Epstein-Barr virus (EBV) infection or reactivation is a frequent complication of hematopoietic stem cell and solid organ transplantation. While EBV-infected B cells are under the control of EBV-specific T cells in immunocompetent individuals, delayed immune reconstitution and immunosuppressive treatment after transplantation can lead to the uncontrolled proliferation of malignantly transformed EBV-infected B cells, resulting in post-transplant lymphoproliferative disease (PTLD). The incidence of PTLD ranges from 1% to 20%, depending on the transplanted organ, the patient’s EBV serostatus, and the intensity of immunosuppression.1 PTLD has a 3-year overall survival rate of only 49% to 80% and significant side effects, emphasizing the need for novel treatment approaches.2 3

Current treatment options, such as reduction of immunosuppression, chemotherapy and B-cell depletion, exhibit varying response rates and are often limited by severe side effects.4–6 Recently, it was shown that adoptively transferred EBV-specific T cells expanded or directly isolated from the blood of EBV-seropositive donors can restore the EBV-specific T-cell immunity and persist for several years.6–9

An immunotherapeutic approach gaining increasing importance in the last years is to genetically redirect T cells by introducing chimeric antigen receptors (CARs) that combine antigen-binding properties with the activating signals of cellular receptors (eg, CD3ζ chain).10 Most often, an antibody-based single-chain variable fragment (scFv) is used as the antigen recognition domain, which mediates binding to extracellular antigens.11 Current limitations to CAR-T cell (CAR-T) applications, such as an immunosuppressive tumor microenvironment and heterogeneous antigen expression, prompted the development of T cells redirected for universal cytokine-mediated killing (TRUCKs), which further endow T cells with the ability to deliver a transgenic effector molecule that accumulates in the target tissue.12 TRUCKs engineered to secrete a proinflammatory cytokine, such as interleukin (IL)-12 or IL-18, on tumor cell engagement exhibit improved effector functions; they act by recruiting and activating innate immune cells (eg, natural killer (NK) cells and macrophages), and by re-programming the immunosuppressive tumor microenvironment. When an inducible expression cassette is used, the cytokine is only released after CAR activation, which reduces systemic toxicity by locally restricting the presence of the cytokine.13 Impressively, TRUCKs were shown to eliminate heterogeneous solid tumor lesions and antigen-loss variants that would otherwise remain invisible to conventional, antigen-dependent CAR-Ts.14

The major challenge in CAR design is to choose an appropriate target antigen only expressed on tumor and not on healthy cells to prevent on-target off-tumor toxicity.15 In order to specifically address malignant EBV-positive PTLD cells and sparing healthy cells, so-called T-cell receptor (TCR)-like monoclonal antibodies (mAbs) can be utilized to direct CARs against an intracellular EBV antigen. TCR-like mAbs specifically bind to an epitope consisting of a defined peptide in complex with an HLA molecule and their usage as therapeutic agents is increasingly being investigated, since tumor or virus-infected cells could be targeted with a high specificity.16–22 Using the binding specificity of TCR-like mAbs for the generation of CARs, researchers could circumvent the need for frequent antibody administration and enable a persisting in vivo response to tumor-associated peptides in the HLA context.19 23–25

In the present study, we generated a CAR based on the scFv of the TCR-like mAb TÜ165. TÜ165 is one of the first TCR-like mAbs with human specificity and exclusively binds to an EBV-infected, HLA-B*35-positive B-cell line but not to cell lines lacking HLA-B*35 or non-EBV-infected HLA-B*35-positive cells.26 It was shown that TÜ165 recognizes LPPHDITPY when bound to HLA-B*35:01, a peptide derived from the Epstein-Barr nuclear antigen (EBNA)−3C, which is expressed in EBV latency type III associated with PTLD.27 The peptide is predicted to bind to the majority of HLA-B*35 subtypes (www.cbs.dtu.dk/services/NetMHCpan; July 7, 202028) and, depending on the geographic region, 11.3% to 37.9% of the population are positive for HLA-B*35 alleles (www.allellefrequencies.net; July 7, 202029). The antibody’s unique TCR-like binding specificity allows the CAR to address an intracellular EBV target antigen, thus sparing healthy B cells in the immunocompromised PTLD patients.

By additionally introducing an inducible IL-12 (iIL-12) expression cassette into the TÜ165 CAR construct, we developed iIL-12-secreting TÜ165 TRUCKs with improved effector functions that recruit and activate bystander immune cells, which otherwise occur rarely in close proximity to lymphoproliferative cells in PTLD patients. Our results demonstrate that TÜ165 CAR-Ts and TÜ165 CAR-based TRUCKs that are equipped with inducible IL-12 expression are able to specifically and effectively respond to an intracellular EBV target and may thereby enable control of EBV-associated lymphoproliferation in immunocompromised patients.

Currently, the two dominant strategies for engineering T cells are to introduce either a recombinant TCR or a CAR into T cells. TCR-like CAR-Ts combine the advantages of TCR- and CAR-engineered T cells. Major concerns when generating TCR-engineered T cells (TCR-Ts) are the development of unexpected specificities due to mispairing of the introduced TCR with the endogenous TCR and low expression of the introduced TCR; however, some strategies were recently developed to circumvent these limitations.37 The use of CAR-Ts, on the other hand, prevents dimer formation with the endogenous TCR. Some studies already showed impressive anti-tumor responses, leading to the recent approval of two CD19-targeting CAR-T therapies for the treatment of different B-cell malignancies.38 Since the depletion of all B cells includes EBV-infected lymphoma cells, CD19 CAR-T transfer for the treatment of PTLD is conceivable. However, as is the case for current PTLD treatment options like chemotherapy and antibody-mediated B-cell depletion, this strategy would be limited by the patient’s poor condition and the high level of immunosuppression needed to prevent organ rejection.4–6 Thus, an approach targeting only EBV-infected B cells and sparing healthy B cells to treat PTLD without diminishing the already low numbers of immune cells in these patients would be very desirable. Adoptively transferred EBV-specific T cells address this problem, but their use is often limited by time constraints or by the lack of an appropriate donor.6 Building on this strategy in the present study, we utilized the TCR-like mAb TÜ165, which binds to a PTLD-associated, EBNA-3C-derived peptide presented in the HLA-B*35 context. TCR-like antibodies have shown promise as therapeutic agents designed for the targeting of tumor and virus-infected cells.16–22 Otherwise, for the HLA-B*35/LPPHDITPY epitope no naturally occurring TCRs could be found (online supplemental figure S5). The absence of T cells with natural specificity against the epitope can be explained by the fact that the epitope is not immunodominant, respective T-cell frequencies are too low to induce a detectable response or that specific T cells have been deleted via tolerance mechanisms due to self-reactivity.

In the present study, TÜ165 bound to EBV-positive B-LCLs from HLA-B*35-positive donors expressing similar EBV proteins as PTLD cells.6 B-LCLs were generated by using the laboratory EBV strain B95.8, but also naturally occurring strains C666-1 (GenBank no. AB828190.1), HKNPC1 (GenBank no. JQ009376.2) and M81 (GenBank no. KF373730.1) encode the LPPHDITPY peptide indicating the recognized sequence is conserved in different EBV strains. Only the laboratory strain AG867 (GenBank no. DQ279927.1) has an amino acid exchange at the first position of the peptide (leucine → phenylalanine; L → F; FPPHDITPY), to which the TÜ165 antibody and CAR binding would need to be assessed in further studies. Besides B-LCLs, the mAb TÜ165 bound to EBNA-3C peptide-loaded PBMCs and equally loaded HLA-B*35-transduced K562s (K562-B*35/pEBV). This confirms the reported capability of TÜ165 to bind to EBNA-3C peptide/HLA-B*35:01 epitopes.26 27 The peptide was predicted to be presented by a majority of HLA-B*35 subtypes (www.cbs.dtu.dk/services/NetMHCpan; 07.07.202028) and HLA-B*35 is a prevalent allele (www.allellefrequencies.net; 07.07.202029). HLA-B*35:03-positive B-LCLs could be stained with the mAb TÜ165 with the same or slightly lower frequency confirming the predicted peptide binding and indicating TÜ165 is also able to recognize the EBNA-3C peptide in context of other HLA-B*35 subtypes than the reported HLA-B*35:01. It would be interesting to assess the binding capability of the TÜ165 antibody and CAR to the EBV peptide in complex with different HLA-B*35 subtypes in further studies. TÜ165 binding was moreover peptide concentration-dependent. EBNA-3C peptide-loaded K562-B*35 were recognized at low loading concentrations, indicating a high binding specificity, whereas other EBV/CMV peptides were not. Importantly, unloaded K562-B*35 and PBMCs can be expected to present a variety of peptides derived from self-proteins on their HLA-B*35 molecules,39 but TÜ165 did not bind to them. Consequently, we conclude that TÜ165 binding is restricted to the EBNA-3C-derived peptide when presented by HLA-B*35 and that binding to other peptides will occur only at unphysiologically high peptide loading concentrations. Overall, the observed differential recognition of B-LCLs and PBMCs as well as the recognition of EBNA-3C peptide-loaded versus unloaded K562-B*35 confirms the suitability of TÜ165 to specifically distinguish between EBV-associated target and healthy cells.

TÜ165-based CAR-Ts and TRUCKs showed a favorable phenotype with low PD-1 expression, indicating that the cells were not exhausted despite being expanded around 100-fold in the first 9 days using an expansion protocol similar to the one used for clinical manufacturing. Only following activation with CD3/CD28 beads, expression of PD-1, TIM-3 and LAG-3 was transiently increased on T cells (data not shown). However, since these markers are also activation markers, only long-term expression would demonstrate an exhausted state.40 Here, after cultivation for 16 days, expression of all markers markedly decreased again (data not shown) and cells were shown to be functional. Therefore, we deduced that T-cell dysfunction might not be a major problem in TÜ165 CAR-T and TRUCK generation. Previous studies have shown that the phenotype of adoptively transferred T cells predicts their long-term persistence and engraftment in vivo.41 42 In this context, less differentiated T_CM cells, which formed the main population after expansion in the present study, showed superior functionality in vivo: they exhibited long-term persistence and occupied memory T-cell niches.

When TCR-like CAR-Ts were first introduced as a new strategy, various problems had to be addressed. For example, one study using CARs based on high-affinity TCR-like antibodies recognizing Wilms tumor suppressor gene 1 (WT1) peptide/HLA-A02 complexes showed a loss of specificity and elimination of HLA-A*02-positive, peptide-negative cells, whereas the low-affinity TCR-like CAR-Ts maintained potent cytotoxicity and specificity.43 Maus et al 44 later confirmed that lowering the affinity of a TCR-like CAR to the TCR affinity range ensures epitope specificity and effective cytotoxicity of the CAR. This is most likely due to the ability of the CAR-Ts to sequentially engage peptide/HLA complexes, as is reportedly necessary for proper T-cell activation. In the present study, the TÜ165 CAR was shown to be activated highly specifically only on recognition of K562-B*35/pEBV in a concentration-dependent manner, whereas unloaded K562, K562-B*35 or K562-B*35 loaded with other peptides did not activate the CAR. Only when transduced into primary T cells, a background activation of TÜ165 CAR-Ts and TRUCKs following stimulation with unloaded K562 and K562-B*35 was detected. A major drawback of this model is allogenicity. Although K562 cells only express low levels of HLA (mainly HLA-C),45 untransduced T cells mildly recognize the cell line, which is shown by a very slight upregulation of CD137 and CD25 and release of TNF-α and IFN-γ in co-cultures with unloaded K562-B*35 (online supplemental figure S3). TÜ165 CAR-Ts exhibit a response towards the unloaded control cells to a similar extent showing that this response most likely is mediated by the TCR and not unspecifically by the introduced CAR towards empty HLA-B*35. TÜ165 TRUCKs, however, showed an even enhanced response to unloaded K562-B*35, which might be due to the fact that allogeneic recognition of HLA on the surface of K562 cells by the TCR induces NFAT expression and a subsequent induction of the IL-12 cassette. This hypothesis could be confirmed by testing the TÜ165 EGFP-TRUCK in JE6-1 Jurkat cells as effector cells. By the use of these cells lacking endogenous TCRs, no background response towards either K562 or K562-B*35 cells was observed. The expression of EGFP and ECFP is only induced upon loading with the EBNA-3C-derived peptide. These findings showed that the recognition of the epitope by the scFv of the CAR is highly specific.

When using primary T cells as effector cells, substracting the activation and cytokine data obtained for co-cultures with unloaded K562-B*35 from the ones in co-cultures with K562-B*35/pEBV could indicate the LPPHDITPY-specific component of the TÜ165 CAR-T-cell recognition. Here, TÜ165 CAR-Ts showed functionality by upregulating the expression of activation markers (CD137 and CD25) and by slightly increasing intracellular IFN-γ production in specific response to K562-B*35/pEBV but not to unloaded K562-B*35 controls. Moreover, they exhibited cytotoxicity by upregulating perforin and granzyme B when co-cultured with EBV peptide-loaded target cells and formed large clusters around them. When cultured with autologous B-LCLs in a model of lymphoproliferation, they increased the expression of activation markers (CD69 and CD25) and released IFN-γ in a range similar to that of untransduced T cells. Of note, expansion and off-target cytotoxicity of TÜ165 CAR-Ts towards unloaded K562 cells and uninfected PBMCs was equal to that of untransduced T cells. This indicates that they did not recognize HLA-B*35 without EBNA-3C peptide or when complexed with other naturally occurring peptides. The maintenance of specificity in TÜ165 CAR-Ts can be explained by the fact that the CAR is based on an immunoglobulin M (IgM) antibody, which normally has 10 antigen-binding sites. When reduced to a single antigen-binding site in the TÜ165 CAR, a lower avidity and weaker binding might be the consequence. This, in turn, allows the aforementioned sequential peptide/HLA engagement and yields functional CAR-Ts. In general, effector functions of the TCR-like TÜ165 CAR were rather mild when compared with other CARs, such as the well-described CD19 CAR. However, CAR activation is known to be highly dependent on the expression levels of target antigen46 explaining that the target of the TÜ165 CAR, a peptide presented in HLA context, induces less CAR activation as compared with other highly-expressed surface molecules such as CD19. Considering their weak target binding and mild functionality, it would need to be evaluated, if TÜ165 CAR-Ts are able to effectively control EBV-associated lymphoproliferation for a potential future clinical use. For this, detailed in vivo testing is now required. Regarding possible models to study TÜ165 CAR-T function, immunocompromised mice previously engrafted with human B-LCLs are a model of EBV-associated PTLD that was recently used to test an EBV-specific, TCR-like mAb for the treatment of PTLD.20 Further, generation and testing of other TCR-like CARs and TRUCKs recognizing EBV epitopes in context of PTLD would be desirable to extend the therapy to HLA-B*35-negative patients. As source for the scFv, TCR-like mAbs were described that recognize EBV-derived peptides in context of HLA-A*02:01,20 47 and more can be generated using conventional hybridoma and phage display technology.16 48

Furthermore, we generated TÜ165 TRUCKs which induced IL-12 secretion only on recognition of the TÜ165 target. We encoded constitutive CAR expression and inducible cytokine expression in a novel ‘all-in-one’ vector to further facilitate the manufacturing of these cells for clinical applications and to reduce the risk of insertional mutagenesis when introduced into other cells as previously described.31 To evaluate functionality of the inducible cassette, TRUCKs with inducible EGFP expression were used. When transduced into JE6-1 reporter cells, EGFP expression was only induced after specific target recognition. It was not induced in 100% of cells; however we only measured it after 48 hours and kinetics of CAR activation and signaling are not fully understood yet.49 There is evidence that EGFP expression might not be completely upregulated in some cells at that time point or even already be downregulated again. Furthermore, Zimmermann et al showed that T cells transduced with a GD₂-specific CAR and the same inducible EGFP cassette did upregulate EGFP expression on 60% of the cells after specific target recognition.31 The slightly reduced EGFP induction of 45% in TÜ165 EGFP-TRUCKs can, most likely, again be explained by the aforementioned more rare expression of the target antigen. IL-12 has been reported to induce T helper 1 differentiation of CD4⁺ T cells and to enhance cellular immunity by increasing IFN-γ release, augmenting granzyme and perforin production, and improving T and NK-cell proliferation.50 In the present study, we used an inducible delivery route of IL-12 by TÜ165 TRUCKs on target recognition. This resulted in highly improved CAR function in specific response to EBV peptide as well as to malignant EBV-infected cells. Also the abovementioned allogeneicity towards unloaded K562-B*35 was enhanced, because, most likely, signaling by the endogenous TCR lead to an induction of IL-12. For a potential future clinical application, this would imply that TRUCKs should not be utilized in the allogeneic setting as they would probably cause an even more severe graft-versus-host disease. However, this should not be a problem in the autologous use and, moreover, attempts to generate universal CAR-Ts with disrupted TCR expression are currently under investigation.51 Compared with CAR-Ts, TÜ165 TRUCKs showed augmented upregulation of CD25 expression, TNF-α and IFN-γ release, and IL-2 secretion. This is in line with previous findings that other IL-12-engineered T cells show substantially increased IFN-γ and TNF-α release, which is mandatory for the accumulation of activated macrophages in the tumor lesion and enables the elimination of antigen-loss cancer cells.14 Both IFN-γ and TNF-α promote resistance to tumor-derived immunosuppressive factors often released in EBV-positive Hodgkin’s disease.52 TRUCKs with other specificities have not shown enhanced cytotoxicity14 or superior CD107a or granzyme B expression53 compared with the respective CAR-engineered or TCR-engineered T cells. Here, interestingly, TÜ165 TRUCKs released slightly more perforin and significantly more granzyme B on recognition of the EBV-associated target than TÜ165 CAR-Ts. Moreover, TÜ165 TRUCKs tended to be more cytotoxic and exhibited cytotoxicity and granzyme B release towards autologous B-LCLs. Thus, besides their usage in the treatment of PTLD, we here provide a proof of principle how to improve the functionality of weakly binding CAR-Ts with mild effector function. As described above, TCR-like CAR-Ts were shown to maintain epitope specificity and effective cytotoxicity by lowering the affinity of a TCR-like CAR to the TCR affinity range.44 Furthermore, for clinical use, one major concern is tonic CAR-T signaling leading to exhaustion.54 Hence, using low affinity CARs in combination with an inducible cytokine might provide a safer and more enduring T cell product with the same or even increased functionality, which would need to be examined in further CAR studies.

The rationale for developing TRUCKs with inducible IL-12 was not only to improve functionality but also, and even more importantly, to modulate the tumor microenvironment. In a study with TCR-Ts against EBV-associated Hodgkin’s disease, the released IL-12 mediated resistance to the anti-proliferative and -cytotoxic effects of TGF-β that normally limit T-cell functionality against these tumors.52 Moreover, IL-12 results in the recruitment of NK cells and macrophages to the tumor lesion, thereby promoting antitumor effects.55 In the present study, TÜ165 TRUCKs releasing IL-12 on target engagement recruited both monocyte and NK-cell lines and enhanced target-cell killing as well as IFN-γ expression of NK-92 cells. PTLD patients display a heterogeneous immune deficiency associated with their immunosuppressive treatment regimen.56 However, the importance of NK cells in EBV-associated malignancies was demonstrated since primary immunodeficiencies in NK cells predispose patients to EBV-driven pathologies57 and a distinct subpopulation of tonsillar NK cells could restrict B-cell transformation by EBV.58 Thus, the use of IL-12-secreting TÜ165 TRUCKs to attract and activate NK cells, which are not normally located in close proximity to PTLD cells, is a new approach to treating immunocompromised patients that not only spares healthy B cells, but also attracts as many other innate immune cells as possible to induce a broad anti-EBV response. An immunocompetent mouse model is required to study attraction and involvement of endogenous immune cells by the engineered T cells in vivo. IL-12-induced toxicities are of particular interest and should be evaluated in future studies. Most of the available studies of inducible IL-12-secreting TRUCKs report that safe adoptive transfer of those cells into mice is possible, but one group observed severe toxicities associated with increased systemic levels of TNF-α and IFN-γ due to the non-specific release of IL-12.59 Although the amounts of IL-12 secreted by unstimulated TRUCKs in vitro in the present study were lower than those in the aforementioned study, evaluation of possible toxicities should definitely be a major focus of prospective preclinical in vivo testing of TÜ165 TRUCKs.

In conclusion, the results of this study demonstrated the suitability of CAR-Ts based on the peptide-selective and TCR-like mAb TÜ165 to specifically recognize and react to an intracellular EBV target associated with PTLD. Equipping the TÜ165 CAR with inducible IL-12 expression to generate TÜ165 TRUCKs improved effector functions and, importantly, led to the recruitment and activation of surrounding immune cells, which normally are not found in close proximity to lymphoproliferative cells. Thus, we suggest this method of specifically targeting EBV-infected cells while sparing and mobilizing the healthy immune cells in already immunocompromised patients as a new and promising approach to PTLD treatment, thereby enabling the control of EBV-associated lymphoproliferation.