Abrogation of HLA surface expression using CRISPR/Cas9 genome editing: a step toward universal T cell therapy

As recent advancements in the chimeric antigen receptor-T cells have revolutionized the way blood cancers are handled, potential benefits from producing off-the-shelf, standardized immune cells entail the need for development of allogeneic immune cell therapy. However, host rejection driven by HLA disparity in adoptively transferred allogeneic T cells remains a key obstacle to the universal donor T cell therapy. To evade donor HLA-mediated immune rejection, we attempted to eliminate T cell’s HLA through the CRISPR/Cas9 gene editing system. First, we screened 60 gRNAs targeting B2M and multiple sets of gRNA each targeting α chains of HLA-II (DPA, DQA and DRA, respectively) using web-based design tools, and identified specific gRNA sequences highly efficient for target deletion without carrying off-target effects. Multiplex genome editing of primary human T cells achieved by the newly discovered gRNAs yielded HLA-I- or HLA-I/II-deficient T cells that were phenotypically unaltered and functionally intact. The overnight mixed lymphocyte reactions demonstrated the HLA-I-negative cells induced decreased production of IFN-γ and TNF-α in alloreactive T cells, and deficiency of HLA-I/II in T cells further dampened the inflammatory responses. Taken together, our approach will provide an efficacious pathway toward the universal donor cell generation by manipulating HLA expression in therapeutic T cells.

The emergence of chimeric antigen receptor (CAR)-T cell therapy has changed the paradigm of cancer immunotherapy by virtue of its durable remission with manageable toxicity profile. Two CD19-targeting autologous CAR-T products have shown high rates of responses in clinical studies and recently received FDA approval for the treatment of B-cell lymphoid malignancies [1][2][3] . However, challenges in the utilization of autologous cells from patients extend from the labor-intensive, high-cost nature of the manufacturing procedures required for individualized therapy to the limited quantity and/or sub-optimal intrinsic quality of patient T cells. Genomic, phenotypic, and functional analyses of CD19 CAR-T cells from treatment-responding patients in chronic lymphocytic leukemia demonstrated that the intrinsic properties of T cells, such as upregulation of IL-6/STAT3 signaling and enhanced transcription of memory T cell-related genes, are key determinants of the efficacy of the therapy 4 . Limited efficacy associated with epigenetic modulation, impaired functionality, and more exhaustion and apoptotic phenotypes of the infused autologous T cell products can preclude durable remission and clinical response following treatment [4][5][6][7] . Furthermore, a recent report in a trial of autologous CAR-T cell therapy documented the unintentional transduction of a single leukemic B cell with anti-CD19 CAR during manufacturing and its product masking the target antigen, thereby escaping CAR-T recognition and resisting the therapy 8 . These limitations could be overcome by employing allogeneic T cells obtained from "fitted" healthy donors as a source of universal CAR-T cell production. But issues related to HLA barriers, such as the risks of graft-versus host disease (GvHD), as well as host-mediated rejection of the infused allogeneic cells, are the major obstacles to this intervention.
Donor allogeneic T cells recognize the "non-self " antigens of the host cells and elicit GvHD by TCR αβ-mediated signaling 9 . Elimination of endogenous TCR from the donor T cells by disruption of the TCR α or β chain using different gene editing technologies, such as zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs), or the CRISPR/Cas9 system, has been shown to prevent alloresponses from donor T Scientific Reports | (2020) 10:17753 | https://doi.org/10.1038/s41598-020-74772-9 www.nature.com/scientificreports/ cells [10][11][12][13][14] . Using γδ T cells, which are unlikely to cause GvHD, to generate CAR-T cells is an alternative approach without genetic manipulation to circumvent GvHD 15 . Conversely, graft rejection is caused by the host immune system recognizing the mismatching HLAs of the infused or transplanted allogeneic cells. One of the approaches to evade host-mediated rejection is to eliminate HLA-I from donor T cells. Ren et al. demonstrated deletion of the B2M gene, an essential component for the cell surface expression of HLA-I, through the CRISPR/Cas9 system in human T cells and showed that HLA-I-deficient T cells resulted in a reduction of surveillance of allogeneic T cells compared to non-gene-edited T cells 13 . Reports by others further indicated that HLA-II molecules are highly expressed on activated T cells as well 16,17 , and HLA-II mismatch can activate the alloreactive CD4 + T cells of the recipient 18,19 . While editing of multiple HLA targets in allogeneic T cells may offer great potential for universal CAR-T cell therapy, no report has yet demonstrated highly efficient methods to simultaneously abrogate the expression of HLA class I and II in T cells and its impact in inducing alloresponse.
In this study, we discovered gRNAs specific to B2M and HLA class II that enabled highly efficient and ontarget genome editing, and delivered those gRNAs simultaneously into primary human T cells. We have targeted α chains of HLA-II genes (HLA-DRA, DQA and DPA) because they are relatively less polymorphic compared with β chains. The gene-edited HLA-I/II-negative T cells retained their T cell functionality and phenotypes upon in vitro stimulation. Additionally, in vitro mixed lymphocyte reactions revealed that alloresponses in responder cells were dampened against HLA-I/II-negative T cells compared to HLA-I-negative cells, solidifying a conceptual framework that narrates the role of HLA-II plays in infused, therapeutic allogeneic cells during host-mediated rejection.

Results
Newly discovered gRNAs for HLA deletion showed high deletion efficiency without off-target effects. In order to ablate the expression of HLA-I and HLA-II on the cell surface, we attempted to target genes encoding β 2 -microglobulin (B2M) and α chains of HLA-II molecules with the CRISPR/Cas9 gene editing system. We employed web-based gRNA designing tools such as CHOPCHOP 20 , E-CRISP 21 and CRISPR-ERP 22 to identify gRNA sequences targeting the B2M, HLA-DRA, HLA-DQA and HLA-DPA gene. Out of hundreds of gRNA candidate sequences per target, we narrowed the lists to 60 gRNA sequences for each target gene (see the Methods section for the criteria and details). Then, the gRNAs were transcribed in vitro and transfected into Raji cells together with the Cas9 protein to validate their target deletion efficiency. 20 gRNAs were tested per experiment, and the gRNAs showing deletion efficiencies exceeding the internal criteria in three independent experiments were selected (Supplementary figure S1). The selected gRNAs were tested again collectively in a single experiment, and the gRNAs highly efficient in target deletion (> 70%) were identified (Fig. 1A). Currently, 29 DRA alleles, 216 DQA1 alleles and 161 DPA1 alleles are assigned in the IPD-IMGT/HLA database 23 , and genomic sequences are available for 28 DRA alleles, 140 DQA1 alleles and 86 DPA1 alleles 23 (Supplementary  table S1). Among the newly identified gRNAs, DQA-40 and DPA-13 gRNA were ultimately selected for further experiments because their target sequences are conserved in all DQA1 or DPA1 alleles whose sequences are publicly available. All known DRA allele sequences were covered by the three selected gRNAs above, so we chose DRA-18 gRNA as a final candidate because it had the highest deletion efficiency.
Genomic mutations in each target were confirmed by mismatch-sensitive enzyme-based assays and deep sequencing (Fig. 1B,C). To analyze possible off-target effects, we sequenced the top five predicted off-target sites for each target after gRNA delivery and observed no detectable off-target events (Fig. 1C).
CRISPR-Cas9-mediated multiplex genome editing efficiently ablates expression of HLA on human primary T cells without affecting effector function. While HLA-I is expressed by all nucleated cells, constitutive HLA-II expression is conventionally thought to be restricted to professional antigenpresenting cells such as dendritic cells, B cells and macrophages. However, previous reports by others as well as our own data showed increased levels of HLA-II in NK cells or T cells when they were expanded or activated ex vivo 16,24 (Supplementary figure S2), and host-derived alloreactive immune rejection is mediated by expression of HLA-I and HLA-II molecules on infused third-party cells.
In order to examine whether selected gRNAs can simultaneously suppress the expression of HLA-I and HLA-II in T cells, we transfected a mixture of four gRNAs each targeting the B2M, HLA-DRA, HLA-DQA and HLA-DPA gene (referred to hereafter as quadruple gRNAs) or non-targeting gRNA together with the Cas9 protein into primary human CD3 + T cells isolated from healthy donor-derived PBMCs. After 13 days of expansion of the transfected cells, the control CD4 + and CD8 + T cells treated with non-targeting gRNA expressed high levels of HLA-I and HLA-II, as previously reported, and this expression was efficiently downregulated by transfection with the quadruple gRNAs ( Fig. 2A,B). The quadruple-gene-edited CD3 + T cells exhibited 62.1% of HLA-I/II-double-negative cells ( To test whether the HLA-negative T cells maintain the CD4/CD8 subset ratio and T cell phenotypes, we measured expression of surface CD4, CD8 and T cell activation/exhaustion markers in the HLA-negative T cells and control T cells. As seen in Fig. 3B,C, there were no differences in CD4 + /CD8 + T cell distribution and phenotypes between the HLA-I/II-negative T cells and the control T cells.
To further analyze the effect of HLA ablation on T cell effector function, we measured proinflammatory cytokine production and CD107a release in the genetically engineered T cells stimulated with PMA/ionomycin or anti-CD3/CD28 beads. No significant changes were noticed in the production of TNF-α and IFN-γ or the release of cytotoxic granules between the HLA-I/II-negative T cells and control T cells in vitro (Fig. 3D www.nature.com/scientificreports/ together, our data demonstrate that simultaneous delivery of multiple gRNAs targeting B2M, HLA-DRA, HLA-DQA and HLA-DPA can efficiently eliminate HLA molecules from human primary T cells without affecting their effector functionality or surface phenotypes.

Alloreactive T cell responses are mediated by expression of HLA-I/II in target T cells. To inves-
tigate whether the deletion of HLA molecules in T cells can alleviate alloreactive immune responses, HLA-I/ II-positive, HLA-I-negative or HLA-I/II-negative cells were sorted from the non-target gRNA, B2M gRNA or quadruple gRNAs-treated CD3 + T cells, respectively (Fig. 4A). Then, the sorted cells were γ-irradiated and cocultured with another set of allogeneic PBMCs stained with a CellTrace Violet (CTV) dye. The level of alloreactivity was determined by intracellular cytokine production in CD3 + T cells gated from the CTV + allogeneic PBMCs. Our data revealed the elimination of HLA-I on the target cell surface diminished IFN-γ and TNF-α secretion from the allogeneic T cells, and the alloresponse was further abrogated by HLA-I/II ablation www.nature.com/scientificreports/ ( Fig. 4B,C). These results indicate that HLA expression serves as an indispensable immune modulator required for eliciting alloreactive T cell responses and that elimination of both the HLA-I and HLA-II molecules on donor cells is crucial to dampen host-mediated rejections in the setting of allogeneic cellular therapy.

Discussion
The current CAR-T cell therapies approved for the treatment of hematological malignancies require genetic engineering of patient-derived T cells. There are issues related to the manufacturing processes and quantity of these T cells in a fraction of pediatric or intensively treated patients, especially due to the heavy lymphocytotoxic chemotherapies the patients often receive 1,25 . Additionally, an increasing amount of data supports that greater clinical efficacy of the CAR-T cell therapy is determined by intrinsic factors of the T cells, such as poly-functionality and memory-like phenotypes 4,6,26 . Off-the-shelf CAR-T cells produced from healthy allogeneic donor T cells with the "fittest" phenotypes can increase the likelihood of therapeutic response and overcome the barriers of the autologous CAR-T cell therapy. The allogeneic CAR-T cell therapy, in order to evaluate its feasibility and broader applicability, must primarily address the HLA barriers that mediate the frequency and magnitude of alloimmune responses driven by the host.
To bypass the recognition of foreign HLA molecules by the host cells, we have identified highly specific gRNA sequences for HLA class I/II deletion on the cell surface and generated T cells lacking surface HLA molecules www.nature.com/scientificreports/ through simultaneous quadruple genome editing. Our in vitro data suggest that the genetically modified T cells retained their functionality and immune phenotypes and that alloresponses against the genetically engineered T cells were markedly reduced compared with control cells. To our knowledge, this is the first report presenting engineered human T cells devoid of both HLA-I and HLA-II molecules. Previously, HLA-A-negative primary human T cells were generated by ZFNs targeting the HLA-A locus and evaded HLA-A2-restricted T cell recognition 27 . Also, HLA-I-deficient human primary T cells generated by B2M gene disruption using the CRISPR/Cas9 gene editing system were shown to reduce alloresponses compared with wild-type control cells 13 . While the expression of HLA-II molecules in T cells was induced after ex vivo expansion (Supplementary figure S2), elimination of HLA-II molecules from the surface of primary T cells has not yet been reported, probably due to the highly polymorphic nature of HLA-II genes. We identified gRNAs able to cover the majority of alleles of each HLA-II α chain gene (HLA-DRA, HLA-DQA and HLA-DPA, respectively), which are less polymorphic than β chains. Among the gRNAs selected by their high target editing efficiency in Raji cells, we could select one gRNA for each target (DRA-18, DQA-40, DPA-13) that could cover all the target alleles whose sequences are known to date 23 . In B lymphoblastoid cell lines or untransformed human endothelial cells, ablation of HLA-II was achieved by a single-gRNA-driven triple-gene knockout of the HLA class II β chain 28 or by targeted disruption of class II transactivator (CIITA), an essential transcription factor for HLA-II genes 19,29 . Recently, several reports have described induced pluripotent stem cells (iPSCs) lacking HLA-II molecules generated by CRISPR/Cas9-mediated CIITA targeting [30][31][32] . Because the purpose of the above studies was to generate HLA-II knockout cells, which could be obtained by expanding a single engineered clone, the genome editing efficiency was of less importance. Indeed, the efficiency of HLA-II deletion before cell sorting or antibiotic selection was either low 28 (shown by 1.5 to 9.7% of HLA-DR negative cells) or not specified 19,[29][30][31][32] . We have tested the CIITA-targeting gRNAs used by the two studies above 29,31 , and obtained much lower efficiency in HLA-II ablation compared with HLA-II α chain gene-targeting gRNAs we have employed (Supplementary figure S6). In generating engineered mature primary cells, higher genome editing efficiency is critical due to the limited expansion potential of primary cells in the culture, implicating potential benefits provided by our gRNA sequences. Additionally, targeting transcription factors may confer unpremeditated transcriptional changes to several genes outside of HLA-II biology. For www.nature.com/scientificreports/ instance, expression of RAB4B, a protein involved in endocytic recycling, is augmented by CIITA, whereas IL4, FasL, and COL1A2 are repressive targets of CIITA [33][34][35][36] . These findings imply that CIITA could play a broader role inside and outside of HLA-II-mediated antigen presentation processes and that targeting such a transcription factor may appeal less attractive for primary cell engineering aimed for clinical use. Several issues still remain to be addressed with allogeneic HLA-negative T cells for therapeutic application. The first concern is that a lack of HLA class I/II molecules on the cell surface can induce the "missing-self " response from host NK cells, resulting in the lysis of the infused cells. To this end, HLA-E or HLA-G, a ligand for the NK inhibitory receptor CD94/NKG2A or ILT2, respectively, can be expressed on the HLA-negative donor cells to mitigate NK cell-mediated cytotoxicity 37,38 . Overexpression of other molecules interacting with NK inhibitory receptors 39,40 or CD47, which transmits a "don't-eat-me" signal and inhibits phagocytosis 30,41 , can be alternative targets to evade NK cell-driven host innate immune responses.
Another concern is off-target nuclease activity driven by highly efficient CRISPR/Cas9-mediated genome editing. Although we have demonstrated that the top five off-target site candidates for each target gene did not exhibit any mutations after transfection, it is worth noting that there could be off-target mutagenesis at sites other than predicted DNA sequences in a given genome, especially one with less than three mismatches. Further investigation is required to evaluate the unbiased identification of off-target cutting using whole-genome sequencing technology such as Digenome-seq 42 . Two Cas9 mutants, eSpCas9 and SpCas9-HF1, have demonstrated improved DNA targeting specificity 43,44 , and those variants could also be employed to avoid any undesirable off-target events if detected. www.nature.com/scientificreports/ A recent paper describing multiple-gene-edited CAR-T cells by three gRNAs demonstrated the presence of chromosomal translocation during cell manufacturing 45 . In accordance with this result, our end-point PCR results using genomic DNA extracted from quadruple-gene-edited T cells from 3 different PBMC donors show that 19 rearrangements were induced among 20 potential rearrangements upon transfection with our four gRNAs (Supplementary figure S7). Most of the detected rearrangements were decreased during the expansion while only one translocation (DQA2:DRA) were slightly increased in 2 donors on expansion day 28. Together with the previous data illustrating declined proliferative capacity of the multiple-gRNA-transfected T cells compared with the controls (Supplementary figure S4), these results implicate that evident chromosomal rearrangement accompanied with transfection would not lead to growth advantages in the HLA-deficient T cells.
Although evaluation of HLA-deficient T cells' efficacy in more clinically relevant settings needs to be preceded, our approach to target HLA expression on allogeneic cell surfaces could be applied more broadly to other therapeutic strategies, as exemplified in our data showing ex vivo stimulated NK cells with high HLA-II expression (Supplementary figure S2). Retinal pigment epithelial (RPE) cells are also explored to be used as allografts for the treatment of ocular diseases, and are reported to upregulate HLA-II in the presence of IFN-γ-mediated inflammatory responses 46 . Our engineering strategy may potentially be considered for the ablation of HLA molecules on those cells to control the host Th1 immune response 18 for successful engraftment without HLA matching and/or conventional immune suppression [47][48][49][50][51] .
Our findings expand upon the increasingly recognized potential of allogeneic cell therapeutics in diverse clinical settings by identifying novel gRNA sequences ablating the surface expression of HLA, including highly polymorphic HLA-II molecules. The HLA-negative human primary T cells generated by delivery of the newly identified gRNAs maintained T cell functionality and activation phenotypes while elicited dramatically diminished alloreactive inflammatory reactions in responders. As our study reminds us of an essential conceptual aspect of HLA barriers in production of universal donor cells, it can ultimately be exploited to design new therapeutic strategies that decrease the risk of graft rejection and improve the clinical outcomes.
Methods gRNA design and synthesis. gRNAs for each target gene were designed using the web-based tools CHOPCHOP 20 , E-CRISP 21 , and CRISPR-ERA 22 . Genomic sequences of the HLA-DRA allele 01:01 and HLA-DQA allele 01:01, known HLA type for Raji, and HLA-DPA allele 01:03 were used for gRNA design. Hundreds of candidate sequences were obtained from the web-based tools, and sixty gRNA sequences for each target gene (B2M, HLA-DRA, HLA-DQA and HLA-DPA) were pre-selected from the list according to the criteria: 1) Sequences which were obtained from multiple tools, 2) Sequences which exist in an exon, and 3) Sequences which have high rank in each tool. gRNAs were transcribed in vitro using a GeneArt Precision gRNA Synthesis Kit (Thermo Fisher Scientific, A29377) according to the manufacturer's protocol. gRNA targeting sequences used in the study are listed in Supplementary Table S2.

Preparation of human PBMCs. The study was approved by the institutional review board of the Mogam
Institute for Biomedical Research (MG-2018-10-01) and conducted according to the Declaration of Helsinki. Human PBMCs were obtained from healthy volunteers by leukapheresis from the National Red Cross Blood Center (Suwon, South Korea). The informed consent form was signed by all subjects. PBMCs were isolated by centrifugation on a Ficoll density gradient and stored in liquid nitrogen.

Human T cell transfection and expansion.
Cryopreserved human PBMCs from healthy donors were thawed, and CD3 + cells were isolated using human CD3 MicroBeads (Miltenyi Biotec, 130-050-101). Upon T cell isolation, the CD3 + T cells were initially cultured in X-VIVO15 (Lonza, BE02-060Q) supplemented with Dynabeads Human T-Activator CD3/CD28 (Gibco, 111.31D), 200 IU/mL IL-2 (Novartis, 502519AF) and 5% human plasma (Valley Biomedical, HP1050) in a culture bag (NIPRO, 87-352). On day 1, 37.5 μg of Cas9 protein and 37.5 μg of gRNA per reaction were mixed and incubated at room temperature for 10 min to form Cas9/gRNA complex. For multiplex genome editing, 37.5 μg of gRNA in total were used. The CD3 + T cells were taken out from the culture bag and 4 × 10 6 cells were transfected with Cas9/gRNA complex by 4D-Nucleofector X Unit (Lonza, AAF-1002X) using P3 Primary Cell 4D-Nucleofector X Kit L (Lonza, V4XP-3024) with the program DN-100. Following transfection, the cells were cultured at 37 °C in 5% CO 2 , and fresh culture medium, X-VIVO15 (Lonza, BE02-060Q) supplemented with 200 IU/mL IL-2 (Novartis, 502519AF) and 5% human plasma (Valley Biomedical, HP1050), was added every 2 to 3 days to reach a density of 1 × 10 6 cells/mL. The cell number and viability were documented using the automated fluorescence cell counter (NanoEntek, ADAM-MC). The Dynabeads Human T-Activator CD3/CD28, which was provided in the culture media at the beginning of the ex vivo expansion was not replenished further.  Mismatch-specific nuclease-mediated mutagenesis analysis. Genomic   www.nature.com/scientificreports/ Supplementary Table S4. All 20 possible primer combinations B2M:DRA, B2M:DQA1, B2M:DQA2, B2M:DPA, DQA:B2M and so forth, were tested to detect rearrangement. HPRT was used as an endogenous control. Synthetic double-stranded DNA containing tandem sequences for the 5 forward primers and for the 5 reverse primers were inserted into pUC19 plasmid and used as a positive control. Samples were loaded on 1.5% agarose gel.