Introduction

The recent renaissance of interest in the idea of immunosurveillance against tumours has arisen as a result of new data requiring reexamination of the reasons for rejecting this theory over the past couple of decades [38]. The evidence against immunosurveillance was derived largely from an inability to detect remarkable differences in cancer occurrence in athymic nude mice compared to their normal wild-type equivalents [146] and because cancer incidence in long-term immunosuppressed organ transplant recipients was thought either not to be increased or to be limited to cancers with a viral etiology. However, athymic mice are not completely devoid of functional T cells [66], and with longer follow-up, a larger range of cancers does appear at increased frequency in transplant recipients [49, 82]. Increasing awareness of the latter among transplant physicians is leading to routinisation of specialist cancer screening follow-up in such programs (for recent examples, see [7, 155, 170]).

For many years now, clinical vaccination trials have sought to trigger or enhance antitumour immunity, but always with rather disappointing results. Many explanations (apart from the nonexistence of anticancer immunity) could account for the unimpressive success rates, from the classical concept of immunoselective pressure giving rise to resistant variants, to the more recent realisation that tumour-induced alterations to the patients’ immune system may subvert anticancer responses. These mechanisms may be classified into the following major groups:

1):

alteration of MHC class I and tumour antigen expression

2):

dysregulated expression of adhesion / accessory molecules by tumour and/or antigen-presenting cells

3):

secretion of immunosuppressive soluble factors either by tumour cells or infiltrating T cells or both

4):

induction of immune nonresponsiveness via anergy induction or clonal deletion of responding T cells

5):

induction of suppressor cells

6):

changes in T-cell signal transduction molecules

7):

tumour utilisation of products of stimulated leukocytes, i.e. immunostimulation of cancer.

The following sections provide brief examples of some key findings that have contributed to our current understanding in each of these categories and, where appropriate, directions to recent reviews on these topics, as well as indications for remediation.

Alteration of MHC and tumour-antigen expression

MHC class I–restricted cytolytic T cells (CTLs) are thought to be major effectors of anticancer immunity. Loss of HLA antigens from tumour cells would therefore prevent their recognition and lysis by such CTLs. There is a lot of data documenting that down-regulation of HLA-A and HLA-B alleles is common and clinically important [47, 76], with reduced levels of class I expression predicting clinical outcome [69]. Even total loss of HLA expression is not uncommon [16]. Several mechanisms are responsible, but total loss of class I expression is usually a result of deletion of β2-microglobulin [108]. Another common reason for decreased class I expression is loss of peptide transporter function, usually regulatory rather than deletional and therefore susceptible to remedial manipulation, e.g. with cytokines, as shown for TAP-1 [136]. Down-regulation of MHC antigen by tumour cells is a powerful strategy to avoid killing by MHC-restricted CTLs; accordingly, poor prognosis has been reported to result from HLA loss [4]. However, tumours lacking MHC class I expression might be expected to become more susceptible to immunotherapy based on NK cells. This phenomenon may underlie reports inconsistent with the above paradigm which suggest that HLA loss may sometimes predict better not worse survival [94], or even the extreme example of uveal melanoma where high HLA expression is seen and correlates with metastatic spread and poor prognosis [67].

In addition to down-regulation of “classical” HLA antigens, tumour cells may up-regulate expression of MHC molecules such as HLA-E, which ligate inhibitory receptors such as CD94/NKG2A that are expressed both by NK cells and CTLs. The expression of these NKIR is up-regulated by cytokines such as IL-15 and TGF-β, both of which can be produced by tumour cells [9, 54]. Moreover, even cytokines such as IFN-γ may act as double-edged swords via up-regulation of NKIR ligands on tumour cells via up-regulation of HLA-G [89]. Consistent with this, there is also a correlation between higher serum levels of soluble HLA-G in melanoma patients and advanced stage of disease and tumour load [154]. For recent reviews on the role of NKIR, see [20, 26].

Loss of tumour-antigen expression may not occur infrequently compared with the loss of the HLA molecule presenting tumour peptide, but it has been difficult to document. Experiments in mice have shown that immunoselection against dominant tumour antigens often but not always results in reduced MHC class I expression [36]. Powerful circumstantial evidence for similar phenomena in humans derive from an extensive clinical study examining 532 melanoma lesions from 204 patients after vaccination with gp100 peptide. The frequency of lesions highly expressing gp100 significantly decreased after therapy whereas the expression of MART-1 was essentially unchanged [126]. HLA loss cannot strictly be excluded but would require that gp100 production was reduced by HLA loss, which is unlikely. More recently, a range of mutations in the PA-1 target antigen in mice in the absence of MHC loss has been documented as a major mechanism accounting for tumour escape after adoptive immunotherapy [8].

The emergence of tumour variants under selective pressure of a specific immune response reflects the usual acquisition of therapy resistance resulting in loss of susceptibility to therapy. A common approach to alleviate this is to increase the dose of therapeutic agent. This may be more feasible for adoptive immunotherapy than chemotherapy and stem cell transplantation. Indeed, in some models, increasing the number of effector cells can destroy tumour cells before they have time to evolve resistant variants [91] – here is therapy as a race against time. In this context, either in vivo, or in in vitro propagation of large numbers of T cells for use in adoptive immunotherapy, a further problem arises: cell senescence caused by extensive replication of somatic cells (for reviews, see [39, 112]).

Expression of adhesion or costimulatory molecules by tumour and/or dendritic cells

Tumour-antigen presentation by dendritic cells in the context of appropriate costimulation is critical for eliciting CTLs, and accessory molecule expression by tumour cells is important in their susceptibility to such CTLs. Both DC and tumour cell adhesion/accessory/costimulatory molecule expression is dysregulated in cancer and contributes materially to tumour escape. Early work showed that tumour-infiltrating DCs were strongly MHC class II+ but failed to express the important costimulatory molecules CD80 or CD86 (thereby inducing T-cell anergy), whereas DCs in inflammatory infiltrates such as in Crohn’s disease lesions were highly functional and all CD80/86+ [23]. Failed attempts to normalise DC function in cancer have included treatment with Flt3-ligand [40], and use of neutralising antisera against IL-10, VEGF, TGF-β or PGE2 [72]. In other model systems in which DCs progressively lost MHC class II expression, Flt3L treatment also proved ineffective at restoring DC integrity [27].

Effector cells, once generated, must interact with their target cells, initially via antigen nonspecific adhesive mechanisms. Tumour cells of different histologies frequently show relatively decreased levels of important adhesion molecules, such as ICAM-1 [158], which may have functional consequences [45]. IFN-γ treatment can up-regulate ICAM-1 on colon carcinoma cells and increase their lytic susceptibility [12]. Cancer cells may also lack expression of other important costimulatory molecules, such as CD40 [62], the absence of which on epidermal tumours has been suggested to facilitate escape [156]. Von Leoprechting et al. reported that advanced stages were CD40− whereas primary tumours and even metastases were CD40+ [157]. Moreover, CD40 ligation on melanoma cells enhanced their susceptibility to lysis by Melan-A/MART-1–specific CTLs, so loss of CD40 expression would prevent this and contribute to escape [157]. CD40 expression may be associated with a more favourable prognosis in some other tumours as well, e.g. diffuse large B-cell lymphoma [83].

As with NKIR mentioned above, costimulatory family receptor/ligand pairs are also present not only as positive but also as negative regulatory effectors. Recent awareness of the widespread expression of ligands such as PD-1L on several different types of cancer and the negative effects that they mediate on antitumour T cells [34] may make these molecules critical targets for immunoregulation therapy. Again, care must be exercised here; commonly employed immunomodulatory cytokines such as IFN-γ increase the level of PD-1L expression [34]. Undoubtedly, yet more negative receptors remain to be discovered [160]; complete knowledge of all such possible interactions might be beneficial in manipulating responses in the desired direction.

Secretion of immunosuppressive substances

It has been known for many years that sera of cancer patients can contain an impressive variety of immunosuppressive proteins, ranging from acute phase reactants with nonspecific inhibitory properties, to adhesion molecules blocking cell interactions or apoptosis. Reactivity not only to tumour but also to nontumour antigens may be depressed in cancer patients and contribute to their increased susceptibility to infection. Soluble forms of adhesion molecules such as CD54, CD58, and others may correlate with disease progression, as has been suggested for plasma sCD54 levels [53, 132]. Gangliosides may be inhibitory at the level of antigen-presenting cell function [113, 138]. Serum levels of soluble Fas may also contribute to tumour escape [13, 68, 153]. Molecules commonly overexpressed in tumour cells, such as MUC-1 and MUC-2, may also be immunosuppressive for T cells in soluble form [73] and associate with poor survival and poor anticancer responses in patients on immunotherapy [87]. Annexin II, overexpressed in several tumours, may also inhibit T-cell proliferation [1]. The human neutrophil proteins known as “defensins” may also fall into this category [59]. Tumours can also exert nonspecific suppressive activity, e.g. by secreting adenosine as a result of their hypoxic metabolism. Adenosine can inhibit IL-12 and stimulate IL-10 production by monocytes, contributing to these suppressive effects [84]. Another simple substance possibly functioning in this way may be tryptophan, secreted by tumour-associated macrophages (for review, see [93]).

Many cytokines secreted by either tumour, immune system or both, can exert immunosuppressive effects. The best known of these are probably TGF-β and IL-10, but a whole range of others, including those most commonly thought of as immunostimulatory, may also have this effect. IL-10 has been shown to hinder a number of immune functions, i.e. T-lymphocyte proliferation, TH1-type cytokine production, antigen presentation, and lymphokine-activated killer cell cytotoxicity. Elevated levels of IL-10 concentrations have been found in patients with various solid tumours, as well as haematological malignancies [41, 133] and may have prognostic significance in a variety of cancers [31]. Many negative effects of IL-10 on the host immune system have been described, including inhibition of proinflammatory cytokines, and down-regulation of both the initiation and effector phase of inflammatory and delayed-type hypersensitivity responses in vivo (for review, see [96]). Kim et al. [73] described secretion of IL-10 by carcinoma cells and showed that intralesional treatment with IFN-α induced tumour regression, associated with down-regulation of IL-10 mRNA. We have shown that CML cells spontaneously secrete large amounts of IL-10 ex vivo and that IFN-α acts to decrease this while increasing IL-1β secretion without altering TNF-α [111]. IL-10 might therefore play a central role as one of the mechanisms responsible for immune dysregulation in cancer patients. Taken together, the majority of reports suggests that when tumours or/and tumor-infiltrating lymphocytes (TILs) express higher levels of IL-10 (and TGF-β), this mostly results in deleterious immunosuppressive effects.

There is also evidence that other cytokines, such as circulating IL-6, are associated with worse survival and greater extent of disease [130]. IL-6 production could contribute to peripheral T lymphocyte dysfunction, enabling tumour cells to escape immune surveillance by preventing the antitumour TH1 immune responses [42]. In lung cancer patients, levels of serum IL-6 are greater even than in patients with chronic obstructive pulmonary disease and acute infection; thus it is unlikely that the increases in IL-6 reflect merely a systemic inflammatory response [35]. In melanoma, patients responding to therapy showed a serum IL-6 level twice that of controls, whereas in nonresponders this factor was 11-fold, suggesting a strong correlation between IL-6 level and clinical status [98]. In breast cancer, patients with more metastases and patients refractory to therapy had higher levels of IL-6 in their serum; they also had poorer survival such that multivariate analysis showed that IL-6 and disease-free interval were the major prognostic factors [172]. On the other hand, the prognostic relevance of single cytokine markers may vary with the type of tumour examined: for example, in hepatocellular carcinoma, it has been reported that serum IL-10 levels but not serum IL-6 levels predict clinical outcome after resection [22]. It is therefore, and is likely to remain, an extremely difficult task to synthesise available data into a common model. Given this complexity, a unifying hypothesis has great attraction; one such is the rule of thumb that TH1 predominance equates with host-mediated tumour rejection, but that TH2 predominance inhibits this process. Although this may end up to be the case more often than not, it is obviously not a universal finding under all circumstances. For example, IL-4–transduced tumour vaccines trigger type 2 polarisation in both CD4 and CD8 cells; the CD8 cells are instrumental in rejecting the tumours, in an IL-4–dependent fashion involving other CD8 cells and probably also NK cells [127]. There is evidence that IL-10 can also enhance tumour rejection [125], elicit cytotoxic immune memory due to the combined action of NK cells, CD8+ T cells and neutrophils [50], and, in conjunction with CD80-CD28 costimulation, can prime tumour-reactive CTLs [166]. This is consistent with the ability of IL-10 to prevent T-cell apoptosis [110, 147]. Nevertheless, particularly for IL-10, many negative effects on the host immune system have been described, including inhibition of proinflammatory cytokine production by macrophages [32], and down-modulation of both the initiation and effector phase of inflammatory and delayed-type hypersensitivity responses in vivo [117]. Moreover, DCs exposed to IL-10 may induce anergy in peptide-specific antitumour CTLs instead of activating them [142]. IL-10–pretreated DCs also tend to prime IL-4–secreting T cells, perhaps by default due to the down-regulation of IL-12 production [85] and lower levels of expression of costimulatory molecules [18]. In addition, IL-10–exposed DCs are more susceptible to lysis by autologous NK cells [18], which would decrease antigen-presenting capability but might also help to prevent anergy induction. Recently the mechanism responsible for the decreased MHC class I presentation in the mouse lymphoma RMA and the mastocytoma P815 was shown to involve a down-regulation of the expression of the TAP-1/2 proteins and of their functions in the IL-10–expressing tumours [129]. An intriguing possibility is therefore that IL-10 might be one of the mechanisms responsible for the finding that TAP-1/2 expression is frequently turned off in human tumours (for review, see [135]), promoting their escape from tumour-specific CTLs, as discussed above. Zeidler et al. [171] confirmed that cellular and viral IL-10 affects antigen presentation and MHC class I expression in EBV-infected human B lymphocytes through its ability to reduce TAP-1 expression. Since not only TH2 cells but also many tumours themselves produce IL-10, this could contribute to immunosuppressive effects and the class I down-regulation (also class II down-regulation) commonly seen in tumour cells.

Considering the complexity and multitude of effects that IL-10 exerts on the immune system, as reviewed above, it could be asked what the net effect of a systemic overexpression of this cytokine on host antitumour surveillance will be. As transgenic mice expressing IL-10 under the control of the IL-2 promoter were found to be unable to limit the growth of immunogenic tumours [56], this is consistent with the argument made above that a deleterious effect of this cytokine on the host resistance to tumour growth will predominate in vivo. On the other hand, IL-10 has been shown to exert antiangiogenic and antimetastatic effects in certain murine models [64], so that it is remains difficult to dissect out the contradictory activities of this cytokine in tumour immunology.

Inducing immune nonresponsiveness in TILs

T-cell destruction

Perhaps the most direct example of tumour escape by induction of unresponsiveness in T cells is the finding that secreted protein from tumours may be presented in the thymus and cause clonal deletion of newly generated T cells, according to the usual paradigm of central tolerance induction [80]. Another possible tumour escape mechanism involving clonal deletion, but this time peripheral deletion, relies on the fas/fas-ligand pathway. Shortly after activation, T cells begin to express fas (CD95). Some time, but not immediately thereafter, CD95+ T cells acquire susceptibility to fas-mediated cell death. Therefore, under certain conditions remaining to be precisely defined, the interaction of fas-ligand with CD95 can induce T-cell apoptosis but other outcomes are also possible. The nonapoptotic consequences of fas signaling may have been relatively overlooked thus far (for review, see [159]). Many types of tumour have been reported to express fas ligand following the initial publication in 1996 [57] (for review, see [123]), although this is controversial [21]. It was proposed [124] that contrary to the prevailing view that tumour cells cause the death of antitumour T cells by expressing FasL, the FasL is in fact expressed by T lymphocytes upon activation after tumour cell recognition, causing them to kill themselves (“suicide”) and each other (“fratricide”) through the same caspase-based mechanism. Some of the disparities in the literature may perhaps be due to variation in tumour stage, within lesions or after selection. For example, in gastric carcinoma it has been argued that fas-ligand positivity of metastatic but not primary tumour is functional, in terms of inducing apoptosis of TILs [78]. In many other reports, the fas ligand expressed was shown to be functional, i.e. its ligation of fas resulted in apoptosis of the target cell: for example, in breast and cervical cancer [28, 99]. In esophageal cancer, up-regulation of fas ligand and down-regulation of fas was found to be an early indicator of progression [52]; in this cancer, fas ligand was also found to be functional in triggering apoptosis of the TILs [11]. Longitudinal studies in melanoma revealed that fas-ligand expression by primary tumour was weak and rare, whereas in metastases, including those of these primary tumours, fas-ligand expression was stronger and commoner [148]. In gastric carcinoma, a correlation between tumour fas-ligand positivity, lymph node metastasis, and level of apoptosis in TILs was found [100]. However, in oral and oropharyngeal squamous cell carcinoma, no correlation was reported between fas or fas-ligand expression by the tumour and clinicopathological factors. Nonetheless, there was a correlation between fas-ligand expression and IL-10 (and GM-CSF) expression [44]. Direct evidence of a role for fas ligand in tumour escape is nonetheless hard to come by; in one attempt intratumourally transfected fas ligand antisense did result in reduction of tumour growth and metastasis, but the exact mechanism was not detailed [104]. Animal models demonstrate a potential clinical impact of these types of findings, where regressor and progressor variants of the same tumour can be distinguished by their level of fas-ligand expression (and MHC class II expression, which may induce anergy) [15]. These important facets of tumour-host interaction continue to be discussed in a lively manner [152, 163].

T-cell anergy

The induction of T-cell nonresponsiveness without destroying them and thereby possibly triggering compensatory responses on the part of the host probably contributes to a large degree to tumour escape. Even highly effective antigen-presenting cells such as DCs can be subverted by tumour products, eg. IL-10, to anergise rather than activate antitumour cells [142, 143], as mentioned above. Anergy induction is antigen-specific and is an early event associated with tumour progression [141]. Naturally occurring peptide sequences from endogenous as well as foreign proteins can act as partial agonists for the melanoma antigen MART-1/Melan-A (27–35) and anergise antitumour T cells by cross-presentation [86]. Moreover, the presence of such anergy-inducing peptides on the melanoma surface can prevent T-cell activation by immunodominant peptides [19].

On the other hand, recent investigations using HLA-A2/Melan-A (27–35) tetramers to visualise antigen-specific T cells revealed that in patients where antigen-specific cells expressed a CCR7+CD45RO− (“naïve”) phenotype, there was a lack of response to the peptide in vitro. In contrast, where the tetramer+ cells were CCR7−CD45RO+ there was a response [37]. This rather suggests lack of activation of antigen-specific cells in nonresponders, not activation and anergy induction. However, potentially dangerous for immunotherapy, under certain circumstances, anergy can also be induced by vaccination with immunogenic peptides representing tumour antigens [150]. This may be prevented by engineering stimulation via CD40-CD40L interactions [33], or other costimuli [164].

The in vivo relevance of these mechanisms in a clinical context has recently become susceptible to analysis by employing tetramer technology to identify tumour antigen–specific T cells in patients. Thus, the use of soluble tetramer/peptide complexes between HLA-A2 and common melanoma antigens such as tyrosinase (368–376) allow the direct demonstration of clonally expanded antigen-specific CD8 cells in melanoma patients. These effector cells were demonstrated to be anergic, being unable to lyse target cells or secrete cytokines on activation [81]. Similar clonal expansions have been found under other conditions of chronic antigenic stress, especially in aging where the driving antigenic force is likely to be persistent herpes virus infection [70, 106, 107].

Induction of suppressor cells

The recent renaissance of interest in suppressor cells has lead to the reinterpretation of many older data, which in the meantime had been dismissed as artifactual (for review, see [97]). The realisation that CD4+CD25+ regulatory cells play an important role in many aspects of immunological tolerance has led to attention being focused on such cells also in cancer (for review, see [128]). Potential treatment modalities may be developed following this realisation. In an animal model, injection of CD25 mAb preferentially depletes CD25+CD4 cells and can prevent tumour progression [105]. Consistent with these data, Shimizu et al. reported that unresponsiveness to a variety of tumours in mice can be prevented by removing “naturally” activated (i.e. CD25+) CD4+ cells [139]. The remaining CD4+CD25− cells were found also to proliferate to MHC class II+ self-peptides on autologous APCs, suggesting that tolerance to self, including tumours, had been broken.

TH2-type cells can also be considered to have suppressive functions in that they secrete IL-4 and IL-10 as potentially down-regulatory cytokines. Such cells have also been identified among melanoma TILs, and could act as negative regulatory cells, although they themselves also specifically lysed autologous tumour cells in 18-h (i.e. long-term) cytotoxicity assays [71]. Along these lines, it has also been found that soluble products from NSCLC can induce IL-10 production in normal human PBMCs [63]. This was found to be caused by the PGE2 secreted by the cancer cells; neutralisation of the secreted IL-10 enhanced IFN-γ production, suggesting a negative autocrine loop triggered by cancer cell–derived PGE2. Intervention with prostaglandin blockers may therefore be beneficial, as has been reported for COX-2 inhibition in a rat model [145].

Alterations in signal-transducing molecules

Direct effects on T cells

The original observation that T-cell signal transduction is compromised in tumour-bearers [95] has subsequently been confirmed and extended to a variety of human tumours, including renal, colorectal, ovarian, liver, gastric, oral, prostate, pancreatic and cervical carcinomas, glioblastomas and melanomas (for review, see [162]). Of particular interest is the repeatedly documented correlation between these alterations and the disease stage in many different cancers. In mice, resection of tumour late after inoculation could still result in reappearance of an immune response, and normalisation of p56lck expression, a hallmark of this dysregulated state [131]. As even large tumour burdens in mice may still have effects limited to the locality, and do not cause systemic suppression [122], there may be hope of manipulation here to restore T-cell responsiveness. Down-regulated TCR signal transduction may be paralleled by down-regulation of CD28 [58]. This may reflect T-cell replicative senescence caused by continuous antigen activation, which could contribute to tumour escape from immunosurveillance [39, 109]. These correlations indicate that loss of ζ chain, or abnormal association between ζ chains and other CD3 components, might explain the observed gradual decline of cell-mediated responses in patients and experimental animals with progressing tumours. However, under certain conditions, extinction of ζ-signalling may not inhibit T-cell responses [6, 137], and decreased levels of CD3ζ chain in cancer patients may not correlate with their proliferative or cytotoxic capacity [25]. Conversely, it has been reported that T cells from PBMCs in early breast cancer patients do not show ζ chain deficiencies but are nonetheless functionally compromised [103]. Hence, there is some controversy still also in this area. However, where observed, the presumption is that prevention or reversal would be a good thing. In CMLs, for example, the majority of patients’ T cells were indeed found to be CD3ζ deficient, which could be at least partially reversed by stimulation with CD3 mAb, IL-2 and IFN-α [24]. Whether the same sort of manipulation would be effective in vivo is also controversial. However, decreased CD3ζ in three patients with myeloid malignancies after successful remission induction showed normalisation of the TCRs [14]. A study of 26 RCC patients treated with IL-2, IFN-α and LAK cells revealed posttreatment improvement of low ζ-chain expression in 62% of the patients [51]. Moreover, 4/5 achieving a complete response normalised ζ (and p56lck), whereas only 2/7 patients with progressive disease did so [51]. These reports therefore suggest that under certain conditions the defect in CD3 ζ expression might be reversible in vivo.

Effects via macrophages

The possibility that a mechanism of action of suppressor macrophages may be to induce alterations in signal transducing molecules has recently emerged from studies in murine and human systems. Macrophages with the capacity to suppress immune responses in tumour-bearing hosts have been extensively described before [2, 3, 90]. Aoe et al. [5] described the ability of tumour-infiltrating macrophages to decrease CD3ζ expression even on freshly isolated normal T cells. This effect is not tumour-specific, but may be a normal consequence of activation; thus zymosan/LPS-activated but not unactivated normal macrophages also induced CD3ζ down-regulation. However, this phenomenon does occur in cancer patients; Kono et al. [75] showed that macrophages isolated from metastatic lymph nodes of melanoma patients were able to down-regulate CD3 ζ levels in autologous PBLs. Again, this was not tumour-specific, because LPS-stimulated monocytes from normal PBMCs did the same. Because treatment with catalase prevented this, and H2O2 duplicated it, it was concluded that reactive oxygen metabolites produced by activated monocytes were responsible for ζ chain down-regulation [75]. The final mechanism of action of this oxidative stress-induced CD3 ζ loss may be via activation of components of the apoptotic pathway. Thus, oxidative stress triggers many cellular responses including proapoptotic factors such as p53; indeed, in a model system, loss of CD3 ζ chain may be directly mediated by one of the enzymes intimately involved in the apoptotic pathway, caspase-3 [48]. Moreover, the induction of apoptosis in T cells by fas-ligand–bearing ovarian carcinoma cells is accompanied by ζ down-regulation, and inhibitors of fas or apoptosis prevent this [121]. Moreover, it may well be the memory-phenotype cells, representing the antitumour effectors, which are most susceptible to oxidative stress [88].

These observations therefore argue that chronic inflammatory conditions in advanced cancer and in certain infectious and autoimmune conditions will alter the redox potential of macrophages, causing them to exert an immunosuppressive effect on the host immune system via secretion of factors such as H2O2. These factors will rapidly shut off the effector functions of CTL and NK cells. Research aimed at developing drugs which can counteract suppression of antitumour activity and which should be given in combination with immunotherapy should provide new and promising avenues for the treatment of cancer.

Immune stimulation

It has been repeatedly suggested that the immune system can exert a bipolar effect on tumours, often encouraging their growth [55, 115, 116, 118]. A striking recent demonstration of this phenomenon is the observation of enhanced growth of tumours in cancer-prone mice immunised with ras mutant peptides [140]. The progression of tumour development may even be dependent upon immune responses in at least some models, as shown by Hammond et al. [60]. Here, the rapidity and progression of carcinogen-induced guinea pig tumours was directly correlated with the immune status of the animals. A fully competent immune system furthered tumour progression, a fully incompetent one did not, and a partly competent one lay between the two extremes. As has been pointed out [119], this kind of phenomenon might explain the paradoxical observations that certain types of human melanoma which are usually curable have a bad prognosis if and only if they show signs of spontaneous regression. A potential clinical relevance for these findings may be found in accumulating data on tumour incidence in immunosuppressed transplant recipients. Stewart et al. [144] found a reduced risk of breast cancer (but not other cancers) in a very large number of women with heart or kidney transplants.

Cytokines produced by T cells reacting to the tumour may well encourage tumour growth, especially in the case of hematopoietic tumours, e.g. B-CLL respond to IL-4 produced by T cells. IL-4 protects the tumour cells from apoptosis and acts as a growth factor [29]. Receptors for IL-4 [120] may be even more widespread than those for IL-2 and be important for either enhancing or inhibitory effects of immune activity on tumour growth [30]. In CMLs, class II–restricted T cells specific for b3/a2 fusion products enhanced tumour cell colony formation, despite their concurrent cytotoxicity [167]. Receptors for IL-10 have been found on melanoma cells and it has been reported that IL-10 may function as a growth-stimulating factor for melanoma as well as reducing cell surface expression of HLA and adhesion molecules [169]. In this case, melanoma cells may also produce the IL-10 themselves, making it an autocrine growth factor; but as alluded to above, infiltrating T cells may provide a rich source of IL-10 as well. A further remarkable example of immunostimulation was reported recently where mucosa-associated B-cell lymphomas develop secondary to H. pylori infection in the stomach; their growth was shown to depend on the presence of H. pylori-specific CD4+ T cells [77]. Similarly, but without direct evidence, it has been suggested that the pathogenesis of hepatocellular carcinoma is dependent upon the immune response to HBV [102]. Novel and unsuspected mechanisms continue to be discovered, as illustrated by the recent finding that a T cell and monocyte-proinflammatory cytokine identified as macrophage migration inhibitory factor (MIF) inhibits the tumour suppressor activity of p53 [65].

Mechanisms other than cytokine production may also be involved in immunostimulation of the tumour. An intriguing report showed that many melanoma cells express CD40, which was on occasion up-regulatable by IFN-γ. While CD40 may engage CD154 on activated T cells and possibly deliver costimulatory signals, it is well known that on B cells CD40 itself delivers stimulatory signals required for target cell activation. According to Thomas et al. [149] the same may be true in melanoma, where ligating CD40 with mAb resulted in enhanced cell division. Thus, antimelanoma cells expressing CD154 (CD40-ligand) may interact with melanoma cells and directly stimulate them via CD40. Another aspect of immunostimulation of tumour growth is represented by the finding that TILs can also secrete angiogenic factors contributing to vascularisation of the tumours (basic fibroblast growth factor) and factors directly stimulating tumour cells (heparin-binding epidermal growth factor-like GF) [114]. TILs may also secrete other factors which indirectly assist the growth of tumour, e.g. vascular endothelial growth factor, which enhances angiogenesis [43]. In this study, in situ hybridisation showed that T cells infiltrating bladder and prostate cancer expressed VEGF mRNA and protein, and that isolated T cells could secrete bioactive VEGF. Activated macrophages may also stimulate enhanced VEGF and IL-8 production by melanoma cells, via a TNF/IL-1–dependent pathway; indeed the degree of macrophage infiltration has been reported to correlate with tumour stage and angiogenesis in melanoma [151]. Certainly, increased serum levels of VEGF, IL-8, bFGF and angiogenin do correlate with advanced disease state and degree of tumour burden [154]. Indeed, one of the requirements for CD4+ cells in tumour rejection may be their production of IFN-γ, which results in blockade of tumour angiogenesis; if the tumour fails to respond to IFN-γ, e.g. by down-regulating IFN-γR, or by expressing factors blocking IFN-γ activity at downstream signalling pathways [165], it can escape this effect [10]. Indeed, IFN-γ may represent a critical regulatory cytokine for tumour control, also contributing to counteracting enhancement of tumour progression by T- and B-cell products. It appears that without IFN-γ, even when antitumour effectors are generated, they may not be able to home in properly to the tumour [101].

Tumours may also subvert the immune response by expressing receptors for T-cell growth factors such as IL-2. Absorption of IL-2 secreted by antitumour T cells could induce anergy. Moreover, tumours themselves might even be able to use IL-2 for their growth [17], so that expression of the IL-2R has been associated with higher levels of proliferation but increased drug resistance in some tumours [79]. The potential clinical impact of these findings is illustrated by a case report of a melanoma patient whose tumour progressed and metastasis increased during IL-2/IFN-α therapy [61]. As IL-2–based immunotherapy is effective in only a small minority of patients, and many reports record progressive disease in the majority, such immunoenhancement may be more common than usually accepted. McMillan et al. [92] previously reported that the majority of solid tumours express IL-2R β but not IL-2R α chains and some can respond to IL-2 by increasing growth rates, and this was blocked by anti-IL-2 Ab. On the other hand, some tumour cells expressing IL-2R can actually be blocked in their proliferation and not stimulated by IL-2 [161, 168]. The reasons for these differences are not clear; possibly they relate to different histological types tested (SCCHN and gastric blocked; melanoma and lung stimulated). In any event, it is clear that use of IL-2–based immunotherapy may be a risky business, and that the clinical experience that only a minority of patients responds while most progress may reflect this variation.

Concluding remarks

Tumours are immunogenic but escape complete destruction by the immune system by the main strategies of invisibility (down-regulation of recognisable targets) and subversion (both by nullifying attacking cells and by utilising them and their products to their own growth advantage). Overcoming the last hurdle to successful cancer immunotherapy will require identifying and abrogating each of these escape mechanisms.