Designing T cells for Cancer Immunotherapy

Abstract

Adoptive transfer of T cells can enhance immune-mediated elimination of tumor cells and provides a specific, non-toxic cancer therapy. This approach has been effective in treating some hematologic and solid malignancies. In addition, the ability to genetically modify T cells to enhance their activity and persistence as well as overcome tumor immune evasion mechanisms has the potential to increase the success of these therapies in a wide range of tumors. In this review we discuss methods for gene transfer and specific modifications that have been made to T cells.

Introduction

The adoptive transfer of T cells expanded ex vivo can restore immune responses to viruses after bone marrow transplantation and augment immune responses to tumors in some circumstances. In fact, transfer of virus-specific cytotoxic T lymphocytes (CTLs) or tumor-specific T cells has proved safe and effective for Epstein-Barr virus (EBV)-associated malignancies, such as Hodgkin’s lymphoma and nasopharyngeal carcinoma (Bollard et al., 2004; Bollard et al., 2006; Heslop et al., 1996; Rooney et al., 1995; Rooney et al., 1998), as well as melanoma (Rosenberg and Dudley, 2009). The broader application and success of this approach for tumor therapy is limited in that many tumors do not induce strong responses from T cells in the first place because of inefficient antigen presentation and/or the presentation of tumor-associated antigens (TAA) that are mostly self-proteins, which are not usually immunogenic. Furthermore, tumors utilize multiple immune evasion strategies to dampen or shut off the activity of the T cells that are capable of mounting a response to the tumor. These immune evasion strategies include secretion of inhibitory cytokines, such as TGF-β and IL-10, expression of inhibitory ligands and molecules, such as PDL-1 and indoleamine 2,3 dioxygenase (IDO), and recruitment of T regulatory cells (Tregs) and myeloid suppressor cells (Stewart and Abrams, 2008). However, recent progress in gene transfer techniques has opened the door to engineering T cells with genetic modifications that can potentially overcome these limitations and provide for enhanced function and efficacy of adoptive T cell therapy in a wide range of tumors.

Methods of Gene Delivery to T Cells

T cells are highly proliferative, and thus, sustained and stable modification of these cells requires vectors that provide for integration of the transgene(s) into the cellular DNA as well as persistent expression of the transgene(s) at moderately high levels. There are several types of vectors that can satisfy these requirements, each having advantages and disadvantages. Most studies to date have utilized vectors based on gamma retroviruses or lentiviruses, but nonviral gene transfer using integrating transposon-based vectors is also being explored.

Gamma retroviruses integrate into the genome and produce reliable gene expression in T cells. These vectors have been used in several clinical trials in which they were shown to have an acceptable profile for efficacy and safety in expressing transgenes in T cells (Brenner and Heslop, 2003) and although there is concern about the ability of these vectors to cause insertional mutagenesis when they integrate into the genome, it is important to note that no adverse effects from insertional mutagenesis have been reported in any patient infused with T cells modified with these vectors in the past 20 years (Brenner and Heslop, 2003). These vectors, however, have limitations with respect to the size of the gene(s) that can be transferred (limited cargo capacity) (Hu and Pathak, 2000), although improvements in vector design have provided for at least three distinct genes to be expressed from one vector (Di Stasi et al., 2009; Quintarelli et al., 2007). In addition, because these vectors can only transduce dividing cells (Hu and Pathak, 2000), T cells must be activated ex vivo, which can be detrimental for their in vivo persistence once returned to the patient. Finally these vectors are costly to produce and test for clinical use.

Lentiviral vectors can transduce nondividing or minimally proliferating T cells (Hu and Pathak, 2000), reducing the requirement for ex vivo activation of the T cells, which could benefit their in vivo persistence by reducing activation induced cell death and cell exhaustion that comes with repeated stimulation. Lentiviral vectors also offer the advantages of accommodating larger genes or gene cassettes with reduced susceptibility to gene silencing as well as the reduced, but not absent, potential for insertional mutagenesis (Montini et al., 2006). However, as with retroviral vectors, there is substantial cost to the production, testing, and clinical use of lentiviral vectors.

Nonviral vectors, such as DNA plasmids, are much less costly to produce and test for clinical use than viral vectors. However, unlike viral vectors, which infect target cells and can produce high transduction rates, nonviral vectors cannot enter target cells on their own and require electroporation or chemical (liposomal)-based transfection methods for cellular entry. Transfection of primary T cells by any method is fairly inefficient but has been improved recently with the use of new devices. Furthermore, DNA plasmids, by themselves, integrate at very low frequency and thus, require selection and long in vitro culture of the modified cells, which can lead to exhaustion of the T cells (Park et al., 2007). However, transposon-based gene delivery systems, such as Sleeping Beauty and PiggyBac, are integrating nonviral gene delivery systems that can overcome low random integration rates and have been evaluated for their ability to modify T cells (Huang et al., 2006; Nakazawa et al., 2009; Singh et al., 2008). In these systems, the transposase (Sleeping Beauty or PiggyBac) is provided on one DNA vector, and the gene of interest flanked by transposon recognition sequences (the transposon) is on a separate DNA vector. The transposase “cuts” the transposon carrying the gene of interest out of its DNA vector and “pastes” it into the genome of the target cell providing for stable modification of the cell. Like the viral vectors, transposase-mediated integration has the potential to disrupt essential cellular genes and cause genotoxicity. However, a recent study by Galvan et al. (2009) suggests that PiggyBac may be less genotoxic than retroviral or lentiviral vectors. This study demonstrated that while PiggyBac integration is nonrandom and shows a preference for transcriptional units, the integration into or near known proto-oncogenes was not statistically different than what is expected from random integration. This result is in contrast with what has been reported for retroviral and lentiviral vectors, which have a higher frequency of integration into or near known protooncogenes (Galvan et al., 2009). Although T cells modified with transposon-based vectors have shown promise in preclinical studies, their safety and efficacy profiles in clinical trials remain to be determined.

Genetic Engineering of T Cells

The ability to transfer genes into T cells opens the door for improving the efficacy and safety of adoptive T cell therapy because these cells can now be genetically modified to improve their biological activity and provide resistance to hostile tumor microenvironments as well as to provide safety switches through which the cells can be eliminated if necessary (Figure 1).

Figure 1.

Potential gene modifications to T cells.

Enhancing the Biological Activity and Persistence of T Cells

Although tumor-infiltrating lymphocytes can be detected in some tumors, their receptors often have a low affinity for TAA, and the generation of tumor-specific T cells from cancer patients is challenging even under the best of conditions (Quintarelli et al., 2008). However, this problem can be overcome by genetically modifying T cells with transgenic αβ T cell receptor chains, which are cloned from high affinity TAA-specific T cell receptors (Uckert and Schumacher, 2009), or by expression of chimeric antigen receptors (CARs) (Dotti et al., 2009), to direct their specificity to tumors. CARs are a fusion of the light and heavy chain variable regions of a monoclonal antibody with the transmembrane and signaling domain of the CD3ζ chain of the T cell receptor and thus, combine the specificity of an antibody with the killing capacity of a T cell. In addition, because CARs recognize target antigens in a human leukocyte antigen (HLA)-independent manner, they also overcome tumor-immune evasion mechanisms that down-regulate antigen presentation, and thus recognition by specific T cells. Clinical studies with “first generation” CARs were disappointing in that there was limited efficacy, and CAR engagement failed to stimulate T cell proliferation and survival in vivo [reviewed in (Dotti et al., 2009)], likely because the T cells do not receive the necessary co-stimulatory signal to optimally activate the cells. Therefore, to improve the expansion and persistence of adoptively transferred T cells, “second-generation” CARs have been developed to include fusion of the CD3ζ chain with a co-stimulatory endodomain derived from a co-stimulatory molecule such as CD28, ICOS, CD134, or CD137. T cells modified with “second-generation” CARs are now being evaluated in clinical trials. “Third-generation” CARs fusing multiple co-stimulatory endodomains are also being developed, but whether “super” activation of T cells leads to increased toxicity of CAR-expressing T cells needs to be determined. Alternatively, CARs have been expressed in antigen-specific T cells, such as EBV-specific CTLs. In this platform, the T cell receives its necessary co-stimulation through activation of its native receptors and the CAR is used as a way to re-direct the T cell to the tumor. Our group has successfully utilized this approach to re-direct EBV-CTLs to the disialoganglioside antigen, GD2a, on neuroblastoma cells, which resulted in longer in vivo persistence of the T cells and complete remission (Pule et al., 2008).

In addition to adequate co-stimulation, T cells require cytokine signaling to promote their proliferation and survival. Although recombinant cytokines can support T cells after adoptive transfer, systemically delivered cytokines can produce significant toxicity as well as provide unwanted support to antagonistic T regulatory cells (Tregs). However, cytokine support can be provided to the adoptively transferred T cells by genetically modifying them to express particular cytokines. For example, antigen-specific T cells have been modified with retroviral vectors encoding the cytokines, IL-2 (Liu and Rosenberg, 2001) or IL-15 (Quintarelli et al., 2007), which increased the in vitro and in vivo expansion of the T cells as well as their anti-tumor activity in a SCID mouse model (Quintarelli et al., 2007). Importantly, these cytokine-transgenic T cells maintain their antigen-specificity, phenotype, and function as well as their dependence on antigen stimulation for continued expansion, all of which are important for preventing an overzealous T cell response and support the idea that cytokine-transgenic T cells are safe.

Alternatively, the T cells may be modified to express a receptor for a cytokine that is present (or can be safely administered) in the environment but to which they may not otherwise be able to respond. For example, our group has modified activated antigen-specific T cells to express the IL-7 receptor (Vera et al., 2009), which is normally down-regulated on T cells once they have been stimulated by antigen. This modification restores the T cell’s ability to respond to IL-7 that is available as a result of cytokine production in response to lymphodepletion of the host or from administration of the recombinant cytokine, which has been administered safely and without expansion of Tregs (Rosenberg et al., 2006).

In order for T cells to kill tumor cells, they must first go to the tumor. T cells express various chemokine receptors that allow them to migrate toward specific chemokines. Tumors often secrete chemokines that are recognized by receptors that are expressed by Tregs and Th2 T cells but not by cytotoxic tumor-specific T cells and so, tumor-specific T cells are not attracted to the tumor. However, tumor-specific T cells can be genetically modified to express the appropriate receptors for the chemokines secreted by the tumor and improve their ability to migrate to the tumor site. Our group has recently shown that this approach can be used to enhance the migration of T cells to Hodgkin’s lymphoma tumors resulting in increased anti-tumor effects (Di Stasi et al., 2009).

Providing Countermeasures to Resist Tumor Immune Evasion

Tumors hijack multiple immune suppressive mechanisms, such as secretion of suppressive cytokines, expression of inhibitory molecules, and recruitment of Tregs, that allow them to evade immune responses induced by endogenous as well as adoptively transferred T cells. However, T cells can be genetically modified to provide resistance to immune evasion strategies and restore their function in a hostile environment. TGF-β is a cytokine that is produced in many tumors and can suppress T cell proliferation and function. T cells modified to express a dominant-negative TGF-β receptor (dnTGF-β-RII), which inhibits TGF-β signaling, maintain their ability to proliferate and function in the presence of the cytokine in vitro and in vivo (Bollard et al., 2002) resulting in increased persistence and anti-tumor activity in mouse models with TGF-β-expressing tumors (Bollard et al., 2002; Foster et al., 2008). Based on these studies, a clinical trial to assess the safety and efficacy of dnTGF-β-RII-modified T cells has been initiated at our institution for the treatment of patients with relapsed/refractory Hodgkin’s lymphoma.

In addition, tumors often express Fas ligand (FasL) that can bind to Fas expressed on activated T cells, triggering apoptosis of the T cells and thus, protecting tumor cells from T cell-mediated destruction. To counter this immune evasion strategy, T cells can be genetically modified to express a small interfering RNA (siRNA) that promotes down-regulation of Fas and renders T cells resistant to apoptosis induced by FasL-expressing tumor cells (Dotti et al., 2005).

T cells deprived of the amino acid tryptophan or arginine fail to proliferate and function normally. Tumors take advantage of this property by expressing (or recruiting cells that express) IDO or arginase, enzymes that degrade tryptophan and arginine, respectively, and starve T cells of these amino acids. T cells deprived of tryptophan or arginine activate the stress response kinase, GCN2; and T cells from GCN2 knockout mice are able to function normally in the presence of IDO (Munn et al., 2005) and arginase (Rodriguez et al., 2007). Therefore, modification of T cells with an siRNA that down-regulates GCN2 could protect the cells from inhibition by IDO and/or arginase activity in tumors.

The modifications mentioned above are just a few examples of how we can, through genetic engineering, potentially improve the function of T cells in a hostile tumor environment. As our understanding of the mechanisms behind tumor immune evasion increases, we will be better able to rationally design tumor-specific T cells with one or more existing or novel countermeasures to provide the best possible anti-tumor activity.

Providing Safety Switches

Genetic modifications providing T cells with enhanced proliferative capacity, survival, and function carry with them an increased risk of direct toxicity and unchecked proliferation as well as the risk of malignant transformation from insertional mutagenesis. Therefore, several groups have evaluated various suicide genes or safety switches to effectively eliminate the T cells in case of unwanted proliferation or toxicity. The thymidine kinase gene from herpes simplex virus I (HSV-tk), which promotes killing of dividing cells in the presence of ganciclovir, has been the most widely used suicide gene, and HSV-tk modified cells are eliminated in vivo after ganciclovir administration (Ciceri et al., 2007). A drawback to this system, however, is that HSV-tk is itself immunogenic, which can result in elimination of the HSV-tk modified cells by reactive T cells (Traversari et al., 2007). Therefore, alternative non-immunogenic suicide genes based on inducible versions of the pro-apoptotic proteins, Fas, Fas-associated death domain-containing protein (FADD), or Caspase 9, have been developed (Straathof et al., 2005; Tey et al., 2007; Thomis et al., 2001). The pro-apoptotic protein is fused with a FK-binding protein (FKBP) variant that can bind a synthetic drug, AP1903, also called a chemical inducer of dimerization (CID), and when present, the CID binds two FKBP variants and causes dimerization and activation of the pro-apoptotic proteins. While inducible FAS or FADD can eliminate up to 90% of transduced cells (Thomis et al., 2001), inducible caspase 9 performs better, eliminating up to 95% of transduced cells (Quintarelli et al., 2007; Tey et al., 2007), and is thus being tested in a clinical trial at our institution as a way to eliminate adoptively-transferred polyclonal T cells after haploidentical transplantation.

Conclusion

Adoptive T cell therapy holds much promise as a specific, nontoxic way to treat cancer but has been effective only in a limited number of settings. However, the ability to genetically modify T cells offers the potential to improve their targeting, persistence, and anti-tumor activity, and the safety and efficacy of T cells with various genetic modifications are being evaluated in current clinical trials for the treatment of a number of different cancers.