Tumor resistance mechanisms and their consequences on γδ T cell activation

Human γδ T lymphocytes are predominated by two major subsets, defined by the variable domain of the δ chain. Both, Vδ1 and Vδ2 T cells infiltrate in tumors and have been implicated in cancer immunosurveillance. Since the localization and distribution of tumor‐infiltrating γδ T cell subsets and their impact on survival of cancer patients are not completely defined, this review summarizes the current knowledge about this issue. Different intrinsic tumor resistance mechanisms and immunosuppressive molecules of immune cells in the tumor microenvironment have been reported to negatively influence functional properties of γδ T cell subsets. Here, we focus on selected tumor resistance mechanisms including overexpression of cyclooxygenase (COX)‐2 and indolamine‐2,3‐dioxygenase (IDO)‐1/2, regulation by tumor necrosis factor‐related apoptosis‐inducing ligand (TRAIL)/TRAIL‐R4 pathway and the release of galectins. These inhibitory mechanisms play important roles in the cross‐talk of γδ T cell subsets and tumor cells, thereby influencing cytotoxicity or proliferation of γδ T cells and limiting a successful γδ T cell‐based immunotherapy. Possible future directions of a combined therapy of adoptively transferred γδ T cells together with γδ‐targeting bispecific T cell engagers and COX‐2 or IDO‐1/2 inhibitors or targeting sialoglycan‐Siglec pathways will be discussed and considered as attractive therapeutic options to overcome the immunosuppressive tumor microenvironment.


| INTRODUC TI ON
The immune system has successfully developed various mechanisms to recognize and respond to foreign antigens including tumor antigens. The different mechanisms include recognition of microbe Although significant improvement has been obtained such novel biological therapies, many patients do not experience clinical benefit due to other tumor resistance mechanisms. [5][6][7][8] These mechanisms negatively influence T cell effector function including the anti-tumor response of γδ T cells.
γδ T cells are a heterogenous subpopulation of lymphocytes that combine innate and adaptive properties. A prognostic significance of γδ T cells in a broad range of human tumor entities and a correlation with patient outcome have been revealed by recent bioinformatic analyses of a meta-genomic datasets. 9 A further improved deconvolution of human cancer microarrays revealed a considerable inter-individual variation of tumor-infiltrating Vγ9Vδ2 γδ T cell abundance across the 50 analyzed types of hematological and solid malignancies. 10 Although the correlative data of γδ T cell infiltration in tumors are encouraging, an immunosuppressive TME can hinder γδ T cells from exerting an effective anti-tumor response and can actually promote γδ T cell differentiation into an immunosuppressive phenotype based on their high plasticity. [11][12][13] Otherwise, the high functional plasticity of γδ T cells offers interesting perspectives for their application as effector cell population for T cell-based immunotherapy. [14][15][16] 2 | γδ T LYMPHO C Y TE S

| Ligand recognition of γδ T cells
Apart from the difference in their mode of antigen recognition, γδ T cells differ from αβ T cells in T cell receptor (TCR) diversity, the structure of TCR-CD3 complex, and their frequency in tissue and blood distribution. [17][18][19][20] γδ T cells are suggested to play an important role in immunity to infection, stress surveillance, and carcinogenesis. [21][22][23][24] Human γδ T cells can be classified in at least three major subsets. [25][26][27][28] Human Vδ1 T cells constitute a main population in the mucosa and skin. Several Vδ1 T cell clones or lines have been shown to recognize microbial and self-lipids bound to non-classical CD1c and CD1d molecules with high affinity, but also CD1d molecules without presented antigens with low affinity. 29,30 Similar to unloaded CD1d molecules, several cancer patients-derived Vδ1 T cell lines directly bind via NKG2D to stress-induced MHC class I-related chain (MIC) A and UL16-binding protein (ULBP)-3 expressed on tumor cells. 31,32 The current knowledge about recognition of MHC-like molecules as well as other recognition concepts for Vδ1 and Vδ1-negative T cells is extensively summarized in recent reviews. 20,33 Vδ1 T cells express a TCR composed of any of the six different variable (V) genes (Vγ2, 3, 4, 5, 8, or 9) paired with Vδ1. 34 Human Vδ2 T cells which co-express Vγ9 are mainly found in the peripheral blood. Vγ9Vδ2 T cells recognize pyrophosphate intermediates of a dysregulated mevalonate pathway in tumor cells and of the prokaryotic non-mevalonate pathway by a butyrophilin (BTN) molecule-dependent mechanism. [35][36][37][38][39][40] A blockade of the isoprenoid pathway with aminobisphosphonates (n-BP) including zoledronic acid, which is used for the treatment of bone-related diseases, inhibits the pathway and induces an Vγ9Vδ2 proliferation via accumulation of pyrophosphate intermediates. 41

| γδ T cells in cancer
Peripheral γδ T cell lymphomas (γδ TCLs) are very rare with <1% of lymphoid neoplasms but aggressive and accompanied with a dismal prognosis. While hepatosplenic γδ TCLs are usually associated with Vδ1 T cells, cutaneous γδ TCLs are associated with Vδ2 T cells. 51 An abundance of Vγ9Vδ2 γδ T cells has been shown in B cell-acute lymphoblastic leukemia, chronic myeloid leukemia, and promyelocytic leukemia by transcriptome analysis. 10 Circulating γδ T cells were increased in chronic lymphocytic leukemia (CLL), and dysfunctional Vγ9Vδ2 γδ T cells were reported to be a negative prognostic factor in CLL. 52 In addition, analysis of paraffin-embedded sections from patients with Epstein-Barr virus (EBV)-associated Hodgkin's lymphoma demonstrated that most of the γδ T cells from these patients expressed Vγ9, whereas γδ T cells expressing Vγ2, 3, or 4 were nearly absent. 53 γδ T cell infiltration and correlation with patient outcome have been also demonstrated in epithelial and solid tumor cells including melanoma, colorectal, breast, pancreatic, and ovarian cancer. [54][55][56][57][58] Immunohistochemistry (IHC)-based analysis allows analysis of tumor-infiltrating γδ T lymphocytes (γδ TIL) within the tumor as well as in the context of the TME or the surrounding tissue. γδ T cells moderately infiltrate in superficial intraepithelial and superficial spreading melanoma, whereas a higher amount of γδ TIL has been demonstrated in ulcerated nodular melanoma. γδ TILs are localized in the periphery of the tumor in all analyzed melanomas. 55 A localization of γδ TILs rather in the tumor stroma than in the intratumoral tissue has also been shown in patients with colorectal cancer, 56 whereas γδ TIL in patients with hepatocellular carcinoma localized in the peritumoral liver tissue. 59 Regarding pancreatic ductal adenocarcinoma (PDAC), our studies initially demonstrated that γδ T cells were mainly localized in the stroma adjacent to the tumor cells or within malignant ductal epithelium in 44% of a cohort of 41 patients. 57 Our further studies revealed a significant correlation of TCRγδ expression with CD3 and CD8. Different stromal localization of γδ T cells in PDAC patients compared with chronic pancreatitis patients may be explained by changes of stromal composition during progression to invasive PDAC characterized by a pronounced immunosuppressive TME. 60,61 Although the frequency of γδ T cells in PDAC tissue is low, γδ T cells accumulate in malignant ductal epithelium in close interaction with regulatory T cells (Treg) and immunosuppressive fibroblasts. 60 By analyzing the distribution of the PDAC γδ TIL subsets, Intratumoral γδ T cells in breast cancer patients are reported to serve as a prognostic marker for an advanced tumor stage and poor clinical outcome, 64 probably due to the enrichment of immunosuppressive CD73-expressing Vδ1 T cells or immunoregulatory γδ T cells. 65,66 In contrast, several other studies with triple negative breast cancer patients revealed that the infiltration of γδ T cells is accompanied with good prognosis 67 which is discussed in more detail in the recent publication of Chabab and colleagues. 54 In addition, adjuvant chemotherapy and physical activity of breast cancer patients did not significantly alter absolute cell number of peripheral blood γδ T cells. 68

| Distribution of tumor-infiltrating γδ T cell subsets
The immunohistological approaches are often combined with flow cytometry analysis to determine the different γδ T cells by anti- esophageal tumors, 72 non-small-cell lung cancer, 73 PDAC, 63 and ovarian cancer. 58,74 The inversion of the Vδ1/Vδ2 ratio is demonstrated by a high percentage of Vδ1 γδ T cells and a reduced percentage of Vδ2 γδ T cells within TIL compared to peripheral blood lymphocytes (PBL) of the same donors. An enriched percentage of Vδ1 and Vδ1/Vδ2-negative γδ TIL co-expressing Vγ2, 3, or 4 is also reported for ovarian cancer patients. 58 Similar to the results in ovarian cancer measured by flow cytometry, a co-expression of Vγ2, 3, or 4 on Vδ1 TIL is also detected by IHC in frozen sections of colon adenocarcinoma patients. 53 In contrast, an enhanced number of Vδ2 was found in glioblastoma multiforme, 75 whereas no significant difference in the percentage of Vδ1 and Vδ2 T cells was observed in prostate cancer tissue in comparison with PBL. 76 In the majority of the above described tumor entities, γδ TIL are described to be in an activated stage accompanied by an increased number of Vδ1 TIL with a T CM or T EM -phenotype and Vδ2 TIL with an T EM or TEMRA-phenotype. 12,77 Although a clear correlation with γδ T cell infiltration and patients' prognosis is missing, 77 there is a tendency that a reduction of Vδ2 γδ T cells correlates with an advanced stage of the disease. An enhanced Vδ2 T cell percentage has been shown to correlate with an early stage of development of melanoma and the absence of metastasis. 55 The abundance of γδ T cells (mainly Vδ2 T cells) with a T helper (TH) 1 phenotype and cytotoxic capacity revealed a significantly longer 5-year disease free survival rate of colorectal cancer patients determined by transcriptome analysis. 56 In contrast to an initial study demonstrating a positive correlation of interleukin (IL)-17 producing γδ TIL with tumor stages and other clinicopathologic features in colorectal cancer patients, 78 Meraviglia and colleagues reported that interferon (IFN)-γ production of Vδ1 and Vδ2 TIL was reduced in colorectal tissue compared to adjacent tissue; however, IL-17 secretion was nearly absent. 56 In this context, it is very interesting that freshly isolated Vδ1 and Vδ2 derived from PBL and TIL of ovarian cancer patients highly express intracellular granzyme A and B without additional phorbolester and ionomycin (PMA/Iono) stimulation suggesting a pre-activated state of both γδ TIL subsets.
In striking contrast, intracellular IFN-γ production was nearly absent but was inducible donor-dependently after PMA/Iono stimulation. 58 Conversely, intracellular IL-17 was detected after PMA/Iono in ovarian γδ TIL in the publication of Chen et al, 74 but not in our own studies. 58 This may suggest a donor-dependent influence by the TME or an inclusion of IL-17-producing non-γδ TIL in the flow cytometric gating-strategy. Interestingly, CD4-expressing γδ T cells were identified in ovarian tumor to be beneficial for the patients' survival. 79 Although the number of Vδ2 TILs is lower than the number of Vδ1 TILs in some tumor entities, Vδ2 TILs are a very interesting effector cell population for T cell-based immunotherapy as extensively summarized in recent reviews. [80][81][82] The interest is based on the capacity of Vδ2 T cells to present soluble proteins and cell debris to CD8 αβ T cells by a proteasome-dependent cross-presentation pathway, [83][84][85] to recognize phosphorylated antigen (PAg) released by tumor cells in a HLA-unrestricted manner and their reduced potential to cause graft versus host disease. 36,86,87 The frequency, phenotype and functions of γδ TILs are presumably influenced by genetic differences between tumor cells, polymorphisms of regulatory genes, exposure to certain pathogens (eg intestinal microbiome) and TME. 11,12 The TME comprises a cellular compartment composed of a variety of immunosuppressive cells including mesenchymal cells (eg, fibroblasts), myeloid-derived suppressor cells (MDSC), tumor-associated macrophages (TAM), and Treg, which release immunosuppressive molecules or express immunosuppressive molecules and thereby prevent the penetration or activation of cytotoxic γδ T cells. This review presents a crucial update of potential molecular mechanisms partially described in recent publications. 11,12,60,81,88 More importantly, it focuses on some novel aspects of selected tumor resistance mechanisms influencing the effector function of γδ TILs. rectal adenocarcinoma as well as pulmonary, mammary, prostate, kidney, ovary, and liver tumors. [92][93][94][95][96] However, this is not a general tumor resistance mechanism of these tumor entities, since COX overexpression is also variable in any tumor entity due to the heterogeneity of tumor cells in different patients. In addition, COX-2 is also overexpressed in cells of the TME including cancer-associated fibroblasts in nasopharyngeal and colorectal cancer as well as mesenchymal stem cells (MSC) in colon-, breast-, and ovarian cancer, glioma, and melanoma. [97][98][99][100] An enhanced COX expression of tumor cells results in an increased release of its metabolite PGE2 which recruits Treg to the tumor site and suppresses the effector function of cells of the innate and adaptive immunity including γδ T cells. 91  An enhanced COX-2 expression has also been shown to alter the cytokine profile of T cells by switching from TH1 to TH2-type cytokines. This could be converted by inhibition of COX-2 activity after in vivo application of selective COX-2 inhibitors or silencing with anti-sense oligonucleotides in an experimental colitis mouse model. 110

| Tumor necrosis factor-related apoptosisinducing ligand (TRAIL)
Ligation of tumor necrosis factor-related apoptosis-inducing ligand (TRAIL or Apo2L) with cognate receptors such as TRAIL-R1

| Liaison between TRAIL-R4 and COX
TRAIL, which can be produced in small amounts by tumor cells, tumor-surrounding cells or Vδ2 T cells, has been shown to upregulate COX-1 by caspase and/or nuclear factor of κB (NFκB) activation. [137][138][139] Cytokines, growth factors, and tumor promoters have been shown to binding to TRAIL-R1-or-2 but sufficient to activate ERK and increase COX-2 expression. 136 An imbalance of PGE2 and granzymes in the coculture depicted by an enhanced PGE2 production and an impaired granzyme secretion results in an impaired γδ T cell cytotoxicity against PDAC cells. Selective COX-1 inhibitor valeryl salicylate as well as selective COX-2 inhibitor DuP697 reduced PGE2 release by tumor cells but did not enhance granzyme production by γδ T cells. To overcome the impaired granzyme release and cytotoxic activity of γδ T cells, a TCR-stimulus is necessary which can be delivered by the use of γδtargeting bispecific T cell engager (bsTCE). 136 The use of γδ-targeting bsTCE as targeted biological of γδ T cell-based immunotherapy is discussed in the following section about bispecific antibodies.

| Cross-talk between IDO-expressing cells and γδ T cells
An accumulation of IDO-1/-2 enzymes in PDAC cells with metastatic origin and an upregulation in primary PDAC cells after co-culture with IFN-γ-producing γδ T cells has been recently shown to inhibit degranulation and cytotoxicity of γδ T cells. 160 Additionally, the IDO-down-stream molecule kynurenine, which is released by tumor and stromal cells in the TME suppresses anti-tumor response of γδ T cells 160 [ Figure 1(G)].

tively. Activated γδ T cells can induce a MSC-mediated immunosup-
pression by their release of IFN-γ which in turn exerted a negative feedback mechanism on proliferation and cytokine production of γδ T cells 161 [ Figure 1(B)]. Similar, an interaction of γδ T cells with human mesenchymal stroma cells propagate an immunosuppressive milieu in the TME. 162,163 A liaison between COX-2 and IDO has been reported in a mammary as well as in lung carcinoma mouse models by an anti-tumor vaccination strategy which prevents IDO activation after application of the selective COX-2 inhibitor Celecoxib. [164][165][166]

| Glycan-lectin interactions-An important role of galectins
Several technology advances including genomic, proteomics, glycomics, and glycoproteomics have developed to optimize personalized medicine. 167,168 Aberrations in protein glycosylation have been implicated as a hallmark of cellular oncogenesis, tumor progression, and poor patients' prognosis, although the complexity of cancerassociated glycosylation mechanisms is not completely understood.

Several efforts have been made to influence glycan-lectin inter-
actions by pharmacological anti-tumor agents that target lectins such as sialic-acid-binding immunoglobulin-like lectins (siglecs) and C-type lectin receptors such as selectins or galectins. 169 Galectins are a family of soluble proteins that bind β-galactoside-containing glycans. 170 Here, a focus is set on galectins including galectin-1, galectin-3, and galectin-9, which so far are described to play a role in the interaction of tumor cells and γδ T cells.
Mammalian galectins are defined by a conserved carbohydrate recognition domain (CRD) and common structural fold. Galectin-1 belongs to the prototype galectin subgroup, galectin-3 is the only chimera-type member of the galectin family and galectin-9 belongs to the tandem-repeat-type galectins. 170 As multifunctional proteins, all three galectins are overexpressed by tumor cells and secreted in the serum of cancer patients. However, an enhanced galectin expression correlates with poor patients' survival in only in a few tumor entities, but not in others. 171 Galectin-1, 3 and 9 are expressed also by immune cells and by tumors. All three galectins have multiple binding partners and can contribute to immune evasion by regulating tumorigenesis and metastasis. 170,172 For instance, galectin-1 is described to serve as negative immune regulator of immune responses by inducing tumor-mediated TH1cytokine suppression, IL-10 release and apoptosis in TH1 and TH17 cells but not in TH2 cells due to differential sialylation of cell surface glycoproteins. [173][174][175][176][177] Targeting galectin-1 might promote anti-tumor response by reduction of T-cell apoptosis, abrogation of tumor and Treg-associated immunosuppression and improvement of T cell infiltration. [178][179][180][181] Galectin-3 mediates both pro-and anti-apoptotic activity depending on the subcellular localization. 182 [191][192][193][194] Regarding the potential therapeutic aspect of glycan-lectin pathways, new approaches have been developed to target sialoglycan-Siglec pathway for improvement of chimeric antigen receptor (CAR) T cell therapy as well as immunotherapeutic nanovaccines of plant-derived glycosylated lectins to induce CLR-dependent immune responses 195,196  γδ T cell-based immunotherapies is summarized extensively in different reviews. 80,82,199,200 Briefly, although the application of n-BP or PAg ± γδ T cells is well tolerated in these clinical trials, limited promising results have been delivered. Strikingly, a complete remission has been shown in 4 patients with hematological malignancies after adoptive transfer of haploidentical γδ T cells, and in one patient with renal-carcinoma-derived lung metastases after adoptive transfer of autologous γδ T cells. 201,202 In contrast to other clinical trials, adoptively transferred γδ T cells were additional stimulated by additional intravenous administration of n-BP together with subcutaneous application of IL-2 in both trials, which enabled a necessary restimulation of γδ T cells initially expanded for 11 to 14 days under GMP-conditions in vitro before they were adoptively transferred.

| Galectin-mediated interaction of tumor cells and γδ T cells
A restimulation of adoptively transferred γδ T cells with n-BP and IL-2 was initially demonstrated in a preclinical study. 203

| Bispecific antibodies
Bispecific antibodies (bsAb) are a family of biological molecules designed in their simplest format with two different antigen-binding sites which are lack a constant immunoglobulin domain. 205,206 Improvement of the design of bsAb has been generated by different means in various formats. 206 Of interest is also a new bispecific nanobody in a different format, which targets Vγ9Vδ2 T cells and has specificity for EGFR expressed on tumor cells. This Vγ9Vδ2 T cell-specific nanobody induced lysis of patient-derived colorectal carcinoma cells in vitro as well as in vivo in a xenograft mouse mode and has minimal reactivity to other EGFR-expressing cells such as keratinocytes. 213,214 An enhanced cytotoxic activity of Vδ1 and Vδ2 T cells toward CD20-expressing lymphoma cells and, more importantly, against patient-derived chronic lymphocytic leukemia cells were induced by two immunoligands, designated [MICA × 7D8] and [ULBP2 × 7D8]. 215 The two immunoligands have specificity for CD20 (7D8) and NKG2D ligand MICA or ULBP-2, respectively, expressed on tumor cells including lymphoma and leukemia cells. 215,216 Beside Vδ1 and Vδ2 T cells, these immunoligands also attract NK cells and synergistically enhanced antibody-dependent cell-mediated cytotoxicity. An additional approach focuses on re-directing CD16-expressing Vδ1 and Vδ2 T cells as well as NK cells to HER-2-expressing PDAC, breast, and ovarian cancer cells through tribody [(HER2) 2 × CD16]. This tribody mediates an enhanced cytotoxicity and increased release of cytotoxic mediators by CD16-expressing innate cells against HER-2-expressing cancer cells. 217 Targeting of different innate cells can be promising with respect to their differential infiltration in different tumor entities.
Overall, these γδ-targeting bsTCE approaches offer new possibilities for γδ T cell-based immunotherapy.

| Effects of γδ T-cell engager on tumor resistance mechanisms
Since the interaction of tumor cells and tumor-surrounding cells with γδ T cells can increase inhibitory mechanisms including overexpression of COX-2 or IDO, downregulation of TRAIL-R4, or release of galectins, the anti-tumor response of γδ T cells is impaired. The application of γδtargeting bsTCE can drastically enhance impaired cytotoxic activity by targeting γδ T cells to tumor cells and increasing the release of cytotoxic mediators. Since γδ-targeting bsTCE are not designed to reduce immunosuppressive molecules such as PGE2 and kynurenine, the application of appropriate inhibitors has to be considered. A combination of γδ-targeting bsTCE with COX-2 or IDO inhibitors completely restored the γδ T cell cytotoxicity against tumor cells resistant against γδ T cellmediated lysis in vitro. 103,218 Interestingly, in a lung and mammary carcinoma mouse model, COX-2 inhibitor Celecoxib prevents IDO activation. [164][165][166] In addition, COX inhibitors improved the anti-tumoral γδ T cell cytotoxicity regulated by TRAIL-R4. 136 So far, it is uncertain whether an in vivo administration of γδ-targeting bsTCE and COX-2 or IDO inhibitors can overcome an immunosuppressive TME. Of course, other inhibitory mechanisms including oxidative and metabolic stress, expression of inhibitory receptors and galectins can inhibit γδ T cell proliferation, 11,77,88,171,218,219 which is only marginally affected by γδtargeting bsTCE. In this context, an activation of γδ T cells by γδ-targeting bsTCE in vivo may not abolish a suppressed proliferation or activate accumulating exhausted Vδ2 TIL. For instance, co-culture of freshly isolated PDAC tissue containing tumor-associated immune cells together with TIL in the presence of n-BP revealed an unresponsiveness of Vδ2 TILs and an enhanced secretion of galectin-3. 218 As a consequence to overcome exhaustion of Vδ2 TIL, an adoptive transfer of Vδ2 γδ T cells together with proliferation-inducing n-BP and enhanced cytotoxicityinducing γδ-targeting bsTCE can be a therapeutic option. If a possible suppressive activity of galectin-9, produced by adoptively transferred Vδ2 γδ T cells, inhibits αβ T cell activation or can be abolished by targeting sialoglycan-Siglec pathway has to be elucidated.
Surely, the data summarized in this review support the idea of an (personalized) assessment of phenotypic and functional characterization of the different T cell subsets of cancer patients. This analysis is important to get insights in the anti-tumor response of circulating and tumor-infiltrating γδ T cells. In addition, γδ-targeting bsTCE provide a valuable tool to analyze the functional capacity of γδ T cell subsets within peripheral blood and tumor-infiltrating cells co-cultured with autologous tumor cells. 58,82,220

| FUTURE PER S PEC TIVE S
To understand the functional properties of γδ T cell subsets and to improve γδ T cell-based immunotherapy, additional studies are required to further determine their frequency and distribution of γδ TIL in human tumor tissues, natural ligands, homing, survival, cytotoxicity, cytokines/chemokine production, requirement of costimulatory molecules, pro-and anti-tumoral activities, and regulatory functions. Additionally, different factors influencing the interaction of γδ TIL and tumor cells have also to be considered in more detail. These factors include genetic differences between tumor cells, oncogenic signaling pathways and inflammatory conditions at the tumor-site, intrinsic tumor escape mechanisms, influence of TMEassociated cells, and cancer-associated glycosylation.
A new wave of innovative strategies, including new design of γδ T cell engagers such as bsAb targeting different tumor-associated tumor antigens, CAR γδ T cell therapy, targeting of tissue-resident Vδ1 T cells, agonistic anti-BTN3A Ab, allogeneic γδ T cell adoptive cell therapy, and combination therapies will certainly help to improve the success of γδ T cell-based immunotherapy in the future. 81,221,222 A further understanding of functional properties of γδ T cell subsets, development of three-dimensional spheroid culture systems, use of non-human primate models, and further clinical studies will hopefully pave the way for the future success. 81,223

ACK N OWLED G EM ENTS
In memoriam, we cordially thank Wendy Havran for her excellent scientific work in the γδ T cell field and for several fruitful discussions during different γδ T cell conference meetings. We gratefully thank Matthias [Correction added on 17 November 2020, after first online publication: Projekt Deal funding statement has been added.]

DK is member of the Scientific Advisory Board of Incysus
Therapeutics, Inc; Imcheck Therapeutics; and Lava Therapeutics BV; all other authors declare no competing interest.