Prospects for immunotherapy of malignant disease

Abstract

The majority of T cell-recognized tumour antigens in humans are encoded by genes that are also present in normal tissues. Low levels of gene expression in normal cells can lead to the inactivation of high-avidity T cells by immunological tolerance mechanisms. As a consequence, low-avidity T cell responses in patients are often inadequate in providing tumour protection. Recently, several technologies have been developed to overcome tolerance, allowing the isolation of high-affinity, HLA-restricted receptors specific for tumour-associated peptide epitopes. Furthermore, transfer of HLA-restricted antigen receptors provides an opportunity to empower patient T cells with new tumour-reactive specificities that cannot be retrieved from the autologous T cell repertoire.

INTRODUCTION

The goal of therapy for malignant disease remains the long-term eradication of tumour cells without adverse effects on normal tissue. Conventional approaches utilizing chemotherapy and radiotherapy are limited by both their lack of specificity and toxicity. Since the first description of a human tumour-associated antigen recognized by cytotoxic T cells (CTL) 10 years ago [1], advances in understanding the nature of tumour-specific immune responses and mechanisms of tolerance induction have encouraged researchers and clinicians alike to develop a more refined approach to immune-mediated therapies. Current immunotherapeutic approaches are many and varied with respect to target specificities, effector mechanisms, mode of administration and prospects for translation into clinical practice.

In this review we will discuss recently developed vaccination strategies that have led to a large number of clinical trials. We will compare the limited clinical success of vaccination with the well-documented curative potential of adoptive tumour therapy with allogeneic T lymphocytes. Finally, we will highlight novel approaches exploiting the allogeneic T cell repertoire and the repertoire of HLA transgenic mice, as well as in vitro selected receptors against defined MHC/peptide combinations, as a basis for antigen-specific gene therapy of cancer.

VACCINATION STRATEGIES

Active vaccines

T cell-recognized tumour antigens can be divided into two main categories. The first are known as tumour specific antigens (TSA), and the genes encoding TSA are present only in tumour cells and not in normal tissues. The second group, called tumour-associated antigens (TAA), are expressed at elevated levels in tumour cells but are also present in normal cells.

Theoretically, TSA represent the most desirable target antigens for anticancer vaccination or adoptive therapy. Their tumour cell-specific expression means there will be no pre-existing immunological self-tolerance, and immune responses directed against TSA will be unlikely to damage normal tissues. Examples of TSA include the products of mutated genes, such as the oncogene RAS and the bcr/abl fusion protein. The major limitation of targeting TSA for immunotherapy is that they are frequently invisible to CTL, due to restrictions dictated by rules governing MHC class I presentation. Presentation of peptides on the cell surface requires proteasome-aided peptide release from larger precursors, peptide transport by TAP molecules and high-affinity peptide binding to MHC class I molecules. Frequently, peptides with specific mutations will not successfully compete against the large number of peptides produced from normal cellular proteins, and hence be poorly presented on the cell surface. This may help to explain the observation that the vast majority of peptides that have been identified as CTL targets in cancer patients are products of normal cellular proteins expressed at elevated levels in tumours, i.e. TAA.

Active vaccination strategies for cancer are usually directed against defined TAA [2], or against an unidentified number of antigens in cases when the vaccine consists of whole tumour cells [3], cell lysates [4,5] or heat shock proteins [6] and RNA [7,8] isolated from tumour cells. Although vaccines based on multiple unidentified antigens have shown promise in the experimental setting [9], the monitoring of vaccine-induced immune responses is difficult without knowledge of relevant target antigens. The progressive, rational improvement of vaccines is more probable with defined antigens, as it provides an opportunity to link the nature of antigen-specific immune responses with clinical outcome. Also, there is a growing consensus that antigen-specific cancer vaccines will prove to be more effective than vaccination strategies based on unidentified antigens [10–12].

Vaccination in the murine model

Prophylactic active vaccination with a range of TAA has been shown to protect from tumour challenge and prevent tumour occurrence in some animal models [13–15]. Murine models have demonstrated that vaccination with melanocyte differentiation antigens can provide protection against melanoma. Bronte et al. [16] have found that vaccination with murine TRP-2 protects mice from challenge with melanoma cells, while Mullins et al. [17] have shown in HLA-A*0201-transgenic mice that vaccination with tyrosinase or gp100 can control melanoma outgrowth. Soares et al. have recently demonstrated in a MUC1 transgenic mouse model that vaccination with dendritic cells pulsed with a 140 amino acid long peptide derived from the TAA MUC1 improved survival of tumour challenged mice. Surprisingly, the level of protection in MUC-1 transgenic mice was similar to that seen in wild-type mice, suggesting that T cell tolerance did not interfere with the development of protective immunity. It is possible that the 140 amino acid long peptide stimulated unconventional, MHC-unrestricted T cells responses [18], which may explain the failure to identify MHC-restricted CTL responses in vaccinated mice [13]. Romieu et al. [19] used the SV40 large T transgenic tumour model to demonstrate that vaccination with MHC class I presented large T epitopes together with heterologous helper peptides stimulated low-avidity CTL which did not prevent the development of large T induced spontaneous tumours, whereas adoptive transfer of higher avidity CTL specific for the same epitope substantially decreased the tumour burden.

Frequently, vaccination can generate antigen-specific CTL, but their avidity is low. However, it may be possible to exploit the low avidity self-specific T cell repertoire for tumour immunotherapy as suggested by experiments in transgenic models [20]. For example, in transgenic mice expressing a CTL-recognized influenza nucleoprotein (NP) epitope in most normal tissues, an oligoclonal population of NP-specific CTL can escape thymic and peripheral deletion and can be activated subsequently by immunization. Although primary CTL were of low avidity, repeated vaccination led to the generation of memory CTL with improved avidity, and these CTL were capable of delaying the growth of tumour cells expressing the NP epitope. Thus, multiple vaccinations with a tumour-associated self-antigen may provide partial tumour protection by improving both the quality (avidity) and quantity (frequency) of antigen-specific CTL [21,22].

In contrast to their success in the prophylactic setting, TAA-based vaccines were largely unsuccessful in the therapeutic setting, suggesting that the quality and quantity of the CTL response is not protective against existing tumours. One possible explanation is that pre-existing tolerance prevents rapid activation and expansion of high avidity CTL required for the inhibition of existing tumours. This explanation is compatible with the observation that vaccination against murine TSAs that are not present in normal cells can be effective in the therapeutic setting and inhibit growth of existing tumours [23–25].

Studies in the context of non-self target antigens have been performed where high avidity-CTL have been shown to provide better protection than low avidity CTL against viral infection [26]. Derby et al. demonstrated that high avidity viral-specific CTL were capable of recognizing lower antigen densities and more rapidly lysing target cells at a given antigen density than the lower-avidity CTL. Similarly, high-avidity CTL for tumour associated self-antigens have been shown to have superior antitumour efficacy both in vitro and in vivo in animal models [27].

Vaccination in humans

In the past decade antigen-specific cancer vaccines have entered phase I and II clinical trials. The majority of trials were performed in melanoma patients using TAA-based vaccine preparations. Many trials have reported detectable vaccine induced T cell responses in the absence of clinical responses [28–30]. For example, in one trial, end-stage melanoma patients were vaccinated with mature monocyte derived dendritic cells pulsed with the MAGE-3A2·1 peptide, which resulted in the generation of MAGE-3 specific T cells in patients without signs of tumour reduction [30]. As discussed above, this may be due to lack of activation of high avidity CTL caused by immunological tolerance mechanisms. Other important factors determining clinical responses are level of antigen expression in the target tumour cells. Expression of a given target antigen is likely to vary between patients and at different time points in the malignant process. Some tumours have been shown to escape immunological surveillance via down-regulation or loss of expression of HLA Class I molecules [31,32].

Nevertheless, it is clear that tolerance to melanoma antigens is incomplete. For example, in a few clinical trials there have been reports of clinical responses including stabilization of disease, partial and even complete tumour regression [33–40]. A statistically significant positive correlation between the percentage of tetramer positive antigen-specific CD8⁺ T cells and the clinical response of patients was observed by Fong et al. [36], indicating that clinical responses correlated directly with detectable immune responses. In contrast, Marchand et al. [37] found clinical responses in the absence detectable T cell responses. Together, the studies in melanoma patients indicate that more trials with well-defined vaccination and monitoring protocols are needed to establish a link between immunological response profiles and clinical outcome.

Immunological tolerance

Negative selection (deletion) of T cells reacting with intrathymic self-antigens and self-MHC is the principle mechanism by which tolerance is achieved [41]. In addition, peripheral tolerance mechanisms exist to prevent effective triggering of self-reactive T cells that have escaped thymic deletion [42]. Peripheral tolerance mechanisms include ignorance [43], anergy [44], TCR down-regulation [45], deletion [46–48], or suppression of autoreactive T cells by CD4⁺/CD25⁺ regulatory T cells [49].

Purging of high-avidity CTL as a consequence of extrathymic antigen expression in a small number of cells in the peripheral tissues has been documented in murine models [50–52]. However, peripheral tolerance mechanisms are leaky, allowing escape of a small number of autoreactive T cells. Using a double transgenic model, Morgan et al. demonstrated that the number of adoptively transferred autoreactive T cells determined the level of tolerance [53]. Transfer of 100 high-avidity CTL specific for a self-antigen expressed in the pancreas resulted in tolerance within 5 days, whereas tolerance induction of 10 000 high-avidity CTL required 120–220 days. This indicates that small numbers of high-avidity CTL, matching naturally occurring numbers that escape thymic deletion, are tolerized efficiently in the periphery, while large numbers of high-avidity CTL, achievable in the adoptive therapy setting, can escape tolerance and may provide tumour protection.

As TSA are neoantigens not present in normal tissue, the level of antigen expression on target tumour cells may be less important in determining the functional avidity of the antigen-specific CTL, as high-avidity CTL are not susceptible to tolerance. However, when the target antigen is an overexpressed TAA, self-restricted CTL will be susceptible to the tolerance mechanisms as described above, resulting in deletion/anergy of high-avidity CTL. Therefore, tumour cells will be required to express higher levels of antigen to trigger low-avidity CTL, which escape tolerance and normally ignore low level expression of the self-antigen in normal tissues.

Adoptive therapy

Reduced immune competence and immunological tolerance are major limitations for vaccination strategies that need to be overcome by alternative approaches. One of the major benefits of adoptive therapy is the ability to transfer antigen-specific CTL of known avidity and specificity. The transfer of antigen-specific T cells has major applications in the therapy of viral infectious diseases, virus-associated malignancies and cancer. The successful translation of adoptive immunotherapy into clinical practice has been achieved first for the management of cytomegalovirus (CMV) and Epstein–Barr virus (EBV) infection complicating haematopoietic stem cell transplantation (SCT). The adoptive transfer of CMV-specific CTL clones has reduced the incidence of CMV disease [54]. Similarly, infusions of donor-derived EBV-specific T cell lines following T cell-depleted SCT have prevented EBV-triggered lymphoproliferative disorder [55,56] and reduced the viral load in patients with high titre EBV DNA post-transplant [57]. For patients undergoing solid organ transplantation the use of EBV-specific CTL has been explored with some success [58,59].

Adoptive transfer of allogeneic T cells is currently the only immunotherapy strategy in humans with proven curative potential against cancers of non-viral aetiology. The first description of the allogeneic graft-versus-leukaemia (GVL) effect was as early as 1956, when Barnes et al. observed the erradication of leukaemia in irradiated mice receiving allogeneic, but not syngeneic bone marrow transplants [60]. GVL is now regarded as a donor T cell-mediated antileukaemic immune response. The first evidence of a similar effect in humans followed the observation that relapse rates were lower following T cell-containing allografts and in the presence of acute and chronic GVHD [61–63].

Donor lymphocyte infusions (DLI) were first used to treat relapse after HLA-identical sibling SCT in the late 1980s [64]. Subsequent work has confirmed the efficacy of DLI for molecular, cytogenetic and haematological relapse of CML following stem cell transplantation [65–67]. The use of DLI is now commonplace in the management of haematological malignancies following allogeneic SCT [68–70]. The major disadvantage of DLI is the development of acute or chronic GVHD, which significantly increases the transplant-related mortality, reduces quality of life and increases infectious complications if long-term immunosuppression is required [71,72]. The risk of GVHD following a given DLI dose increases with the degree of HLA disparity between donor and recipient and is greater if DLI is given early post-transplant.

The target antigens involved in both GVL and GVHD have not yet been elucidated fully, and it is likely that some are shared. For example, the minor histocompatibility antigen 1 (HA1) has been implicated in GVHD [73,74], and more recently in GVL [75]. CTL directed against minor histocompatibility antigens expressed exclusively in haematopoietic tissues are currently in clinical studies to test if they can mediate GVL without GVHD. It is anticipated that adoptive CTL transfer will be beneficial in patients after SCT, as normal haematopoietic cells will be antigen-negative because they are of donor origin, whereas residual leukaemia cells will be antigen-positive and thus function as CTL targets [76,77].

The role of CD4⁺ CD25⁺ suppressor T cells in tumour immunobiology is not yet well understood, although there is increasing evidence that they inhibit tumour protective immune responses [78]. It has been shown recently that lung tumours contain large numbers of CD4⁺ CD25⁺ cells expressing high levels of CD152 (CTLA-4). These regulatory T cells have been shown to inhibit the proliferation of autologous but not allogeneic T cells [79], although the mechanism leading to the selective suppression is unclear. The implications of this study are that resident CD4⁺ CD25⁺ cells suppress tumour-infiltrating autologous T cells, and that adoptively transferred CTL might escape this suppression.

Future of antigen-specific immunotherapy

A number of novel approaches have been developed recently to circumvent the tolerance mechanisms responsible for limiting efficacy of CTL responses to self-antigens (Fig. 1). Although this had led to the isolation of high avidity CTL with killing activity against malignant cells, the expression levels in normal cells were insufficient to trigger CTL killing.

Fig. 1.

Novel approaches for the generation of high-avidity CTL specific for tumour-associated antigens. Overview of four recently developed technologies to isolate high-affinity receptors specific for HLA class I presented peptide epitopes (HLA-A2 is shown because it is one of the most frequent class I alleles in humans). (a) and (b) take advantage of the TCR repertoire of healthy individuals or transgenic mice that are not tolerant to A2-presented peptide epitopes of tumour-associated antigens (TAA); (c) and (d) take advantage of in vitro mutagenesis and selection of high affinity TCRs (c) or high-affinity single chain antibodies specific for the peptide/HLA combination. (e) Strategies to transfer high avidity HLA-restricted receptors into patient T cells or stem cells to produce tumour-reactive T cells.

One approach, the allorestricted strategy, has grown from a greater understanding of the alloresponse and the observation that T cell tolerance is self-MHC-restricted [41,80]. Using this approach in a murine model, CTL have been generated against the mdm2 protein that is over-expressed in a large number of tumours [81]. Subsequently, allorestricted tumour-specific T cells were generated from normal human donors [82]. PBMCs from HLA-A2 negative donors were stimulated with APCs presenting synthetic peptides derived from cyclin-D1 in the context of HLA-A2. Cloning of bulk T cell cultures was used to isolate peptide-specific CTL lines that killed HLA-A2 positive breast cancer cells over-expressing cyclin-D1.

Similarly, HLA-A2-restricted CTL specific for the Wilm's tumour antigen-1 (WT1), over-expressed in most leukaemias, have been shown to inhibit leukaemic stem cells, while ignoring normal stem cells [83,84]. The allorestricted approach has also been used to produce CTL against the minor histocompatibility antigen HA1, for possible therapy of patients after HLA-mismatched SCT [76].

Allorestricted approaches allow isolation of high avidity antigen-specific CTL when the target antigen is an overexpressed self-antigen. There are limitations, however, which may reduce its applicability outside the context of HLA-mismatched allogeneic SCT. Allorestricted CTL generated from an HLA-mismatched donor are likely to be rejected rapidly by the recipient unless they are profoundly immunosuppressed. Therefore, the transfer of an allorestricted TCR into patient T cells provides a technology to overcome these limitations.

The feasibility of TCR transfer has been demonstrated recently using CTL specific for an mdm2 peptide presented by HLA-A2 molecules. To bypass self-tolerance to this tumour antigen the CTL were isolated from HLA-A2 transgenic mice and from A2-negative donors via the allorestricted strategy [85]. Murine HLA A2-restricted CTL were generated by immunizing the mice with a human mdm2 peptide epitope differing by a single amino acid to the naturally presented murine peptide. A high-avidity TCR was cloned from murine CTL, partially humanized and retrovirally transferred into human T cells to produce CTL capable of high-avidity killing of mdm2-expressing cells. Both avidity and antigen-specificity were reconstituted in the human T cells after retroviral transduction of TCR genes. Such therapeutic TCR transfer provides a method of rescuing the self-MHC-restricted human T cell repertoire by producing high-affinity, broad-spectrum tumour reactive TCRs that are usually deleted by tolerance mechanisms.

Alternative approaches of generating high-affinity receptors for tumour-associated peptide epitopes have been described recently. For example, in vitro mutagenesis followed by tetramer selection can be used to isolate TCRs with improved antigen-specific binding affinity [86]. This technology can be used to convert low-affinity TCRs specific for tumour-associated self-peptides into high-affinity TCRs and thus improve tumour cell recognition [87]. Phage display of single chain antibody fragments provides another technology for the isolation of antibody molecules that bind specifically to a given MHC/peptide complex, thus mimicking the binding specificity of TCRs. Recently, phage display was used to isolate a single chain antibody with high-affinity binding to a melanoma-associated MAGE-3 peptide presented by HLA-A1 class I molecules. The single chain antibody fragment was fused to the FcɛR1 gamma-chain and introduced into human T cells via retroviral gene transfer. The transduced T cells showed specific killing activity against HLA-A1⁺, MAGE-3⁺ melanoma cells, illustrating that the chimeric receptor can elicit antigen-specific effector function in human T cells [88].

In addition to receptor transfer into mature T cells, it is also possible to target the haematopoietic stem cell. It has been demonstrated recently that TCR transfer into murine haematopoietic stem cells produced mature T cells that responded to antigen-specific challenge in vivo[89]. A significant advantage of this approach is the potential to continually generate antigen-specific T cells from a self-renewing stem cell pool.

Together, several new technologies are now available to isolate high-affinity TCRs and TCR-like molecules to overcome immunological tolerance. This provides a platform to overcome one of the major limitations of autologous immune responses against tumours, namely low-avidity T cell responses against tumour-associated antigens.