Generation of iPS-fibro and ΔiPS-fibro
Dermal fibroblasts of donors A and B (48 and 26 y.o. females, respectively) were reprogrammed into iPSCs using the delivery of Yamanaka’s factors by integration-free Sendai virus. We obtained the iPSCs cell lines IPSFF1S and IPSFD4S for donors A and B, respectively. In this paper, we use designations according to the donor: iPSC-A and iPSC-B. The iPSC-A and iPSC-B were characterized according to standard criteria. Using CRISPR/Cas9 genome editing for the B2M gene knockout, we obtained ΔiPSC-A and ΔiPSC-B lines lacking HLA-I expression. A detailed description of ΔiPSC-A was previously published [20]. The iPSC-A and its subclone ΔiPSC-A are registered in hPSCreg database [47]. The ΔiPSC-B was derived upon the same procedure described briefly below.
For CRISPR-Cas9 genome editing, we used the PX458 vector [48] with a gRNA targeting the first exon of the B2M gene, which encodes a signal peptide of the B2M protein. A schematic illustration of the gRNA-targeted sequence of the human B2M gene is shown in Fig. 1B. After the cell sorting, selected clones were analyzed for B2M and HLA-I expression by flow cytometry. Expression of B2M and HLA-ABC was not detected in ΔiPSC-B even after IFNγ treatment, confirming that the functional knockout of the B2M gene led to a complete deficiency of HLA-I proteins on the cell surface (Fig. 1C). In addition, B2M knockout in ΔiPSC-B was validated by Sanger sequencing. We revealed deletions of 11 bp and 2bp in alleles of the B2M gene in ΔiPSC-B, both causing a frameshift mutation (Fig. 1B). ΔiPSC-B displayed typical pluripotent stem cell morphology and maintained normal karyotype 46, XX (Additional file 2: Fig. S1D). They expressed key markers for pluripotency in nuclei (OCT4, SOX2, NANOG) and cell surface (SSEA-4, TRA-1-60) (Additional file 2: Fig. S1A, S1C). Upon spontaneous in vitro differentiation, ΔiPSC-B derivatives displayed markers of all three germ layers, i.e., ectoderm (CK18), mesoderm (CD31), and endoderm (HNF4A) (Additional file 2: Fig. S1B). Thus, ΔiPSC-B maintained pluripotency after genome editing.
The wild-type iPSCs and ΔiPSCs were differentiated into fibroblast-like cells (iPS-fibro and ΔiPS-fibro) through the stage of 3D spheroids. The main inductors of differentiation were EGF, BPM-4 and bFGF. The differentiation protocol is shown in Fig. 1D. In brief, 3D spheroids formed from iPSCs were cultured in a dynamic suspension for 14 days. They were then transferred to Matrigel-coated plates, where cells migrated, forming a monolayer at the dish bottom. The differentiated cells had fibroblast-like morphology (Fig. 1E) and expressed markers specific for fibroblasts: CD73 (ecto-5'-nucleotidase), CD90 (Thy-1), and CD105 (endoglin) (Fig. 1F). Neither B2M nor HLA-ABC expression was detected in ΔiPS-fibro-A and ΔiPS-fibro-B (Additional file 2: Fig. S1E).
iPSC-derivatives did not elicit increased T-cell responses compared with somatic cells.
First, we compared allogeneic and autologous T-cell responses promoted by isogenic dermal fibroblasts and iPS-fibro of donors A and B. Upon cocultivation a 2.5 times higher percentage of allogeneic compared to autologous T-lymphocytes upregulated surface CD69, regardless of whether primary fibroblasts or IPS-fibro were used as targets (Fig. 2A, 2B). A similar T-cell activation level against fibroblasts and iPS-fibro was also observed for autologous T-cells (Fig. 2A). These results indicate that no immunogenic neoepitopes emerged during reprogramming, cultivation, and subsequent differentiation in the iPS-fibro of both donors.
In addition, we showed that the absence of foreign HLA class I molecules reduced the activation of allogeneic T-cells against ΔiPS-fibro (Fig. 2B). Meanwhile, as we expected, the absence of “self” HLA class I molecules did not affect the immune response of autologous T-lymphocytes (Fig. 2A). Moreover, for ΔiPS-fibro, we did not observe a difference in the activation of autologous or allogeneic effector cells (Fig. 2A). This data proves that manipulating HLA expression in hPSCs can lead to the immunological tolerance of hPSC-derivatives to allogeneic T-cells.
iPSC-derivatives were vulnerable to NK-cell degranulation and cytotoxicity regardless of HLA-I status.
Next, we examined NK-cell immune responses to isogenic dermal fibroblasts and iPSC-derivatives using CD107a mobilization assay and LDH cytotoxicity test. In the analysis of T-cell activity, we observed low variation among donors in the expression of the CD69 activation marker (Fig. 2B). The low variation allowed for the direct comparison of CD69+ T-cells activated by co-culturing with analyzed cells. On the contrary, in the CD107a mobilization assay, we observed substantial variation among different donors in the number of degranulated NK-cells (Additional file 2: Fig. S2A). Furthermore, considerable variation was also observed in independent experiments with NK-cells from the same donor (Additional file 2: Fig. S2B). Therefore, we introduced the degranulation index calculated as a ratio of the number of CD107a+ NK-cells to the number of CD107a+ NK-cells cocultured with the positive control’s K562 cells:
$$\text{Degranulation index}\text{ }\text{= }\frac{{\text{CD107a}}^{\text{+}}\text{ NK cells in analyzed samples}}{{\text{CD107a}}^{\text{+}}\text{ NK cells in positive control}}$$
According to the “missing-self” hypothesis, one of the major functions of NK-cells is the recognition of cells lacking self HLA class I molecules. As expected, NK-cell activity against dermal fibroblasts was relatively low, and the degranulation index of allogeneic NK-cells was insignificantly higher than that of autologous NK-cells (Fig. 3A). Surprisingly, we observed that iPS-fibro provoked the aggressive degranulation of allogenic and even autologous NK-cells. The degranulation index of allogeneic NK-cells against iPS-fibro was 1.7 times higher compared to that of isogenic dermal fibroblasts. That difference was even higher for autologous NK-cells. The autologous NK-cell response on iPS-fibro was 2.7 times higher compared to isogenic dermal fibroblasts (Fig. 3A). Moreover, the ΔiPS-fibro lacking HLA-I proteins provoked the response of allogenic and autologous NK-cells to the same extent as isogenic wild-type iPSC-derivatives (Fig. 3B). CD107a mobilization assay data was consistent with LDH cytotoxicity tests. On all effector/target ratios, the level of NK-cell cytotoxicity was higher against iPSC-derivatives but not against dermal fibroblasts (Fig. 3D). We also tested NK-cell response on other iPSC-derived cells: retinal pigment epithelium (iPS-RPE) and cardiomyocytes (iPS-CM). These cells were also susceptible to the cytotoxic properties of NK-cells (Additional file 2: Fig. S2F, S2I). The differentiation protocols as well as characteristics of iPSC-derived RPE and CM are shown in Additional file 2: Fig. S2C-E and S2G-H.
Gene expression profiling revealed the imbalance of NK-cell ligands in iPSC-derivatives.
The NK-cell can respond to increased signals from activating receptors and decreased signals from inhibitory receptors. We proposed that there was no proper balance of NK-cell ligands in iPSC-derivatives that provoked an NK-cell response. To test this, we performed gene expression profiling in isogenic dermal fibroblasts, iPS-fibro, and ΔiPS-fibro from donors A and B. To increase the resolution of the analysis, we included analogous samples from donor C. Our main goal was to identify differentially expressed genes encoding ligands for activating and inhibitory NK-cell receptors.
We first assessed whether iPS-fibro derivatives were similar to primary human fibroblasts. We selected three publicly available datasets consisting of the transcriptome data (GSE61390, GSE62772, GSE73211) on fibroblasts reprogrammed to iPSCs, iPSCs, and iPSC-derived fibroblast-like cells [49–51]. The analysis of RNA-sequencing data demonstrated that our iPS-fibro and ΔiPS-fibro were largely like primary human fibroblasts (Fig. 4A), with a correlation over 0.9 between our iPSC-derivatives and human fibroblasts (Fig. 4B). The highest level of correlation was observed between our iPS-fibro and other iPSC-derived fibroblast-like cells (Fig. 4B). These results indicate that interlaboratory differences in reprogramming, differentiation, and cultivation did not affect the transcriptomic signature of fibroblast-like iPSC-derivatives.
We then identified differentially expressed genes in our iPS-fibro compared to parental fibroblasts (Additional file 2: Fig. S3A). Transcriptomic analysis revealed that 1670 genes were downregulated. Among them were 13 genes encoding ligands for NK-cell receptors or other molecules necessary for NK-cell activation (GO:0030101) (Additional file 3). Similarly, 1597 genes were upregulated in iPS-fibro, with 12 of them encoding molecules necessary for NK-cell activation. Gene Ontology enrichment analysis showed that genes downregulated in iPS-fibro were significantly enriched in several immunological pathways, including immune response, immune effector response, and inflammatory response (Additional file 2: Fig. S3C). Gene Set Enrichment Analysis revealed that genes from sets “Hallmark interferon-gamma response” and “Hallmark interferon-alpha response” were significantly downregulated in iPS-fibro (Additional file 2: Fig. S3E).
Next, we focused on the pattern of NK-cell ligands expression (Fig. 4C; Additional file 2: Fig. S3D). We found that many genes encoding ligands for inhibitory and activating receptors were differentially expressed in iPSC-derivatives compared with primary fibroblasts. Alongside this, there was a consistency in the expression of ligands for NK-cell receptors between iPS-fibro and ΔiPS-fibro (Additional file 2: Fig. S3B).
HLA class I molecules serve as ligands for two main classes of inhibitory NK-cell receptors: the KIR (Killer-cell immunoglobulin-like receptors) family and the CD94-NKG2A heterodimer. This interaction underlies the molecular basis of “missing-self” recognition. Compared to their parental fibroblasts, the expression of all classical HLA class I transcripts (HLA-A, HLA-B, and HLA-C) was down-regulated in iPS-fibro with fold change > 2. Likewise, the expression of the light chain of HLA class I molecules, B2M, was also down-regulated in iPS-fibro. The downregulation of HLA-A, HLA-B, HLA-C, and B2M genes for iPS-fibro-A and iPS-fibro-B was validated by RT-qPCR (Additional file 2: Fig. S4A). The difference between fibro-B and iPS-fibro-B was more drastic than the other two isogenic systems. This phenomenon may be explained by the comparatively higher expression of HLA-I molecules for fibro-B and may be a donor-specific feature.
The level of non-classical HLA class I transcripts was also reduced in iPS-fibro. Non-classical HLA-I includes HLA-E, HLA-G, and HLA-F genes that exert immunomodulatory properties in NK-cells. Accordingly, the decline in their expression might also promote the manifestation of NK-cell cytotoxic functions. On the other hand, no differentially expressed genes were found among the other ligands for minor inhibitory receptors (such as PD-1, NKRP1A, CEACAM1, CD96, TIGIT, KLRG1, and TIM-3). Hence, we assumed that a relatively low level of HLA class I molecules in the iPS-fibro led to the deficiency in inhibitory signals that might tip the balance towards the activation of the cytotoxic program of NK-cells.
Almost half of the genes encoding the key ligands for activating NK-cell receptors were differentially expressed in iPS-fibro (Fig. 4C). Upregulated genes included the genes of ligands for dominant activating receptors: NKG2D, DNAM-1, and natural cytotoxicity receptors (NCRs). Compared to their parental fibroblasts, the stress-induced molecule’s MICA (NKG2D ligand) gene expression was more than 1.5 times higher in iPS-fibro. The DNAM-1 ligand, NECTIN2 (CD112) and PVR (CD155), and the NKp30 ligand, NCR3LG1 (B7-H6), underwent a more noticeable increase in gene expression with fold-change of > 3 in iPS-fibro. The upregulation of MICA, ULBP1, NECTIN2, and PVR genes for iPS-fibro-A and iPS-fibro-B was validated by RT-qPCR (Additional file 2: Fig. S4B). Finally, some genes such as CADM1 (CRTAM ligand) and CD70 (CD27 ligand) were expressed only in iPSC-derivatives but not in parental fibroblasts. Notably, an imbalance in the expression of ligands for activating NK-cell receptors was also observed in publicly available RNA-seq datasets. In particular, NECTIN2, PVR, CADM1, and CD70 gene expression was upregulated in independently derived fibroblast-like cells (data not shown). In this regard, we supposed that it might be an intrinsic feature of iPSC-derivatives, at least for fibroblast-like cells.
In addition, we analyzed the gene expression of adhesion molecules. The interaction of adhesion molecules with their receptors on NK-cells contributes to firming NK-cell adhesion to the target cell and leads to the assembly of immunological synapses essential for target cell killing [52]. ICAM-1 (LFA-1 ligand) and VCAM-1 (VLA-4 or integrin α4β1 ligand) genes were upregulated in iPSC-derivatives (Fig. 4C). The same change was observed in publicly available RNA-seq datasets (data not shown). The overexpression of some adhesion molecules might also contribute to NK-cell mediated cytotoxicity against iPSC-derivatives.
Various factors have affected the balance regulating the response of NK cells to iPS-fibro. First, we observed a relatively low gene expression of HLA-I molecules, major inhibitory ligands, in iPS-fibro. Second, genes coding for main activating NK-cell ligands were upregulated in iPS-fibro. Third, the genes of some adhesion molecules were also overexpressed in iPS-fibro.
IFNγ treatment increased HLA-I expression and reduced NK-cell-mediated cytotoxicity towards iPS-fibro.
The diminished HLA-I expression leads to a decisive advantage of activating NK-cells. Therefore, to bring the balance of activating and inhibitory ligands into an equilibrium state, it is necessary to boost the HLA class I expression level. A shift in the balance towards inhibition should lead to a decrease in NK-cell activation and cytotoxicity. Therefore, we analyzed whether changing the proportion of activating and inhibitory ligands for NK-cell receptors in iPS-fibro was possible.
The IFNγ treatment enhanced the HLA-I gene and protein expression more than twice in fibroblasts and over 6 times in iPS-fibro (Fig. 5A; Additional file 2: Fig. S5A). The HLA-ABC gene expression was higher in iPS-fibro pretreated with IFNγ than in intact parental fibroblasts (Additional file 2: Fig. S5A). IFNγ treatment also slightly affected the gene expression of activating NK-cell receptors (Additional file 2: Fig. S5B, S5C).
Next, we assessed whether IFNγ-pretreatment could alter the NK-cell response. The response of autologous and allogenic NK-cells declined by about half toward IFNγ-pretreated iPS-fibro compared to their unstimulated counterparts (Fig. 5B). Interestingly, the NK-cell response to IFNγ-pretreated iPS-fibro dropped nearly to values typical for intact isogeneic fibroblasts. In contrast, IFNγ-stimulation did not alter NK-cell responses to ΔiPS-fibro with the knockout of the B2M gene. Both the degranulation index and NK-cell cytotoxicity remained at a level comparable to untreated samples (Fig. 5B, 5C). NK-cell activation against autologous fibroblasts also changed only slightly. However, allogeneic NK-cells significantly reduced the release of cytotoxic granules after IFNγ treatment of parental fibroblasts (Fig. 5C). Thus, there was no mitigation in cytotoxicity against parental fibroblasts pretreated with IFNγ. The reduced cytotoxic effect was observed only when IFNγ-pretreated fibro-A were cocultured with NK-cells of donor 2 (Fig. 5C).
Since the low expression of HLA class I molecules is a common feature of hPSCs, the low expression of HLA class I molecules by intact iPS-fibro might be associated with immaturity. Therefore, we evaluated the HLA-ABC and B2M expression by parental fibroblasts and iPS-fibro at different passages. We showed that the most significant difference in the expression of HLA-ABC and B2M was observed between parental fibroblasts and iPS-fibro, on the second passage, i.e., the “youngest” iPSC-derived cells (Fig. 5D; Additional file 2: Fig. S5D). Further in the process of cultivation and passaging, a significant increase in the HLA-ABC and B2M expression was observed (Fig. 5D; Additional file 2: Fig. S5D).