of different NK cell subsets could represent a valid strategy to increase immune evasion of MHC-I-deprived iPSC derivatives. By the generation via CRISPR-Cas9 technology of B2M−/−/B7-H3−/−, B2M−/−/CD155−/−, and B2M−/−/B7-H3−/−/CD155−/− cell lines, we could quantify the effects of knocking out both B7-H3 and CD155 on cytotoxic activity exerted by NK cells on an MHC-I-null cell line. We observed a significant reduction of NK-mediated lysis in all three NK ligand-engineered cell lines compared with B2M−/− cells, suggesting that the downregulation of one of these ligands prevented the stimulation threshold from being reached. This is in line with the hypothesis that proper NK cell activation requires the integration, or the sum, of signals propagated by multiple counterreceptors (Long et al., 2013). NK cells have both a critical physiological role in immune defense, surveillance, and homeostasis and emerging functions in tolerance, autoimmune disease, and immunotherapy, thus providing a link between the innate and the adaptive immune systems (Trinchieri, 1989; Xie et al., 2017; Gianchecchi et al., 2018). Their greatly diversified physiopathological behavior is reflected in a substantial number of NK subsets identified according to the differential expression of the adhesion molecule CD56 and the activating receptor CD16 (Freud et al., 2017; Zimmer, 2020). In particular, the CD56dim/CD16dim NK subset is characterized by lower inhibitory KIR expression, and higher degranulation and cytotoxic activity than CD56dim/CD16+ cells (Amand et al., 2017). By comparing the receptor repertoire of sorted CD16+ and CD16dim NK cells we found that CD16dim cells were characterized by the highest expression of both NKG2C and NKp30 and, in agreement with previous findings (Amand et al., 2017), by a higher percentage of NKG2A+ cells. Wild-type iPSCs are slightly more susceptible to killing by CD16dim NK cells compared with the effects exerted by their CD16+ counterparts, whereas surface abrogation of MHC-I did not significantly increase killing, suggesting that the mechanisms involving HLA class I-mediated inhibition could be partially missing in CD16dim NK cells, probably due to decreased expression of inhibitory KIRs (Amand et al., 2017). Interestingly, HLA-E∗01:03 did not confer to iPSCs or their pancreatic derivatives any further protective effects against the CD16dim NK subset, according to the higher expression of the activating complex CD94:NKG2C that recognizes HLA-E molecule as well (Pegram et al., 2011). In contrast, targeting by single and double knockout of B7-H3 and CD155 on iPSC-derived β cells and the other related pancreatic differentiation by-products was the most effective strategy to hinder the immune recognition of CD16dim NK subpopulations.

While downregulation of B7-H3 or CD155 was sufficient to promote escape from the CD16+ NK response to the same extent of HLA-E+ cells, B7-H3 and CD155 double-knockout seemed to minimize the risk of rejection mediated by other HLA-E-insensitive NK subsets, such as CD16dim and, in particular, NKG2C+ cells. The B7 family includes different surface molecules that regulate immune responses, delivering either stimulatory or inhibitory signals on both T lymphocytes and NK cells (Collins et al., 2005). Knockout of B7-H3 did not alter the expression of other B7 family ligands on iPSCs or insulin-producing cells. While the abrogation of B7-H3 alone is sufficient to elicit an immune-tolerant phenotype, we cannot exclude that knockout of other B7 family genes, expressed in differentiated cells, may enhance its effects also on other immune system cells. In particular, the activating ligand B7-H2 is known to act as a co-stimulatory molecule for CD8+ T lymphocytes, as it is a ligand for CD28 and CTL4A (Yao et al., 2011). Knockout of B7-H2 may therefore represent an alternative to the CTL4A-Ig knockin approach and/or a synergic strategy to further enhance immune escape.

All these in vitro findings were sustained by results obtained in a mouse model of NK-mediated acute rejection, showing that double knockout of B7-H3 and CD155 enhances immune tolerance of MHC-I−/− PPs. We used iPSC-derived PP cells because they have shown the capacity to be easily expanded and cryopreserved, promoting both scalability and clinical translation (Memon and Abdelalim, 2020). Moreover, iPSC-derived PPs present lower heterogenicity compared with the iBeta cell stage, and several works demonstrated that PPs are able to better differentiate into functional endocrine cells in vivo than in vitro (Rezania et al., 2013, 2014). Three phase 1/2 clinical trials led by ViaCyte proposed the use of stem cell-derived PPs in T1D therapy (NCT04678557, NCT03163511, and NCT02239354). By in vivo assessments of wild-type, B2M−/−, B2M−/−/HLA-E+/+, and B2M−/−/B7-H3−/−/CD155−/− luciferase-expressing PP cells transplanted in NSG mice infused with allogeneic donor-derived NK cells, we highlighted the role of B7-H3 and CD155 as key ligands for NK-mediated recognition of grafts with an MHC-I-null background, supporting the proposal for a B2M−/−/B7-H3−/−/CD155−/− stem cell line as an alternative cell source for β cell replacement strategies. In conclusion, B7-H3 and CD155 deletion can offer prospects for creating a next-generation platform of cell therapy for the treatment of T1D.

Limitations of the study

The editing of NK-activating molecules on stem cells and their derivatives represents a key tool to elucidate the role played by NK cells in rejection upon allotransplantation of MHCI-null grafts. In the present work, we edited just a single cell line as a putative cell model recapitulating common features of different human stem cells. However, the immunogenic profile of MHC-I−/− grafts could be affected by line-dependent cell-surface expression of different activating ligands and/or by intrinsic heterogenicity of stem cell derivatives, thus requiring additional editing steps to obtain the desired immune-escape phenotype. Moreover, our in vivo model focused on investigating acute rejection, emphasizing the missing-self recognition mechanism mediated by NK cells as it occurs in humans. However, such model may not properly recap the fine process that ensues in an autochthonous context, nor can it infer the impact of a complete immune system on graft survival and functions. Finally, since the in vivo follow-up was only 2 weeks long, further investigations and appropriate in vivo assessments will be necessary to fully evaluate the long-term survival and functionality, as well as the safety of gene-engineered iPSC-derived β cells.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests resources and reagents should be directed to and will be fulfilled by the lead contact, Prof. Lorenzo Piemonti (piemonti.lorenzo@hsr.it).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Cell culture

Human iPSC clone CGTRCiB10 was obtained from Cell and Gene Therapy Catapult (London, UK). Human iPSC CB4 and NeoFiWT were obtained from the Institute of Experimental Neurology (INSPE) at IRCCS San Raffaele Scientific Institute (Milan, Italy). Human iPSC DRI1 clone 11 (DRI1#11) and DRI2 clone 3 (DRI2#3) were generated by our laboratory following previous described Sendai virus-based protocol (Pellegrini et al., 2018). H1-hESC were supplied by Ditadi’s Lab at San Raffaele Telethon Institute for Gene Therapy (SR-Tiget) (Milan, Italy). Wild type iPSCs and CGTRCiB10-derived gene engineered stem cell lines were cultured in complete Essential 8 Flex Medium (Gibco, Thermo Fisher Scientific) added with 1% penicillin/streptomycin (P/S) (Lonza) on plates coated with recombinant human vitronectin (Thermo Fischer Scientific). Passages were performed when cell density reached 70–80% of confluence and detached using 0.5 mM pH 8.0 EDTA (Thermo Fisher Scientific) or 1X Trypsin/EDTA (Lonza). iPSC karyotyping, aCGH characterization and validation of pluripotency were described in supplementary methods. Jurkat (Clone E6-1, ATCC TIB-152), MOLT-4 (ATCC CRL-1582) and JY (obtained from the SR-TIGET, Telethon Institute for Gene Therapy, Milan, Italy) cell lines were cultured in RPMI-1640 (Lonza), added with 10% FBS (Lonza), 1 mM L-glutamine (Lonza) and 1% P/S. CD8+ T and NK cells were isolated from PBMCs derived using Ficoll-Paque separation method on whole blood of healthy donors (see Human subject section below). CD8+ T cells were cultured in X-VIVO (Lonza) supplemented with 5% FBS (Euroclone) and 1000 U/ml IL-2 (Peprotech). NK cells used for in vitro cytotoxicity test were maintained in culture for 2–4 days using RPMI-1640 (Gibco, Thermo Fisher Scientific) added with 10% FBS (Euroclone), 1000 U/ml IL-2 (Peprotech), 20 ng/mL IL-12 (Peprotech), 1 mM sodium pyruvate (Lonza), 50 μM 2-mercaptoehanol (Gibco, Thermo Fisher Scientific) and 1% P/S (Lonza). For in vivo experiments primary NK cell were first expanded for 12 days by culturing into NK MACS Medium (Miltenyi Biotec), added with 5% human AB Serum (Corning), 140 ng/mL IL-15 (Peprotech) and 500 U/ml IL-2 (Peprotech) followed by 24 hours of incubation in activation medium [1000 U/ml IL-2 (Peprotech) 20 ng/mL IL-12 (Peprotech), no IL-15].

Human subject

All PBMCs used in this study were derived from buffy coats purchased by the Immunohematology and Transfusion Medicine Service (ITMS) of Garbagnate, Milan, Italy, and managed (accepted, registered, and distributed) by the ITMS of San Raffaele hospital, Milan, Italy. Buffy coats were derived from healthy donors who donate blood and sign an informed consent in accordance with the D.M. November 2nd 2015 entitled "Provisions relating to the quality and safety requirements of blood components". In accordance with the ministerial provisions and the IOG 364 institutional procedure "Request and delivery of buffy coats for research purposes", it is not possible to retrieve any type of information (gender, age, HLA typing) of buffy coat donors.

Animals

Experiments involving mice were performed under protocols approved and monitored by Animal Care and Use Committee of San Raffaele Scientific Institute. The NOD-scid IL2Rgammanull (NSG) mice [sex: female; age: 6–8 weeks; weight: 20–24 g] used as recipients were obtained from Charles River Laboratories, Calco, Italy.

Method details

iPSC karyotyping and aCGH, characterization, validation of pluripotency and trilineage differentiation test

Large chromosome rearrangements and/or short chromosomal copy number changes at high-resolution scale were assayed by performing Q-banding karyotype analysis and array comparative genomic hybridization (aCGH) with over 600 probes and a median probe spacing of 41 kb (Agilent Technologies). Karyotiping and aCGH experiments and accompanying reports were fully managed by the Integrated System Engineering (ISENET Biobanking, Milan, Italy). iPSCs were evaluated on their morphological parameters and tested for pluripotency markers by flow cytometry (OCT4), qRT-PCR (OCT4, SOX2, NANOG) and immunofluorescence (OCT4, SOX2, NANOG, SSEA4), following the below protocols. Functional analysis to assess wild type iPSCs capability to correctly differentiate into all three germ layers was performed by using STEMdiff Trilineage Differentiation Kit (Stemcell Technologies), according to manufacturer’s instructions. Outcome of trilineage differentiation was evaluated by testing committed cells for pluripotency and endo-, meso- and ectodermal markers by flow cytometry (OCT4, CD56, CXCR4, NESTIN) and qRT-PCR using Fast 96-well panel of TaqMan hPSC Scorecard Assay (Life Technologies).

iPSC differentiation into pancreatic β cells, characterization and maintaining

Human wild type and genetic engineered iPSCs were differentiated into insulin-producing β cells following a previously described protocol (Pellegrini et al., 2018). Differentiation started with cells in adhesion, at 70–80% of confluence. Morphological changes were followed by using digital phase contrast inverted microscope (Invitrogen EVOS XL; Thermo Fisher Scientific). During differentiation, cells were sampled at specific time points to evaluate stage-specific pancreatic markers by qRT-PCR and flow cytometry. Terminally differentiated cells were maintained in CMRL-1066 medium (Thermo Fisher Scientific), added with 10% FBS (Euroclone), 10 μM Alk5i II (Selleckchem), 1 μM L-3,30,5-Triiodothyronine (T3) (Sigma-Aldrich), 10 mM nicotinamide (Sigma-Aldrich), 1 mM L-Glutamine (Lonza), 1% P/S (Lonza). In preparation for cytotoxicity test, cells were detached by gently pipetting and cultured in suspension by agitation at 60 rpm, adding to complete medium 10 ng/mL IFN-γ (Peprotech), 50 μg/mL DNAse I (Roche) and 10 μM Y27632 (Stemcell Technologies).

Alloreactive CD8+ T cell isolation

6x108 PBMCs isolated from healthy donors were co-incubated with 50 Gy-irradiated MOLT-4. Co-incubation was performed for 4 weeks in X-VIVO (Lonza) supplemented with 5% FBS (Euroclone) at 100:1 PBMC/MOLT-4 ratio on flat-bottom 6-wells plates. After the first week, all cells in suspension were recovered and transferred in new plates to eliminate monocytes and other adherent cells. Additional 1x104 irradiated MOLT-4 were added to culture after 14 days. From the third week, clusters of naïve T lymphocytes primed against MOLT-4 antigens could be appreciated by inverted microscope. Finally, CD8+ T cells were isolated using magnetic separator and Naïve CD8+ T Cell Isolation Kit (MACS, Miltenyi Biotec), following producer’s protocol. Enriched CD3+ CD8+ T cells were resuspended in complete culture media or frozen in FBS with 10% DMSO (Sigma-Aldrich). Four days before cytotoxicity test, frozen CD8+ T cells were thawed and activated with 1000 U/ml of IL-2 (Peprotech).

NK cell subsets isolation

PBMCs isolated from healthy donors were processed by using magnetic separator and CD56+ CD16+ NK Cell Isolation Kit (MACS, Miltenyi Biotec). NK cell isolation was performed in a two-step procedure: non-NK cells were labeled with a cocktail of biotin-conjugated antibodies and removed by negative selection. Pre-enriched fraction contained both CD56dim/CD16dim and CD56dim/CD16+ subsets. During the second step, pre-enriched fraction was labeled with CD16 MicroBeads following incubation for 5′ at 4°C, then loaded into the separation MS columns. While CD56dim/CD16+ cells were retained into columns, unlabeled CD56dim/CD16dim was recovered from washing flow-through. Finally, CD56dim/CD16+ NK cells were isolated by flushing out magnetically labeled fraction. Identity and purity of NK subsets were evaluated by flow cytometry, using anti-human CD56 and CD16 antibodies. Both NKG2A- and NKG2C-positive NK cells were purified by cell sorting, by using FACSAria Fusion (BD Biosciences). NK cells used for in vitro cytotoxicity tests were maintained in culture for 2–4 days under stimulation with IL-2 (Peprotech) and IL-12 (Peprotech). NK cell used for in vivo experiments were expanded in vitro for a total of 12 days into NK MACS Medium (Miltenyi Biotec), added with 5% human AB Serum (Corning), 140 ng/mL IL-15 (Peprotech) and 500 U/ml IL-2 (Peprotech). Since the reactivity profile of each NK cell line can vary from donor to donor and merging them could result in confounding results (or induce cross-alloreactivity reactions), NK cells from different donors were never pooled to perform in vitro or in vivo cytotoxicity experiments.

gRNA IVT, plasmid vector and genome editing

Gene editing on iPSCs was achieved by applying CRISPR/Cas9 technology. All guide RNAs (gRNAs) were designed and chosen depending on the number of off-target sites using CHOPCHOP v.3 tool (Labun et al., 2019). We performed in vitro assembly of gRNAs for B2M (one sense gRNA on exon 1: 5′-AGTAGCGCGAGCACAGCTA-3′), B7-H3 (two sense gRNAs on exon 3: 5′-GCTGGTGCACAGCTTTGCTG-3′, 5′-CTGGTGCACAGCTTTGCTGA-3′; one anti-sense gRNA mapping on exons 3 and 5: 5′-ACCGGCCACCTGCAGGCTGA-3′) and CD155 (three sense gRNAs on exon 2: 5′-GGCCGTCTTCCACCAAACGC-3′, 5′-GAATTCGTGGCAGCCAGACT-3′, 5′-GATGTTCGGGTTGCGCGTAG-3′). Primer pairs were annealed to assemble by PCR the gRNA DNA template; then gRNAs were synthetized by in vitro transcription (IVT) and finally purified using the GeneArt Precision gRNA Synthesis Kit (Thermo Fisher Scientific), according to manufacturer’s protocol. gRNAs concentration was measured by spectrophotometer (Epoch, Gen5 Software; BioTek). Targeting efficiencies reached up to 50–60%, 30% and 20% for B2M, B7-H3 and CD155 genes, respectively. Knock-in of HLA-E was performed using a plasmid vector as donor DNA, including right/left homology arms spanning exon 1 of B2M gene (chr15:44710847-44711568; chr15:44711569-44712268), with coding sequence for the HLA-G signal peptide VMAPRTLFL, PBL (G4S)3 flexible linker, wobbled B2M protein, BHL (G4S)4 flexible linker and HLA-E∗01:03 heavy chain containing R107G modification to increase its thermal stability, as previously described (Celik et al., 2015). An IRES-eGFP-bGH poly(A) cassette was also cloned downstream the pep-B2M-HLA-E construct to facilitate selection of recombinant clones. Knock-out of B2M, B7-H3 and/or CD155 was set up delivering into wild type iPSCs 10 μg of TrueCut Cas9 protein v2 (Invitrogen, Thermo Fisher Scientific) and 5 μg of corresponding gRNAs as ribonucleoprotein (RNP) complex by using 4D-nucleofector (Lonza) with P3 electroporation solution and CB-150 program, according to manufacturer’s instructions. For knock-in, Cas9/gRNA RNP complex was electroporated with 1 μg of donor DNA into iPSCs treated for 16 hours with 1 μg/mL of Nocodazole (Sigma-Aldrich) to increase homology-directed repair (HDR) events from ∼0.1% to >7%. After 7 days from electroporation, gene disruption and recombination efficiencies were evaluated by flow cytometry, then single clones of biallelic knock-out population of interest, as well as knocked-in HLA-E/GFP-expressing cells, were purified following cell sorting by using FACSAria Fusion (BD Biosciences).

RNA isolation, retro-transcription and qRT-PCR

Total RNA was extracted with mirVana Isolation Kit (Ambion) and quantified by using spectrophotometer (Epoch, Gen5 Software; BioTek). Up to 2 μg of total RNA was treated with 2 U/μl TURBO DNAse (Ambion), then retrotranscribed by using the SuperScript IV RT (Thermo Fisher Scientific), according to manufacturer’s protocol. Both homemade primer pairs for SYBR Green-based protocol and commercial pre-designed gene-specific primer-probe sets from TaqMan Gene Expression Assays (Applied Biosystems) were used for the gene expression analysis. Quantitative RT-PCR were performed by using TaqMan Universal PCR Master Mix (Applied Biosistems) or PowerUp Green Master Mix (Applied Biosistems), according to manufacturer’s protocol. Expression levels were normalized applying 2−ΔCt method, using RPS18, RPS27 and GAPDH as housekeeping genes.

Immunofluorescence

For pluripotency test, cell lines were cultured in 4-Well Culture Slide (Falcon) until 70% confluence was reached, then fixed with 4% paraformaldehyde. Terminally differentiated β cell aggregates were embedded in 2–4% of low gelling temperature agarose (Sigma-Aldrich), fixed with 10% zinc formalin (Sigma-Aldrich), then included in paraffin. For intracellular staining cells were permeabilized (PermWash 0.4% Triton X-100 in PBS) and stained by using the antibodies listed in the key resources table. All images were acquired using Confocal UltraVIEW ERS microscope (PerkinElmer Life Sciences) and DeltaVision Ultra microscope (GE Healthcare Life Sciences), then analyzed by using Fiji/ImageJ v.1.52p.

Flow cytometry

Flow cytometry was performed with FACSCanto (BD) flow cytometer using FACS Diva Software. Cell surface staining was performed incubating cells for 20′ at RT in PBS supplemented with 2% FBS (Euroclone) and 0.1% sodium azide (Sigma-Aldrich). Cells were then analyzed immediately or after fixation with Cytofix/Cytoperm (BD), containing 4,2% formaldehyde. Intracellular staining required cell permeabilization with ice-cold Cytofix/Phosflow perm Buffer III (BD), then incubation for 30′ at 4° with conjugated antibodies. Titration experiments were carried out following UWCCC Flow Cytometry Laboratory guidelines (Telford et al., 2009). For HLA and NK ligands overexpression induction experiments, staining was performed after overnight treatment with 10 ng/mL IFN-γ (Peprotech), 50 ng/mL TNF-α (Peprotech) or both. Results were analyzed by using FlowJo software v.10 (FlowJo).

Dynamic perifusion of iPSC-derived β cells

Dynamic stimulation of terminally differentiated β cells was performed by using high-capacity low-pulsatility peristaltic pump-based automated perifusion system (BioRep Perifusion V2.0.0), as previously described (Pellegrini et al., 2018). Secretion was sampled every minute, collecting perifusate in round-bottom 96-wells plates. Released insulin was quantified by ELISA Kit (Mercodia), according to manufacturer’s instructions. ELISA plates were analyzed by ELISA Reader (MicroPlate Reader, Model 680, BioRad).

In vitro cytotoxicity assay with CD8+ T and NK cells

Effector (CD8+ T or NK) cells were labeled following incubation for 10′ at 37°C with 5 μM Cell Proliferation Dye eFluor670 (Thermo Fisher Scientific). Labeling was stopped by adding 5 volumes of cold complete medium (X-VIVO with 5% FBS) and incubating cells in ice for 5’. Finally, after 3 washings with complete media, effector cells were recollected by centrifugation (5′ x 400 g) and resuspended in new fresh medium with 200 U/ml IL-2 (Peprotech) and 10 ng/mL IFN-γ (Peprotech). iPSC lines were detached by incubation with 1X Trypsin/EDTA for 5′ at 37°, to reduce number and dimension of cell aggregates. Target (iPS cell lines, Jurkat, Molt4 or JY) cells were washed with PBS and counted. Target cells were stained with LIVE/DEAD Fixable Violet Dead Cell Stain (Thermo Fischer Scientific), following manufacturer’s protocol, to discriminate cells that die before the cytotoxicity test and exclude them from the analysis. LIVE/DEAD-stained target cells were resuspended in X-VIVO (Lonza) with 5% FBS (Euroclone), 200 U/ml IL-2 (Peprotech) and 10 ng/mL IFN-γ (Peprotech), then plated in round-bottom 96-well plate at a seeding density of 4x104 cells/well. Effector cells were added to target cells with different ratios (0.5:1, 1:1 and 1:5), mixing by pipetting to ensure that target and effector cells were homogeneously distributed. For each target cell type at least 3 wells were left without effectors, to measure spontaneous/basal cell death levels. Media volume was adjusted up to 50 μL and 96-wells plate was centrifugated for 1′ at 1000 rpm. Co-incubation proceeded for 4 hours at 37°C and 5% CO2. At the end of incubation time, cells were stained with 20 μg of propidium iodide (PI) (Signa-Aldrich) and quickly analyzed by flow cytometry to identify dead cells. For cytotoxicity tests performed on terminally differentiated cells, about 20–30 suspension cultured iPSC-derived pancreatic β cell aggregates of 70–100 μm diameter were co-incubated with 50x104 NK cells for 4 hours. An aliquot of effector/target cell mix was used to quantify NK response and activation by cytofluorimetric analysis, performing surface staining for degranulation marker CD107a and intracellular staining for IFN-γ and TNF-α. For the blocking antibodies experiments, effector NK cells were pre-incubated for 20′ at 4° in PBS added with 2% BSA (Sigma-Aldrich) and mock unrelated antibody or anti-human NK receptor-specific antibodies. NK cells from different donors were never pooled together to perform cytotoxicity assays.

In vivo reactivity with donor-derived primary NK cells

Wild type, B2M−/−, B2M−/−/HLA-E+/+ and B2M−/−/B7-H3−/−/CD155−/− iPSC lines were stably transduced with a luciferase-expressing Red-FLuc-NGFR lentivirus vector (supplied by Lombardo’s Lab at the SR-Tiget, IRCCS San Raffaele Scientific Institute, Milan, Italy) and the NGFR-positive cells were used for in vivo experiments. Luciferase-expressing cells were differentiated to pancreatic progenitors (PPs) following differentiation protocol as described above, then 2x106 of PPs were embedded in 4% alginate-matrix by dripping cell/alginate mixture into a gelling solution containing 100 mM of CaCl2. NSG mice were divided into 2 groups: NK cell-injected group (N = 7) and control group (N = 7). NK cell-injected group’s mice were infused intravenously with 106 pre-activated expanded primary NK cells 24 hours before and 24 hours after PPs transplantation. Each NSG mice of the NK cell-injected group received NK cells derived from a single different donor. NK cells from different donors were never pooled or infused together within the same animal. PPs were transplanted sub-dermally into NSG mice and each recipient received all the four PP lines. In vivo monitoring of luciferase-expressing cells was done by the Lumina II IVIS imaging system (Caliper Life Science), at day of transplant and after 24 and 72 hours. Data analysis was done using the Living Image software v. 4.2 (Caliper Life Sciences). After seven days from last image acquisition, mice were sacrificed, and graft collected for H&E and IF staining. The histological procedures were conducted in the Animal Histopathology facility at San Raffaele Scientific Institute.

Quantification and statistical analysis

Data are represented as means ± standard deviation. Statistical analysis was carried out by using PRISM v.8 (GraphPad). Numerosity indicated in figure legends refers to independent experiments carried out according to figure-related description. In particular, quantification of NK activating ligands on iPSC cell lines was performed by measuring mRNA and protein levels on batches of same cell lines at different passages and stage of confluence. Biological replicates for experiments on differentiated cells are intended performed on different independent in vitro differentiation, starting from undifferentiated iPSC at different passages. Each cytotoxicity experiment was replicated by using effector cells derived from different donors. As reported in Method Details section, NK cells from different donors were not pooled together, neither for in vitro nor for in vivo experiments. One-way analysis of variance (ANOVA) was performed for comparison of data belonging to more than two groups, followed by either Tukey’s honestly significant difference post hoc test for multiple comparison, or Dunnett’s test to compare means from several experimental groups against a single control group. For two groups comparison, assuming normal distribution of data, a two or one-tailed Student’s t-test was used. The specific test used to compute the significance is appropriately indicated in the text or in the figure legends. A p-value < 0.05 was considered significant (∗ for p < 0.05; ∗∗ for p < 0.01; ∗∗∗ for p < 0.001, ∗∗∗∗ for p < 0.0001).

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
APC anti-human β2-microglobulin antibody [2M2]	BioLegend	Cat#316312;RRID: AB_10641281
PE anti-human HLA-A,B,C antibody [W6/32]	BioLegend	Cat#311406;RRID: AB_314875
PE anti-human HLA-E antibody [3D12]	BioLegend	Cat#342604;RRID: AB_1659249
Monoclonal Anti-HLA-G PE [MEM-G/9]	Quimigen	Cat#1P-292-C100;RRID: AB_10736086
Alexa Fluor 647 anti-human HLA-DR, DP, DQ Antibody[Tu39]	BioLegend	Cat#361704;RRID: AB_2563169
PE-Cy7 Mouse anti-human CD3 [SK7]	BD Biosciences	Cat#557851;RRID: AB_396896
Alexa Fluor 647 Mouse Anti-Human CD16 [3G8]	BD Biosciences	Cat#557710;RRID: AB_396819
PE-Cy5 Mouse Anti-Human CD56 [B159]	BD Biosciences	Cat#555517;RRID: AB_395907
Human CD155/PVR Alexa Fluor 488-conjugated antibody[300907]	R&D Systems	Cat#FAB25301G;RRID: AB_2269068
Human B7-H3 PE-conjugated antibody [185504]	R&D Systems	Cat#FAB1027P;RRID: AB_2073697
Human B7-H6 PE-conjugated antibody [875001]	R&D Systems	Cat#FAB7144P; RRID: AB_2636810
Human Nectin-2/CD112 APC-conjugated antibody [610603]	R&D Systems	Cat#FAB2229A;RRID: AB_10973290
Human MICA Alexa Fluor 488-conjugated antibody [159227]	R&D Systems	Cat#FAB1300G;RRID: AB_10891134
Human MICB PE-conjugated antibody [236511]	R&D Systems	Cat#FAB1599P;RRID: AB_10972308
Human ULBP-1 APC-conjugated antibody [170818]	R&D Systems	Cat#FAB1380A;RRID: AB_2687471
Human ULBP-2/5/6 PE-conjugated antibody [165903]	R&D Systems	Cat#FAB1298P;RRID: AB_2214693
Human ULBP-3 Alexa Fluor® 488-conjugated antibody[166510]	R&D Systems	Cat#FAB1517G;RRID: AB_10719122
Human ULBP-4/RAET1E Alexa Fluor 750-conjugated antibody [709116]	R&D Systems	Cat#FAB6285S;RRID: AB_10888662
Human NKG2A/CD159a Alexa Fluor 488-conjugated [131411]	R&D Systems	Cat#FAB1059G:RRID: AB_2132978
Human NKG2C/CD159c APC-conjugated [134591]	R&D Systems	Cat#FAB138A;RRID: AB_416838
Human NKG2D/CD314 [149810]	R&D Systems	Cat#MAB139; RRID: AB_2133263
Human NKG2D/CD314 PE-conjugated [149810]	R&D Systems	Cat#FAB139P;RRID: AB_2133264
Human NKp30/NCR3 [210845]	R&D Systems	Cat#MAB1849; RRID: AB_2149446
Human NKp30/NCR3 PE-conjugated [210845]	R&D Systems	Cat#FAB1849P;RRID: AB_2149447
Human NKp46/NCR1 APC-conjugated [195314]	R&D Systems	Cat#FAB1850A;RRID: AB_416861
Human DNAM-1/CD226 Fluorescein-conjugated [102511]	R&D Systems	Cat#FAB666F;RRID: AB_2072626
Human CD96 v2 Alexa Fluor 488-conjugated [628211]	R&D Systems	Cat#FAB6199G; RRID: AB_2608850
Human LAMP1/CD107a Alexa Fluor 488-conjugated antibody [508921]	R&D Systems	Cat#IC4800G; RRID: AB_2296836
Anti-Interferon gamma antibody [4S.B3]	Abcam	Cat#ab234193; AB_2747846
Anti-TNF alpha antibody [EPR20972]	Abcam	Cat#ab225576; RRID: AB_2893364
Alexa Fluor 647 Mouse Anti-OCT3/4 [40/OCT-3]	BD Biosciences	Cat#560307;RRID: AB_1645319
PE Mouse anti-CD184 [12G5]	BD Biosciences	Cat#557145;RRID: AB_2292652
Alexa Fluor® 647 Mouse Anti-Nestin [25/NESTIN]	BD Biosciences	Cat#560393;RRID: AB_1645170
Alexa Fluor 488 Mouse Anti-PDX-1 [658A5]	BD Biosciences	Cat#652274;RRID: AB_10611998
PE Mouse Anti-NKX6.1 [R11-560]	BD Biosciences	Cat#563023;RRID: AB_2716792
Alexa Fluor 647 Mouse Anti-Insulin [T56-706]	BD Biosciences	Cat#565689;RRID: AB_2739331
Mouse IgG1 Isotype Control	R&D Systems	Cat#MAB002;RRID: AB_357344
Mouse IgG2A Isotype Control	R&D Systems	Cat#MAB003;RRID: AB_357345
Mouse IgG2B Isotype Control	R&D Systems	Cat#MAB004;RRID: AB_357346
Goat Anti-Mouse Ig, Human ads-Alexa Fluor 488	Southern Biotech	Cat#1010-30;RRID: AB_2794130
Mouse F(ab)2 IgG (H+L) PE-conjugated Antibody	R&D Systems	Cat#F0102B;RRID: AB_622014
Mouse F(ab)2 IgG (H+L) APC-conjugated Antibody	R&D Systems	Cat#F0101B;RRID: AB_622013
Human/Mouse/Rat BAT3/BAG6 antibody	Biotechne	Cat#AF6438;RRID: AB_10717417
Human/Mouse/Rat Vimentin antibody	Biotechne	Cat#AF2105;RRID: AB_355153
Human/Mouse/Rat Vimentin antibody [280618]	R&D Systems	Cat#MAB2105;RRID: AB_2241653
Human Properdin antibody [10-18]	Novus Biologicals	Cat#NB100-64749;RRID: AB_963443
Human OCT4 antibody [GT486]	Novus Biologicals	Cat#NBP2-15052; RRID: AB_2895225
Human NANOG antibody	R&D Systems	Cat#AF1997;RRID: AB_355097
Human/Mouse/Rat SOX2 antibody [245610]	R&D Systems	Cat#MAB2018;RRID: AB_358009
Anti-SSEA4 antibody [MC813-70]	Abcam	Cat#ab16287;RRID: AB_778073
Anti-PDX-1 antibody [EPR3358(2)]	Abcam	Cat#ab134150; RRID: AB_2631338
Anti-Insulin antibody	Abcam	Cat#ab63820;RRID:AB_1925116
Goat anti-Mouse IgG (H+L), Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Thermo Fisher Scientific	Cat#A11001;RRID: AB_2534069
Donkey anti-Sheep IgG (H+L), Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Thermo Fisher Scientific	Cat#A11015;RRID: AB_141362
Rabbit anti-Goat IgG (H+L), Superclonal Recombinant Secondary Antibody, Alexa Fluor 488	Thermo Fisher Scientific	Cat#A27012;RRID: AB_2536077
Biological samples
Healthy donor PBMC-derived CD8 naïve cells	San Raffaele Immunohematology andTransfusion Medicine Service	N/A
Healthy donor PBMC-derived NK cells	San Raffaele Immunohematology and Transfusion Medicine Service	N/A
Chemicals, peptides, and recombinant proteins
Calcium chloride	Sigma-Aldrich	Cat#C1016
Sodium alginate	Sigma-Aldrich	Cat#W201502
rhIL-2	Peprotech	Cat#200-02
rhIL-12 p80	Peprotech	Cat# 200-12p80H
rhIL-15	Peprotech	Cat#200-15
Alk5i II	Selleckchem	Cat#S2750
L-3,30,5-Triiodothyronine (T3)	Sigma-Aldrich	Cat#T2877
Nicotinamide	Sigma-Aldrich	Cat#0636
rhIFN-γ	Peprotech	Cat#300-02
rhTNF-α	Peprotech	Cat#300-01A
Y27632 (ROCK inhibitor)	StemCell Technologies	Cat#72304
IBMX 3-Isobutyl-1-methylxanthine	Gibco	Cat#PHZ1124
Nocodazole	Sigma-Aldrich	Cat#M1404
Cell Proliferation Dye eFluor670	Thermo Fisher Scientific	Cat# 65-0840-85
LIVE/DEAD Fixable Violet Dead Cell Stain	Thermo Fisher Scientific	Cat#L34955
Propidium Iodide (PI)	Sigma-Aldrich	Cat#P4170
Critical commercial assays
STEMdiff Trilineage Differentiation Kit	StemCell Technologies	Cat #05230
TaqMan™ hPSC Scorecard™ Kit, Fast 96-well	Thermo Fisher Scientific	Cat#A15871
Naïve CD8+ T Cell Isolation Kit	MACS, Miltenyi Biotec	Cat#130-093-244
CD56+ CD16+ NK Cell Isolation Kit	MACS, Miltenyi Biotec	Cat#130-092-660
BD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit	BD Biosciences	Cat#554714
GeneArt Precision gRNA Synthesis Kit	Thermo Fisher Scientific	Cat#A29377
SuperScript IV First-Strand Synthesis System	Thermo Fisher Scientific	Cat#18091050
TaqMan Universal PCR Master Mix	Applied Biosystems	Cat#4305719
PowerUp Green Master Mix	Applied Biosystems	Cat#A25741
Deposited data
Single cell RNAseq datasets of hiPSC during differentiation into pancreatic beta cells	Pellegrini et al., 2020	GEO: GSE149613
Experimental models: Cell lines
Human: iPS cells	Cell and Gene Therapy Catapult, London, UK	CGTRCiB10
Human: iPS cells	Institute of Experimental Neurology (INSPE) - IRCCS San RaffaeleScientific Institute	CB4
Human: iPS cells	Institute of Experimental Neurology (INSPE) - IRCCS San RaffaeleScientific Institute	NeoFiWT
Human: iPS cells	Diabetes Research Institute (DRI) - IRCCS San Raffaele Scientific Institute	DRI1#11
Human: iPS cells	Diabetes Research Institute (DRI) - IRCCS San Raffaele ScientificInstitute	DRI2#3
Human: H1-hESC	San Raffaele Telethon Institute for Gene Therapy (SR-TIGET)	N/A
Human: Jurkat [Clone E6-1]	ATCC	TIB-152
Human: MOLT-4	ATCC	CRL-1582
Human: JY	San Raffaele Telethon Institute for Gene Therapy (SR-TIGET)	N/A
Experimental models: Organisms/strains
Mouse: NSG: NOD.Cg-PrkdcSCID Il2rgtm1Wjl/SzJ	Charles River Laboratories	JAX: 005557
Oligonucleotides
PVR (Hs00197846_m1)	Thermo Fisher Scientific	Cat#4331182
CD276 (Hs00987207_m1)	Thermo Fisher Scientific	Cat#4331182
NECTIN2 (Hs01071562_m1)	Thermo Fisher Scientific	Cat#4331182
MICA (Hs00741286_m1)	Thermo Fisher Scientific	Cat#4331182
MICB (Hs00792952_m1)	Thermo Fisher Scientific	Cat#4331182
ULBP1 (Hs00360941_m1)	Thermo Fisher Scientific	Cat#4331182
ULBP2 (Hs00607609_mH)	Thermo Fisher Scientific	Cat#4331182
ULBP3 (Hs00225909_m1)	Thermo Fisher Scientific	Cat#4331182
ULBP4 (Hs01026643_g1)	Thermo Fisher Scientific	Cat#4331182
ULBP5 (Hs01584111_mH)	Thermo Fisher Scientific	Cat#4331182
Other oligonucleotides: Table S1
Recombinant DNA
Plasmid : B2M-ETrimer-IRES-GFP	This paper	N/A
Lentiviral vector : PKG-FLuc-mCMV-NGFR	This paper	N/A
Software and algorithms
CHOPCHOP v.3 tool	Labun et al., 2019	https://chopchop.cbu.uib.no/
FlowJo software v.10	Ashland	https://www.flowjo.com/solutions/flowjo
Fiji/ImageJ v.1.52p	Schindelin et al., 2012	https://imagej.net/Fiji/Downloads
PRISM v.8	GraphPad Software	https://www.graphpad.com/scientific-software/prism/
R v.4.0.3	R Foundation forStatistical Computing	https://www.R-project.org/
Other
High-capacity low-pulsatility peristaltic pump-based automated perifusion system v.2.0.0	BioRep	https://doi.org/10.1177/0963689718798564