Introduction

Immune therapy is often considered to be very “T cell chauvinistic”, with a major focus on strategies that efficiently will activate cytotoxic T lymphocytes (CTLs). There are several reasons for this; one being that CTLs are the only cell type which in a highly specific and efficient manner can kill their target cells unaided by other cell types. The focus on cell-mediated effector mechanisms also has a historic background, since for decades it was argued that the humoral immune system and so-called “enhancing antibodies” are interfering with T cell–mediated cytotoxicity [1]. This negative view on antibodies in tumor immunology initially slowed down the introduction of monoclonal antibodies (mAb) in tumor therapy. However, during the past couple of decades, this view has been replaced by a tremendous excitement in the area of mAb tumor therapy. Antibody therapies were initially mainly used for the treatment of hematological malignancies, but have lately also been used for therapy of solid tumors, where mAb’s such as Herceptin and Ipilimumab have been shown to prolong survival of breast cancer and advanced melanoma patients, respectively, particularly when combined with chemotherapy in a dual-targeted approach [2, 3]. Another historic reason for the focus on CTLs in immune therapy is the pioneering work on the identification of human tumor antigens, where tumor-specific CTLs were critical tools [4]. Early attempts to utilize the tumor-specific peptide epitopes from these antigens in pilot phase I clinical trials instilled great optimism in the field of tumor vaccines [5]. However, the over-all sentiment of this field is now that of disappointment, and in most of the clinical cancer vaccine trials, no or only a small number of cancer patients have demonstrated objective tumor regression [6]. This poor clinical outcome can be explained by sophisticated tumor immune escape mechanisms, which can be categorized into two major types. One category of escape mechanism is active at the level of the tumor and can be ascribed to abnormalities in the expression or processing of major histocompatibility complex (MHC) class I restricted antigens, enabling tumors to take on a “stealth” phenotype and hiding from detection by CTLs. The other category results from the ability of the progressing tumor to “sabotage” the host immune system. We will here exemplify both types of mechanisms by examples from the literature and recent observations in our laboratory.

Camouflage of the tumor by immune selection or as a consequence of malignant transformation?

Loss of MHC class I

Human tumors frequently have defects in MHC class I antigen presentation [7]. It has been argued that immune selection, favoring the outgrowth of MHC class I deficient tumor types in a manner analogous to selection of drug resistant tumor variants, could be a driving force behind this phenomenon. While this interpretation may be valid in individuals undergoing immune therapy, such as a in a study where progressing metastatic melanoma lesions were found to express low levels of MHC class I while regressing lesions expressed high MHC class I [8], it is unlikely to explain the majority of observations of defects in MHC class I expression in human tumors. As an example, defects in MHC class I antigen presentation have been described also among carcinomas, such as breast carcinoma [9], in which there is little if any evidence for immune surveillance. Thus, in a recent meta-analysis of cancer incidence in immunosuppressed transplant recipients and HIV/AIDS patients [10], cancer incidence was found to be generally higher mainly for cancers with a known or suspected viral etiology, while most common epithelial cancers such as breast cancer did not occur at increased rates.

Alternatively, defects in MHC class I expression and/or in the antigen processing and presentation machinery (APM) may be a direct consequence of the oncogenic process. In favor of this alternative, many oncogenes have been shown to interfere with antigen processing and presentation [11–14]. The oncogene Her2/neu (HER2) represents an example of this as tumors expressing high levels of HER2 have a “stealth” phenotype with generally low MHC class I expression and antigen presentation [15–18]. We have shown that silencing HER2 by siRNA in HER2+ breast carcinoma cell lines increased the expression of surface MHC class I [19]. In line with this, we found that the HER2 induced defects in APM components hamper the generation of HLA-A2-restricted HER2-derived epitopes in vivo and thereby impede recognition of the tumor by CTLs in a human HLA-A2 transgenic mouse model [20]. In addition, Herrmann et al. [21] found an inverse correlation between HER-2/neu and the peptide transporter associated with antigen processing (TAP) protein expression as determined by immunohistochemical analysis of a series of HER-2− and HER-2+ breast cancer specimens and proposed a functional link between deficient APM component expression and HER-2/neu overexpression.

One important question which follows from these observations is to what extent the HER2 high expressing tumor cell lines would be less visible for CTLs in general, i.e., also those which recognize tumor antigens other than those derived from HER2. We addressed this by transfecting HER2 low-expressing human melanoma and breast carcinoma lines with a full-length HER2 construct and confirmed that high expression of HER2 leads to markedly reduced expression levels of MHC class I and APM components [22]. As a novel finding and of considerable relevance to immunotherapy, tumors with high expression of HER2 were found to have a globally reduced capacity of being recognized by tumor antigen-specific CTLs. This was demonstrated using a HER2 high expressing melanoma transfectant, which was found to be relatively resistant to recognition by T cells specific for an MHC class I restricted epitope of the tyrosinase molecule. From these findings, we conclude that attempts to treat HER2 high carcinomas with CTLs, adoptively transferred or activated by tumor vaccines, would be less likely to succeed.

In spite of this, it is important to note that experimental models have shown that HER2 vaccines based on defined CTL epitopes can induce tumor protection in vivo [23, 24] and ongoing clinical trials based on HER2-derived CTL epitopes have shown some promising results [25, 26], although these need to be confirmed in larger controlled studies. It is therefore possible that vaccination methods optimized to yield MHC class I restricted T cells in great numbers or of high avidity can overcome the relatively poor antigen presentation in HER2 over-expressing tumors. Nevertheless, we are of the opinion that a safer approach of targeting HER2 high expressing tumors would be by activating an integrated immune response relying not only on CTLs but also on antibodies, CD4+ T cells, and NK cells.

Overcoming the camouflage strategy by activating an integrated immune response

A case in point was observed by us when testing an adenovirus (Ad)-based HER2 vaccine construct derived from the same kinase deficient HER2 DNA sequence (E2A) previously demonstrated to confer protection in experimental models [27, 28] and recently also tested by us in a clinical pilot trial as a plasmid DNA (pDNA) vaccine in patients with advanced breast cancer [29]. This Ad-E2A construct was able to induce tumor protection, most efficiently in wild type mice but also in mice transgenic for human HER2 and shown to be partially tolerant to this antigen [30]. Ad-E2A vaccine-induced tumor protection was independent of CD8+ T cells and was inactive in mice lacking B-cells or Fc-receptors. Purified syngeneic NK cells when admixed with sera from the vaccinated mice could efficiently kill HER2 expressing tumor targets but not HER2 negative control cells. Our conclusion was that NK cell-mediated antibody dependent cellular cytotoxicity (ADCC) played a major role in tumor protection. This pivotal importance of antibodies is in agreement with other reports on HER2 vaccine-induced tumor protection based on Ad-constructs or pDNA [31]. Thus, it seems that this mode of vaccination induces a protection analogous to that induced in a mouse pre-clincal model of Herceptin treatment, where the in vivo therapeutic effect of this antibody is diminished in the absence of Fc receptor signaling [32] which argues for a role for ADCC. The role of ADCC in the Herceptin-induced anti-tumor effect is also supported by results showing that Fc receptor (FcR) polymorphism is related to clinical outcome in Herceptin treated breast cancer patients [33] and that the quantity and lytic efficiency of CD16+ lymphocytes are important factors for ADCC induction by Herceptin [34].

It is important to underline, however, that the critical role for ADCC as an efficient effector mechanisms in anti-HER2 antibody-induced tumor protection does not preclude that also an adaptive immune response can be induced both by passive anti-HER2 antibody administration and by active HER2 vaccination. As for anti-HER2 antibody therapy, it was recently confirmed that the mechanisms of tumor regression require FcR as well as activation of innate immunity and T cells, initiated by the antibody treatment [35]. For efficient anti-HER2 tumor vaccination, there are recent examples of CD8+ T cell responses induced also in the HER2 transgenic mice [36]. Overcoming CD8+ T cell tolerance to HER2 may pose a greater challenge as compared to breaking tolerance in CD4+ T cells and inducing a humoral response, as also concluded in the Balb/neu T-model transgenic for mutant rat neu [37].

Sabotage of anti-tumor immunity by tumor-mediated immune suppression

In addition to tumor intrinsic immune escape mechanisms such as loss of MHC I, many tumors highjack parts of the immune system to protect themselves against immune attack. To this end, they induce or recruit immune-suppressive myeloid cells or regulatory T cells (Tregs), which normally serve as safeguards against overwhelming inflammation or autoimmunity. By turning the immune system against itself, tumors can gain an impressive arsenal of weapons to hamper the induction as well as the exhibition of anti-tumor immune activity.

Myeloid-derived suppressor cells and tumor associated macrophages

Myeloid cells with suppressive functions, such as myeloid-derived suppressor cells (MDSC) or tumor associated macrophages (TAM), have recently received much attention in the field of tumor immunology, as they have emerged as potent suppressors of T cell- and potentially NK cell-mediated immunity with a wide repertoire of effector mechanisms.

MDSC are a heterogeneous population of immature cells of myeloid origin that are characterized by their suppressive potential [38]. In mouse models, MDSC are typically identified as CD11b+GR1+ cells and separate detection of two epitopes covered by the GR1 antibody can be used to further define monocyte-like and granulocyte-like MDSC populations as Ly6C+Ly6G− and Ly6G+Ly6Clow, respectively. Additional markers sometimes used for the characterization of mouse MDSC include IL-4Rα, S100, and F4/80−/lo [39–41].

While some aberrant myeloid population can be detected in virtually every cancer that has been studied, the heterogeneity of MDSC in human malignancies is striking. To characterize them, most researchers could agree on myeloid markers such as CD33 and CD11b, but the described human MDSC phenotypes range from (i) CD34+ or Lin−DR−CD33+ myeloid precursors [42–44] to (ii) CD15+ granulocyte-like cells [45, 46] and (iii) cells resembling monocytes (CD14+HLADR−/low) [47–49].

One common denominator of this cell type in humans and mice is their dependence on tumor-derived factors, usually causing them to increase with tumor progression and decrease with successful treatment [50, 51]. Naturally, the combinations of cytokines produced by different tumors are diverse, which can help to explain the heterogeneous appearance of MDSC induced by various cancers. In addition, the number of potential MDSC-inducing and MDSC-activating factors is large, including vascular endothelial growth factor (VEGF), interleukin (IL) 6, prostaglandin E2, IL-1β, stem cell factor (SCF), macrophage- + granulocyte macrophage colony–stimulating factors (M-CSF and GM-CSF), and interferon (IFN) γ, IL-4, IL-13, transforming growth factor (TGF) β or toll-like receptor ligands [52], respectively.

MDSC can exhibit an array of suppressive functions including (i) depletion of amino acids essential for T cell function [45, 46, 53, 54], (ii) production of suppressive cytokines [55–57], (iii) interference with T cell homing [58], (iv) collaboration with Tregs [48, 59–61], (v) contribution to tumor angiogenesis [57] as well as (vi) production of reactive nitrogen and oxygen species (ROS) [62, 63]. There are also some reports indicating that MDSC can suppress NK cell function [56, 64, 65], while Nausch et al. [66] reported that MDSC stimulate NK cell IFNγ production.

MDSC appear to be closely related to TAMs, which usually exhibit M2 polarization and can contribute to tumor progression and immune suppression by producing IL-10, TGFβ, and pro-angiogenic factors such as matrix metalloproteases, VEGF, and platelet-derived growth-factor (PDGF) [67]. Recent evidence from mouse models suggests that MDSC can differentiate into TAMs upon reaching the hypoxic environment of the tumor and thereafter display distinct phenotypical and functional characteristics [68].

We were among the first to show that myeloid cell-derived ROS inhibit tumor-specific T cell- and NK cell-mediated cytotoxicity [69]. We and others had previously observed that lymphocytes from tumor bearing mice and cancer patients frequently exhibited functional defects due to the down-regulation of the ζ-chain [70–73], which is important for signal transduction through the T cell receptor (CD3ζ) and FcR (CD16ζ). Decreased expression of the ζ-chain could be reproduced in vitro by co-culturing lymphocytes with macrophages derived from metastatic lesions of melanoma patients or activated monocytes and could be prevented by scavenging myeloid cell ROS production through addition of the H2O2 neutralizing enzyme catalase [74].

Further, recently we observed that addition of catalase to peripheral blood mononuclear cells derived from patients with advanced malignant melanoma could restore T cell proliferation in some, but not all cases [49]. We found that these patients had increased frequencies of CD14+HLA-DR−/low cells compared to healthy controls. These CD14+HLA-DR−/low cells, hereafter referred to as mel-MDSC, were potent suppressors of autologous T cell proliferation and IFNγ production. They exhibited signs of high oxidative stress and increased mRNA levels of Arginase1 (Arg1). Arg1 is an arginine utilizing enzyme which, when highly expressed, can induce T cell dysfunction by causing CD3ζ-chain down-regulation as a result of arginine starvation. In addition, the nitrogen radical NO is a byproduct of Arg1 metabolism and ROS can be produced when Arg1 is co-expressed with inducible nitric oxide synthase (iNOS), another arginine metabolizing enzyme. Consequently, mel-MDSC-mediated suppression of T cell proliferation could also be inhibited by adding the Arg1 inhibitor nor-NOHA. In agreement with the notion that MDSC often utilize several suppressive mechanisms at once, neither ROS scavenging nor Arg1 inhibition could completely restore proliferation or improve IFNγ production. On the other hand, blocking the Stat3 signaling pathway abolished the ability of mel-MDSC to interfere with T cell proliferation and function. Interestingly, we found that total and phosphorylated Stat3 was increased in mel-MDSC and observed a correlation of Stat3 activation and oxidative stress levels. Together with evidence identifying Stat3 is a key molecule for MDSC differentiation and suppressive function in mouse models [75, 76], these findings highlight the Stat3 pathway as a promising therapeutic target to relieve MDSC-induced T cell suppression in melanoma patients.

Regulatory T cells

Regulatory T cells (Tregs) were initially described in the 1970s [77] but assigning them a characteristic phenotype as CD4+CD25+ cells [78] was necessary for their recognition as key players in tumor immunology. Since then, increased levels of Tregs have been duly noted in many types of cancer (for a recent review see [79]) and were often found to correlate with a poor disease course and prognosis. It is widely accepted that tumor immunosurveillance regularly resembles autoreactivity, and Tregs, a dominant factor in peripheral tolerance, can act as a double-edged sword [80]. Eliminating Tregs in vivo in several pre-clinical cancer models as well as in vitro in patient’s specimen has boosted inherent anti-tumor immunity and unmasked several TAA directed responses [81], but could potentially even result in autoimmune side effects [82].

To date based on their ontogeny, Tregs are subdivided into two major groups: the naturally occurring forms deriving from the thymus and the adaptive variants generated in the periphery. The list of suppressive lymphocyte subsets is steadily growing and phenotype as well as mode of action can vary significantly among them [83]; the CD4+CD25+CD127low/negFOXP3+ cells of thymic origin and the induced IL10+ Tr1 and TGF-β+ Th3 as of yet represent the best characterized subpopulations. Consensus exists that the conversion of conventional T cells into Tregs requires a certain cytokine composition (mainly TGF-β and IL-10) and is often preceded by a suboptimal antigen presentation (by, e.g., immature dendritic cells (DCs) or MDCS). Even though we did not observe a correlation between the frequency of mel-MDSC and Tregs in the blood of melanoma patients, nor an expansion of this population after mel-MDSC T cell co-culture (unpublished), other reports indicate that MDSC and Tregs might collaborate in different ways. Hoechst et al. [48] have shown that MDSC isolated from hepatocellular carcinoma patients favor in vitro expansion of Tregs and several investigators reported that antigen-specific Tregs were induced by MDSC [59, 60, 84] or aided MDSC suppressiveness [61] in tumor bearing mice.

The suppressive repertoire of Tregs is very broad and ranges from contact-dependent mechanisms to the secretion of suppressive molecules by which they efficiently blunt T cell-, NK cell-, and DC-mediated immune responses (reviewed in [85]). A novel approach is undertaken by investigating so-called Treg-mediated metabolic disruption. Two very prominent paradigms for this concept include the depletion of ATP and IL-2 by Tregs, where Tregs remove a physiological stress signal (the ATP [86]) and at the same time outcompete activated conventional T cells for IL-2, driving them to starvation and ultimately cell death [87, 88]. Tumor cells in conjunction with their cellular bystanders, including TAMs and MDSCs, generate vast amounts of ROS [89]. Several effector cells of the immune system demonstrate a remarkable sensitivity toward ROS, which is reflected by an increased induction of anergy and apoptosis [90–92]. This phenomenon marks a substantial hindrance for the inherent immunosurveillance as well as the efficiency of immunotherapies [93]. On the other hand, we recently observed that human Tregs are remarkably resilient toward ROS-mediated impairments, which consequently presents them with a major survival advantage within the tumor microenvironment [94]. One of the underlying mechanisms for their reduced sensitivity is the possession of a more robust Thioredoxin (Trx) system [95]. Trx is a key cellular antioxidant acting as a direct scavenger of ROS and also regulating other redox-relevant molecules. Interestingly, Trx helped the Tregs to maintain their density of surface thiols that represent the first and pivotal line of defense against ROS. Notably, all cell subsets that exhibit an increased resistance to ROS, like CD56bright NK cells [96], Tregs [97], and myeloid DCs [97] share a superior surface thiol-“shield” as a common attribute. Recent studies have also shown that Tregs require oxidative stress to be suppressive [98] and actively disturb the redox-balance of other immune cells by inter alia inhibiting the synthesis of non-enzymatic antioxidants [99, 100]. These findings could explain how Treg suppression can be so potent in a milieu that is hostile for conventional T cells. Furthermore, they suggest that in addition to recruitment, expansion and conversion from conventional T cells selective enrichment might be one key mechanism responsible for the Treg accumulation observed in many malignancies. Future studies are needed to show whether a targeted redox-modulation of the tumor microenvironment may skew immune responses from tolerance toward reactivity.

Tumor-mediated immune suppression: a challenge to successful immunotherapy

The findings presented above and many other studies emphasize the need to acknowledge immune-mediated escape mechanisms as a significant hurdle for anti-tumor immunity. Great efforts have been undertaken to deplete Tregs either by chemotherapy [101, 102] or by direct depletion [103, 104], which is hampered by the lack of truly Treg specific markers. Recently, studies attempting to unleash the immune system by blocking inhibitory checkpoints such as CTLA-4 with the monoclonal antibody ipilimumab have shown promising results [105–107].

We believe that additional efforts to either directly target MDSC or their suppressive mechanisms such as oxidative stress will aid the development of successful combination-immuno-therapies. Due to their immature nature, a potential way to remove MDSC is to force them to differentiate, for example, by using all-trans retinoic acid (ATRA) or vitamin D3, which promotes myeloid differentiation and has been clinically applied [108–111]. Another option is to deplete MDSC, which again is difficult to achieve due to the lack of a specific marker. However, MDSC are sensitive to certain types of chemotherapy [112, 113] which can be used to kill MDSC at the cost of wiping out additional parts of the myeloid compartment. Of course many approaches are under way to (i) prevent MDSC expansion, e.g., by targeting VEGFR or SCF1 [114–117] (ii) inhibit MDSC function by preventing the upregulation of Arg1 or iNOS [118–120] or decreasing ROS production [121].

In order to generally address the immune deficiency associated with oxidative stress in cancer, we have treated colorectal cancer patients with the antioxidant vitamin E. Short-term supplementation with high dose dietary vitamin E boosted immunity by increasing production of Th1 cytokines and NK cell cytolytic activity in the majority of patients and might therefore be supportive of immunotherapeutic regimens [122, 123].

Interestingly, in an effort to improve DNA vaccination approaches, we found that mice vaccinated with DNA-dependent activator of interferon regulatory factors (DAI) plus antigen displayed lower frequencies of splenic MDSC than animals vaccinated with the antigen plasmid alone [124]. DAI is a cytosolic pattern recognition receptor, conferring adjuvant activity by inducing proinflammatory cytokines and improving T cell responses, but potentially also by relieving immune suppression.

In view of the many ongoing clinical trials using genetically modified T cells, direct engineering of effector T cells to become resistant to oxidative stress becomes an interesting option. We demonstrated that retroviral transduction with the ROS scavenger catalase made effector T cells more resistant to hydrogen peroxide [125]. Catalase expressing T cells remained better able to respond to their cognate antigen when exposed to oxidative stress compared with untransduced T cells, suggesting a strategy to arm tumor antigen-specific T cells prior to adoptive T cell transfer.

Conclusion

The difficulties in developing efficient immune therapy against cancer are caused by two distinct phenomena; one acting at the level of the tumor target and the other at the level of the host immune system. We have pointed out that immune escape by the tumor cells may not necessarily be due to “editing” of the tumor by the immune system, but could rather be a consequence of the tumor transformation. If tumor immune escape is at least partially driven by oncogens such as HER2, the key to novel and more effective anti-cancer therapies will be to induce integrated immune responses which do not rely on one single effector mechanism. A clinical trial in that direction is the aforementioned pDNA pilot trial in patients with advanced breast cancer [29]. All three long-term survivors in this pDNA trial developed a “late” CD4+ T cell response as measured against 4 HER2 HLA-DR epitopes and some of the patients also produced endogenous anti HER2 antibodies. As an alternative to this approach and if alternations in the tumors are “soft,” meaning reversible and not hard-wired in the tumor genome [126], one may try to combine immune therapies with approaches that make the tumor visible to the immune system. This could be done, e.g., by inducing re-expression of co-stimulatory molecules or MHC class I molecules on the tumor through cytokines treatment.

Yet another approach would be to pre-select only those patient with tumors that express low or intermediate levels of HER2, avoiding those where high HER2 has down-regulated MHC class I which would be less likely to respond to treatment. Arguing for the validity of this pre-selection approach, Bebavides et al. [127] reported that low HER2 expressing patients (IHC: 0 or 1+) respond better to the HER2 E75 vaccine than overexpressing patients (IHC: 2+ or 3+), because of elements of immunologic tolerance in the latter group. Concurrently, HER2 low-expressing patients with higher expression of MHC class I may be more suitable for peptide vaccination than HER2 over-expressing patients, since vaccination therapy requires moderate HER2 expression for CTL induction and low levels or even intracellular localization of HER2 suffice for CTL recognition [127]. We suggest that the HER2 low-expressing group, which represent >50% of breast cancer patients, will have greater clinical benefit from the HER2 E75 vaccine.

One may question if the second set of phenomena, when tumors “highjack” parts of the immune system to protect themselves against immune attack, should be discussed in the context of immune escape. These phenomena become stronger as the tumor becomes clinically established and are usually reversible upon removal of the tumor [50, 51]. Rather than being a mechanisms by which early clonally expanding tumor cells escape immune eradication these phenomena should rather be looked upon as a consequence of the “smoldering” inflammation caused by the progressing tumor. The term “tumor-induced immune suppression” is an appropriate term for these mechanisms we have discussed above, represented by MDSC’s, Tregs, and oxidatives stress. The examples given clearly demonstrate that such tumor-induced immune suppression represents a major hurdle for immunotherapy. A number of ways to target any of the reported suppressive mechanisms exist and are being explored, but the variety of cellular populations and their suppressive mediators involved will make it difficult for any one strategy to be truly efficient. Also, suppressive mechanisms vary between cancers and might even be different between individuals, requiring truly tailor-made combinations of immunotherapeutic and conventional anti-cancer agents.

In conclusion, we suggest that immunotherapeutic approaches need to consider the molecular makeup of the tumor as well as the tumor micro-environment and aim to trigger a multi-facetted immune response involving humoral, cellular, and innate immunity.