Magnitude of Off-Target Allo-HLA Reactivity by Third-Party Donor-Derived Virus-Specific T Cells Is Dictated by HLA-Restriction

T-cell products derived from third-party donors are clinically applied, but harbor the risk of off-target toxicity via induction of allo-HLA cross-reactivity directed against mismatched alleles. We used third-party donor-derived virus-specific T cells as model to investigate whether virus-specificity, HLA restriction and/or HLA background can predict the risk of allo-HLA cross-reactivity. Virus-specific CD8pos T cells were isolated from HLA-A*01:01/B*08:01 or HLA-A*02:01/B*07:02 positive donors. Allo-HLA cross-reactivity was tested using an EBV-LCL panel covering 116 allogeneic HLA molecules and confirmed using K562 cells retrovirally transduced with single HLA-class-I alleles of interest. HLA-B*08:01-restricted T cells showed the highest frequency and diversity of allo-HLA cross-reactivity, regardless of virus-specificity, which was skewed toward multiple recurrent allogeneic HLA-B molecules. Thymic selection for other HLA-B alleles significantly influenced the level of allo-HLA cross-reactivity mediated by HLA-B*08:01-restricted T cells. These results suggest that the degree and specificity of allo-HLA cross-reactivity by T cells follow rules. The risk of off-target toxicity after infusion of incompletely matched third-party donor-derived virus-specific T cells may be reduced by selection of T cells with a specific HLA restriction and background.


INTRODUCTION
Adoptive transfer of autologous or human leukocyte antigen (HLA)-matched patient-specific T-cell products, including antigen-specific T cells, Chimeric Antigen Receptor (CAR) T cells and T-cell receptor (TCR) modified T cells are clinically applied and show feasibility and safety (1)(2)(3)(4)(5). Nevertheless, the complex logistics and delays associated with the generation of these products for adoptive immunotherapy strategies are hampering easy broad application. Off-the shelf T-cell products, generated from cells of healthy third-party donors and suitable for treatment of multiple patients, may be an elegant solution, but such products are often only partially HLA-matched with the recipient.
In our study we focused on virus-specific T-cell products derived from healthy third-party donors that can be used for the treatment of uncontrolled viral reactivations and/or viral disease in patients without easy access to autologous or donorderived virus-specific T cells. Reactivations of cytomegalovirus (CMV), Epstein-Barr virus (EBV) and adenovirus (AdV) are frequently seen and associated with high morbidity and mortality in immune-compromised patients (6,7), like patients after allogeneic stem cell transplantation (AlloSCT), but also patients after solid organ transplantation. For patients transplanted with stem cells from a virus-non-experienced donor or in general for solid-organ donors there is no easy access to (HLA-matched) memory virus-specific T cells. Adoptive transfer of partially HLA-matched virus-specific T cells from healthy third-party donors is a potential strategy to temporarily provide anti-viral immunity to these patients. However, these third party donorderived virus-specific T cells have not been tolerized by thymic negative selection to the non-matched HLA molecules that are present within the patient (8,9), thereby implying the risk of offtarget toxicity due to allo-HLA cross-reactivity directed against the mismatched HLA alleles (10).
It was demonstrated that third-party derived virus-specific T cells can exert allo-HLA cross-reactivity directed against mismatched HLA alleles in vitro (11)(12)(13)(14). The viral specificity as well as the allo-HLA cross-reactivity was shown to be mediated by the same T-cell receptor (TCR) complex (11,14). Additionally, TCR cross-reactivity could be a major trigger of graft rejection, as shown by the association between viral reactivation and graft rejection in recipients of solid organs (15,16). Despite the clearly documented allo-HLA cross-reactivity of virus-specific T-cell populations documented in vitro, only low rates (∼5%) of offtarget toxicity/de novo Graft vs. Host Disease (GVHD) were observed in stem cell recipients that were treated with partially HLA-matched virus-specific T cells (17)(18)(19)(20)(21). There are several potential reasons for this discrepancy: (1) the specific allogeneic peptide/HLA complex recognized by the cross-reactive virusspecific T cells is not present in the patient, (2) removal of the in vitro off-target (>10% cytotoxic) virus-specific T cells from the product prior to administration to the patient and/or selection of T-cell products that do not show in vitro allo-HLA reactivity (18), (3) low T-cell numbers of cross-reactive virusspecific T cells administered and/or limited in vivo proliferation, (4) Rejection of the partly HLA-matched third party virusspecific T cells by the recipient (22). In the last example, such rejection prevents toxicity, but it also diminishes the short-term protection afforded by the third-party derived T cells. In a recent phase I/II clinical study by Neuenhahn et al., survival/persistence was only demonstrated for adoptively transferred virus-specific T cells of the original stem cell donor (8/8 HLA-matched), but not for virus-specific T cells derived from third-party donors with a higher degree of HLA-mismatch (22).
It would be useful if we could predict which non-matched HLA molecules are recognized by third-party derived T cells so that specific donors and/or specific T-cell populations can be selected with a low likelihood of exerting off-target reactivity. Thus far, recurrent off-target reactivity toward the same nonmatched HLA molecule was only found for T-cell populations isolated from different individuals using the exact same TCR (public TCR) (14, 23,24). A classic example of such public cross-reactivity is the HLA-B * 08:01-restricted EBV-EBNA3A FLRspecific T-cell population that contains a dominant public TCR showing cross-reactivity against non-self HLA-B * 44:02 (23,24). Importantly, this public TCR is not found in the T-cell repertoire of HLA-B * 08:01/HLA-B * 44:02 positive individuals, demonstrating the deletion of this otherwise potentially autoreactive public TCR during in vivo thymic selection. Many antiviral T-cell responses are, however, not so clearly dominated by a single dominant public TCR, making predictions of crossreactivity more difficult.
The aim of this study was to investigate whether we could identify and predict allo-HLA cross-reactivity patterns by thirdparty donor-derived T cells, using virus-specific T cells as a model. We investigated whether the allo-HLA cross-reactivity by third-party donor-derived virus-specific T cells was influenced by virus-specificity, HLA-restriction and/or HLA background of the donors. Our data show that the level of allo-HLA crossreactivity is not affected by viral-specificity, but surprisingly strongly associated with HLA restriction and influenced by the HLA background of the donors.

Collection of Donor Material
After informed consent according to the Declaration of Helsinki, healthy donors (homozygously) expressing HLA-A * 01:01 and HLA-B * 08:01 or HLA-A * 02:01 and HLA-B * 07:02 were selected from the Sanquin database and the biobank of the department of Hematology, Leiden University Medical Center (LUMC). Two donors expressing HLA-A * 02:01/HLA-B * 07:02 were not homozygous. Peripheral blood mononuclear cells (PBMCs) were isolated by standard Ficoll-Isopaque separation and used directly or thawed after cryopreservation in the vapor phase of liquid nitrogen. Donor characteristics (HLA typing, CMV and EBV serostatus) are provided in

Isolation and Expansion of Virus-Specific T Cells
Phycoerythrin (PE), allophycocyanin (APC), BV421, BV510 and/or peridinin-chlorophyll-protein (PerCP)-labeled pMHCtetramer complexes were used for fluorescence-activated cell sorting (FACSorting). The pMHC-tetramers used (for generation see Supplementary Material and Methods) are shown in Supplementary Table 1. PeptideMHC-tetramer positive, CD8 pos /CD4 neg T cells were sorted and seeded at 10,000 cells per well in U-bottom microtiter plates for the generation of bulk T-cell populations. After 2 weeks of culture, pMHC-tetramer pos T-cell populations were considered pure if they contained ≥97% pMHC-tetramer pos cells. Polyclonality of the sorted virus-specific T cells was assessed by T-cell receptor-variable β (TCR-Vβ) family analysis using the TCR-Vβ kit (Beckman Coulter, Fullerton, USA). Sub-populations expressing a single TCR-Vβ family were sorted from the bulk using monoclonal antibodies from the TCR-Vβ kit. Sub-populations were then non-specifically expanded. Subpopulations using one specific TCR-Vβ family were considered pure if ≥95% of the population was positive for that TCR-Vβ family. Sorting was performed on a FACS ARIA (BD) and analyzed using Diva software (BD). All analyses were performed on a FACS Calibur (BD), and analyzed using Flowjo Software (TreeStar, Ashland, USA). Procedures to isolate and expand virus-specific T cells are described in the Supplementary Material and Methods.

Selection and Generation of Stimulator Cells for Functional Analyses
EBV-transformed lymphoblastoid cell-lines (EBV-LCLs) were generated according to standard protocols (25). EBV-LCLs were selected to cover a total of 116 frequently occurring HLA molecules, as listed in  The panel of HLA-mismatched EBV-LCLs was composed of high resolution HLA-typed EBV-LCLs that together covered all HLA-class-I and almost all frequently HLA-class-II molecules that are frequently (2%) occurring in the Caucasian population. The HLA typing was determined molecularly. XX indicates that only the allele group could be determined (2 digit resolution). Blanks indicate homozygosity for the given allele.

Cytokine Production Assays to Determine T-Cell Reactivity
Interferon-γ (IFN-γ) production by virus-specific T cells was quantified using standard enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions (Sanquin Reagents, The Netherlands). Responder T cells were co-cultured with stimulator cells at a ratio of 1:10 (responder: stimulator) for 16 h at 37 • C in T-cell medium used for expansion of T-cell populations as described in Supplementary Material using 25 IU/ml Interleukin-2 (IL-2) instead of 100 IU/ml IL-2. Recognition of HLA-mismatched EBV-LCLs, HLA-matched peptide-pulsed EBV-LCLs and K562 cells transduced with specific HLA molecules was defined as production of ≥200 pg/ml of IFN-γ.

Statistical Analysis
Statistical analyses were only performed on quantitative data and were performed using non-parametric tests. The Fisher's-Exact-test was used to assess the differences in cross-reactivity (present or absent) of HLA-mismatched EBV-LCLs between groups (i.e., HLA-A * 01:01-and HLA-B * 08:01-restricted virusspecific T cells). Differences in the numbers of recognized HLA-mismatched EBV-LCLs was first assessed by the Kruskal-Wallis test. Differences between two groups were then further assessed with the non-parametric Mann-Whitney U test. pvalues were adjusted by the Bonferroni correction for multiple testing. Statistical analyses were conducted using GraphPad Prism (GraphPad Software, version 8).

Virus-Specific T-Cell Populations Show Profound and Diverse Cross-Reactivity Against a Panel of HLA-Mismatched EBV-LCLs
To study the influence of HLA restriction and antigen specificity on the level of allo-HLA cross-reactivity mediated by virus-specific T cells, bulk virus-specific T-cell populations targeting single epitopes were isolated from total PBMCs of (homozygous) HLA-A * 01:01/HLA-B * 08:01 pos or HLA-A * 02:01/HLA-B * 07:02 pos healthy donors. Two donors did not homozygously express HLA-A * 02:01/HLA-B * 07:02. All donors were EBV seropositive and 5 out of 12 HLA-A * 01:01/HLA-B * 08 :01 pos donors and 9 out of 12 HLA-A * 02:01/HLA-B * 07:02 pos donors were CMV seropositive ( Table 1). The serostatus for AdV was unknown for all donors. Virus-specific T cells were isolated by FACS using pMHC-tetramers for various peptides (n = 21) from CMV, EBV, and AdV (Supplementary Table 1). In total,   Table 3). These T-cell populations were analyzed for allo-HLA cross-reactivity using a panel of HLA-mismatched EBV-LCLs, expressing the most frequent (>2%) HLA-class-I and class-II antigens in the Caucasian population ( Table 2). EBV-specific T-cell populations were only tested against HLA-mismatched EBV-LCLs that did not express the specific restriction molecules to avoid recognition of EBVderived peptides in self-HLA. In total, 65 out of 164 (39%) virusspecific T-cell populations produced interferon-γ in response to stimulation with at least one HLA-mismatched EBV-LCL (Supplementary Figure 1A). Next, we investigated whether the T-cell populations that did not recognize any HLA-mismatched EBV-LCL contained smaller sub-population(s) of T cells that could recognize HLAmismatched EBV-LCLs, but were missed in the initial bulk analysis. Sub-populations were sorted based on expression of single TCR-Vβ families. Twenty-four different TCR-Vβ families can be identified with the provided monoclonal antibodies in the kit that was used for flow cytometry, covering around 70% of the human TCR-Vβ repertoire (26). Sub-populations that could not be stained with the antibody kit could not be separated from the bulk populations using this strategy and were not analyzed for recognition of HLA-mismatched EBV-LCLs. In total, 165 sub-populations expressing a single TCR-Vβ family were isolated from the 99 bulk T-cell populations that initially did not show reactivity against the EBV-LCL panel. These sub-populations were subsequently analyzed for their capacity to recognize HLA-mismatched EBV-LCLs. We observed that 31 of these isolated sub-populations contained T cells that were capable of exerting allo-HLA cross-reactivity (Supplementary Figure 1B). Additionally, 193 sub-populations were sorted from bulk T-cell populations that did already demonstrate HLA-mismatched EBV-LCL recognition in the initial analysis (derived from 65 initial bulk populations). Eighty-six of these isolated sub-populations contained T cells that recognized HLA-mismatched EBV-LCLs. Recognition of additional HLA-mismatched EBV-LCLs could be observed that were not detected in the initial analysis of 25 different bulk T-cell populations (Supplementary Figure 2). In summary, a total of 83  sub-populations derived from the 18 bulk T-cell populations that did not recognize any HLA-mismatched EBV-LCLs in the initial analysis). In these analyses, HLA-B * 08:01-restricted virus-specific T cells exhibited a significantly broader crossreactivity pattern, illustrated by reactivity against a median of 6 different HLA-mismatched EBV-LCLs, whereas HLA-A * 01:01, HLA-A * 02:01 and HLA-B * 07:02-restricted virus-specific T-cell populations showed reactivity against a median of only 2, 2 and 3 HLA-mismatched EBV-LCLs, respectively (Figure 1C). Within the different HLA-B * 08:01-restricted T-cell populations, similar high frequencies of T cells capable of exerting cross-reactivity against HLA-mismatched EBV-LCLs were observed, regardless of viral antigen-specificity ( Figure 1D). These results show that the occurrence and frequency of cross-reactivity against HLAmismatched EBV-LCLs is highly affected by HLA-restriction and not by virus-specificity (CMV, EBV or AdV).

Cross-Reactivity Against HLA-Mismatched EBV-LCLs Is Mediated by Recognition of Allogeneic HLA Molecules
To investigate if the recognition of HLA-mismatched EBV-LCLs was indeed caused by recognition of allogeneic HLA molecules, HLA-deficient EBV neg K562 cell-lines transduced with single HLA-class-I molecules were used as stimulator cells (12,27). T-cell populations exhibiting a clear pattern of EBV-LCL recognition, corresponding with the expression of a single HLA allele, were tested against K562 cells transduced with the respective HLA-molecule. For example, a population of EBV-EBNA3A QAK -specific T cells recognized HLA-mismatched EBV-LCL ABF, which uniquely expressed HLA-A * 30:04 and HLA-B * 55:01 (Figure 2A). Recognition of K562 cells that were transduced with HLA-B * 55:01 confirmed part of this respective cross-reactivity pattern ( Figure 2B). In another example, a population of EBV-BRLF1 YVL -specific T cells recognized HLAmismatched EBV-LCLs ACD, WKD, AVZ and UBM that all expressed HLA-A * 24:02 (Figure 2A) and this was confirmed by recognition of K562 cells transduced with HLA-A * 24:02 ( Figure 2B). Some virus-specific T-cell populations (especially HLA-B * 08:01-restricted T cells) showed more complex reactivity patterns when tested against the EBV-LCL panel, that could not be (fully) attributed to recognition of a single allogeneic HLA-molecule. The first representative example shows CMV-pp65 RPH -specific T cells that recognized multiple different EBV-LCLs, not allowing direct complete elucidation of the HLA allele(s) being recognized ( Figure 2C). Only part of the reactivity could be explained by the unique shared expression of HLA-B * 40:01 in EBV-LCLs WKD and RHP, that were both recognized. EBV-LCL UCE was the only EBV-LCL expressing HLA-B * 27:05. However, the HLA alleles underlying the recognition of EBV-LCLs GML, GMS, and MWX could not be deduced. Similarly, EBV-LMP2 CLG -specific T cells recognized 4 EBV-LCLs with unique shared expression of HLA-B * 35:01 or HLA-B * 35:03, while the recognition of EBV-LCL GMK could not be traced back to a specific HLA allele (Figure 2C and  Supplementary Figure 3). Recognition of the HLA molecules that were anticipated to partly underlie the cross-reactivity patterns was confirmed using K562 cells transduced with the respective HLA molecules (Figure 2D). No recognition was observed of K562 cells transduced with irrelevant HLA molecules, whereas recognition of K562 cells transduced with the HLA restriction molecule of the respective virus-specific Tcell population only occurred upon exogenous peptide loading (Figures 2B,D). Allo-HLA cross-reactive virus-specific T cells also showed to be able to lyse HLA-mismatched target cells (Supplementary Figure 4), in line with previous studies (11, 14). These results show that recognition of HLA-mismatched EBV-LCLs can be mediated by recognition of single or multiple allogeneic HLA-molecules.  Table 2). With this panel we covered 63% of the HLA-A, 73% of the HLA-B and 37% of the HLA-C alleles present in our EBV-LCL panel. Testing the HLA-B * 08:01-restricted T-cell populations against this K562 panel revealed recognition of multiple specific groups of allogeneic HLA alleles by single T-cell populations, which could in part explain the cross-reactivity patterns observed when tested against the EBV-LCL panel (Figures 3B,C).

HLA-B
Next, we determined if the cross-reactivity of HLA-B * 08:01restricted T-cell populations was skewed toward HLA-A, B, or C molecules. In total, 22 HLA-B * 08:01-restricted bulk or sub-populations (derived from the 11 HLA-B * 08:01 pos donors that contained complex and extensive cross-reactive virusspecific T-cell populations) were tested against the K562 panel expressing a selection of HLA-A, B, and C alleles.    Table 1). Contrary, HLA-B * 13:02 was never recognized by T cells from HLA-B * 08:01 pos donors. Therefore, we also isolated T-cell populations (n = 10) with the same specificities from 3 HLA-B * 08:01 pos donors, heterozygous for HLA-B * 13:02 (Table 1) (Figure 6B). These results show that the occurrence and broadness of allo-HLA cross-reactivity by virusspecific-specific T cells is influenced by the HLA background of the donors.

DISCUSSION
In this study, we demonstrated that 50% (83/164) of virusspecific T-cell populations contained T cells that cross-reacted against HLA-mismatched EBV-LCLs, in line with previous findings (11). We showed that the level of allo-HLA crossreactivity is highly influenced by HLA restriction and not by the viral specificity of the virus-specific T-cell populations. HLA-B * 08:01-restricted virus-specific T cells showed the highest frequencies and diversities of allo-HLA crossreactivity compared to the HLA-A * 01:01, HLA-A * 02:01 or HLA-B * 07:02-restricted virus-specific T-cell populations. Cross-reactivity against HLA-mismatched EBV-LCLs was shown to be mediated by recognition of allogeneic HLA molecules, which was confirmed by recognition of EBV neg K562 cells transduced with specific HLA-class-I molecules, illustrating that the peptides presented by these allogeneic HLA molecules were not EBV or B-cell-derived. HLA-B * 08:01restricted virus-specific T cells showed a skewed pattern of recognition of a group of allogeneic HLA-B alleles, with HLA-B * 35:01 being recognized most often. We demonstrated that cross-reactivities against multiple allogeneic HLA-class-I molecules by HLA-B * 08:01-restricted EBV-EBNA3 QAK -specific T cells could be mediated by single T-cell clones. Finally, heterozygosity for HLA-B * 35:01, but not HLA-B * 13:02 significantly reduced the degree of HLA cross-reactivity by HLA-B * 08:01-restricted T cells, demonstrating that the HLA background of donors influences the off-target reactivity of virus-specific T cells.
Several groups have investigated whether the allo-HLA crossreactive risk of virus-specific T cells could be predicted. In most of these studies, allo-HLA cross-reactive patterns could only be predicted when a T-cell population used a public TCR (14, 23,24). Public T-cell populations could often be found by analysis of sub-populations of T cells expressing a single TCR-Vβ family. However, no pattern of allo-HLA cross-reactivity could be  observed in our study, except for HLA-A * 02:01-restricted EBV-LMP2 CLG -specific T cells sorted for expression of TCR-Vβ5.1 ( Figure 2C). Although virus-specific T cells often expressed the same TCR-Vβ family, differences in the Complementary Determining Region 3 (CDR3) or a different TCR-alpha chain could result in variation in the allo-HLA cross-reactivity patterns. Allo-HLA cross-reactivity can therefore not be predicted based on TCR-Vβ-family usage alone and may only result in clear patterns if the TCR-Vβ family consist of a public TCR (27).
Similar to other studies we observed that part of the allo-HLA cross-reactive T-cell populations showed recognition of HLAmismatched EBV-LCLs, but no recognition of our panel of single HLA-class-I transduced K562 cells expressing 58% (n = 37/64) of the HLA-class-I molecules present in the EBV-LCL panel (11). The scope our current study was not to fully unravel the recognized allogeneic peptide in allo-HLA molecules. However, this may demonstrate that allo-HLA cross-reactive T cells do not solely recognize an household peptide in the context of allogeneic HLA, but potentially also lineage-specific peptide-allo-HLA cross-reactivity exists (28). Also recognition of HLA-class-II molecules by HLA-class-I-restricted CD8 pos virus-specific T cells has previously been described (11). However, in our study we did not see a correlation with the pattern of recognition against the EBV-LCL panel and the expression of specific HLA-class-II molecules. Therefore, HLA-class-II-restricted cross-reactivity was not further analyzed in depth in our current study.
Finding third-party donors with anti-viral T cells that are fully (HLA-class-I) matched to the recipient patients is probably difficult. When allo-HLA cross-reactive T cells targeting HLA alleles expressed on cells of the patient or (organ) donor are present in the virus-specific T-cell product, acute graft vs. host disease (GVHD) or graft rejection could occur. Strikingly, only a Six T-cell clones per donor were stimulated with a panel of single HLA-B molecule-transduced K562-cell lines (x-axis) for 16 h and IFNγ production (y-axis) was measured by ELISA to analyze which HLA molecules were being recognized. Reactivity was defined as production of >200 pg/ml IFNγ. One representative T-cell clone is shown for each donor. T-cell clones from donor #4 expressed TCR-Vβ14, T-cell clones from donor #8 expressed TCR-Vβ4 and T-cell clones from donor #12 expressed a TCR-Vβ family that could not be determined by the TCR-Vβ flow cytometry kit. Shown are means with standard deviations of 1 experiment carried out in triplicate. TCR, T-cell Receptor; Vβ, Variable Beta Chain; NGFR, Nerve growth factor receptor; Resp, responder. very low incidence of de novo acute GVHD or graft rejection has been observed in clinical trials analyzing the effect of adoptive T-cell therapy with third-party donor-derived products, either in the setting of HLA-mismatched stem cell transplantation or of solid organ transplantation (18,29). It has therefore been assumed that third-party virus-specific T cells do not mediate GVHD or graft rejection (18). It is not clear whether, the observed absence of GVHD or graft rejection in these cases was the result of: (1) no expression of the particular mismatched HLA allele recognized by the transferred virus-specific T cells, (2) removal of the in vitro off-target (>10% cytotoxic) virus-specific T cells from the product prior to administration to the patient and/or selection of T-cell products that do not show in vitro allo-HLA reactivity (18), (3) extensive culturing of the virus-specific T cells prior to adoptive transfer, leading to senescence and impaired cytokine production (30), 4) weak adhesion molecule expression (i.e., ICAM-1) by the target organ (31), (5) biased production and administration of HLA-A * 02:01-restricted virus-specific T cells with an intrinsic low risk of off-target toxicity, as shown in this study, (6) low T-cell numbers of cross-reactive virus-specific T cells administered and/or limited in vivo proliferation, or (7) rapid rejection of the virus-specific T cells (22).
Here, we demonstrated that around 40% of HLA-A * 01:01, HLA-A * 02:01, or HLA-B * 07:02-restricted T-cell populations recognized one or more HLA-mismatched EBV-LCLs. For each T-cell population this recognition was found to be limited to only a few HLA-mismatched EBV-LCLs and could be attributed to recognition of one or a couple of allogeneic HLA alleles. The risk for accidentally mismatching for the particular allogeneic HLA allele(s) cross-recognized by the virus-specific T cells would be low, but studies do report cases of GVHD after infusion of virusspecific T cells derived from the SCT donor (32,33) or derived from a third-party donor (33)(34)(35). Importantly, we found that HLA-B * 08:01-restricted T cells isolated from donors that were homozygous for HLA-B * 08:01 or heterozygous for HLA-B * 08:01 and a specific HLA-B allele (e.g., HLA-B * 13:02) showed abundant allo-HLA cross-reactivity in vitro and are therefore likely to cause graft rejection or GVHD in vivo. Since in the majority of studies so far, the adoptive transfer of third-party donorderived virus-specific T cells was focused on HLA-A * 02:01and/or HLA-B * 07:02-restricted virus-specific T cells (36), the effect of HLA-B * 08:01-restricted virus-specific T cells has not been extensively studied (37). Our results on the higher incidence of HLA-cross-reactivity by HLA-B * 08:01-restricted compared to HLA-A * 01:01, HLA-A * 02:01, or HLA-B * 07:02-restricted virusspecific T cells may have important value for the design of future clinical trials. Since the specificity did not contribute to the allo-HLA cross-reactivity, these results have also important value for third-party derived CAR-T cell therapies or in the field of organ transplantations. Intriguingly, studies in the field of organ transplantations show a significant increase of acute graft rejections in recipients that express HLA-B * 08:01, HLA-C * 07:01, and HLA-DRB1 * 03:01 (38,39). These three HLA molecules are part of a common haplotype (40) and the homozygous donors used in our study have the same haplotype, suggesting that these rejections are mediated by HLA-B * 08:01-restricted T cells. Altogether, these results imply that the HLA background of the donor is important for the broadness of the allo-HLA crossreactivity. Therefore, the most compatible HLA background of the donor should be aimed for and homozygous donors should not be used despite the lower chance of rejection.
Since we only analyzed virus-specific T cells restricted to four different HLA molecules, it remains unclear whether T cells with another HLA restriction could show similar reactivity patterns as HLA-B * 08:01-restricted T cells. However, we hypothesize that these findings might only be restricted to a few HLA molecules since the peptidome of HLA-B * 08:01 shows an unique pattern, that is specific for only HLA-B * 08:01 and HLA-B * 08:02. Based on binding data and sequence information, Sidney J. et al. classified the majority of HLA-B molecules into 9 super families (41). We hypothesized that super families with only a few HLA-B alleles, have unique peptidomes and T cells with this specific HLA background are likely to be cross-reactive against HLA molecules from other HLA super families, since negative thymic selection for these peptide-HLA complexes has not taken place. In the present study, virus-specific T cells isolated from donors that expressed HLA-B * 08:01 and HLA-B * 35:01 proved to be less allo-HLA cross-reactive than those from donors that were homozygous for HLA-B * 08:01 or heterozygous for HLA-B * 08:01 and HLA-B * 13:02. We hypothesize that HLA-B * 35:01 may elicit thymic negative selection for all HLA molecules present in the B07 superfamily to which it belongs (e.g., HLA-B * 07:02, HLA-B * 35:03, HLA-B * 42:01). Being heterozygous for any of the HLA molecules from this B07 superfamily would then presumably result in the same outcome as heterozygosity for HLA-B * 35:01. HLA-B * 13:02 could not be assigned to a particular HLA superfamily (41), possibly explaining why it did little to the level of allo-HLA cross-reactivity of the HLA-B * 08:01restricted repertoire in our study. Therefore, if full matching for HLA-B is not possible, we propose that donors should be used that express HLA-B molecules that are part of different superfamilies to reduce the chance for a broad off-target toxicity in clinical application of third-party donor-derived Tcell products.
Altogether, our results indicate that selection of virus-specific T-cells with specific HLA restrictions and donors with specific HLA backgrounds may decrease the risk of developing GvHD or (organ) graft rejection after infusion of third-party donorderived virus-specific T cells into patients with uncontrolled viral reactivation. Ideally, if complete HLA-class-I matching is not feasible, donor and recipient should at least be fully matched for HLA-B or matched for HLA-B alleles from the same HLA-B superfamily. Mismatching of HLA-B alleles that are unclassified should be avoided, because the peptides presented by these HLA-molecules are unique and could mediate allo-HLA cross-reactivity.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in Supplementary Figure 1, further inquiries can be directed to the corresponding author/s.

AUTHOR CONTRIBUTIONS
WH, DL, LM, and LH, performed experiments. WH analyzed results and made the figures. WH, JF, DA, and IJ designed the research and wrote the paper. All authors contributed to the article and approved the submitted version. This study was in part also supported by research funding from Stichting den Brinker (The Netherlands, Zeist) that made a donation to the Leukemia fund from the Bontius Foundation (Leiden University Medical Center). These sponsors are non-profit organizations that supports science in general. They had no role in gathering, analyzing, or interpreting the data.