One of the most controversial issues in immunology for over a century has been whether an effective immune response can be elicited against malignant tumours. Whether the immunology community has believed cancer immunotherapy is feasible or impossible has been largely determined by the prevailing immunological paradigms at that time. In fact, during the last 110 years it is possible to trace at least five dramatic fluctuations in attitude towards cancer immunotherapy. It now appears, however, that overwhelming evidence is available to support the view that both the innate and adaptive immune responses can recognize and eliminate tumours. On the other hand, it remains to be seen if these immune responses can be harnessed to control cancer as, at the time of diagnosis, many tumours have already been immunoselected to be highly resistant to immune elimination. Based on these observations it is argued that immunotherapy approaches, other than the generation of tumour-specific cytotoxic T lymphocytes, must be explored. Alternative strategies include recruiting tumouricidal myeloid cells into tumours, generating antiangiogenic immune responses and directing innate immunity to hypoxia-induced ligands on tumour cells.
One of the most controversial questions in immunology for over a century is ‘Can the immune system recognize and eliminate malignant tumours?’ Whether the answer to this question has been in the negative or the affirmative has been largely dependent on prevailing immunological theories at the time. In this review I will trace the changing attitudes to cancer immunotherapy during the last 100 years, discuss current attitudes to the problem and describe future approaches that may overcome some of the difficulties associated with immunotherapy of malignancy.
Cancer immunotherapy: a historical overview
Table 1 summarizes the extraordinary fluctuations in attitudes towards cancer immunotherapy during the last 110 years. The first indications that the immune system might respond to cancerous tissue actually appeared in the eighteenth century when it was noted that feverish infections in cancer patients were occasionally associated with cancer remission.1 In the 1890s William Coley (1862–1936), a leading New York surgeon, began to seriously investigate this phenomenon. Coley's interest in the area stemmed from a cancer patient who had a complete remission of their cancer following two attacks of erysipelas caused by acute infection with the bacteria Streptococcus pyogenes. Subsequently Coley injected streptococcal cultures provided by Robert Koch into cancer patients and observed tumour regression in some cases. His findings were published in 1893, with his paper being the first that describes a serious attempt at cancer immunotherapy.2
During the next 43 years Coley treated almost 900 cancer patients with his bacterial preparation which became known as ‘Coley's toxin’. Most of the treated patients had inoperable sarcomas, with the bacterial toxin achieving a cure rate of over 10%.1 Despite these successes ‘Coley's toxin’ was not widely accepted by the scientific and clinical communities of the time, possibly due to the severe fever induced by the treatment and the perceived low cure rates. Also early studies suggesting that it was possible to immunize against transplantable tumours3 fell into disrepute when it was realized that, due to the outbred nature of the animals being used, the putative tumour-specific immune responses observed were actually against antigens expressed on normal tissues.4 In fact, with no theoretical framework suggesting that tumours could be rejected (Table 1), the general feeling amongst immunologists was that it would be impossible for the immune system to recognize and respond to malignant cells. Woglom expressed this view in dramatic terms in a review in 1929 by stating that ‘It would be as difficult to reject the right ear and leave the left ear intact as it is to immunize against cancer’!4 Nevertheless, Coley's early studies led to the use of bacille Calmette-Guérin (BCG) for cancer immunotherapy, with this treatment being used to the present day as the most effective therapy against superficial bladder cancer.5, 6
In 1949 Burnet published his theory of acquired immunological tolerance which proposed that lymphocytes that were able to respond to self tissues were deleted in prenatal life during the development of the immune system.7 This theoretical model was verified experimentally by Medawar and colleagues soon afterwards.8 Such a model clearly reinforced the view that the immune system is incapable of responding to malignant cells as it was assumed that transformed cells are indistinguishable from healthy self tissues. During the 1950s it became apparent, however, that in the case of transplantable tumours induced by various carcinogens it is possible to immunize syngeneic animals against their tumours.9, 10, 11, 12 The classic experimental approach was to excise a tumour from an animal and show that, in many cases, the animals could reject a second injection of the same tumour cells. Such experiments implied that there must be antigens associated with tumour cells that can be recognized by the immune system, these antigens being often called tumour-associated antigens (TAA) or tumour-specific transplantation antigens (TSTA).
By the 1960s immunological opinion began to clearly swing in favour of cancer immunotherapy. Ironically, Burnet played an important role in this change in attitude. Whereas Burnet's theory of immunological tolerance had convinced immunologists in the 1950s that immune recognition of tumours was unlikely, in the 1960s Burnet began to champion the view that a major function of the immune system is to eliminate malignant cells (Figure 1).13 Thomas proposed a similar hypothesis about the same time, suggesting that homograft rejection is an aberrant example of processes that normally eliminate tumours.14 In contrast, Burnet provided a more general hypothesis, suggesting that lymphocytes are continually patrolling tissues and eliminating transformed cells, presumably via recognition of TAA, a process he termed ‘immunosurveillance’.13 During this period the search for TAA also began in earnest. It is interesting to note that tumour immunosurveillance had been proposed, albeit in a very primitive form, by Ehrlich in 1909 when he suggested that cancers may occur at an ‘overwhelming frequency’ if it were not for the immune system,3 although the idea appears to have been not adopted by his peers.
The favourable attitude towards cancer immunotherapy that emerged in the 1960s was short lived, with there essentially being an abandonment of Burnett's immunosurveillance concept from the early 1970s to the mid-1990s. There were a number of reasons for this change (Table 1). First, on purely theoretical grounds, it was thought to be highly unlikely that the immune system has evolved to recognize and reject malignant tumours. It was argued that acute infections are much more life threatening, particularly for young members of a species, and thus provide a much stronger selective pressure for the evolution of the immune system. Second, experimental evidence began to emerge that did not support the immunosurveillance hypothesis. The most telling was the observation that T cell deficient athymic nude mice have a comparable incidence of primary tumours as syngeneic wild type mice.15, 16, 17 It was only some years later that the validity of these experiments was questioned when it was realized that athymic nude mice still possess a significant population of functional T cells.18 Third, the process of thymic selection of T cells began to be delineated.19, 20 Based on these studies it became clear that thymic deletion of auto-reactive T cells is extremely efficient. Such data were consistent with Burnet's theory of immunological tolerance and resulted in a return to the 1950s view that few, if any, tumour-specific lymphocytes are present in the periphery. Finally, it was proposed that the rejection of transplanted tumours is artefactual as it represents an aberrant immune response to tumour-associated viruses. In this regard a particularly influential paper was published in 1976 where it was claimed that, unlike chemically induced transplantable tumours, spontaneously arising tumours are not recognized by the immune system.21 In this paper the authors state that, based on isotransplants of 27 different spontaneously occurring tumour types, no evidence of spontaneous tumour regression was observed, despite approximately 20 000 tumour transplants being performed. I distinctly recall that by 1980 cancer immunotherapy was generally regarded as an approach with little or no chance of success.
Remarkably, by the mid-1980s the tide of opinion began to move in the opposite direction (Table 1). Initially it became apparent that some auto-reactive T cells can escape thymic deletion, raising the possibility that some self-reactive T cells in the periphery could be recruited to respond to transformed self cells.22, 23, 24 The successful characterization of relatively immunogenic TAA in mice and humans25, 26, 27, 28, 29 further supported the concept that cancer immunosurveillance may occur. In my opinion, however, the single most powerful piece of evidence in favour of tumour-specific immunity was the demonstration that highly malignant tumour cells are profoundly genetically unstable.30, 31 In fact, as a result of genomic instability it has been estimated that a single carcinoma can exhibit 11 000 genomic alterations!31 As a result of this genetic instability large numbers of novel TAA are generated that have never been seen by the host's immune system. Under these circumstances a lack of tumour-specific T cells is no longer an issue. In fact, by the early 1990s the tumour immunotherapy debate had gradually shifted to other areas, notably the inability of tumour cells to initiate an immune response due to their lack of costimulator molecules.32, 33, 34, 35, 36, 37 It was suggested that this lack of costimulation would result in ‘peripheral T cell tolerance’ being induced by the tumour cells.
Since 1995 the evidence in favour of effective tumour-specific immunity has become increasingly compelling (Table 1). There have been a large number of studies since 1995 that indicate that dendritic cells (DC), when appropriately activated and induced to present tumour-derived peptides, can very effectively elicit tumour-specific T cell immunity. Following successful preclinical studies38, 39, 40 a large number of pilot trials have been performed in patients with a wide range of cancer types.41, 42, 43, 44 Many of these trials have demonstrated the induction of antitumour immune responses with clinical responses in some cases. Further support for the old immunosurveillance hypothesis has come from analyses of cancer incidence in a range of immunodeficient knock out mice. For example, recombinase-activating gene (RAG−/−),45 STAT1−/−,45 perforin−/−,46, 47 IFN-γ−/−48 and IFNR-γ−/−45,49 mice have all been shown to have higher incidences of carcinogen-induced tumours and, in some cases, spontaneous epithelial carcinomas. Recent studies have also unequivocally shown that the innate immune system plays a key role in tumour immunosurveillance, with NK cells,50, 51, 52, 53 NKT cells50, 51, 53 and γδT cells53, 54 being implicated. Perhaps the most crucial conclusion from all these studies is that many of the tumours that emerge in immunocompetent animals have been selected to evade the host's immune system. There is no doubt that recent research in the field has resulted in a renaissance of the tumour immunosurveillance hypothesis.53
Cancer immunotherapy: The current state of affairs
Table 2 summarizes some of the central aspects of our current understanding of tumour immunity. Based on recent immunological developments discussed earlier there is no doubt that the immune system can recognize and eliminate malignant cells. In fact, it is now generally accepted that tumour immunosurveillance occurs, with the ‘successful’ tumours that emerge in cancer patients having been immunoselected to evade immune elimination. It is also abundantly evident that both the innate and adaptive immune responses play a role in tumour clearance. The emergence of innate immunity as a potent antitumour response is, in hindsight, not that surprising. It has been estimated that only approximately 1.4% of all multicellular animal species on this planet posses an adaptive immune system.55 Thus if the immune system is to control tumour development in the 98.6% of animal species that lack adaptive immunity it stands to reason that the innate response must be involved. Another interesting implication of the involvement of innate immunity in cancer clearance is that it provides an explanation for the antitumour properties of Coley's toxin and has resulted in a reassessment of Coley's approach to cancer immunotherapy.1, 56 In this regard, it is now known that bacterial DNA is immunostimulatory via bacterial specific CpG motifs that are recognized by the innate immune system.57, 58, 59, 60, 61 It appears highly likely that these motifs may have contributed to the therapeutic effects of Coley's bacterial extracts. Similarly the array of Toll-like receptors that can recognize several unique molecular patterns associated with bacteria and activate immune responses may also be involved.60, 61
Another important point is that immune recognition of malignant tumours is not unique but involves the same immunological processes that are used to combat pathogens.50 This revelation resolves the paradox of the tumour immunosurveillance hypothesis discussed earlier, namely that it is hard to imagine how the immune system could have evolved to eliminate tumours when pathogens would have provided much stronger selective pressures.
One paramount question that remains to be answered is how are tumours recognized as abnormal by the immune system and eliminated? Unlike pathogens, which can provide ‘danger’ signals to the immune system via pathogen recognition receptors,24, 62, 63, 64, 65 tumour cells are unable to provide such signals.66 Recent studies indicate that a number of stress-induced molecules, that structurally resemble MHC class I molecules (i.e. in humans MICA/B and the ULBP family and in mice H60 and the RAE-1 family), are expressed on the surface of virus infected and heat-shock treated cells.52, 67, 68, 69, 70, 71, 72 The NKG2D receptor has been shown to recognize these stress-induced molecules and provide critical costimulatory signals to NK cells, γδT cells and CD8+ T cells, these signals being able to trigger cell-mediated cytotoxicity.52, 67, 68, 69, 70, 71, 72 The presence of NKG2D on CD8+ T cells is particularly intriguing as it raises the possibility that T cells with low affinity for self antigens can effectively recognize stressed self cells via stress-induced ligands for NKG2D. It is assumed that the same recognition processes operate for virus-infected cells and tumour cells. But the questions still remains what are the stimuli that induce the expression of the stress-induced molecules? In the case of tumour cells I would like to suggest that hypoxia may be an important stimulus. Figure 2 depicts the remarkable network of blood vessels that exists in a healthy tissue, in this case adipose tissue. It is well documented that such a vascular network is essential to supply tissues with oxygen and nutrients and to remove toxic byproducts of metabolism. In the case of a solid tumour, due to the lack of a vascular network the microenvironment of the tumour rapidly becomes hypoxic, a situation that is only partially relieved by the induction of angiogenesis.73, 74, 75, 76 Thus hypoxia appears a likely stimulus for the expression of NKG2D ligands on the surface of tumour cells and would also provide an explanation for the selective recognition of tumours by both the innate and adaptive immune systems.
Cancer immunotherapy: Future directions
As discussed earlier in this review, there is now ample evidence that many tumours express antigens that can be recognized by the adaptive immune system and which potentially can be used to induce an antitumour immune response. Currently the most popular approach to cancer immunotherapy has been the generation of CD8+ CTL that recognize the tumour antigens in association with MHC class I molecules on tumour cells. Unfortunately, as outlined in Table 3, there are a number of problems associated with this approach. First, one would anticipate that there will be a low CTL precursor frequency for most tumour antigens.29 Furthermore, since patients receiving cancer immunotherapy will usually have a substantial tumour mass that needs to be cleared by the immune system it may be difficult to generate adequate numbers of effector T cells to eliminate the tumour. Second, the tumour-specific CTL must efficiently localize within the tumour and eliminate the malignant cells via direct cell contact and delivery of cytotoxic signals.29 This may often be difficult to achieve, particularly in a large tumour mass that is poorly vascularized. Third, there is mounting evidence to suggest that many tumours have already been immunoselected to be CTL resistant by the time the cancer is first diagnosed.77, 78 In fact, the much greater frequency of tumours in many immunodeficient mice45, 46, 47, 48, 49, 50, 51 supports the view that immunoselection is a common feature of tumour progression. Mechanisms of immune evasion include down-regulation of MHC molecules or the tumour antigens themselves by the tumour,77, 78 the production of immunosuppressive cytokines such as IL-10 and TGF-β79 and, more controversially, the elimination of tumour-specific CTL by the expression of the apopotosis-inducing molecule, FasL, on tumour cells.80 Furthermore, there is every likelihood that even if a tumour is CTL sensitive at the commencement of immunotherapy there will be rapid selection of tumour variants that are CTL resistant. Fourth, as with all tumour immunotherapy approaches, there is the potential of inducing inadvertent autoimmunity, as usually the tumour immune response will be directed against abnormally expressed or slightly modified self antigens. In this regard antimelanoma immunity has been sometimes associated with vitiligo81 and animal models of tumour immunity have been shown to produce severe autoimmune disease.82 Finally, with DC based vaccines being a popular means of generating tumour-specific CTL, an additional problem is the use of appropriately activated DC for CTL immunotherapy. There is a wealth of evidence to indicate that DC, depending on their activation and maturation state, can be either tolerogenic or immunogenic.83, 84, 85 Identifying and targeting DC that are appropriate for the induction of optimum tumour immunity is a major challenge in the future.
With the substantial number of problems associated with CTL-based immunotherapy I believe it is time to seriously consider other cancer immunotherapy strategies that may overcome some of the inherent drawbacks of tumour-specific CTL. An obvious approach is to harness the innate immune response in a much more controlled manner than by William Coley. A dramatic example of this is the use of the glycolipid, α-galactosylceramide, to polyclonally activate NKT cells and induce tumour rejection responses.86, 87 Clearly this form of cancer immunotherapy is in its infancy with much to be learned. Of particular interest will be whether the innate immune system can be used to target putative stress-induced ligands that are expressed by tumour cells and not by ‘normal’ cells. One would imagine that this strategy would overcome the problems of autoimmunity and the low frequency of tumour-specific effector cells associated with conventional adaptive immune responses.
Another approach, which my laboratory has been pursuing, is to generate tumour-specific CD4+ T cells. The use of CD4+ T cells to directly eliminate tumours has been largely ignored in the past, with these cells being thought to play only a secondary role in tumour clearance by providing ‘help’ for the development of optimum numbers of tumour-specific CD8+ CTL.88, 89, 90 Nevertheless, there are reports suggesting that CD4+ T cells can directly eliminate tumours, without the participation of CTL.91, 92 There are potentially a number of advantages associated with the use of tumour-specific CD4+ T cells, these advantages being listed in Table 4. The most obvious advantage is that, since TAA can be presented by bystander APC to CD4+ T cells, these T cells can still respond to tumours that lack MHC molecules and are resistant to CTL lysis. Unlike CTL the CD4+ T cells can also indirectly eliminate tumours by the recruitment of tumouricidal myeloid cells into the tumours,91, 93, 94 or by the secretion of antiangiogenic cytokines.95 Such an effect greatly amplifies the local antitumour activity of the CD4+ T cells. Furthermore, this form of tumour immunity would still be effective against subpopulations of cells in a solid tumour that lack the antigen against which the antitumour response is directed. Thus the emergence of tumour variants that are resistant to immune elimination would be less likely to occur. Of course, the effectiveness of this type of immune response to tumours may be dependent on the availability of appropriate APC within tumours or in the vicinity of tumours, with recent studies suggesting that DC within tumours may be poor antigen presenters.44
Recently my laboratory has examined the ability of tumour-specific CD4+ T cells, that have been polarized to produce a cytokine profile characteristic of Th1 or Th2 cells, to eliminate a B16 melanoma that is highly CTL resistant.96 It was found that the tumour-specific CD4+ Th2 cells very effectively cleared established lung and visceral metastases of this tumour, whereas the tumour-specific Th1 cells were ineffectual. Histological analyses revealed that the tumour-specific Th2 cells were recruiting large numbers of eosinophils into the tumours, with degranulating eosinophils appearing to rapidly eliminate the tumours. In fact, a series of knockout mice demonstrated that tumour clearance was totally dependent on the chemokine, eotaxin, which recruits eosinophils into tissues.97 Paradoxically, the tumour-specific Th1 cells, despite recruiting considerable numbers of macrophages into the tumours, had no effect on tumour growth. This finding may be explained by the observation that tumour infiltrating macrophages, rather than being tumouricidal, are often pro-angiogenic and enhance tumour growth.98, 99 In fact, clinical studies have shown that a high number of tumour infiltrating macrophages in the tumours of melanoma and breast cancer patients is associated with a poor prognosis.99, 100 Collectively these data suggest that recruitment of eosinophils into tumours by tumour-specific CD4+ Th2 cells is a viable approach to cancer immunotherapy, particularly against tumours that are resistant to CTL attack.
In this short review I have attempted to highlight the changing attitudes to cancer immunotherapy during the last 100 years. Based on the evidence currently available it now appears likely that both the adaptive and innate immune systems can recognize and eliminate tumours. The problem we face, however, is that immunoselection continually occurs during tumour development with the malignant tumours that eventually emerge in patients often being resistant to elimination by the immune system. The challenge for tumour immunologists in the future is identifying patients for which immunotherapy will be efficacious101 and devising means of reactivating the immune response against tumours.