The structure of the T lymphocyte antigen receptor resembles in many features that of an antibody Fab fragment. The T cell receptor (TCR) is a heterodimer consisting of either a TCRα-chain and a TCRβ-chain (for αβ T cells) or a TCRγ-chain and a TCRδ-chain (for γδ T cells) which are covalently linked by a disulfide bond. The majority of T cells express αβ TCRs. γδ T cells only make up about 5% of the peripheral T lymphocyte population and their precise role in the immune system and in tumor immunology is still under debate (reviewed in [1 ,2 ]). In this study, only the structure and function of αβ TCRs will be discussed.

The TCR chains are expressed on the cell surface and each consists of an N-terminal variable part (V) that mediates the binding to antigen and the major histocompatibility complex (MHC), a constant region (C) that harbors the inter-molecular disulfide bond, a charged transmembrane domain and a short cytoplasmic tail that is involved in signal transduction (Figure 1A). The variable and constant parts are linked by a joining region (J) in the case of TCRα or a diversity region (D) and a joining region in the case of TCRβ. The existence of various V, J and D segments, which are recombined to one Cα segment or one of the two Cβ segments (Cβ1 and Cβ2) during T cell maturation, and the presence of three hypervariable complementarity determining regions (CDRs) in the variable domains of both TCR chains are the basis for the high diversity of TCRs. Each T cell expresses only one type of TCR and it was estimated that 10⁸ – 10⁹ different T cell clones circulate in any individual allowing the recognition of a multitude of pathogens.

Analysis of various crystallized TCR fragments (reviewed in [3 ,4 ,5 ]) revealed that each variable and constant domain display an immunoglobulin (Ig)-like “β-barrel” structure that consists of three to four anti-parallel β-sheets facing three similar sheets on the other side (Figure 1B). Only Cα diverges from this predicted Ig fold as its outer strands exhibit a random coil structure rather than β-sheets which may be the reason for an observed higher lability of the TCRα-chain [6 ,7 ,8 ]. 

On the cell surface, the TCR chains are expressed in complex with the CD3 subunits γ, δ, ε and ζ which associate to covalently linked CD3ζζ homodimers and non-linked CD3γε and CD3δε heterodimers [9 ]. The extracellular parts of CD3ε and CD3γ are predicted to adopt an Ig-fold [10 ]; and the cytosolic region of all CD3 molecules harbor immuno-receptor tyrosine-based activation motifs (ITAMs) which can be phosphorylated and play a role in recruiting downstream signal transducers (Figure 1A). The stoichiometric composition of the TCR/CD3 complex is still controversially discussed. Some studies suggest that two TCRαβ heterodimers are clustered with one CD3γ and CD3δ chain, two CD3ε chains and two CD3ζ chains (Figure 1C), others report that only one TCRαβ molecule is involved (reviewed in [11 ,12 ]. However, there is compelling evidence that the formation of multivalent TCR/CD3 complexes upon MHC binding is required for full T cell activation [13 ,14 ,15 ].

Figure 1 :   Schematic and crystal structure of the αβ   TCR complex. (A) A schematic drawing of a TCR, showing the TCRα-chain (purple) and the TCRβ-chain (blue) linked by a dilsulfide bond (yellow). Accessory CD3 molecules (green) mediate the signal transduction via ITAM motifs (red boxes). (B) Crystal structure of the extra-cellular part of the murine 2C TCR. The numbers in the V regions indicate the CDR loops. Adapted from [6 ]. (C) Presumed model of a TCR-CD3 cluster.

Several binding sites in the TCR chains have been identified that are crucial for the association with CD3 molecules. The FG loop in the Cβ domain – a large solvent-exposed protrusion – forms a cavity that is predicted to interact with one CD3ε subunit [16 ]. Also, critical amino acids in the extracellular and transmembrane parts of Cα and Cβ have been found that mediate the assembly with CD3ζ [17 ].

In contrast to antibodies, TCRs cannot recognize free antigens; instead they bind small peptide fragments that have been processed intracellularly and loaded on MHC molecules. Cytotoxic CD8-positive cells recognize antigens bound to MHC class I, whereas CD4-positive T helper cells are specific for MHC class II-presented antigens. The peptide-MHC class I complexes are found on the surface of nearly every cell and represent the cell’s whole proteome. Upon binding of the TCR to its cognate peptide-MHC complex, an immunological synapse between the T cell and the target cell is formed that leads to recruitment of specific molecules, e.g. adhesion proteins, to lipid rafts. The affinity of TCRs for their specific antigen is rather weak when compared to that of an antibody. Therefore, the interaction is further stabilized by coreceptors, such as CD4 and CD8 which bind to invariable parts of the MHC molecule.

The role of the immune system in tumor development has been discussed for more than a century. Ehrlich postulated already in 1909 that the immune system might protect the host from cancer [18 ]. Some fifty years later, Thomas and Burnet formulated the cancer immuno-surveillance hypothesis [19 ,20 ] saying that lymphocytes were eliminating continuously developing transformed cells. Though this hypothesis in the context of non-virally induced tumors is still highly debated today [21 ,22 ] numerous studies have demonstrated that cancer cells have an antigen pattern which is distinct from that of normal cells and that T cells can specifically recognize these tumor antigens (a list of known human tumor antigens can be found in [23 ]). 

Although tumor-infiltrating lymphocytes (TILs) have been isolated from many human or animal malignancies, their presence alone is obviously not sufficient to reject an established tumor. One reason for this may be that many tumor antigens are auto-antigens that are aberrantly expressed in cancerous tissue, and therefore, may be only of low immunogenicity. T cells recognizing auto-antigens are usually deleted in the thymus during negative selection and auto-reactive lymphocytes that escape this central tolerance machinery are mainly of low affinity. Furthermore, it has been shown by Willimsky et al. that spontaneously developing tumors – although expressing a tumor-specific transplantation rejection antigen – can induce general T cell unresponsiveness and tolerance [24 ].

Nevertheless, the potential of T cells to infiltrate into tissues and to specifically recognize and destroy a cell presenting a foreign antigen make them a useful tool in anti-cancer therapy. The obstacles that inhibit naturally occurring tumor-specific lymphocytes might be overcome by adoptive T cell transfer as this therapy provides the means to inject a large number of cells and to select or engineer T cells with a high affinity to one or several tumor antigens. Additionally, previous irradiation or chemotherapy of the patient can give the transferred T cells a proliferative advantage and augment the presentation of tumor antigens [25 ].

Transfer of unmodified T cells

The principle of adoptive T cell therapy is to isolate T cells either from the patient or an allogeneic donor, expand and activate them in vitro and transfer them back into the patient. To date, the most successful clinical application of T cell therapy is the treatment of chronic myeloid leukemia by MHC-matched allogeneic stem cell transfer [26 ,27 ]. It is assumed that donor T cells that are transferred along with the stem cells are the main effectors in preventing relapse. Most likely, disparities in minor histocompatibility antigens between donor and recipient are recognized and account for the elimination of residual malignant cells [28 ,29 ]. A hint to that is the observation that the anti-tumor effect is markedly decreased when T cells are depleted from the graft or when bone marrow derived from a genetically identical twin is transferred [30 ]. Still, an increased graft-versus-leukemia (GvL) effect also seems to correlate with increased incidence of graft-versus-host disease (GvHD) – an often lethal complication that occurs when the transplanted T cells attack normal tissue.

A second successful approach of transplanted T cells has been the treatment of virally-induced tumors which are frequent in immuno-suppressed patients, e.g. after stem cell or organ transplantation. Reinfused autologous or donor T cells effectively restored immunity against Epstein-Barr virus (EBV) [31 ,32 ,33 ,34 ] or cytomegalovirus (CMV) [35 ,36 ,37 ] and prevented lympho-proliferative disorders. A strong advantage in this type of application is the possibility to select, expand, clone and characterize antigen-specific T lymphocytes in vitro before transfer. 

The translation of this approach to non-virally induced malignancies, however, was only of limited success (reviewed in [38 ]). This is most likely due to the lower immunogenicity of non-viral tumor antigens and the lower precursor frequency of TILs compared to virus-specific and allo-reactive T cells. One exception to this rule seems to be melanoma from which frequently large numbers of highly lytic TILs can be isolated [39 ]. Redefined culture and treatment conditions recently led to objective responses in 30 to 80% of melanoma patients in recent clinical trials [40 ,41 ]. Still, given the required laborious and time-consuming ex vivo expansion of TILs, this therapy will only be available for a limited number of patients.

Transfer of gene-modified T cells

The genetic modification of T lymphocytes in vitro before adoptive transfer allows endowing these cells with enhanced properties or defined antigen receptors and might overcome many obstacles observed with unmodified T cells. Engineering T cells with a desired specificity by gene therapy has the advantage that (i) large numbers of therapeutic lymphocytes can be created in relatively short time compared to the long and cumbersome expansion of unmodified tumor-reactive T cells, (ii) specificities other than naturally occurring can be employed and (iii) treatment will also be feasible for patients from which TILs could not be isolated.

CARs are fusion proteins of antibody Fv fragments and a TCR signaling domain [42 ] that recognize antigens independent from MHC. They have a high, antibody-like affinity and T cells transduced with CARs have been shown to kill tumor cells in vitro (reviewed in [43 ]). In a first clinical trial the in vivo function of CAR-transduced T cells specific for an epitope of carboxy-anhydrase-IX (CAIX), but also auto-immunity, was reported [44 ]. An important drawback in the use of CARs, though, is their restriction to cell surface tumor antigens which greatly limits their broad application. Furthermore, their high affinity bears the risk of auto-immunity when the tumor-antigen is also expressed on a normal cell. 

Another strategy to confer a T cell with tumor-specificity is the transfer of TCR genes. High-affinity TCRs for human tumor-associated antigens can either be isolated (i) from rare, highly reactive human TILs, (ii) by generation of peptide-specific, allo-reactive T cells [45 ], (iii) from HLA-transgenic mice immunized with human tumor antigens [46 ,47 ] or (iv) by in vitro mutagenesis using yeast or phage display techniques [48 ,49 ,50 ]. To date, a large panel of TCRs against viral and tumor antigens has been isolated which, when genetically transferred into a T cell, were able to redirect its specificity (reviewed in [51 ,52 ,53 ,54 ]). Genetic modification can either be transient through RNA electroporation or stable through the use of retroviral vectors. An initial clinical trial using PBLs transduced with a melanoma-antigen specific TCR has led to partial remission in some patients [55 ]. Still, many hurdles regarding the efficiency of TCR gene-modified T cells in tumor therapy have to be overcome. One is the high-level expression of the transgene, which can be hampered by competition with the endogenous TCR or by formation of mispaired TCR heterodimers composed of one endogenous and one transgenic TCR chain. Furthermore, the choice of the retroviral vector and composition of the transgene cassette have been shown to be important [7 ,56 ]. A second hurdle is the selection of the target antigen. In general, two groups of tumor antigens have been identified: (i) tumor-specific antigens (TSAs) that are only expressed on tumor tissue and are caused by random mutations of different cellular genes, and (ii) tumor-associated antigens (TAAs) that are found over-expressed in cancer cells, but also – albeit at a lower level – on normal tissue [23 ]. As TSAs are usually not shared between different patients, TAAs have been more extensively studied for use in immuno-therapy but bear the risk of therapy side effects when non-tumor tissue is damaged. Ideally, one would find a TCR specific for an antigen which is only expressed on the tumor, but shared between patients such as viral tumor antigens, p53 or ras mutation hotspots, or newly created epitopes of fusion proteins such as bcr-abl. Finally, it has been shown that recognition of tumor cells alone is not sufficient to lead to tumor rejection due to the expansion of antigen-loss variants. To allow efficient elimination of the tumor, it is also necessary to target the tumor stroma [25 ,57 ,58 ]. In line with this, anti-angiogenic treatment or tumor site irradiation can augment T cell therapy response. While the first leads to shortage of blood supply of tumor cells, the latter promotes the cross-presentation of tumor-antigens by the stroma cells which thus become a target of adoptively transferred T cells.

Despite first clinical responses achieved with genetically engineered T lymphocytes, several risk factors in the treatment of patients have to be considered. These can either results from the choice of the target antigen or the genetic manipulation of the T cell itself.

Recognition of tumor-associated antigens on self-tissue

Tumor-associated antigens are the most thoroughly studied targets in T cell based immuno-therapy. The clinical application of a TCR which recognizes a TAA, however, bears the risk of auto-immunity when the engineered T cells damage TAA-expressing non-tumor tissue. In some cases, mild auto-reactivity may well be tolerated as a side effect of the therapy. If, however, essential tissue is affected, it is desirable to have a possibility to eliminate the transferred T cells.

Adverse effects by adoptive T cell therapy have already been observed in some clinical trials and mouse models. In two independent studies, the destruction of melanocytes (vitiligo) and uveitis was seen in patients which had received melanoma antigen-specific T cells [40 ,41 ]. Vitiligo was also observed in tumor-bearing mice which received T cells specific for the melanoma antigen gp100 for treatment [59 ]. In another trial, patients with renal cell carcinoma were treated with PBLs genetically modified with a CAR specific for an epitope of CAIX. Here, liver cytotoxicity occurred due to specific reaction of the infused CAR-modified T cells with CAIX-expressing epithelial cells of the bile ducts and the study had to be discontinued [44 ]. Such auto-immune reactions in adoptive immuno-therapy most likely depend on the antigen level [60 ]. They can hardly be predicted, may vary from patient to patient and will become very important when TCR-redirected T cells are routinely used for therapy.

Formation of heterodimers by endogenous and transgenic TCR  

It has been observed that the introduced TCR chains can individually form mixed heterodimers with the α- and β-chains of the endogenous TCR [61 ,62 ]. The extent of mispairing seems to both depend on the stability of the inter-chain interaction (preferential pairing), and on the intrinsic stability of the respective TCR chain and its ability to compete with the endogenous chain for export and accessory proteins (weak and strong TCRs [61 ]). The occurrence of mixed heterodimers has two principal consequences. First, it reduces the amount of correctly paired therapeutic TCR on the cell surface, and thereby most likely the functional reactivity of the engineered T cell. Second, if a polyclonal pool of T cells is TCR-modified, the specificity of the mispaired TCRs cannot be predicted and they might have auto-reactive capacity.

Several strategies have been tested that promote preferential pairing of the desired TCR chains. For example, the introduction of an additional disulfide bond [63 ,64 ], leucine zipper motifs [65 ] or the inverse exchange of an amino acid pair in the interface of the TCRα and TCRβ constant regions [66 ] stabilized the interaction of the transgenic TCR chains. Also, the use of murine constant regions instead of human both enhanced the binding to human CD3 molecules and supported preferential pairing [67 ,68 ]. Willemsen et al. constructed chimeric single chain and two chain TCRs which did not pair with the endogenously expressed TCR chains [69 ]. Still, it remains unclear whether these amino acid modifications will be immunogenic and cause unwanted elimination of the transferred T cells by the host’s immune system. Van der Veken et al. introduced a TCR into γδ T cells, whose endogenous TCRs are unable to form heterodimers with αβTCRs [70 ]. However, it is not known whether engineered γδ T cells will exhibit the same anti-tumor function in vivo as αβ T cells. Another option would be the use of RNA interference (RNAi) to down-regulate the endogenous TCR chains and the generation of a codon-modified transgenic TCR which is not influenced by the RNAi mechanism.

Activation of an endogenous auto-reactive TCR

Another safety concern is the possible activation of the endogenous TCR. While some data show that signaling through one TCR in dual-specific T cells is receptor-specific [71 ], others demonstrate that activation of the introduced TCR may also induce a response of the endogenous receptor [72 ]. Most likely the observed cross-activation varies in different model systems and cannot be predicted when polyclonal T cells are transduced. Although clonal deletion in the thymus eliminates the majority of T cells with high-affinity auto-reactive TCRs, low-affinity auto-reactive T lymphocytes escape central tolerance mechanisms [73 ,74 ]. If these T cells are transduced with a second TCR and become activated, they may react against self-tissue. Just as for the prevention of TCR heterodimers, the use of RNAi to modulate the expression of the endogenous TCR, or the pre-selection of T cells with a defined, non-auto-reactive specificity before transduction [75 ,76 ,77 ] might avoid this risk.

Insertional mutagenesis and transformation

It is assumed, that for prevention of relapse the long-term persistence of infused tumor-specific T cells is necessary. For this, stable TCR expression in the transduced T cells is essential. So far, all techniques that support stable expression require integration of the transgenic DNA into the host genome. Most effective delivery systems (e.g. retroviruses), however, allow only non-site-specific insertion which bears the risk of malignant transformation if the integration affects the expression of an oncogene. Until recently it has been assumed that integration occurs randomly. Considering the small proportion of gene-encoding regions in the human genome, it seemed very unlikely that oncogenes will be a target site for the vector.

Reports of serious adverse events in an otherwise successful clinical trial of gene therapy for X-linked severe combined immuno-deficiency (X-SCID), however, demonstrated retrovirus-induced lympho-proliferative disease in 4 of 9 treated patients in one study and 1 of 10 treated patients in a second study [78 ,79 ,80 ]. In four of the cases it was reported that retroviral integration activated expression of the proto-oncogene lmo2 which finally led to the oncogenic transformation [79 ]. Since then, numerous studies analyzed retrovirus integration sites in human and murine cells and showed that integration occurs non-random and preferentially in the 5’ region of transcriptionally active genes [81 ,82 ,83 ,84 ,85 ]. In the X-SCID trials, the fact that hematopoietic progenitor cells were transduced, which are probably more prone to transformation due to deregulation of expression, might have contributed to development of leukemia. In the case of retroviral transduction of T lymphocytes with a suicide gene, also preferential integration sites and deregulated expression profiles were found. This, however, seemed to have no consequences for the T cell biology and no clonal selection in patients was observed [86 ,87 ]. Whether genetic modification of T cells or hematopoietic stem cells using a TCR influences the expression pattern and leads to loss of polyclonality still needs to be analyzed. Besides, many attempts have been made to construct self-inactivating retroviral vectors, non-integrating vectors or vectors with site-specific integration and high efficiency [88 ,89 ,90 ] which in the future might overcome the obstacle of insertional mutagenesis.

In sum, immuno-therapy with TCR-modified T cells bears the potential risk of auto-immune side effects and malignant transformation of the T cell. Although approaches have been tested that avoid some specific risk factors, a general safety strategy allowing in vivo depletion the adoptively transferred T cells – thereby terminating the therapy – is desirable.

In the case of severe side effects, several treatments allow the in vivo suppression of T cells. The administration of T cell-specific antibodies or corticosteroids can block auto-immune reactions, but also abolishes desired – e.g. anti-viral – immune responses. Therefore, a way to specifically eliminate the transferred lymphocytes is preferable. For this, several strategies have been suggested. Among them are (i) suicide genes of viral or bacterial origin which drive the cell into apoptosis upon application of a prodrug, (ii) fusion proteins of cellular apoptosis-inducing molecules that cluster upon administration of a dimerizer drug, (iii) transgenic cell surface molecules which can be targeted by specific depleting antibodies, and (iv) MHC multimers that deliver cytotoxic substances to a specific T cell pool. The feasibility of these strategies in the context of therapy with TCR-modified T cells is discussed below.

Suicide gene/prodrug systems

The most thoroughly studied suicide gene so far is the thymidine kinase of Herpes Simplex Virus type 1 (HSV-TK). This enzyme plays a key role in the phosphorylation of thymidine to produce dTMP (Figure 2A). Unlike cellular kinases, the viral enzyme has very broad substrate specificity and also converts pyrimidine and purine analogs such as acyclovir and ganciclovir (GCV) (Figure 2B). The mono-phosphorylated forms of these analogs are further phosphorylated by cellular kinases and incorporated into nascent DNA leading to an arrest of DNA synthesis, DNA fragmentation and finally apoptosis of the cell.

Figure 2 : HSV-TK can phosphorylate thymidine and nucleotide analogs. (A) Thymidine is phosphorylated by cellular kinases or HSV-TK to thymidine-mono-phosphate (dTMP) and further to thymidine-tri-phosphate (dTTP) which can be incorporated into a replicating DNA strand. To the free OH- group, further nucleotides can be attached. (B) HSV-TK also phosphorylates nucleotide analogs like GCV, which in their tri-phosphorylated form are integrated in the newly synthesized DNA. As no further nucleotides can be attached to GCV, the replication process is discontinued.

HSV-TK in combination with the prodrug GCV has already been applied as a suicide gene in a variety of animal tumor models and for some human cancers (reviewed in [91 ,92, ,93 ]). More recently, it was suggested as a safety modality in patients who develop GvHD after donor lymphocyte infusion following T cell-depleted allogeneic hematopoietic cell transplantation (reviewed in [94 ,95 ,96 ]).  Several clinical phase I-II trials have tested the safety and efficacy of this approach in humans [97 ,98 ,99 ,100 ]. Though application of HSV-TK in T lymphocytes was demonstrated to be safe and partially successful, the trials also revealed significant disadvantages of the strategy. Major limitations of the HSV-TK system are: (i) immunogenicity of the HSV-TK gene product resulting in immune responses and the elimination of transferred gene-modified T cells [99 ,101 ,102 ], (ii) transgene silencing or inhomogeneous transgene expression which leads in cells expressing low levels of HSV-TK to the development of ganciclovir-resistance [103 ,104 ], (iii) the prodrug ganciclovir cannot be used to treat upcoming viral infections (e.g. of CMV or EBV) in patients, and (iv) the restriction to proliferating cells which impedes elimination of slowly dividing T cells usually present in chronic GvHD. Apart from HSV-TK a number of other suicide gene of prokaryotic and eukaryotic origin are known. Their high expected immunogenicity, however, has restrained their use in therapy.

Apoptosis-inducing fusion genes and dimerizer prodrugs

An alternative to the HSV-TK-induced cell cycle arrest is to employ the cell’s own mechanisms of apoptosis (illustrated in Figure 3A). Several different strategies attempted to construct fusion proteins in which a signaling domain of a protein involved in the apoptosis pathway (e.g. the DED domain of FADD, the FasR intracellular domain or a modified caspase 9) was linked to a FK506 binding protein (FKBP) [105 ,106 ,107 ,108 ]. Administration of a synthetic dimerizer prodrug leads to cross-linking of the FKBP domains and activates apoptosis (Figure 3B).

Figure 3 : Mechanism of apoptosis induction by a caspase 9 fusion construct. (A) Pathway of Fas-induced apoptosis. By binding of Fas ligand (FasL) to the Fas receptor (FasR) the receptor molecules form a cluster bringing the death effector domains (DED, red boxes) in close proximity. This recruits FADD which subsequently leads to activation of downstream caspases. Cytochrome C release from mitochondria mediates the conversion of pro-caspase 9 into its activated form. (B) Structure of the caspase 9-FKBP fusion construct. Upon application of the prodrug the molecules multimerize which mimics the process of caspase 9 activation.

Compared to HSV-TK, apoptosis-inducing fusion genes comprise a lower risk of evoking an immune response against the transgene products as all of the fusion protein components are of human origin. Still, it cannot be excluded that an immunogenic peptide at the fusion sites of the several domains is generated, finally leading to an unwanted elimination of the transferred T cells. Besides, in vitro experiments with the caspase 9 transgene and a first in vivo study with the Fas-linked constructs in non-human primates revealed that a small proportion of T cells showed resistance to elimination and that in these cells the expression of the transgene was reduced [108 ,109 ,110 ]. Most likely all of the approaches depend on high levels of transgene expression and comprise the risk that silencing of the fusion construct abrogates the possibility of elimination. Furthermore, the use of apoptosis-inducing molecules like Fas or FADD that act upstream of many apoptosis inhibitors, such as bcl-2, bcl-x_L or c-FLIP, might lead to inhibition of T cell depletion when these molecules become up-regulated. This, however, is not the case for caspase 9, which is a late-stage apoptosis pathway molecule and maintains its function in T cells over-expressing anti-apoptotic proteins.

CD20 and CD20-depleting antibodies

CD20 is a cell surface molecule mainly expressed by B lymphocytes. It acts as a calcium channel in the cell membrane and is presumably involved in B cell activation. Its precise function, however, is still not known. Introna et al. and van Meerten et al. used retroviral or lentiviral vectors to transduce T lymphocytes with the CD20 cDNA [111 ,112 ,113 ]. They could show that CD20 can be expressed on a high level on primary human T cells and T cell lines and that the transduced cells can be enriched via CD20 surface expression. Furthermore they demonstrated that CD20-modified T cells can be depleted by incubation with a CD20-specific antibody and complement factors.

One advantage of this strategy is the availability of a CD20-specific antibody which is well characterized and approved for clinical use. This monoclonal antibody (Rituximab) has originally been developed for the treatment of CD20-positive B cell lymphomas, but is now also used in some B cell-associated auto-immune diseases. The large clinical experience with Rituximab would ease the implementation of this safety approach in a clinical setting. Compared to the above described safety approaches, modification of T cells with CD20 has the advantage that the molecule is very unlikely to be immunogenic as the entire transgene is of human origin.

Still, administration of the antibody would also lead to the unwanted elimination of the patient’s B cells. Furthermore, resistance of lymphoma cells to Rituximab treatment has been reported [114 ] and the efficacy of depletion has been demonstrated to depend on the level of transgene expression [115 ]. As survival of CD20-modified T lymphocytes will most likely not depend on functional CD20 expression, mutations in the transgene or silencing of the transgene may occur, rendering the T cells resistant to elimination. An additional concern is that the presence of the B cell-specific CD20 molecule in T lymphocytes might alter their phenotype. Although there is a small subset of T cells expressing CD20 naturally [116 ], the function of the molecule in these cells remains unclear. Serafini et al. analyzed CD20-modified T cells in vitro with respect to antigen-specific and allo-induced cytokine release, chemotaxis and the expression of activation markers [117 ]. They observed no differences between mock-transduced and CD20-transduced T lymphocytes. However, it cannot be excluded that in vivo the CD20 molecule might affect the homing behavior or function of the T cell.

Cytotoxic tetramers

MHC multimers are broadly used for the detection and quantification of specific T cell populations in vitro and in vivo. By coupling a cytotoxic agent to the MHC molecule, it is also possible to deplete T cells of certain specificity. Yuan et al. fused MHC tetramers to the short-range, alpha-emitting isotype ²²⁵Ac which specifically killed their cognate T lymphocytes in culture [118 ]. In a different approach, Hess et al. coupled MHC tetramers to the type I ribosome-inactivating protein saporin [119 ]. When these tetramers bind to the specific TCR, the entire complex becomes internalized into the cell where saporin inhibits protein synthesis finally leading to apoptosis. When injected into mice, the saporin-MHC tetramers were able to deplete about 75% of the target T cells. However, in this experiment also mild, transient cytotoxic side effects like loss of body weight and hepatopathy were observed. A similar strategy was used by Casares et al. who constructed MHC class II chimeras coupled to doxorubicin, an antimitogenic drug [120 ].

Although the clonal deletion of T cells by cytotoxic tetramers seems advantageous, this approach also comprises several drawbacks. First, MHC tetramers coupled to cytotoxic agents have low structural stability. Though they might be useful in depleting T cells ex vivo (e.g. from grafts), their in vivo application might be hampered by their short half-life and their inability to infiltrate poorly vascularized tissues. Second, the injection of cytotoxic substances into patients might lead to unwanted bystander effects and their specificity needs to be carefully evaluated. Third, the in vivo administration of MHC multimers has been shown to modulate immune function. Some studies found that injection of soluble MHC induced antigen-specific T cell unresponsiveness [121 ,122 ], while others revealed that it has a T cell-activating effect [123 ,124 ]. Finally, the production and safety testing of individualized suicide tetramers for each tumor antigen-specific TCR would be very cost-intensive.

In sum, all of the described safety mechanisms comprise several limitations rendering them inappropriate for T cell therapy with TCR gene transfer. Combining two different safety approaches may overcome some of the disadvantages [125 ,126 ]; however, this very much complicates the generation of TCR-modified T cells. Apart from depleting tetramers, all strategies require introduction of at least one additional gene into T cells. Retroviral vectors – the most commonly used system to stably transduce T cells – only have a limited transgene capacity. Considering the size of the TCRα- and TCRβ-chain genes it is unlikely that vectors that carry an additional gene can efficiently transduce T cells. Hence, PBLs will necessarily have to be independently transduced with a TCR and a second gene-encoding vector. This increases the number of retroviral integrations into the host cell genome, and thus the risk of insertional mutagenesis [127 ]. Also, the purification and analysis steps needed to ensure that all TCR-redirected cells express the safety modality will prolong the in vitro culture time and decrease their functionality [128 ]. Table 1 summarizes the advantages and disadvantages of each approach. 

Safety approach

Advantages

Drawbacks

HSV-TK / GCV

• clinically approved prodrug

• safe and partially efficient in clinical trials

• immunogenicity

• slow response

• gene silencing or deletion

• no GCV treatment possible

• dependence on expression level, purification, insertion of additional gene

Apoptosis-inducing fusion proteins

• fast response

• low expected immunogenicity

• possible cytotoxicity

• dependence on expression level, purification, insertion of additional gene

CD20 / CD20-mAb

• clinically approved antibody 

• no immunogenicity expected

• fast response

• unwanted elimination of B cells

• possible change of phenotype

• dependence on expression level, purification, insertion of additional gene

Cytotoxic tetramers

• not dependent on expression level, purification, insertion of additional gene

• structural instability

• individualized, cost-intensive production

• possible cytotoxic bystander effects

• immunomodulation

Table 1 : Safety modalieties for adoptive T cell transfer

The aim of this thesis was to develop a method for the specific depletion of adoptively transferred TCR-gene modified T cells. Such safeguard has to meet several criteria: (i) it should not interfere with TCR function; (ii) it should be specific, efficient and rapid; and (iii) the implementation in a clinical setting should be feasible.

With respect to this, the objective was to select a short amino acid sequence (tag) and introduce it into the TCR structure so that it can be recognized by a tag-specific antibody. In vivo, the binding of a depleting tag-specific antibody would then lead to a specific elimination of a T cell expressing the tag-modified TCR.

For this purpose, the myc-tag – a peptide derived from the human c-myc protein – was inserted into various positions of the model murine TCR P14 gene. The wild-type and myc-tag-modified TCRs were cloned into retroviral vectors which were used to transduce murine T lymphocytes. The properties of the expressed TCR molecules were analyzed in vitr o:

• First, it was tested whether introducing a myc-tag allowed expression and assembly of the TCR on the cell surface and if so, whether the expression level was influenced by the myc-tag.

• Second, it needed to be investigated whether the myc-tag was inserted in a conformation in which it could be bound by a myc-specific antibody.

• Third, complement-mediated depletion assays were performed to see whether the binding of a myc-specific antibody was able to induce depletion of the T cells in v i tro.

• Fourth, the function of T cells transduced with a myc-modified TCR – including antigen binding and cytokine secretion upon antigen stimulus – was analyzed and compared to T cells transduced with a wild-type TCR. 

Having found a position for myc-tag insertion which allowed functional expression of the TCR and supported lysis by a myc-specific antibody, the objective was to test the universality of this approach when applied to other TCRs. Therefore, a second murine TCR (OT-I, specific for an ovalbumin-derived peptide) and one human TCR (gp100, specific for a common melanoma antigen) were modified with a myc-tag in the same position. Murine and human T cells transduced with these TCRs were analyzed as described above.

Finally, the application of the safeguard was investigated in vivo. For this, T cells transduced with a myc-tag-modified OT-I TCR were transferred into RIP-mOVA mice which express ovalbumin in the pancreatic islet cells. As without treatment the mice succumb to auto-immune diabetes, it was analyzed whether application of a myc-specific antibody allowed the specific depletion of the auto-reactive OT-I T cells and hence a rescue of the animals from the otherwise lethal disease.