1.2.2 Non-apoptotic Cell Death
Anti-cancer therapies can also result in non-apoptotic cell death , including necrosis , autophagy and mitotic catastrophe.
Necrosis was historically considered a ‘messy’ form of cell destruction, morphologically characterized by random degradation of nuclear DNA, organelle degeneration and swelling and rupture of the cell membrane with release of intracellular components . Recent work, however, suggests that necrosis can occur in a regulated manner, which is dependent on specific signalling modules such as RIP1 activation and can be triggered by a number of processes including alkylating DNA damage and death receptor ligation [35–38]. Zong et al.  first demonstrated that in vitro cell death caused by the alkylating agents nitrogen mustard and MNNG (N-methyl-N-nitro-N-nitrosoguanidine) occurred independently of apoptotic factors, but required activation of the DNA repair protein poly (ADP-ribose) polymerase 1 (PARP1), suggesting a necrotic form of programmed cell death. More recently, the alkylating agent cyclophosphamide was shown to cause tumor regression in a xenograft mouse tumor system in both apoptosis-competent and apoptosis-deficient tumor cells, with the observation of sporadic necrosis in both groups, as identified using cell morphology, high mobility box group 1 (HMBG1) extracellular release and activation of innate immune cells . In an effort to define the molecular signalling network that regulates necrotic pathways, Hitomi et al.  carried out a genome-wide siRNA screen and identified 432 genes that regulate necroptosis of which 32 acted downstream of regulators of RIP, 32 genes were required for death-receptor-mediated apoptosis and 7 genes that were involved in both pathways. Together, these data suggests that necrosis can occur in a regulated manner that is independent of apoptosis and may play a role in several physiological and pathological settings, but its role in tumorigenesis, in chemotherapy-induced cell death , or in the antitumor immune response is yet to be fully elucidated.
Autophagy , or ‘autophagic cell death’, is a process that leads to cytoplasmic vacuolization without features of apoptosis . It often constitutes a cytoprotective response activated by dying cells as a defense against acute stress, and inhibiting autophagy can actually mediate death rather than prevent it [40–42]. It is therefore unclear whether autophagy can be responsible for cell death , or rather is a process that can accompany cell death, without participating in the lethal catabolic cascade .
A third death mechanism, ‘mitotic catastrophe’, refers to cell death triggered by aberrant mitosis and occurring during mitosis or in subsequent interphase . A number of chemotherapeutic agents, namely spindle poisons or microtubule inhibitors (taxanes and vinca alkaloids) interfere with the function of microtubules during mitosis, resulting in activation of the spindle assembly checkpoint (SAC) and causing mitotic arrest. The result is caspase-mediated cell death, but the molecular mechanism connecting initial SAC activation to caspase activation has not been clearly defined . Interestingly, recent work has indicated that these mitotic aberrations can lead to cell senescence or cell death, which can occur either through apoptosis or necrosis [45, 46]. It was therefore proposed that mitotic catastrophe may not be a ‘pure’ cell death pathway but an oncosuppression mechanism that ultimately leads to cell death or senescence. Whether the duration of mitotic arrest truly has an influence on the cell fate remains unclear [11, 47].
Cellular senescence is a form of permanent cell cycle arrest with typical morphological and biochemical changes including the induction of senescence-associated b-galactosidase activity. It can be induced by cellular stresses such as DNA damage and oncogenic stress, which can be regulated by tumor suppressors including p53, p16INK4a and retinoblastoma protein (RB) . Cellular senescence may play a role in suppressing tumorigenesis by disabling apoptosis , as suggested by the early cancer development in murine cells with senescence defects [48, 49]. Moreover, senescence may contribute to chemotherapy sensitivity in primary murine lymphomas, where cyclophosphamide can lead to p16INK4a activation, p53-dependent cell-cycle arrest and premature senescence in vivo .
1.2.3 Morphological vs Functional Classification of Cell Death
Morphological classification of cell death dominated the cell death literature for several decades before the development of many of the biochemical tests available today. However, it is becoming increasingly clear that morphologically similar cell deaths may demonstrate significant functional, biochemical and immunological heterogeneity [43, 51, 52]. In addition, a specific morphological appearance may not necessarily be sufficient to establish a link between a causative process and subsequent cell death . Conversely, biochemical classifications also have drawbacks. For example, a cell death pathway frequently associated with a process may still occur in the absence of that process. Phosphatidylserine (PS) exposure, an early marker of cell death, was shown not to occur in autophagy-deficient cells undergoing apoptosis . The Nomenclature Committee on Cell Death (NCCD), which has previously published two rounds of recommendations in 2005 and 2009, has formulated a novel systematic classification of cell death based on biochemical features intended for both in vitro and in vivo applications .
1.3 Immunogenicity of Cell Death
The immune system has long been thought to play a role in cancer development and surveillance . The multi-step process of carcinogenesis can lead to the accumulation of mutations that may affect the antigenic properties of proteins within tumor cells, which can in turn be presented by MHC class I molecules to CD8 + T cells . In addition, the oncogenic stress occurring in early tumor development can also lead to activation of immune responses by the innate and/or adaptive immune system that may destroy cells prior to development of overt malignancy. It is now accepted that the ability of tumor cells to evade immune destruction is one of the hallmarks of cancer [54, 55].
The original model of immune function described by Burnet  suggested that the immune system functions by differentiating between ‘self’, which is defined early in life, and ‘non self’, which represents anything that comes later. The model evolved over the years to accommodate new data and address incompatible findings such as somatic hypermutation of B cells and transplant rejection, and consequently the ‘danger model’ was proposed . This model suggests that it is the damage or ‘danger’ that is of greater importance to the immune system, rather than ‘foreignness’. As such, established tumors that do not cause damage, i.e. do not elicit endogenous or exogenous danger signals through cellular stress or injury such as ‘immunogenic cell death ’, will not elicit an immune response and conversely ‘self-ness’ of healthy tissues does not guarantee tolerance. Further incarnations of the danger model lead to the current understanding that the immune system would seek to answer two questions when faced with a threat: firstly whether to respond or not—and this is determined by the nature of the threat as determined by ‘danger signals ’; and second, is what kind of immune response is elicited, a decision that will be primarily determined according to the tissue where the response occurs, rather than the nature of the invading pathogen [57, 58].
When we consider the immune response to a tumor that has escaped immune surveillance and destruction, the antitumor immune response can involve components of the innate immune system as well as the humoral (antibody-mediated) and cellular (T cell mediated) arms of the adaptive immune system. However, CD8 + cytotoxic lymphocytes are generally considered the most effective of the anti-tumor immune responses for a number of reasons. Firstly, the importance of CD8 + effector cells is highlighted by the observation that greater proportions and activation of CD8+ tumor infiltrating lymphocytes (TILs) correlates with a better prognosis in several cancers, as discussed in Sect. 18.104.22.168. Secondly, both treated and untreated tumors have been shown to grow more rapidly when CD8 + cells, but not NK cells or B cells are depleted from the tumor microenvironment [59, 60]. Moreover, some chemotherapy can selectively deplete B cells while preserving CD8 + function without obvious detriment to antitumor immunity . The demonstrated efficacy of adoptive T-cell transfer therapy in some patients with advanced melanoma, and the use of graft versus leukaemia effect with donor leucocyte infusions for relapse after allogeneic bone marrow transplantation also highlight the prominent role for CD8 + T cells for antitumor immunity [62–64].
An understanding of the steps required to elicit an antitumor immune response, especially a specific CD8+ antitumor response, as well as the effect of various anticancer therapies on antitumor immunity is necessary to enable the interpretation as well as the rational design of preclinical and clinical studies. This will enable us to harness the immune response and utilize the expanding armament of immunotherapies that can be combined with chemotherapy to address immune evasion resulting from tumor or chemotherapy i.e. the ‘brakes’ on the immune response; and/or to synergize with the immune-stimulatory effects of treatment discussed below to achieve better outcomes with chemo-immunotherapy .
There are six key steps postulated for an effective antitumor CD8 + T-cell response:
Tumor antigens must be present;
Antigens must be seen as ‘dangerous’ and acquired by professional antigen presenting cells (APCs);
Tumor-specific CD8 + T-cells recognize antigens and respond by proliferation;
Circulating CD8 + T-cells must reach the tumor;
CD8 + T-cells must overcome immune-suppressive signals within the tumor microenvironment ;
And memory cells should be generated for a durable response.
The induction of immunogenic cell death is an important goal for anticancer chemotherapy. Apoptotic cell death, the most common modality of chemotherapy-induced cell death, has classically been assumed to be non-immunogenic (or tolerogenic), whereas necrotic cell death was considered truly immunogenic. As millions of cells die every second by ‘physiological’ apoptosis, PS residues are exposed on the cell surface, which leads to the rapid and silent clearance of apoptotic bodies by macrophages and stimulates the production of immunosuppressive cytokines [65, 66], thus protecting the host from overwhelming inflammation and autoimmunity.
Other studies, however have challenged the theory that apoptosis is uniformly non-immunogenic, and demonstrated that the pre-apoptotic cellular responses to stress, the surface characteristics of apoptotic bodies and the process of apoptosis itself can be quite heterogeneous biochemically, despite apparently similar morphologies [28, 60, 67–69]. We have shown that in a murine model of malignant mesothelioma, chemotherapy-induced apoptosis increased antigen cross-presentation and cross-priming rather than tolerance, and demonstrated a synergistic effect between immunogenic chemotherapy and immunotherapy [25, 70].
Kroemer and colleagues demonstrated that although the chemotherapeutic drugs mitomycin C and doxorubicin both cause caspase-dependent apoptosis, only cell death caused by doxorubicin is immunogenic and mediated by dendritic cells (DCs) . They found that caspase inhibition by the broad-spectrum caspase inhibitor Z-VAD-fmk, or using cells transfected with the baculovirus caspase inhibitor p35 delayed, but did not inhibit doxorubicin induced death; and inhibited uptake of dead tumor cells by DCs, but had no effect on their ability to elicit DC maturation. Moreover, the selective depletion of dendritic cells (DCs) or CD8 + T cells , but not natural killer (NK) cells, abolished the antitumor immune response .
1.4 Biochemical Features of Immunogenic Cell Death
Although it has been shown that chemotherapy-induced cell death can be immunogenic, it is less clear which anti-cancer agents, and in what settings, can trigger danger signals that would be seen by the immune system and elicit an antitumor immune response. A number of features of dangerous cell death have been well described, and various chemotherapy agents have been analyzed in vitro and in vivo to determine their relative danger and immunogenicity .
1.4.1 Calreticulin Exposure
Calreticulin (CRT) is a 46 kDa protein previously thought to be an obligate endoplasmic reticulum (ER) protein, but later found to have broad localization within the cell and to participate in multiple processes . When CT26 colon cancer cells were treated with ~ 20 distinct inducers of apoptosis, including doxorubicin and other anthracyclines, then injected into immunocompetent BALB/c mice, CRT was found to translocate to the cell surface from the ER rapidly, and prior to exposure of PS on the outer leaflet of the plasma membrane. There was a strong, positive linear correlation between the surface exposure of CRT and the protection against tumor growth, which was interpreted as a sign of immunogenicity and antitumor vaccination . In addition, the knockdown or blockade of CRT abolished the phagocytosis of anthracycline-treated cells by DCs, while the addition of recombinant CRT protein (rCRT) reversed the defect induced by CRT blockade in anthracycline-treated cells, and conferred an immunogenic cell death in cells treated with etoposide and mitomycin C, which are regarded as non-immunogenic .
The pre-apoptotic exposure of CRT was shown to be accompanied by the co-translocation of ERp57, another ER protein , whereas other cell death inducers cause CRT to be exposed at a later apoptotic stage when several other ER proteins are exposed concomitantly with PS [73, 74].
Together, these data suggest that CRT exposure is a major determinant of immunogenic cell death , that its exposure occurs upstream of apoptosis or necrosis and that it can confer immunogenicity when added to non-immunogenic cell death inducers.
1.4.2 Heat Shock Proteins (HSPs)
Examining the molecular bases of the ‘danger model’, one class of endogenous danger signals , or danger associated molecular patterns (DAMPs ) are heat shock proteins (HSPs) . This family of chaperone proteins is involved in the correct folding (or refolding) of proteins in conditions of cellular stress. HSPs play an important role in driving immune function and can be released from dead cells after either primary or secondary necrosis . The frequent expression of HSPs such as HSP70 and HSP90 inside tumor cells, presumably as a result of stress, can have apoptosis-inhibitory and cytoprotective effects and has been associated with chemotherapy resistance in vitro and in vivo,and associated with poor prognosis in patients with several cancers , including stage II/III breast cancer treated with neoadjuvant chemotherapy [76–78].
Paradoxically, whilst HSP70 and HSP90 can be immune inhibitory when overexpressed inside the cell, they can actually have an immune stimulatory effect when expressed on the cell surface . Scavenger receptors (SR) family members are often involved in HSP70-mediated cross-presentation in DCs and cytotoxic T cell activation. For example LOX-1, a major SR can bind avidly to HSP70 and deliver tumor-specific antigens to cell surface major histocompatibility complex (MHC) class I molecules. Similarly, an HSP90-antigen complex (HSP90.PC) was internalized by DCs and antigens cross-presented to induce a CTL response . Therefore, HSPs can play an important role in cross-presentation of tumor antigens on MHC class I molecules resulting in specific CD8 + T-cell responses [81–83].
In vivo, the proteasome inhibitor bortezomib was shown to induce the surface expression of heat shock protein 90 (HSP90) on dying human myeloma cells resulting in the generation of antitumor T cells through gedlanamycin-inhibitable tumor cell recognition by DCs . This suggests that HSP overexpression inside the cell is different immunologically from their exposure on the cell surface, where they facilitate the immune recognition of stressed or dying cells .
Clinically, the immunogenicity of HSP90 has been explored in two phase I/II trials of autologous tumor-derived HSP90 in the form of a Gp96 peptide vaccine. One study enrolled patients with malignant melanoma with detectable tumor, and the second study recruited patients with colorectal carcinoma after complete resection of liver metastasis. Both trials reported that the autologous vaccine can successfully induce in vivo antitumor immunity and a clinical benefit in a percentage of patients [85, 86].
1.4.3 HMGB1 and TLR Interactions
High-mobility group Box 1 (HMGB1) is a pro-inflammatory cytokine secreted by activated macrophages, NK cells and mature dendritic cells , and can also be released from dying cells . Bianchi and co-workers were first to report that necrotic cell death leads to the release of HMGB1, which remains attached to the nucleus during apoptotic cell death, even after rupture of the plasma membrane . However, apoptotic cells under some circumstances can also release HMGB1. Chemotherapeutic agents such as alkylators that activate poly-adenosylribosyl polymerase (PARP) can release HMGB1from its association with chromatin allowing it to be released from the nucleus, binding to TLR4 and activating macrophage cytokine release [21, 89, 90].
Although HMGB1 was recently identified as an endogenous danger signal [91, 92], its interaction with the innate immune system is controversial, particularly with regards to TLR dependency . HMGB1 can bind to several distinct surface receptors than can be found on DCs. These include the receptor for advanced glycosylation products (RAGE), toll-like receptor 2 (TLR2) and toll-like receptor 4 (TLR4) [93, 94].
TLRs play an important role in the immune response by recognizing both exogenous and endogenous signals and upon ligation, can activate pro-inflammatory gene expression via pathways involving two distinct adaptors, Toll/IL-1R domain-containing adaptor inducing IFNa (TRIF) and myeloid differentiation primary response protein 88 (MyD88) [21, 87].
Apetoh et al. have shown that TLR4 (but not TLR2) was required for bone marrow derived DCs (BM-DCs) to efficiently present antigen from dying tumor cells in vitro, and that dying tumor cells failed to elicit a tumor-specific immune response in TLR4 knockout mice (but not with knockdown of TLR2 or other TLRs) . HMGB1 was also detected in the supernatant of cultured cells that underwent secondary necrosis following immunogenic cell death caused by the anthracycline doxorubicin, and depletion of HMBG1 using specific small interfering RNA (siRNA) abolished the ability of doxorubicin-treated cells to prime T cells in vivo . Moreover, the authors demonstrated that knockdown of either MyD88 or TLR4, but not TRIF, reduced the efficacy of chemotherapy, pointing to a contribution of TLR4/MyD88-dependent immunity to chemotherapy efficacy. Importantly, the TLR4 polymorphism Asp299Gly, which is found in 8–10 % of Caucasians, was revealed to reduce the interaction between HMGB1 and TLR4 and abolish the ability of myeloid derived dendritic cells (MD-DCs) to cross-present dying melanoma cells in vitro, a defect that was reversed by the addition of chloroquine.
Similarly, in a cohort of 280 patients with early breast cancer receiving anthracycline-based chemotherapy, patients carrying the mutated TLR4 Asp299Gly allele had a higher frequency of metastases at 5-years compared to patients without the mutation (40 % vs 26.5 %; p < 0.05) . These data suggested HMGB1as an important DAMP that dictated a TLR4/MyD88-dependent immune response to dying tumor cells, as well as the relevance of HMGB1/TLR4/MyD88 in the efficacy of anticancer drugs.
In contrast, Tian et al. have demonstrated that HMGB1 can interact with CpG-containing oligodeoxynucleotides (ODNs) to form HMGB1-DNA immune complexes that can activate plasmacytoid DCs (pDCs) and augment IFN-a production in a TLR9/MyD88-dependant mechanism, which is mediated by the adaptor RAGE, but is independent of TLR2 or TLR4 .
1.4.4 ATP Release and Activation of the NLRP3 Inflammasome
Caspase-1 (also known as IL 1-b-converting enzyme—ICE) is the first identified member of the caspase family, and a prototypic member of the subclass of inflammatory caspases that also includes caspase-4, -5, -11 and 12. Members of the NOD-like receptor family (NLRs) promote the assembly of multi-protein complexes known as ‘inflammasomes’, and these are required for the activation of inflammatory caspases . In response to danger signals , NLRP3 (NOD-like receptor family, pyrin domain containing-3 protein, also called NALP3 or cryopyrin) forms the caspase-1 activation complex ‘inflammasome’ through interaction with an adaptor molecule, apoptosis-associated speck-like protein (ASC) .
Extracellular ATP has been identified as another signal emitted by tumor cells as they undergo immunogenic cell death , and is released in the blebbing phase of apoptosis [97, 98]. ATP released from dying tumor cells was shown to act on a purinergic receptor (P2RX7) on the surface of DCs which triggers the NLRP3 inflammasome and results in caspase-1 activation and consequent secretion of IL-1b which is required for polarizing CD8 + T cells towards IFN-g production. The knockdown or inhibition of NLRP3, P2RX7, IL-1b or the IL-b1 receptor abolished the ability of mice to mount an immune response against dying tumor cells. Moreover, mice deficient for P2RX7, NLRP3, caspase-1, IL-1 or the IL1 receptor were able to grow tumors normally, but failed to respond to anthracycline chemotherapy. Reminiscent of mutant TLR discussed above, a single nucleotide polymorphism affecting P2RX7 (Gly496Ala) was also shown to have a negative impact on progression free survival of breast cancer patients receiving adjuvant anthracycline-based chemotherapy .
These data further highlight the ability of chemotherapy to trigger the release of immunogenic signals from dying cancer cells that are detected by DCs and can result in a specific anti-tumor immune response.
1.4.5 Autophagy-Dependent Immune Responses
Michaud et al. have recently demonstrated a potential role for autophagy in immunogenic signalling in the context of chemotherapy-induced cell death . They showed that tumor cells that were rendered autophagy-deficient by the knockdown or knockout of essential autophagy proteins such as Atg5 and Atg7 and subsequently treated with chemotherapy, released lower amounts of ATP compared to autophagy-competent tumor cells. In response to chemotherapy, both autophagy deficient and autophagy competent tumor cells underwent apoptosis, exposed CRT and released HMGB1, but only autophagy competent tumor cells attracted dendritic cells and T lymphocytes into the tumor bed. Moreover, the inhibition of autophagy in tumor cells with siRNA or short hairpin RNA (shRNA) abolished their ability to mount an antitumor immune response in vivo. Importantly, intratumoral injection of the ecto-ATPase inhibitor ARL67156 in chemotherapy treated, autophagy-deficient tumors enhanced the recruitment of DCs and IFN-g producing T cells; and restored T cell priming, as well as significantly reducing tumor growth. These finding were thought to be potentially translatable to humans through the development of autophagy inducers or compensatory strategies to enhance peri-tumoral ATP pharmacologically .
1.5 Effects of Chemotherapy on the Immune System
Conventional cytotoxic chemotherapy has long been assumed to have either a neutral effect or, more commonly, a negative effect on the immune system. This was based on the inhibitory effect on proliferation of immune constituents, observations of clinical lymphopenia, and the production of apoptotic cell death, which was thought to be immunologically ‘bland’, as discussed above. In line with this, conventional cytotoxic drug development has involved the selection of agents that are active against human tumor cells in vitro and against transplanted xenografts in immunodeficient mice .
There is now a significant body of existing and emerging evidence that chemotherapy can have both immune-stimulatory and immune inhibitory effects over and above the distinctive properties of immunogenic vs. non-immunogenic cell death discussed above.
1.5.1 Immune-Suppressive Effects
Neutropenia and lymphopenia are well-recognized side effect of cytotoxic chemotherapy, and are common dose-limiting toxicities of many cytotoxic regimens, including anthracyclines, alkylating agents, antimetabolites and others [100–102]. The tyrosine kinase inhibitor imatinib has been shown to suppress T cell proliferation and activation in vitro , and to inhibit expansion of memory cytotoxic T lymphocytes (CTLs) but not the primary immune response in vivo . More recently however, Balachandran et al. demonstrated a significant contribution of the immune system to the antitumor effects of imatinib in a mouse model of spontaneous GIST, through the activation of CD8 + T cells and apoptosis of regulatory T cells (Tregs) .
22.214.171.124 Non-immunogenic Cell Death
While a number of chemotherapy agents have been shown to induce immunogenic cell death by eliciting some of the hallmarks of danger discussed in Sect. 1.3, other agents are capable of inducing cell death without these associated danger signals , and therefore do not induce an effective antitumor immune response. For example, treatment of murine colon cancer cells with mitomycin C, etoposide or camptothecin resulted in apoptotic cell death, which resulted in a weaker antitumor immune response compared to that of immunogenic agents such as the anthracycline doxorubicin. Of note is that while a positive correlation was observed between CRT exposure and immunogenicity , there was a discrepancy between low CRT and low or intermediate mitomycin C immunogenicity . This highlights that multiple parameters may determine immunogenicity of cell death, resulting in a spectrum from none or low, to higher immunogenicity .
1.5.2 Immune-Stimulatory Effects
There is an abundance of evidence that chemotherapy can promote development of an antitumor immune response through immune-stimulatory effects on both innate and adaptive anti-tumor immunity, potentially influencing one or several of the six steps required for induction of CD8-mediated antitumor immunity highlighted in Sect. 1.3, and reviewed in .
126.96.36.199 Enhanced Antigen Delivery and Presentation
The importance of tumor-specific antigens has evolved over several decades , with the recognition of an increasing number of antigens that may be either ‘self’ antigens such as un-mutated self-proteins, or neoantigens such as mutated proteins, or oncogenic viruses [105, 107–111]. Conventional and targeted therapies can result in the release of antigen from dead or dying tumor cells, which are then available for presentation to the immune system. Novel, targeted therapies are also now increasingly used in the clinic in a paradigm shift towards personalized cancer medicine, guided by some of the recently discovered molecular biomarkers such as HER2 amplification (trastuzumab), BCR-ABL translocation (imatinib), BRAFV600E mutation (Vemurafenib), EGFR mutation (erlotinib and gefitinib) and EML4-ALK fusion (crizotinib) [112, 113]. These targeted approaches, in selected patients, can produce tumor regression that far exceeds what was previously seen with conventional chemotherapy for metastatic solid tumors [114–117], and this in turn can potentially improve delivery of tumor antigen for presentation.
Once antigen concentration is above the threshold required to elicit an immune response, it must then be presented to the immune system in a context that is seen as ‘dangerous’, otherwise there may be no immune response or tolerance may ensue . “Cross-presentation” is the process by which exogenous antigens ‘cross’ from the typical major histocompatibility complex class II pathway, into the class I pathway which is typically responsible for expression of endogenous antigens [118, 119]. We have shown that apoptosis induced by gemcitabine chemotherapy in vivo significantly increased cross-presentation of tumor antigens but did not induce tolerance of tumor-specific CD8+ T cells . Moreover, HSPs have also been proposed to play a role in MHC class I-mediated cross-priming of CD8+ T cells by APCs, as discussed in Sect. 1.4.2.
188.8.131.52 Homeostatic Proliferation
The release and presentation of tumor antigen to the immune system in a context that leads to priming of CD8+ T cells must be followed by antigen recognition and proliferation of tumor specific T cells .
The observation of clinical lymphopenia resulting from chemotherapy is generally considered an immunosuppressed state. However, the induction of transient lymphopenia therapeutically in the context of adoptive transfer treatment and vaccination strategies is thought to enhance the effectiveness of these therapies through homeostatic mechanisms. This lymphoid reconstitution could overcome cancer-induced defects in T-cell signalling and increase cytokine production, resulting in enhanced T-cell activity [120, 121]. Clinically, there was a demonstrated benefit from adoptive cell transfer therapy following lymphodepleting chemotherapy in patients with refractory metastatic melanoma .
Recent work by our group studying the effects of chemotherapy on the antitumor immune response in patients with mesothelioma and NSCLC, we found that while chemotherapy depleted proliferating CD8+ and CD4 + T cells (identified via intracellular Ki67 staining) one week after chemotherapy, these proportions significantly increased by the end of the treatment cycle above baseline levels. Importantly, we found the increase in proportion of proliferating CD8 + T cells after one cycle of chemotherapy to be an independent predictor of patient survival (unpublished data).
184.108.40.206 Regulatory T Cells
CD4 + CD25 + regulatory T cells (Tregs) composing 5–10 % of the peripheral lymphocyte pool, typically express the forkhead box-binding protein 3 (FoxP3), cytotoxic T lymphocyte antigen-4 (CTLA-4) and the glucocorticoid-induced tumor necrosis factor receptor family-regulated gene (GITR). They secrete transforming growth factor b (TGFb) and IL-10, and serve to down regulate the normal immune response and prevent autoimmunity through either direct cell-cell contact or via the effects of TGFb and IL-10 [123–125].
The prognostic value of tumor infiltrating lymphocytes has been a subject of increasing interest over the past decade. High numbers of lymphocytes, especially T lymphocytes have been associated with a better prognosis in patients with several malignancies including non-small cell lung cancer (NSCLC), colorectal cancer, ovarian cancer, oesophageal cancer and head and neck cancers [126–130]. The earliest reports date back more than three decades , and were confirmed again recently with studies that demonstrated an important prognostic role for CD8 + T cells  and memory T cells  for both progression free survival (PFS) and overall survival (OS) in colorectal cancer.
In contrast to CD8 + T cells , Treg infiltration has been associated with a worse prognosis when found in malignant ascites of patients with ovarian carcinoma correlating with higher stage and reduced survival . This association was also found in patients with pancreatic as well as hepatocellular carcinoma [135, 136]. However, high intratumoral Tregs is not always associated with a poorer prognosis, but can in some cases correlate with good prognosis, as demonstrated in several cancers including colorectal cancer, NHL and head and neck cancer [137–139]. Therefore more studies will be needed before the role of intratumoral Tregs is more clearly defined, especially given the added complexities of heterogeneity within tumors with regards to Treg location (center vs periphery), the ability of Tregs to lose FoxP3 expression and change phenotype or normal T cells to acquire FoxP3 without adopting a regulatory phenotype [140–142].
Low-dose cyclophosphamide has been shown to inhibit the regulatory function of CD4 + CD25+ Tregs and enhance the antitumor immune response in preclinical models by promoting CD4 + T helper type 1 immunity  and by facilitating the recruitment of latent CD8 + T cells when given before vaccination, mediating tumor rejection . Treg activity was also inhibited in mice using metronomic cyclophosphamide, paclitaxel or temozolamide [145–147].
The successful depletion of Treg using chemotherapy has also been demonstrated in clinical studies using metronomic cyclophosphamide in patients with late stage cancers  and paclitaxel in NSCLC . In patients with metastatic colorectal cancer, treatment with gemcitabine and FOLFOX4, followed by GM-CSF and IL-2 resulted in significant Treg reduction in 65 % of patients and was associated with a 70 % objective response rate to therapy [150, 151].
220.127.116.11 Myeloid Derived Suppressor Cells (MDSCs)
MDSCs are a diverse population of progenitor cells and immature myeloid cells that expand in cancer patients and can significantly inhibit T-cell responses  . MDSC have been shown to increase in patients with early breast cancer treated with dose-dense doxorubicin and cyclophosphamide (AC) , while gemcitabine eliminates MDSCs in tumor-bearing mice, enhancing CD8 + and NK cell activity [154, 155].
18.104.22.168 DNA-Damage Response and Activation of the Innate Immune System
NKG2D is an activating receptor involved in immune-surveillance by NK cells, NKT cells, gd T cells and CD8 + T cells . DNA-damaging agents, such as topoisomerase inhibitors, can initiate a complex DNA-damage response that involves the activation of tumor-suppressor proteins such as ataxia-telangiectasia mutated (ATM), checkpoint kinase 1 (CHK1) and the transcription factor p53. This can result in the ATM and CHK1-dependent, but p53 independent expression of NKG2D ligands . Although p53 is not required for the expression of NKG2D ligand in response to DNA damage , recent work has demonstrated that reactivation of p53 in hepatocellular cancer lead to expression of pro-inflammatory cytokines , chemokines and adhesion molecules which may contribute to p53-induced recruitment of NK cells, neutrophils and macrophages. This highlights the ability of chemotherapy-induced DNA damage to trigger an innate immune antitumor response .
So far, we have discussed the ability of gemcitabine to suppress Tregs and MDSCs, and to elicit immunogenic cell death and enhance DC-dependent cross-presentation of tumor antigens to T cells . Professional antigen presenting cells (APCs) include, in addition to DCs, B cells and macrophages. B cells are known to have a limited ability to internalize antigen as they do not exhibit a significant capacity for endocytosis . Moreover, there is some data to suggest that the immunogenicity of some tumors is limited by B cells that maybe competing for tumor derived antigen with other APCs such as DCs, therefore interfering with the generation of CD4 + help for cytotoxic lymphocyte mediated tumor immunity .
Although the role of B cells in tumor immunity remains elusive, we have found that in the context of gemcitabine treatment of tumor-bearing mice, gemcitabine can result in significant reduction of total lymphocytes, but is selectively detrimental to humoral immune responses by inhibition of B-cell proliferation and antibody production while it spares antigen-specific cellular immunity .
22.214.171.124 5-Fluorouracil (5FU)
The fluoropyrimidine 5FU is commonly used in gastrointestinal malignancies and in breast cancer. It has been shown to induce HSP expression in tumor cells in vitro, thereby promoting antigen uptake and cross-presentation by DCs . In vivo,the intratumoral inoculation of DCs after 5FU-based chemotherapy resulted in T-cell dependent eradication of the injection site as well as other sites and lead to long term survival of the treated mice . In another mouse model, 5FU enhanced the efficacy of a thymidilate synthase-directed vaccine .
1.5.3 Immune-Modulatory Effects of Targeted Agents
An increasing number of targeted therapies are now in clinical use in many solid tumors and haematological malignancies, with many more in clinical development. The different modes of action of these drugs compared to conventional chemotherapy has led to the observation of different side effect profiles as well off-target effects that may be contributory, or deleterious to their antitumor immune effects. There is accumulating evidence that many of these agents can modulate immune response by various mechanisms, including effects on innate immunity , on immunosuppressive cells such as MDSCs or Tregs, as well effector T cells and DCs.
Imatinib mesylate (Gleevec) is a small molecule tyrosine kinase inhibitor of KIT, PDGFR, ABL and BCR-ABL with demonstrated activity and improved survival benefit in patients with advanced gastrointestinal stromal tumor (GIST) . In mice, a combination of imatinib and IL-2 resulted in the expansion of a population of effector cells that were termed IFN-producing killer DCs (IKDCs) as they shared properties of both NK myeloid DCs and produce IFN-g. These CD11c + B220 + NK1.1 + IKDCs, unlike B220− NK cells, were able to lyse various target cells in the absence of NKG2D ligands or MHC class I molecules. Adoptive transfer of IKDCs but not B220− NK cells delayed tumor growth .
In patients with GIST that did not have KIT or PDGFRA mutations but still responded to imatinib, the secretion of NK-cell IFNg was found to constitute a positive prognostic factor, suggesting that NK cell-dependent antitumor effects may play a role in the efficacy of imatinib .
A recent study by Balachandran et al.  elegantly demonstrated the central role for the antitumor CD8 + T cell responses in imatinib-treated GIST tumors. In a model of transgenic GIST mice that develop spontaneous GIST due to an activating mutation in the Kit gene, treatment with imatinib resulted in an increase in CD8 + T cell frequency, proliferation, activation as well as cytolytic capacity within the tumor and an increase in tumor-specific CD8 + T cells within the draining, but not the non-draining lymph nodes. The antitumor effect was reduced in mice depleted of CD8 + but not CD4 +, NK cells or myeloid cells, and knockout mice lacking mature T and B cells had larger tumors than controls, whereas those lacking B cells only were not. Moreover, untreated mice depleted of CD8 + , but not of CD4 + , NK cells or B cells had larger tumors after 4 weeks. This demonstrated the pre-existing role of CD8 + -mediated immune response , which is enhanced by imatinib therapy. The authors also demonstrated a suppressive effect on Tregs within tumors, but not the draining lymph nodes, resulting from imatinib-induced apoptosis of Tregs through inhibition of tumor cell expression of the immunosuppressive enzyme indoleamine 2,3-dioxygenase (Ido). Importantly, a synergistic effect was seen in mouse GIST treated with imatinib and CTLA-4 blocking antibody compared to either drug alone, suggesting that additional immune activation could augment the antitumor effect of imatinib. Finally, the authors demonstrated a correlation between the findings in mouse GIST and human GIST by analyzing the blood and freshly obtained tumor GIST tissue form 36 patients undergoing surgery followed by either imatinib therapy or observation. A greater frequency of CD8 + , but lower of Tregs and of Ido mRNA was found in sensitive tumors compared to resistant tumors, therefore correlating with the preclinical data in mouse GIST .