Primary tumours are currently treated by a combination of therapies, in most cases including surgery, local radiotherapy and chemotherapy. Even when the tumour has apparently been defeated, micrometastases of dormant tumour cells (or cancer stem cells) frequently lead to tumour relapse and therapeutic failure. To win the fight against cancer, it is necessary not only to develop strategies to kill all cancer (stem) cells efficiently, by using the correct combination and schedule of chemotherapeutic agents, but also to attempt to stimulate an immune response so that the immune system can keep residual tumour cells in check.

Cancer is widely considered to be a cell-autonomous genetic disease that results from alterations in oncogenes, tumour-suppressor genes and genome-stability genes. However, the tumour-cell microenvironment, the stroma and immunity also have a major role in cancer. Indeed, for the development of full-blown neoplasia, cancer cells have to overcome intrinsic (cell autonomous) and extrinsic (immune mediated) barriers to oncogenesis1. Only when tumour cells overcome immune control can they progress and kill the host. Accordingly, the increased incidence of some solid tumours in immunosuppressed patients, reports of spontaneous tumour regression, and the positive prognostic impact of tumour-specific cytotoxic T lymphocytes (CTLs) or antibodies support the idea that the immune system has an effect on tumour progression in humans1.

For historical reasons, drug discovery programmes for cancer therapy have neglected the possibility that immune reactions might contribute to the efficacy of treatment. For example, since 1976, the National Cancer Institute, USA, has used a drug screening and validation strategy in which human tumour cells are xenotransplanted into immunodeficient mice2 (Timeline). Although cancer chemotherapy and radiotherapy is often viewed as a strategy that mainly affects tumour cells, accumulating evidence indicates that cytotoxic drugs also affect the immune system to contribute to tumour regression.

Historical evolution of anticancer drug development and the concept of tumour immunosurveillance

In this Review, we summarize current knowledge on the contribution of the immune system to conventional cancer therapies. The immune system is elicited in two ways by conventional therapies. Some therapeutic programmes can elicit specific cellular responses that render tumour-cell death immunogenic. Other drugs may have side effects that stimulate the immune system, through transient lymphodepletion, by the subversion of immunosuppressive mechanisms or through direct or indirect stimulatory effects on immune effectors. Moreover, vaccination against cancer-specific antigens can sensitize the tumour to subsequent chemotherapeutic treatment. The challenge is to hijack the host immune system so that it can control any residual disease, or as stated by Prendergast and Jaffee, to stop “segregating cancer immunology from cancer genetics and cell biology”3. We anticipate that the physician of the future will need to integrate parameters that pertain to the host and the tumour in order to provide optimal management of the disease.

Theoretical bases for antagonistic effects

Many of the therapeutic procedures that are used in oncology today can curtail the immune response against tumour cells. For example, although it is well established from animal studies that the lymph nodes provide the optimal environment for T-cell priming, surgical oncologists frequently re-sect the tumour-draining lymph nodes, with the goal of performing histological staging of the tumour and removing local metastases4. More importantly, many chemotherapeutic drugs have notable immunosuppressive side effects, either directly, by inhibiting or killing effector cells, or indirectly, by provoking anergy or immune paralysis (Table 1).

Table 1 Deleterious effects of cytotoxic compounds

Immunosuppressive side effects of chemotherapeutics. Several of the cancer chemotherapeutics that are used today are also used as immunosuppressants for the treatment of severe systemic autoimmune diseases. This applies to cyclophosphamide5 and methotrexate6 (Box 1), which impair the proliferative and/or effector functions of peripheral T cells. Protein tyrosine kinase inhibitors may also affect the T-cell arm of adaptive immunity. At high doses, imatinib mesylate (Gleevec; Novartis), which specifically blocks signalling through the receptors KIT, c-ABL (cellular Abelson leukaemia-virus protein) and its oncogenic fusion derivative BCR (B-cell receptor)–ABL, suppresses T-cell proliferation and activation, presumably through inhibition of the protein tyrosine kinase LCK7. Studies that were carried out in mouse models have revealed that imatinib mesylate selectively curtails the expansion of memory CTLs, but does not affect primary T- and B-cell responses8. Treatment of patients with imatinib mesylate may also suppress the graft-versus-leukaemia effect from allogeneic transplantations and increase susceptibility to viral and bacterial infections9.

Owing to their capacity to trigger lymphocyte apoptosis, glucocorticoids are also important components of the chemotherapeutic cocktails that are used in treating several lymphoproliferative diseases. High doses of glucocorticoids are prescribed to patients with cancer to attenuate chemotherapy-associated nausea and vomiting. Glucocorticoids suppress the production of pro-inflammatory cytokines (such as interferon-α (IFNα), IFNβ, interleukin-1α (IL-1α) and IL-1β) and chemokines (such as CXC-chemokine ligand 8 (CXCL8), CC-chemokine ligand 7 (CCL7), CCL8, CCL11, CCL13, CCL17, CCL19 and CCL20) by blood mononuclear cells from healthy donors10. Although glucocorticoids induce the expression of the pattern-recognition receptors Toll-like receptor 2 (TLR2) and TLR4, they severely impair the differentiation of and antigen presentation by dendritic cells (DCs) in vitro and in vivo11. Moreover, glucocorticoids are known to repress the expression of genes that are associated with the adaptive immune response (including those that encode components of the MHC class II machinery, the T-cell receptor (TCR) genes TCRA, TCRB, TCRD and TCRG, signal transducer and activator of transcription 1 (STAT1) and CD40), alter T-cell development and function, suppress the development of TH1 cells (T helper 1 cells) and bias responses towards the TH2-cell type, thereby precluding the elicitation of effector and memory antitumour immunity. Several members of the transforming growth factor-β (TGFβ) family that suppress T-cell and natural killer (NK)-cell effector functions are upregulated by glucocorticoids. Glucocorticoids inhibit the cell-surface expression of the main natural cytotoxicity receptors (NKp30 and NKp44) and reduce IL-2- and IL-15-triggered NK-cell proliferation while compromising the NK-cell cytotoxicity that is mediated by the NKp46, NKG2D (natural-killer group 2, member D) or 2B4 receptors12.

Cell-death modalities: apoptosis and tolerance. Chemotherapy-induced tumour-cell death occurs frequently (but not exclusively) through apoptosis, a cell-death modality that is reputedly non-immunogenic or tolerogenic (Box 2). Based on the assumption that apoptosis, a morphologically defined phenomenon (with nuclear condensation and fragmentation as prominent hallmarks), is biochemically homogeneous, it is assumed that chemotherapy would elicit non-immunogenic cell death. Accordingly, cancer cells that are treated with mitomycin C, which induces classical apoptosis, fail to elicit an immune response when injected subcutaneously into histocompatible hosts in the absence of adjuvant. Similarly, 20 apoptosis-inducing compounds that operate through distinct modes of action failed to induce immunogenic cancer-cell death. This applies to drugs that kill cancer cells through mitochondrial, lysosomal or endoplasmic reticulum stress (ER stress), as well as protein tyrosine kinase inhibitors, proteasome inhibitors or DNA-damaging agents (such as alkylating agents or topoisomerase inhibition)13. One explanation of the non-immunogenic nature of apoptosis resides in the way in which cell corpses are handled by the immune system. For example, apoptosis is accompanied by the exposure of phosphatidylserine in the outer leaflet of the plasma membrane. Phosphatidylserine exposure functions as an 'eat-me' signal, which triggers phagocytosis by macrophages, and stimulates the production of anti-inflammatory cytokines. Likewise, the standard apoptotic programme is linked to the absence of immunostimulatory signals, as discussed in a later section.

Immunosuppression by massive tumour-antigen release. The sudden and systemic release of numerous dying tumour cells resulting from chemotherapy might have deleterious consequences on subsequent tumour-specific immune responses. Regionally advanced melanomas of the limbs can be treated by isolated limb perfusion (ILP) with tumour-necrosis factor (TNF) and the drug melphalan (Alkeran; Celgene) (Box 1). This results in local responses that are up to 80% complete, but does not prolong the survival of the patients14. So, despite the marked cell death that is induced by vasodilatation (vasoplegia), the induction of pro-inflammatory mediators, the expression of the adhesion molecules on endothelial cells, and the recruitment of lymphocytes, granulocytes and macrophages into infused tumour beds15, these patients fail to mount a protective immune response. Instead, pro-inflammatory cytokines (such as IL-1β, TNF and IL-6) that are produced by either the tumour or host immune cells may promote tumour progression through a molecular signalling pathway that involves nuclear factor-κB (NF-κB)16.

In addition, although this is theoretical, the massive release of tumour antigens during ILP may cause a sort of high-dose antigen-mediated tolerance. Indeed, the delivery of large amounts of antigen could be deleterious for mounting reactive effector T cells. It has been shown that there is a direct correlation between the amount of antigens that are expressed in the periphery and the degree of T-cell proliferation and number of tolerogenic antigen-specific CD8+ T cells in the draining lymph nodes17. Studies to evaluate melanoma-specific CD8+ T-cell responses after ILP are underway and may confirm these hypotheses.

Therefore, owing to their chemical nature (for example, as alkylating agents and glucocorticoids), their mode of action (for example, in which tumour-cell death is not preceded by cellular stress, and vasoplegia) or their handling (for example, their dosing and route of administration), several components of the therapeutic arsenal can be detrimental to the immune system.

Agonistic interplay

Accumulating evidence has indicated that radiotherapy and some cytotoxic compounds promote specific anticancer immune responses that contribute to the therapeutic effects of conventional therapy.

DNA damage: a fast track to immunity? DNA-damaging agents, such as ionizing irradiation or topoisomerase inhibitors, stimulate a complex response that involves the activation of tumour-suppressor proteins, such as the protein kinases ATM (ataxiatelangiectasia mutated) and CHK1 (checkpoint kinase 1), as well as the transcription factor p53. This DNA-damage response induces the expression of NKG2D ligands on tumour cells in an ATM-dependent and CHK1-dependent (but p53-independent) manner18. NKG2D is an activating receptor that is involved in tumour immunosurveillance by NK cells, NKT cells, γδ T cells and resting (in mice) and/or activated (in humans) CD8+ T cells.

Although p53 is not required for the expression of NKG2D ligands in cells undergoing DNA damage, a recent study has highlighted an important cooperation between p53-induced tumour-cell senescence and the innate immune system19. It was found that restoration of p53 function in established liver cancers led to tumour regression, but only when the mice had functional NK cells, neutrophils and macrophages. Reactivation of p53 in hepatocellular cancers induced the expression of pro-inflammatory cytokines (such as IL-15 and macrophage colony-stimulating factor (M-CSF)), adhesion molecules (such as intercellular adhesion molecule 1 (ICAM1) and vascular cell-adhesion molecule 1 (VCAM1)) and chemokines (such as CCL2 and CXCL1), which could contribute to the p53-induced recruitment of neutrophils, macrophages and NK cells into tumour beds19. This example illustrates how molecules that are activated during the DNA-damage response (ATM, CHK1 and p53) can alert the innate immune system to mediate an antitumour response.

Distant effects of radiotherapy. Besides its direct cytotoxic properties towards tumour cells, irradiation mediates many effects on cells and tissues, some of which may stimulate an immune response. Low doses of ionizing irradiation upregulate the expression of MHC class I molecules, tumour-associated antigens (carcinoembryonic antigen and mucin 1)20 and CD95 (also known as FAS) by tumour cells, as well as adhesion molecules by endothelial cells21, thereby boosting CTL activity22 and T-cell trafficking towards irradiated tumour sites23. Local irradiation of a single tumour site can reduce the size of non-irradiated metastases that are located at a distant site, a phenomenon known as the 'abscopal effect', when mediated by the immune system. In an elegant study, Reits et al.24 showed that irradiation enhances tumour antigenicity in three ways by modulating the repertoire of tumour-derived peptides that are presented to the immune system. First, irradiation enhances the degradation of existing proteins, thereby increasing the intracellular pool of peptides for MHC class I presentation. Second, activation of mammalian target of rapamycin (mTOR) in irradiated tumour cells stimulates protein translation and increases peptide production. Third, irradiation stimulates the synthesis of new proteins and therefore antigenic peptides, which can be presented for recognition by different T-cell repertoires24. Moreover, potent synergistic effects against established tumours between passively transferred CTLs24 or TLR9 ligands25 and ionizing radiation have been reported24.

High doses of alkylating agents for lymphoablation. The therapeutic induction of lymphopaenia has raised considerable interest in the context of adoptive transfer therapies and vaccination against melanomas26. Transient lymphopaenia is thought to enhance the efficiency of these therapies by activating homeostatic mechanisms that stimulate the tumour-reactive effector T cells and by counteracting tumour-induced suppression. Common methods to induce lymphopaenia include low-dose total body irradiation for reversible myelosuppression, or treatment with cyclophosphamide alone or in combination with fludarabine, which promotes long-term lymphopaenia in humans27. In addition to eradicating the cells that may suppress antitumour responses, such as regulatory T cells28, lymphoid reconstitution could overcome cancer-induced defects in T-cell signalling, endow host antigen-presenting cells with co-stimulatory functions, and increase the production and availability of cytokines, such as IL-7, IL-15 and IL-21, resulting in enhanced CD8+ T-cell activity28. Lymphodepletion also enhances T-cell homing into tumour beds and intratumoral proliferation of effector cells27,29. Finally, total body irradiation could cause mucosal barrier injury, resulting in microbial translocation and systemic release of the TLR4 ligand lipopolysaccharide (LPS), which has been shown to increase DC and T-cell activation and to promote tumour regression30. Animal studies have shown that lymphoablation enhances the effectiveness of adoptively transferred tumour-specific CD8+ T cells31. This strategy has been successful in a clinical trial that involved 35 patients with advanced metastatic melanomas that were refractory to conventional treatments29.

Lymphodepletion can also be combined with vaccination strategies that promote the differentiation of central memory T cells that recognize self tumour antigens32,33. Moreover, lethal myeloablation followed by autologous haematopoietic stem cell (HSC) rescue further enhances the efficacy of adoptive transfer therapies for tumours33. Profound lymphodepletion results in the expansion of both adoptively transferred T cells and host naive T cells.

So, chemotherapy-induced or irradiation-induced lymphodepletion, combined with additional manipulations, such as adoptive cell transfer, vaccination or HSC rescue, may be useful approaches for generating antitumour immune responses.

Low-dose cyclophosphamide for suppression of regulatory T cells. Although high doses of cyclophosphamide have direct tumoricidal effects and cause immunosuppression by lymphoablation (see above), low doses of the same compound mediate immunostimulation. Low doses of cyclophosphamide can potentiate delayed-type hypersensitivity (DTH) responses34 by acting on a cyclophosphamide-sensitive regulatory T-cell subset28. Low doses of cyclophosphamide decrease the number and inhibitory function of CD4+CD25+ regulatory T cells (TReg cells) by downregulating the expression of key functional markers of TReg cells, forkhead box P3 (FOXP3) and glucocorticoid-induced TNF-receptor-related protein (GITR)35. The effects of cyclophosphamide on TReg cells and cyclophosphamide-stimulated IFNα production36 might account for the augmented antibody responses and the persistence of memory T cells. All these effects contribute to the eradication of immunogenic tumours in synergy with specific immunotherapies37,38. However, the ablation of regulatory cells is likely to be of varying importance, depending on the tumour type, stage and location39.

In small clinical studies, the combination of low doses of intravenous cyclophosphamide with vaccines has been shown to augment DTH responses40,41, decrease the proportion of CD4+2H4+ (CD45+) suppressor T cells42 and prolong the survival of patients with metastatic cancer40. A daily dose of oral cyclophosphamide, also referred to as a metronomic programme, given for 1 month to patients with end-stage cancer could suppress TReg-cell inhibitory functions, restore the proliferative capacity of effector T cells and restore the cytotoxicity of NK cells43,44. However, the optimized schedule of cyclophosphamide yielding immunostimulatory effects in patients with cancer awaits further randomized investigation.

Anthracyclines as immunostimulators? Immune modulation by the anthracycline doxorubicin has been studied for decades45,46. Recently, it has been shown that doxorubicin enhances the antitumour potency of the GM-CSF-transfected tumour-cell vaccine when given before or after immunization47. The antileukaemic effects of a combination of IL-12 and various cytotoxic agents (cisplatin, cyclophosphamide, paclitaxel or doxorubicin) were evaluated in immunocompetent mice with syngeneic L1210 leukaemia cells, and it was demonstrated that significantly prolonged survival of the mice is only achieved with a combination of IL-12 and doxorubicin48. The therapeutic effect of IL-12 plus doxorubicin was lost when the mice were irradiated or injected with silica, indicating that radiosensitive immune cells (possibly T cells) and macrophages are required for the therapeutic effect. As a possible mechanism to account for the immune-mediated anticancer effects of anthracylines, it has recently been discovered that anthracyclines — unlike many other cytotoxic agents — can elicit immunogenic cell death (see below)13,49.

Taxanes: TLR4 ligands and beyond. The taxanes paclitaxel (Taxol; Bristol–Myers Squibb) and docetaxel (Taxotere; Sanofi–Aventis) both bind to the β-subunit of tubulin and affect microtubule polymerization, leading to cell-cycle arrest at the G2/M stage and subsequent cell death. Paclitaxel binds to mouse TLR4 (but not to human TLR4) and so can mimic bacterial LPS by activating mouse macrophages and DCs through a pathway that requires TLR4, MD2 (a key component of the LPS receptor complex) and myeloid differentiation primary-response gene 88 (MyD88)50.

In mice that are transgenic for human epidermal growth-factor receptor 2 (HER2; also known as NEU and ERBB2), which is overexpressed in some breast cancers, and that are therefore tolerant to HER2, treatment with paclitaxel increases the efficacy of tumour vaccines that express HER2 and GM-CSF51. Similarly, mice that have a mammary adenocarcinoma transfected with a candidate tumour antigen respond more efficiently to intratumour DC inoculations when they receive paclitaxel52. The optimal scheduling of docetaxel treatment combined with a GM-CSF-producing tumour vaccine was assessed in the 3LL tumour model53. When mice that have established 3LL tumours were pre-treated with docetaxel, followed by vaccination with irradiated GM-CSF-transfected 3LL tumour cells, significant tumour regression and prolonged survival were observed in comparison with chemotherapy alone. If docetaxel was administered after tumour-cell vaccination, treatment with taxane abrogated the antitumour effects of the vaccine, presumably by negatively affecting the dividing effector T cells. Mice that survived the treatment developed memory CTLs specific for a 3LL epitope (MUT-1) and resisted rechallenge with live 3LL tumour cells. In humans, there is circumstantial evidence that taxanes can stimulate the anticancer immune response. T-cell proliferation and NK-cell cytolysis assays have revealed that a cohort of patients with breast cancer (stage II/III) treated with taxanes showed enhanced T-cell and NK-cell functions compared with patients treated without taxanes54. Therefore, experimental data obtained in mice and humans contravene traditional thinking that taxanes suppress immune-cell functions.

Multiple immunostimulatory effects of gemcitabine. Gemcitabine (Gemzar; Celgene) is a synthetic pyrimidine nucleoside analogue that is efficient in the treatment of pancreatic, breast and lung cancers (Box 1). Its phosphorylated metabolites inhibit ribonucleotide reductase and DNA polymerase-α, which results in the depletion of the deoxynucleotide pool, thereby stopping DNA synthesis. Gemcitabine inhibits B-cell proliferation and antibody production in response to tumour antigens55, a phenomenon that may skew antitumour immunity towards beneficial T-cell responses56. Moreover, gemcitabine reduces the frequency of CD11b+GR1+ myeloid suppressor cells57. Gemcitabine-induced apoptosis of established tumours may enhance the DC-dependent cross-presentation of tumour antigens to T cells58. Consistent with these data, gemcitabine can function in synergy with CD40 stimulation of T cells to cure established mouse tumours59. The immunostimulatory effects of gemcitabine have been confirmed in patients. In patients with pancreatic60, non-small-cell lung61 or colon62 cancer, gemcitabine combined with recombinant cytokines or vaccines could enhance the frequency of tumour-specific CTL precursors and result in objective response rates.

5-Fluorouracil. 5-Fluorouracil is a fluoropyrimidine that is commonly used against breast cancer and gastrointestinal malignancies (Box 1). In vitro, 5-fluorouracil induces the expression of heat-shock proteins (HSPs) in tumour cells and facilitates antigen uptake by DCs and subsequent cross-presentation of tumour antigens63. Human colon carcinoma cell lines that were treated with 5-fluorouracil acquired CD95 and ICAM1 expression and became more sensitive to lysis by CTLs. In a mouse model, intratumoral inoculation of DCs after systemic chemotherapy that was based on 5-fluorouracil induced the T-cell-dependent eradication of the DC-treated site and, importantly, also of tumours located at a distant site, resulting in the long-term survival of the mice. Such effects can only be obtained by treatment with a combination of DCs and 5-fluorouracil, not with each treatment alone64. In another model, 5-fluorouracil was also able to augment the immunizing efficacy of a peptide-based vaccine directed against thymidilate synthase, its own molecular target65. These mouse data have prompted the design of a sizeable randomized clinical trial in patients with metastatic colon cancer (X. Cao, personal communication).

5-Aza-2′-deoxycitidine. The DNA methyltransferase inhibitor 5-aza-2′-deoxycitidine (DAC) is used in the treatment of leukaemias and myelodysplastic syndrome66. The antineoplastic effect of DAC is thought to involve reversing the hypermethylation of cancer-associated promoters, which inhibits gene transcription. Treatment with DAC restores the expression of MHC class I molecules and cancer testis antigens on tumour cells, rendering the tumour cells susceptible to CTL attack67. Moreover, hypermethylation of the tumour-suppressor gene that encodes death-associated protein kinase (DAPK) can be suppressed by treatment of leukaemic cells with DAC, thereby restoring the IFNγ-mediated apoptotic-cell-death pathway66. Cytokines that induce IFNγ secretion could therefore potentially be in synergy with DAC. Indeed, combining IL-12 with DAC was synergistic in a T-cell-dependent manner against L1210 leukaemia and B16F10 melanoma in mice68.

Antivascular flavonoids. Vascular-disrupting agents are distinct from anti-angiogenic agents in that they target pre-existing tumour vessels rather than neoangiogenesis. Flavone acetic-acid derivatives are known to have antivascular properties69. However, the flavonoid 5,6-dimethylxanthenone-4-acetic acid (DMXAA) has been shown to have immunological side effects and to elicit type-I-IFN-dependent NK-cell responses, as well as CD4+ T-cell-dependent antitumour responses69. In models of large thoracic tumours, DMXAA was able to promote the infiltration of CD11b+ cells and CD8+ T cells into tumour beds that release immunostimulatory cytokines and chemokines (CXCL10, RANTES, IL-6, IFNγ, TNF, CCL3, CCL2, CXCL9 and inducible nitric-oxide synthase). As the antineoplastic effects of DMXAA depend on perforin and CD8+ T cells, DMXAA could induce the activation of tumour-associated macrophages that contribute to the recruitment of other antitumour effectors70. Recently, DMXAA has been shown to be a novel and specific activator of the TANK-binding kinase 1 (TBK1)–IFN-regulatory factor 3 (IRF3) signalling pathway. Therefore, treatment of primary mouse macrophages with DMXAA has resulted in robust IRF3 activation and a marked increase in the production of the mRNA encoding IFNβ (Ref.71).

Immunostimulatory side effects of imatinib mesylate. Imatinib mesylate is a clinically approved drug for the treatment of gastrointestinal stromal tumours (GISTs), which are associated with activating mutations in KIT or PDGFRA (platelet-derived growth factor receptor-α). This treatment is based on the rationale that imatinib mesylate directly targets the pathognomonic mutations in these receptor protein-tyrosine kinases. However, a subset of patients with GISTs who have no KIT or PDGFRA mutation still respond to imatinib mesylate72. These antitumour effects were found to be related to NK-cell activation that is promoted by the side effects of imatinib mesylate on immune cells43,72. Indeed, imatinib mesylate mediated NK-cell-dependent antitumour effects in vivo in mice with cancer cells that did not respond to imatinib mesylate in vitro73. Imatinib mesylate blocks KIT signalling in host DCs, thereby triggering DC-mediated, NK-cell activation and NK-cell-dependent antitumour effects. In patients with GISTs, NK-cell IFNγ secretion that is induced by imatinib mesylate constitutes a positive prognostic parameter, suggesting that NK cells also contribute to the therapeutic effect of imatinib mesylate in humans72. So, depending on the local concentrations and the duration of exposure to imatinib mesylate, one might expect either inhibitory functions of imatinib mesylate on the differentiation of DCs and memory T cells or stimulatory effects on NK cells.

In mice, the combination of imatinib mesylate plus IL-2 induces the expansion of a unique population of tumour-infiltrating effector cells. These effector cells have been termed IFN-producing killer DCs (IKDCs)74 because they share phenotypic and functional properties with myeloid DCs and NK cells and can produce IFNγ. Unlike B220 NK cells, these CD11c+B220+NK1.1+ cells can lyse various target cells regardless of the expression of MHC class I molecules or NKG2D ligands. IKDCs also upregulate MHC class II expression on contact with transformed cells and produce large amounts of IFNγ, which is in contrast to B220 NK cells74. Adoptive transfer into B16F10 tumour-bearing immunodeficient (Rag−/− Il2rg−/−) mice has revealed that IKDCs but not B220 NK cells significantly delay tumour progression. Together, these data indicate that imatinib mesylate may exert part of its antitumour effects through the expansion and stimulation of IKDCs.

Mechanisms of T-cell-mediated immunity

The immuno-adjuvant effects of cytotoxic compounds discussed in this article may primarily rely on the capacity of antigen-presenting cells to engulf dying tumour cells and then to process and present tumour antigens to naive and/or central memory T cells. Therefore, and as previously reviewed75, the signals that are delivered by stressed or dying tumour or stromal cells under the influence of cytotoxic compounds can be expected to regulate antigen uptake (eat-me signals and antigen transfer), as well as antigen processing and presentation, thereby influencing DC maturation, co-stimulation, polarization and trafficking (Fig. 1; Table 2). In addition, stressed tumour cells may upregulate stimulatory ligands for NK cells (for example, NKGD2 ligands after DNA damage) or death receptors (for example, CD95 and TRAILR), thereby increasing their susceptibility to lysis by endogenous immune effectors (Fig. 2; Table 2).

Figure 1: Sequential events that link tumour-cell stress with activation of antigen-presenting cells.
figure 1

Following tumour insult by cytotoxic agents, tumour cells rapidly translocate intracellular calreticulin (CRT) to the cell surface (within 1 hour in vitro), which acts as a mandatory eat-me signal for dendritic cells (DCs) and induces immunogenic cell death. By 12 hours, molecular chaperones, such as heat-shock protein 90 (HSP90), can appear on the tumour-cell surface, contributing to tumour-cell–DC adhesion and the first steps of DC maturation. Next, by 18 hours, the release by dying tumour cells of the chromatin-binding protein high-mobility group box 1 protein (HMGB1) is required for optimal Toll-like receptor 4 (TLR4)-dependent processing of the phagocytic cargo by DCs. Finally, maturation signals (possibly type I interferons (IFNs) or interleukin-1 (IL-1)) that are delivered by stressed tumour cells complete the maturation programme that is elicited by the earlier events on DCs.

Table 2 Immunostimulatory properties of cytotoxic drugs
Figure 2: Linking tumour-cell stress and effector-cell killing.
figure 2

Tumour cells may be lysed by activated cytotoxic cells, such as natural killer (NK) cells, NKT cells, interferon-producing killer dendritic cells (IKDCs), γδ T cells and cytotoxic T lymphocytes (CTLs). However, this direct killing activity may be indirectly facilitated by the intervention of various effector mechanisms that sensitize tumour cells to express stress or danger signals that promote their recognition by particular effector cells. The DNA-damage response promotes the expression of ligands for NKG2D (NK group 2, member D), while restoration of functional p53 might induce the recruitment and activation of innate immune cells through the induction of CC-chemokine ligand 2 (CCL2) and interleukin-15 (IL-15). Inhibitors of histone deacetylases (HDACs) and inhibitors of the phosphatase PP1 contribute to the expression of NKG2D ligands and of cell-surface calreticulin, respectively, eliciting innate immune responses and phagocytosis, respectively. The release of high-mobility group box 1 protein (HMGB1) by tumour cells can be promoted by anthracyclines, oxaliplatin and irradiation, as well as by soluble TRAIL (tumour-necrosis factor-related apoptosis-inducing ligand) (and potentially activated NK cells and IKDCs that express TRAIL), and is mandatory for dendritic cell (DC)-mediated cross-presentation of apoptotic tumours to T cells. Proteasome inhibitors, such as bortezomib, induce cell-surface expression of heat-shock protein 90 (HSP90) by tumour cells, facilitating antigen uptake and maturation of DCs. Many chemotherapeutic agents can upregulate the expression of MHC class I molecules, tumour antigens and CD95 on tumour cells, leading to enhanced susceptibility to CTLs. ATM, ataxia-telangiectasia mutated; CRT, calreticulin; DAC, 5-aza-2′-deoxycitidine.

Chaperones. A common response to cell stress, including that induced by chemotherapy, is the transcriptional activation of a series of molecular chaperones that belong to the class of inducible HSPs. Such HSPs protect against cell death by re-folding damaged proteins and by directing damaged proteins to proteasome-mediated degradation. In addition, they inhibit apoptosis at the pre- and post-mitochondrial levels. HSPs can also stimulate the immune system at several levels.

First, HSP70 and HSP90 (as well as gp96 and calreticulin) can act on the scavenger receptor CD91 on the surface of DCs, thereby transmitting a maturation signal76. As well as binding damaged proteins, HSPs may bind peptides, including tumour-specific antigens, and may direct them to MHC class I and II pathways for presentation to T cells. Purified peptide–HSP complexes can induce protective immunity against the tumour from which they derive77. Peptide–HSP complexes may be particularly efficient as a source of antigen for cross-priming, as has been shown in vivo in suitable mouse models78. Moreover, encouraging results from HSP-based vaccine trials have been reported79.

Second, in addition to their function as vehicles for peptide antigens, chaperones may have an important role in DC activation as eat-me signals. For example, HSP90 appears on the surface of human myeloma cells after treatment with the proteasome inhibitor bortezomib (Velcade; Millennium Pharmaceuticals)80 and then serves as a contact-dependent activation signal for autologous DCs. Of note, this HSP90 exposure depends on the cytotoxic agent bortezomib and does not occur after γ-irradiation or treatment with steroids, indicating that there are stimulus-dependent differences in the type of stress response induced.

Third, another chaperone, calreticulin, can also be exposed on the surface of dying cells and act as an eat-me signal for macrophages, presumably by interacting with CD91 on the engulfing cell81. In response to some cell-death inducers, and in particular anthracyclines and ionizing irradiation, calreticulin can translocate from the lumen of the ER to the surface of cells at an early pre-apoptotic stage13,82. This early exposure of calreticulin and its aggregation in discrete foci on the cell surface (ecto-calreticulin) is not a general feature of apoptosis (Fig. 1). When tumour cells are treated for a few hours with anthracyclines and then injected subcutaneously into mice, they are highly efficient in inducing a specific DC and antitumour T-cell response49. Ecto-calreticulin is crucial for the recognition and engulfment of dying tumour cells by DCs. Thus, anthracyclines and γ-irradiation that induce ecto-calreticulin cause immunogenic cell death, whereas other pro-apoptotic agents (such as mitomycin C and etoposide) neither induce ecto-calreticulin nor immunogenic cell death. Depletion of calreticulin expression by transfection with specific small interfering RNAs (siRNAs) abolishes the immunogenicity of cell death that is elicited by anthracyclines, whereas an exogenous supply of calreticulin (as a recombinant protein) or the use of pharmacological agents that favour calreticulin translocation (such as PP1 phosphatase inhibitors) can enhance the immunogenicity of cell death13,80. Such PP1 inhibitors (for example, tautomycin and caliculin A) can increase the therapeutic efficacy of mitomycin C and etoposide in vivo by eliciting a specific immune response83.

Therefore, calreticulin exposure by tumour cells may allow prediction of the therapeutic outcome, and studies are ongoing to apply this knowledge to human disease82. The pharmacological reestablishment of calreticulin exposure may ameliorate the efficacy of chemotherapy, and the development of non-toxic PP1-inhibitory molecules that can induce calreticulin exposure is in progress.

HMGB1 and TLR4. Therapies that are based on anthracycline, irradiation and oxaliplatin (Eloxatin; Sanofi–Aventis) are more efficient at inhibiting the growth of established syngeneic tumours in immunocompetent mice than in athymic (nude) littermates, indicating that an intact immune system enhances the therapeutic efficacy of conventional anticancer treatments84,85. This result has prompted the search for a link between innate and adaptive immune responses and has led to the systematic screening of the role of TLRs in the efficacy of chemotherapy.

Surprisingly, the only TLR deficiency that compromises the efficacy of chemotherapy or radiotherapy in vivo is TLR4 and its downstream effector MyD88 (Refs 84, 85). Indeed, dying tumour cells can be cross-presented by DCs to naive T cells in vitro and in vivo, and they promote the differentiation of tumour-specific CTLs only when host DCs harbour a functional TLR4–MyD88 pathway. TLR4-deficient DCs are unable to present antigen from dying cells (taken up by phagocytosis) although they are normal in their ability to present antigen from soluble proteins (taken up by pinocytosis)84,85, which suggests a specific defect in antigen presentation after phagocytosis. TLR4 has been reported to inhibit the lysosome-dependent degradation of phagosomes in macrophages86. Accordingly, the kinetics of fusion between phagosomes and lysosomes is slower in wild-type DCs than in TLR4-deficient DCs loaded with dying tumour cells, which suggests that the TLR4 defect causes rapid lysosomal degradation of phagocytic material. Antigen presentation by TLR4-deficient DCs can be restored by inhibiting the activity of lysosomes, either with chloroquine (a lysosomotropic alkaline) or bafilomycin A1 (a specific inhibitor of the vacuolar ATPase responsible for lysosomal acidification). In addition, chloroquine can reverse the adverse effects of TLR4 deficiency on chemotherapeutic responses in vivo84,85.

Further studies have revealed that the ligand of TLR4 that is produced by the dying tumour cells is high-mobility group box 1 protein (HMGB1), a nuclear protein that is released from dying cells during late-stage apoptosis84,85 (Figs 1, 2). Depletion of HMGB1 from dying tumour cells with siRNAs or neutralization of HMGB1 with specific antibodies abolishes the TLR4-dependent, DC-mediated presentation of antigens from dying tumour cells in vitro and in vivo84. So, HMGB1 release is required for the immunogenicity of cell death through its effect on TLR4. However, neither HMGB1 nor calreticulin (nor a combination of both) can promote complete DC maturation, indicating that the search for immunostimulatory molecules produced by dying cells must continue.

Importantly, a polymorphism in TLR4 (rs4,986,790), which leads to a single-nucleotide exchange (A896G) and a single amino-acid substitution (Asp299Gly) in the extracellular domain of TLR4, was found to reduce the binding of HMGB1 to human TLR4. Accordingly, DCs that are derived from patients who carry the TLR4 Asp299Gly allele were far less efficient in cross-presenting antigens from dying melanoma cells to CTLs than were DCs from patients who carry the normal TLR4 allele84,85. This DC defect could be overcome by adding chloroquine. In a retrospective study, we analysed the time to metastasis in a cohort of 280 patients who had been treated for breast cancer with local lymph-node invasion, following a standard protocol of local surgery, local radiotherapy and systemic anthracycline. Patients who had the TLR4 Asp299Gly allele developed metastasis more quickly than patients who had the normal TLR4 allele, establishing TLR4 Asp299Gly as an independent predictive factor of early disease progression84. These results support the concept that a selective immune defect (in the DC-mediated presentation of antigen from dying cells) can compromise the response to anticancer chemotherapy85.

Clinical trials of combination therapies

Numerous clinical studies have combined conventional anticancer treatments with immunotherapies (Table 3). Pending confirmation through randomized, controlled clinical trials, some of these studies suggest that chemotherapy and immunotherapy can create a synergy.

Table 3 Chemo-immunotherapy protocols reported since 2000*

Two clinical studies in which DC vaccines were administered before salvage chemotherapy have indicated that synergistic antitumour effects between vaccination and cytotoxic compounds can be achieved in late-stage cancers. In one study, 29 patients with end-stage small-cell lung cancer that was resistant to first-line platin-based therapy were enrolled in a vaccination protocol that involved autologous DCs infected with adenoviral vectors encoding p53 (Ref. 87). The tumour progressed in 23 patients, who then received salvage chemotherapy with paclitaxel or carboplatin. The response rate to these second-line chemotherapies was 61.5% and 38%, respectively, survival at 1 year post-vaccination, a result that was not expected from historical controls. Another, retrospective study examined the impact of vaccination with peptide-loaded or lysate-loaded DCs on the efficacy of conventional therapy for glioblastoma. Chemotherapy acted in synergy with previous therapeutic vaccination to extend patient survival. The authors suggested that the immunological targeting of tumour antigen tyrosinase-related protein 2 (TRP2) expressed by glioblastoma cells increased the sensitivity to inducers of cell death87.

Similar intriguing synergistic effects were reported when combining DNA vaccines and chemotherapy in prostate cancers and other end-stage tumours88,89,90. Patients who succeeded in mounting vaccine-specific immune responses responded to salvage chemotherapy, whereas those who failed did not. The aforementioned clinical studies and those summarized in Table 3 illustrate that the combination of chemotherapy and immunotherapy may provide a valid therapeutic option for the treatment of neoplastic disease. However, defining the appropriate doses and schedule for the optimal synergy between the two strategies, as well as the key indications and tumour types53, represents our current challenge.

Concluding remarks and future outlook

There is accumulating evidence that conventional therapy for cancer may profit from the participation of the immune system. This immune contribution is elicited in two ways by conventional therapies. On the one hand, some therapeutic programmes can elicit specific cellular responses — beyond the stereotypical apoptotic pathway — that render tumour-cell death immunogenic. These immunogenic modifications include: pre-apoptotic calreticulin translocation in response to anthracyclines; induction of expression of MHC molecules, tumour-specific antigens or death receptors in response to epigenetic modifiers; HSP90 expression in response to bortezomib; and post-apoptotic HMGB1 secretion in response to anthracyclines and oxaliplatin. Other chemotherapeutic agents that induce immunogenic tumour-cell death may elicit yet more mechanisms. On the other hand, some drugs may have side effects (beyond their effect on the tumour itself) that stimulate the immune system, through transient lymphodepletion, the subversion of immunosuppressive mechanisms and the direct or indirect stimulatory effects of immune effectors.

In addition, many experimental therapies that have been tested in mice and a few clinical trials suggest that vaccination against cancer-specific antigens can sensitize the tumour against subsequent chemotherapeutic treatment. Although the mechanisms that underlie such a synergistic effect have not been elucidated, we speculate that the vaccination-induced increase in the frequency of primed T cells, although by itself irrelevant for tumour progression, may constitute a major advantage as soon as the tumour is insulted by cytotoxic drugs.

Based on these premises, we anticipate that, in the future, physicians will need to integrate four parameters to design an optimal therapeutic programme (Fig. 3) — the first three pertaining to the host and the last one to the tumour.

Figure 3: Reconciling host and tumour requirements for optimal management of cancer patients.
figure 3

For improved clinical management of cancer, four parameters should be integrated in cancer diagnosis and treatment. First, cancer immuno-epidemiology indicates that polymorphisms in the genes that dictate immune responses, such as Toll-like receptor 4 (TLR4) and interleukin-1β (IL1b), should influence the design of the therapy. Second, tumour-induced tolerance relies on suppressor cells or their mediators, such as transforming growth factor-β (TGFβ) and IL-10, which could be antagonized by dedicated therapies that have the aim of restoring antitumour immune responses. Careful immunomonitoring of patients with cancer should reveal which among these immunosuppressive pathways needs to be counterbalanced. Third, the deleterious effects of conventional therapies (including the use of steroids, lymph-node resections and escalating doses of chemotherapeutic agents with immunosuppressive and myelosuppressive side effects) should be avoided. Fourth, by examining the intrinsic characteristics of tumour cells using 'omics' approaches, one should be able to identify the immunosuppressive characteristics of the tumour, its antigenic profile and its capacity to succumb to immunogenic cell death. FcγR, Fc receptor for IgG; IDO, indoleamine 2,3-dioxygenase; ILT4, immunoglobulin-like transcript 4; iNOS, inducible nitric-oxide synthase; KIR, killer-cell immunoglobulin-like receptor; NKG2D, natural-killer group 2, member D; PDL1, programmed cell death ligand 1; STAT3, signal transducer and activator of transcription 3.

First, the emerging field of cancer immuno-epidemiology91 will determine which polymorphisms and mutations dictate therapeutic outcome in patients with cancer. As it stands, it seems that polymorphisms in the genes that encode TLR4, IL-10 and IL-18 may affect the therapeutic response in breast cancer84, lymphoma92 and ovarian cancer93, respectively. Similarly, polymorphisms in the Fc receptor for IgG affect the response to therapeutic monoclonal antibodies94.

Second, mechanisms of tumour-induced tolerance, including suppressor cells (such as TReg cells, T regulatory 1 cells, myeloid suppressor cells and tolerogenic DCs) or their effector mechanisms (such as TGFβ, arginase and IL-10) have to be overcome using specific therapies that may include depleting antibodies, cytokine antagonists or small molecules with immunostimulatory properties.

Third, deleterious effects that are provoked by chemotherapy or radiotherapy on the host's defence system have to be prevented or avoided. For example, the prescription of glucocorticoids during chemotherapy, as well as the ablation of tumour-draining lymph nodes, may be counterproductive because they suppress or avoid the immune responses. Similarly, a neoadjuvant therapy of radiochemotherapy may be immunologically more relevant than adjuvant therapies, and chemotherapies may have to be scheduled so that they liberate a maximum of tumour antigen while avoiding unwarranted immunosuppression.

Fourth, the intrinsic characteristics of tumour cells will have to be determined using 'omics' methods, not only to find the tumour's Achilles' heel for an optimal cytotoxic drug combination (or ideally a 'targeted' therapy) but also to determine the immunological characteristics of the tumour. Precise knowledge on the capacity of the tumour to translocate calreticulin, to expose HSP90, and to express antigen and ligands for recognition by cytotoxic effector cells may yield invaluable information for the optimal design of immunochemotherapies.