Molecular immunology lessons from therapeutic T-cell receptor gene transfer

Abstract

The T-cell receptor (TCR) is critical for T-cell lineage selection, antigen specificity, effector function and survival. Recently, TCR gene transfer has been developed as a reliable method to generate ex vivo large numbers of T cells of a given antigen-specificity and functional avidity. Such approaches have major applications for the adoptive cellular therapy of viral infectious diseases, virus-associated malignancies and cancer. TCR gene transfer utilizes retroviral or lentiviral constructs containing the gene sequences of the TCR-α and TCR-β chains, which have been cloned from a clonal T-cell population of the desired antigen specificity. The TCR-encoding vector is then used to infect (transduce) primary T cells in vitro. To generate a transduced T cell with the desired functional specificity, the introduced TCR-α and TCR-β chains must form a heterodimer and associate with the CD3 complex in order to be stably expressed at the T-cell surface. In order to optimize the function of TCR-transduced T cells, researchers in the field of TCR gene transfer have exploited many aspects of basic research in T-cell immunology relating to TCR structure, TCR–CD3 assembly, cell-surface TCR expression, TCR-peptide/major histocompatibility complex (MHC) affinity and TCR signalling. However, improving the introduction of exogenous TCRs into naturally occurring T cells has provided further insights into basic T-cell immunology. The aim of this review was to discuss the molecular immunology lessons learnt through therapeutic TCR transfer.

Introduction

Retroviral T-cell receptor (TCR) gene transfer was first demonstrated 10 years ago in studies using a melanoma antigen-specific TCR.¹ This and other initial studies generated only small numbers of redirected T cells with relatively poor function.^2,3 Over the last decade, substantial progress has been made in the field of TCR gene transfer, with improved vectors and transduction protocols for TCR gene delivery and, more recently, with additional modification of the TCR genes to improve specific pairing and function.

Detailed studies have demonstrated that the peptide specificity and avidity of TCR-transduced T cells can be equivalent to the parental T-cell clone from which the TCR was isolated.⁴ In animal models, TCR-transduced T cells have been shown to protect against viral infection and to confer protection against tumours.^5–7 A small number of phase I/II clinical studies have been completed and have confirmed that sufficient numbers of genetically modified T cells can be generated ex vivo, that TCR-transduced autologous T cells can persist after adoptive transfer and that anti-tumour activity in melanoma patients was feasible.⁸ However, further improvements are required to optimize the efficacy of TCR gene transfer in the clinical setting.

The efficiency of TCR gene transfer, and the subsequent function of the TCR-transduced T cell, is influenced by the vector delivery system, the TCR transgenes and the transduction conditions. To date, most TCR gene-transfer protocols have utilized gamma-retroviral vectors. Stable genomic integration of retroviral vectors requires full T-cell activation and proliferation during the transduction process. This process requires stimulation through the TCR complex using antibodies against CD3, with or without anti-CD28, in order to stimulate progression through the cell cycle, followed by a period of in vitro expansion in the presence of interleukin (IL)-2. During this in vitro activation process, T-cell differentiation occurs and cell-surface molecules important for homing to secondary lymphoid organs (i.e. CD62L) or costimulation (i.e. CD28) are down-regulated. There are theoretical advantages to redirecting the antigen specificity of less-differentiated cells and this can be achieved using lentiviral vectors, which permit gene transfer into non-dividing T cells.^9,10 These approaches are currently being explored by a number of research teams, together with TCR transfer into selected central memory or naïve T cells and co-transfer of specific homing molecules.

A number of challenges remain, including: (i) to maximize the cell-surface expression of the introduced TCR; (ii) to minimize or eliminate the mispairing of introduced TCR-α and TCR-β chains with endogenous TCR chains; (iii) to improve the association of the introduced TCR with molecules of the CD3 complex; and (iv) to enhance the functional avidity of the TCR-transduced T cells.

The relevant steps in the generation of antigen-specific T cells by TCR gene transfer are indicated in a schematic representation (Fig. 1).

Figure 1.

Schematic representation of critical processes in T-cell receptor (TCR) gene transfer. MHC, major histocompatibility complex.

Strategies to improve TCR expression

TCR assembly and expression is a complex process.¹¹ Before cell-surface expression, the TCR-α and TCR-β chains have to form a heterodimer. This process is influenced by the secondary and tertiary structures of both the variable and constant domains. The TCR-αβ then associates with the CD3 complex within the endoplasmic reticulum (ER), which involves interactions between the TCR constant domain (both intracellular and intramembrane portions) and the CD3 molecules. Finally, the TCR–CD3 complex is released from the ER and translocates to the cell membrane. Surprisingly, murine TCRs have been shown to be more efficiently expressed in human T cells than human TCRs, and this is probably because of a stronger association with human CD3 molecules.¹²

It is clear from murine models of tumour protection that antigen recognition correlates with the TCR expression level. Elegant experiments performed in transgenic mice expressing controllable amounts of cell-surface TCR demonstrated that a reduced density of TCRs on the T-cell surface resulted in reduced proliferation, and in the secretion of interferon-γ (IFN-γ), IL-2 and IL-4 in response to in vivo vaccination with cognate peptide,¹³ which could be overcome in part by stimulation with saturating doses of peptide. Of importance to the field of TCR transfer, the threshold of TCR density required for antigen responsiveness was relatively low (< 1000 surface TCRs per cell), but was significantly affected by the concentration of antigen ligands.

Extensive research is ongoing in the field of vector development to enhance transgene delivery into T cells, but this is outwith the scope of the present review. However, the impact of TCR transgene modifications and vector configuration on the subsequent expression in the transduced cell will be discussed.

Codon-optimization of TCR-α chain and TCR-β chain transgenes

Codon optimization of the TCR-α chain and TCR-β chain transgenes relies on the replacement of infrequently used codons with synonomous codons frequently encountered in the human genome. There is now a substantial body of evidence demonstrating that for multiple TCR specificities the introduction of codon-optimized TCR genes results in higher TCR expression levels in transduced T cells compared with wild-type TCR genes and subsequently improved in vivo function.^14–16 There is a theoretical risk that codon optimization will generate potentially immunogeneic TCRs, resulting in anti-TCR immune responses, as the process of optimization may generate alternative open reading frames, with alteration of peptide sequences; however, this has not yet been reported.

TCR-α/TCR-β vector configuration

For TCR gene transfer it is preferable to use a single viral vector encoding both TCR chain genes, as this limits the risk of insertional mutagenesis and the number of transduced T cells expressing only the introduced α chain or β chain. The introduction of only one TCR chain because of the successful transduction with only one of two vectors would increase the risk of the introduced chain mispairing with the reciprocal endogenous TCR chain (see below). TCR heterodimer assembly and cell-surface expression will be impaired if there is a limiting supply of one or the other chain. Therefore, currently used viral vectors link the TCR-α and TCR-β chain genes with either an internal ribosomal entry site (IRES) sequence or the 2A peptide sequence derived from a porcine tsechovirus.^17,18

Vectors using the IRES sequence result in the expression of a single messenger RNA (mRNA) molecule under the control of the viral promoter within the transduced cell. Translation of the second gene is mediated by the IRES element. The efficiency of this system is limited by the fact that the gene downstream of the IRES sequence is typically expressed at lower levels than the upstream gene.¹⁷ Conversely, the 2A peptide linker results in a single mRNA molecule, but during translation ribosomal skipping generates two separate proteins from the single mRNA.¹⁸

The majority of constructs currently in clinical and preclinical development use the 2A sequence to link the TCR-α and TCR-β chains as a result of the improved equimolar expression of both genes, compared to vectors with an IRES element separating the TCR genes. Importantly, it has been shown by ourselves and others that T cells transduced with constructs containing the TCR genes linked by a 2A sequence express higher levels of cell-surface TCR and demonstrate improved antigen-specific function, as measured by IFN-γ secretion, compared with constructs containing identical TCR sequences separated by an IRES element.¹⁹

Competition between introduced TCRs and endogenous TCRs: limiting CD3 and the concept of ‘strong’ TCRs

Efficient cell-surface TCR expression requires the formation of a stable TCR–CD3 complex.¹¹ In the absence of CD3, TCRs do not assemble properly and are degraded. Therefore, the availability of CD3 molecules for TCR–CD3 complex assembly is a major rate-limiting effect when introducing additional exogenous TCRs into T cells. Competition may reduce cell-surface expression of the introduced TCR and impair the avidity of antigen recognition of the transduced cells. We have recently demonstrated that the double transduction of CD8⁺ T cells with a vector encoding the desired TCR-α and TCR-β chain genes, together with a second vector encoding the CD3 gamma, delta, epsilon and zeta genes (linked by 2A sequences), can enhance the avidity of CD8⁺ T cells (King J, Ahmadi M, personal communication). This may be a mechanism to enhance the functional avidity of transduced T cells expressing low-affinity TCRs.

It is common for the introduced TCRs to be expressed at lower levels than the endogenous TCRs, which may impair the ability of the transduced T cell to respond to low concentrations of the TCR-recognized antigen, as discussed above. This observation is consistent with the introduced TCR competing with the endogenous TCR for limited CD3 molecules. Heemskerk et al.²⁰ have recently shown that the expression levels of the introduced TCR can be influenced by the ‘strength’ of the endogenous TCR by introducing the same TCR into different antigen-specific T-cell clones. It is currently unclear whether TCR-specific molecular motifs exist to determine the ‘competitiveness’ of a given TCR-αβ chain.

Strategies to improve specific pairing of the introduced TCR-α and TCR-β chains

Primary T cells transduced with exogenous TCRs have the potential to express four different TCR-αβ heterodimers on the recipient T-cell surface: (i) the endogenous αβ heterodimer; (ii) the introduced αβ heterodimer; (iii) the endogenous α chain paired with the introduced β chain; and, finally, (iv) the introduced β chain paired with the endogenous α chain. These possibilities are indicated in the schematic diagram shown in Fig. 2.

Figure 2.

Schematic representation of the T-cell receptor (TCR)-transduced T cell.

As the TCR can only be expressed on the cell surface as an αβ heterodimer, we and others have demonstrated that the transduction of either the α or the β TCR chain alone into primary T cells results in the formation of mixed dimers between the exogenous TCR chain and the reciprocal endogenous TCR chain. This has been demonstrated by increased cell-surface expression of the introduced α or β chains.^2,20–22

Mixed αβ TCR dimers are of concern for two main reasons. First, incorrect pairing of the introduced αβ TCR chains causes reduced specific pairing on the cell surface of the desired TCR. This will have a detrimental affect on the avidity of the resultant T cell. Second, and perhaps more importantly for the clinical setting, the formation of mixed dimers has been perceived as a possible safety concern. Such mixed TCR dimers have undefined antigen specificity and because they have bypassed in vivo thymic selection it is postulated that the mismatched TCRs could recognize self-tissue or self-major histocompatibility complex (MHC), leading to autoimmunity. Although off-target autoimmune pathology was not observed in the Rosenberg phase I clinical trial,⁸ it has been reported that TCR-transduced T cells expressing novel mixed TCR dimers can be autoreactive and/or demonstrate alloreactivity in vitro.²³

However, the tendency to form mixed dimers varies between differing TCRs. It is likely that specific sequences within both the variable and constant domains of the TCR dictate whether a given α or β chain has a tendency to behave promiscuously and readily dimerize with reciprocal endogenous β or α chains, respectively.

Murinization of human TCR-constant regions improves cell-surface TCR expression

As a continuation of the observation that murine TCRs can readily replace human TCRs on the T-cell surface, as discussed above,¹² it has been shown that human TCRs which have been modified such that their constant domains are replaced with murine sequences preferentially dimerise with their murinised counterparts in preference to fully human TCRs. Compared with their human equivalent, murinised human hybrid TCRs show increased cell-surface expression immediately after T-cell transduction, which translates into enhanced T-cell function.^12,22 It is hypothesized that the improved function of T cells transduced with the human–murine hybrid TCR is not only caused by the reduction of mispaired TCR dimers, but by the increased efficiency of TCR expression on the cell surface because the constant domain of the murine TCR interacts and competes more efficiently than the human constant domains with endogenous CD3.

Cysteine modification of TCR chains improves specific pairing through the generation of a new intermolecular disulphide bond

The addition of an exogenous disulphide bond in the constant domain of the TCR has also been demonstrated to reduce TCR mispairing and therefore also to increase the functional avidity of the resultant T cells.^22,24,25 Unpublished work from our laboratory, and from others, has demonstrated that the combination of the murinisation and the addition of a cysteine bond in the constant domain are additive on their effect on TCR cell expression, and therefore T-cell functional activity, in comparison to their sole components. However, it must be emphasized that neither murinisation of TCR constant regions nor cysteine modification of TCR constant regions completely eliminates mispairing of the modified TCR chains to reciprocal endogenous human TCR chains. Evidence suggests that the level of TCR mispairing is also affected by the variable region of the endogenous TCR chains (Fig. 3).¹²

Figure 3.

Heterodimer formation, CD3 association and cell-surface expression of modified T-cell receptor (TCR) constructs in TCR-αβ⁻ Jurkat cells. (a) Schematic representation of Wilms' tumour antigen 1 (WT1)-specific human leucocyte antigen-A2 (HLA-A2)-restricted TCR constructs. SS, disulphide bond. (b) Cell-surface expression of the WT1 TCR variants in CD3⁺ TCR-αβ⁻ Jurkat cells is shown. Mock and TCR-transduced Jurkat cells were stained with antibodies to the TCR and to the CD3 before fluorescence-activated cell sorter (FACS) analysis. CD3 association and cell-surface TCR expression was observed with the wild-type, Cysteine-1 and Hybrid TCR constructs. Biochemical analysis of Jurkat cells transduced with the Cysteine-2 TCR construct demonstrated no association of the Cysteine-2 β chain with either the Cysteine-2 α chain or CD3, resulting in no cell-surface CD3 expression or T-cell functional activity (the Cysteine-2 variant was identical to variant 1, except that the cysteine residues responsible for the formation of the natural disulphide bond between the TCR-α and TCR-β chains were removed).

Modifications to the secondary structure of TCR constant chains can improve specific pairing

An additional approach to prevent TCR mispairing, as demonstrated by Voss et al.,²⁶ was the identification and inversion of a pair of specifically interacting amino acids in the TCR-α and TCR-β constant-domain interface. Mutational inversion of these two amino acids changed a ‘knob-into-hole’ configuration into a charged ‘hole-into-knob’ configuration and by so doing increased the preferential pairing of the transduced mutated TCRs. This approach was effective in both human and murine TCR gene-transfer systems.

Chimeric antigen receptors as alternatives to conventional αβ TCRs

An alternative method to completely abolish TCR mispairing is the development of chimeric antigen receptors (CARs), which consist of a single chain Fv fused to CD3 signalling elements. However, the functional activity of CARs is dependent on the sensitivity of the signalling elements, which in some constructs contain additional costimulatory molecules and/or cytokines. Early research with CAR-expressing T cells suggested that they were less sensitive to peptide than T cells expressing αβ TCR heterodimers.^27,28

It is possible that the described modified TCRs will be immunogenic in an immunocompetent host, resulting in reduced persistence or elimination of the transduced T cells. Whilst the lymphodepleting regimens currently used before adoptive T-cell transfer are likely to permit T-cell engraftment, it is still necessary to consider strategies to minimize the possible immunogenicity of the modified TCRs.

Introduction of TCR-αβ chains into alternative effector cells

An alternative and novel method of eliminating TCR mispairing is to transduce TCR-αβ genes into γδ T cells. Using this system, T cells must either be transduced with CD8 or CD4 co-receptor independent TCRs, or TCRs and co-receptors must be co-transferred together. These TCR-transduced γδ T cells demonstrate peptide-specific lysis and cytokine release in vitro and also peptide-specific proliferation, persistence and recall responses in vivo.^29–31

Strategies to improve TCR affinity and avidity

Achieving T cells with a high functional avidity is one of the major challenges in current TCR gene-therapy protocols. One means of attaining T cells capable of recognizing and effectively killing tumour cells is to confer the manipulated T cells with TCRs with a high affinity. As the majority of currently available tumour-associated antigens (TAAs) are self-antigens that are expressed at elevated levels in tumours, T cells expressing high-affinity TCRs to tumour antigens may be deleted in the thymus or rendered unresponsive by central or peripheral tolerance. As a result, TAA-specific T cells identifiable within the autologous repertoire are often of low frequency and low-to-moderate functional avidity. Therefore, the challenge of attaining high-affinity TCRs has been addressed by investigating a means to provide sources of high-avidity T cells as a supply of high-affinity TCRs or by manipulation of the TCR itself to increase its affinity.

The allo-restricted approach to selecting TCRs from high-avidity cytotoxic T lymphocyte clones

A number of different approaches have been used to produce and isolate high-avidity T cells, from which TCRs can be cloned for TCR transfer. Our laboratory has used the allorestricted cytotoxic T lymphocyte (CTL) approach to produce high-avidity T cells which have the added benefit of bypassing T-cell tolerance. High-avidity self-peptide-specific allorestricted T cells have not been subject to tolerance because they are non-self-reactive in the autologous repertoire. For this technique, peripheral blood lymphocytes from a human leucocyte antigen (HLA)-mismatched donor were used to select T cells that recognized a WT-1 antigen expressed on HLA-A2. T cells transduced with TCRs isolated from the allorestricted CTLs demonstrated peptide specificity in vitro and in vivo.^32,33 An alternative method to produce high-affinity TCRs is to immunize HLA-transgenic mice with human peptides. Murine T cells are therefore produced that recognize peptides presented on human HLAs. The TCRs from these cells can then be isolated and transferred into human T cells. This approach has been used by others to isolate TCRs that recognize human murine double minute protein-2 (MDM2)⁶ and p53.³⁴

TCR affinity maturation

Whilst the above approaches rely on selecting and then isolating TCRs from high-avidity T cells, an alternative method is to use an in vitro system to directly mutate the TCR to increase its affinity. It is known that the third complementarity-determining regions (CDR3s) of both antibodies and TCRs play a major role in antigen binding and specificity. In this scenario, TCRs are subjected to in vitro mutagenesis followed by selection of TCR sequences with improved binding affinity for the specific MHC–peptide combination. DNA libraries of TCR variants can be produced by using polymerase chain reaction (PCR) mutagenesis to introduce random mutations, usually in defined TCR regions that are associated with either peptide or MHC recognition. These libraries can be displayed on yeast, bacteriophage or T cells, and are then screened for increased binding affinities to the peptide–MHC complex. The TCRs from selected clones can then be sequenced and transduced into T cells for further analysis. Outside the context of TCR transfer, a number of researchers have studied, in detail, the participation of the TCR CDR1, CDR2 and CDR3 regions in the determination of binding kinetics and peptide specificity. In a simplified model, CDR1 and CDR2 bind to MHC helices and CDR3 binds to the presented peptide. Surpisingly, affinity-matured TCRs with mutants in all three CDRs retained peptide specificity, suggesting that in addition to amino acid sequence, electrostatic forces and the TCR conformation may be important in determining peptide specificity.³⁵

Using affinity-maturation techniques, TCRs can be produced that have affinities at the super physiological level and are several logs greater than the parental wild-type TCR.^36–39 In the field of TCR gene transfer, this approach has been used to target viral-escape mutants occurring in chronic viral infections. Recently, Varela-Rohena et al.⁴⁰ used phage display to generate affinity-matured TCRs specific for an HLA-class I-presented human immunodeficiency virus (HIV)-derived SL9 peptide epitope. When variant α and β chains were combined, the affinities, as determined by surface plasmon resonance, were increased markedly, with one mutated TCR binding to the peptide–MHC complex with a half-life in excess of 2·5 hr. Following transduction of the mutated TCRs into CD8 T cells, antigen specificity was retained and the TCR-transduced T cells produced a greater range of cytokines and increased amounts of IL-2 in response to HIV-infected target cells compared with the CTL line from which the wild-type TCR was isolated.

A number of concerns exist regarding the generation of TCRs with supraphysiological peptide–MHC complex affinities. It is likely that there is an affinity threshold for optimal TCR function. For example, the serial triggering model suggests that a peptide–MHC complex molecule can consecutively interact with several TCRs, resulting in a signal amplification mechanism.⁴¹ This requires a balance between TCR/affinity and the on/off rate. Serial triggering is facilitated by a relatively fast off rate of the TCR-MHC/peptide interaction. It is conceivable that in vitro-selected TCR molecules, achieving affinities far above the affinity window of natural TCR repertoires, and markedly extended off rates, upset this balance and may fail to deliver appropriate signals required for T-cell activation and memory development in vivo.

Furthermore, it has been reported that CD8 T cells transduced with the high-affinity TCRs show a lack of peptide fine-specificity⁴² and as the affinity of a TCR is increased, the number of stimulatory peptides it can recognize also increases.⁴³ There is therefore concern that these T cells will show cross-reactivity with the self-peptide–MHC complex. Interestingly, CD4 T cells transduced with the high-affinity TCRs continue to show peptide specificity, and the increase in TCR affinity is accompanied by an increase in peptide recognition and T-cell avidity.^44,45 This technique could therefore prove to be a valuable means to genetically modify CD4 T cells in order to acquire T-cell help in adoptive cancer T-cell therapies.

TCR deglycosylation

A recently published method of increasing TCR affinity has arisen from data which suggest that increased glycosylation of T-cell-surface proteins is associated with an increased activation threshold, and vice versa. Kuball et al.⁴⁶ demonstrated that deletion of defined N-glycosylation sites in the constant domains of the TCR-α and TCR-β chains increased the functional avidity of T cells transduced with these modified TCRs. This affect was demonstrated for a number of human and murine TCRs specific for different tumour antigens and it is therefore thought that this technique can be readily translated to any TCR. Owing to the ability of TCRs above an affinity threshold level to recognize self-protein, caution must be observed, and it is therefore necessary for all TCRs that have an increased affinity to undergo extensive in vitro and in vivo screening before reaching the clinical setting.

Conclusion

This review has described areas of basic T-cell immunology of fundamental importance to the field of TCR gene transfer and T-cell immunotherapy. However, the ability to transfer TCRs of known affinity and specificity into human or murine T cells ‘at will’ can facilitate further studies into the critical steps of TCR pairing and assembly, antigen recognition, T-cell signalling and function of self-reactive T cells, amongst others. Current research is focused on improving the function of TCR-transduced T cells, but also on exploring the introduction of TCR-αβ chains into alternative T-cell subsets, such as CD4⁺ helper T cells,⁷ CD4⁺ CD25⁺ regulatory T cells^47,48 and γδ T cells,²⁹ to generate specialized antigen-specific T cells.

Disclosures

EM and HS are members of the Scientific Advisory Board of CellMedica Ltd.