Classification of current anticancer immunotherapies

During the past decades, anticancer immunotherapy has evolved from a promising therapeutic option to a robust clinical reality. Many immunotherapeutic regimens are now approved by the US Food and Drug Administration and the European Medicines Agency for use in cancer patients, and many others are being investigated as standalone therapeutic interventions or combined with conventional treatments in clinical studies. Immunotherapies may be subdivided into “passive” and “active” based on their ability to engage the host immune system against cancer. Since the anticancer activity of most passive immunotherapeutics (including tumor-targeting monoclonal antibodies) also relies on the host immune system, this classification does not properly reflect the complexity of the drug-host-tumor interaction. Alternatively, anticancer immunotherapeutics can be classified according to their antigen specificity. While some immunotherapies specifically target one (or a few) defined tumor-associated antigen(s), others operate in a relatively non-specific manner and boost natural or therapy-elicited anticancer immune responses of unknown and often broad specificity. Here, we propose a critical, integrated classification of anticancer immunotherapies and discuss the clinical relevance of these approaches.


INTRODUCTION
Our perception of cancer has changed dramatically during the past 3 decades. For instance, it has been appreciated that tumors are not a purely clonal disorder, although in some cases they do evolve from a single (pre-)malignant cell [1][2][3]. It is now clear that established neoplasms do not consist only of transformed cells, but contain an abundant and heterogeneous non-transformed component, including stromal, endothelial and immune cells [4][5][6]. We no longer consider the metabolism of cancer cells as completely distinct from that of their normal counterparts [7][8][9]. We have shown that the survival of transformed cells can critically depend on adaptive responses that per se are non-tumorigenic, establishing the concept of non-oncogene addiction [10,11]. We discovered mechanisms other than intrinsic apoptosis that may be harnessed for therapeutic applications, such as several forms of regulated necrosis [12][13][14]. Finally, we obtained evidence indicating that the host immune system can recognize (and sometimes react against) (pre-) malignant cells as they transform, proliferate, evolve and respond to therapy, founding the theoretical grounds of anticancer immunosurveillance [15][16][17]. These conceptual shifts have profound therapeutic implications, some of which have already been translated into clinical realities. For instance, several anticancer agents that are now approved by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) for use in cancer patients inhibit tumor-associated angiogenesis, perhaps the best characterized interaction between malignant and nonmalignant components of the tumor microenvironment [18,19].
Over the last decade, great efforts have been dedicated to the development of interventions that mediate antineoplastic effects by initiating a novel or boosting an existing immune response against neoplastic cells (Table  1) [20][21][22][23][24][25][26][27][28][29][30][31][32]. This intense wave of preclinical and clinical investigation culminated with the approval of various immunotherapeutic interventions for use in humans ( Table 2). In 2013, the extraordinary clinical success of immunotherapy was acknowledged by the Editors of Science Magazine with the designation of "Breakthrough of the Year" [33]. Nonetheless, we have just begun to unravel the therapeutic possibilities offered by anticancer immunotherapy. Clinical studies are being initiated at an ever accelerating pace to test the safety and efficacy of various immunotherapeutic regimens in cancer patients, either as standalone interventions or combined with other antineoplastic agents [34]. The hopes generated by this approach are immense, and several other forms of immunotherapy are expected to obtain regulatory approval within the next few years ( Figure 1). Anticancer immunotherapies are generally classified as "passive" or "active" based on their ability to (re-) activate the host immune system against malignant cells [35]. From this standpoint, tumor-targeting monoclonal antibodies (mAbs) and adoptively transferred T cells (among other approaches) are considered passive forms of immunotherapy, as they are endowed with intrinsic antineoplastic activity [23,24,36,37]. Conversely, anticancer vaccines and checkpoint inhibitors exert anticancer effects only upon the engagement of the host immune system, constituting clear examples of active immunotherapy [22,27,28,32,38]. An alternative classification of immunotherapeutic anticancer regimens is based on antigen-specificity. Thus, while tumortargeting mAbs are widely considered antigen-specific interventions, immunostimulatory cytokines or checkpoint blockers activate anticancer immune responses of unknown (and generally broad) specificity [27,[39][40][41][42]. Herein, we critically revise these classifications while discussing the clinical relevance of various forms of anticancer immunotherapy.

Tumor-targeting mAbs
Tumor-targeting mAbs are the best-characterized form of anticancer immunotherapy, and perhaps the most widely employed in the clinic [43][44][45][46]. The expression "tumor-targeting" refers to mAbs that (1) specifically alter the signaling functions of receptors expressed on the surface of malignant cells [47][48][49]; (2) bind to, and hence neutralize, trophic signals produced by malignant cells or by stromal components of neoplastic lesions [50, 51]; (3) selectively recognize cancer cells based on the expression of a "tumor-associated antigen" (TAA), i.e., an antigen specifically (or at least predominantly) expressed by transformed cells but not (or at least less so) by their non-malignant counterparts [30,52]. Tumor-targeting mAbs exist in at least 5 functionally distinct variants. First, naked mAbs that inhibit signaling pathways required for the survival or progression of neoplastic cells, but not of their non-malignant counterparts, such as the epidermal growth factor receptor (EGFR)-specific mAb cetuximab, which is approved by the US FDA for the treatment of head and neck cancer (HNC) and colorectal carcinoma (CRC) [47,48,53]. Second, naked mAbs that activate potentially lethal receptors expressed on the surface of malignant cells, but not of their non-transformed counterparts, such as tigatuzumab (CS-1008), a mAb specific for tumor necrosis factor receptor superfamily, member 10B, (TNFRSF10B, best known as TRAILR2 or DR5) that is currently under clinical development [49,54]. Third, immune conjugates, i.e., TAA-specific mAbs coupled to toxins or radionuclides, such as gemtuzumab ozogamicin, an anti-CD33 calicheamicin conjugate currently approved for use in acute myeloid leukemia patients [55,56]. Fourth, naked TAA-specific mAbs that opsonize cancer cells and hence activate antibodydependent cell-mediated cytotoxicity (ADCC) [44, 57-59], antibody-dependent cellular phagocytosis [60], and complement-dependent cytotoxicity [61], such as the CD20-specific mAb rituximab, which is currently approved for the treatment of chronic lymphocytic leukemia (CLL) and non-Hodgkin lymphoma [62,63]. Fifth, so-called "bispecific T-cell engagers" (BiTEs), i.e., chimeric proteins consisting of two single-chain variable fragments from distinct mAbs, one targeting a TAA and one specific for a T-cell surface antigen (e.g., blinatumomab, a CD19-and CD3 BiTE recently approved for the therapy of Philadelphia chromosome-negative precursor B-cell acute lymphoblastic leukemia) [64][65][66][67][68][69].
The therapeutic activity of opsonizing mAbs and BiTEs clearly relies on the host immune system, implying that these molecules should be considered active immunotherapeutics. Conversely, tumor-targeting mAbs of the first two classes are endowed with intrinsic antineoplastic activity, and have been considered for a long time as passive forms of immunotherapy. However, growing evidence indicates that the actual antineoplastic potential of these molecules does not simply reflect their direct tumor-inhibitory activity, but also involves (at least to some degree) the activation of an anticancer immune response. For instance, cetuximab does not only inhibit EGFR signaling [53], but also promotes ADCC [70], and mediates immunostimulatory effects [71,72]. Similarly, bevacizumab, a vascular endothelial growth factor A (VEGFA)-neutralizing mAb approved for the treatment of glioblastoma multiforme, CRC, as well as cervical carcinoma, renal cell carcinoma (RCC) and lung carcinoma, not only exerts anti-angiogenic effects [50,73], but also boosts tumor infiltration by B and T lymphocytes, [74,75], while inhibiting CD4 + CD25 + FOXP3 + regulatory T cells (Tregs) [76]. Moreover, polymorphisms in the genes coding for the receptors mainly responsible for ADCC, i.e., Fc fragment of IgG, low affinity IIa, receptor (FCGR2A, also known as CD32) and FCGR3A (also known as CD16a), have been shown to influence the response of cancer patients to most tumor-targeting mAbs [77]. Thus, it is possible (although not formally demonstrated) that tumor-targeting mAbs operate as active immunotherapeutics. Irrespective of this possibility, 18 distinct tumor-targeting mAbs are currently approved by the US FDA for use in cancer patients (source http://www.

Adoptive cell transfer
The term "adoptive cell transfer" (ACT) refers to a particular variant of cell-based anticancer immunotherapy that generally involves: (1) the collection of circulating or tumor-infiltrating lymphocytes; (2) their selection/ modification/expansion/activation ex vivo; and (3) their (re-)administration to patients, most often after lymphodepleting pre-conditioning and in combination with immunostimulatory agents [23, 24, 78-80]. Other anticancer (immune)therapies involving the (re) infusion of living cells, such as hematopoietic stem cell transplantation (HSCT), conceptually differ from ACT. ACT involves the (re-)introduction of a cell population enriched in potentially tumor-reactive immune effectors [23,24,81]. HSCT is employed as a means to reconstitute a healthy, allogeneic (and hence potentially tumorreactive) immune system in patients with hematological malignancies previously subjected to myelo-and lymphoablating treatments (which aim at eradicating the majority of neoplastic cells) [82]. Dendritic cell (DC)-based interventions should also be conceptually differentiated from ACT for two reasons. First, (re-) infused DCs are not endowed with intrinsic anticancer activity, but act as anticancer vaccines to elicit a tumortargeting immune response [83,84]. Second, DCs are not administered in the context of lympho/myeloablating chemo(radio)therapy [85][86][87].
Several strategies have been devised to improve the therapeutic potential of ACT [79,80,88]. For instance, genetic engineering has been employed to endow peripheral blood lymphocytes (PBLs) with features such as a unique antigen specificity [89], an increased proliferative potential and persistence in vivo [90-93], including tumor-targeting and immunomodulatory monoclonal antibodies (mAbs); dendritic cell (DC)-, peptide-and DNA-based anticancer vaccines; oncolytic viruses; pattern recognition receptor (PRR) agonists; immunostimulatory cytokines; immunogenic cell death inducers; inhibitors of immunosuppressive metabolism; and adoptive cell transfer. 1MT, 1-methyltryptophan; APC, antigen-presenting cell; IDO, indoleamine 2,3-dioxigenase; IFN, interferon; IL, interleukin; IMiD, immunomodulatory drug; NLR, NOD-like receptor; TLR, Toll-like receptor. Since (re-)infused T cells are endowed with intrinsic antineoplastic activity, ACT is generally considered as a passive form of immunotherapy. However, the survival, expansion, migration and cytotoxic activity of adoptively transferred T cells rely on several cytokines, some of which are supplied by the host immune system. Current ACT protocols involve indeed the administration of exogenous interleukins (ILs), including IL-2, IL-15 or IL-21 [126-130], but these stimulate a cytokine cascade in the host that sustains the survival and activity of adoptively transferred cells. Thus, ACT may not represent a bona fide paradigm of passive immunotherapy.

Oncolytic viruses
The term "oncolytic viruses" refers to nonpathogenic viral strains that specifically infect cancer cells, triggering their demise [131][132][133]. Oncolytic viruses must be conceptually differentiated from so-called "oncotropic viruses", i.e., viruses that exhibit a preferential tropism for malignant cells but no (or very limited) cytotoxic activity [134,135]. The antineoplastic potential of oncolytic viruses can be innate and simply originate from the socalled cytopathic effect, i.e., the lethal overload of cellular metabolism resulting from a productive viral infection [136,137]. As an alternative, these viruses can mediate an oncolytic activity because of (endogenous or exogenous) gene products that are potentially lethal for the host cell, irrespective of their capacity to massively replicate and cause a cytopathic effect [131,132]. Of note, genetic engineering has been successfully employed to endow oncolytic virus with various advantageous traits, including sequences coding for (1) enzymes that convert an innocuous pro-drug into a cytotoxic agent [138-143]; (2) proteins that (at least theoretically) trigger lethal signaling cascades in cancer cells only [144-146]; or (3) shorthairpin RNAs that target factors that are strictly required for the survival of transformed, but not normal cells [147,148]. Of note, no oncolytic virus has been approved by the US FDA for use in cancer patients (source http:// www.fda.gov). Conversely, a recombinant adenovirus (H101, commercialized under the name of Oncorine ® ) has been approved by the regulatory authorities of the People's Republic of China for the treatment of HNC (in combination with chemotherapy) as early as in November 2005 [149, 150].
As oncolytic viruses are endowed with intrinsic anticancer activity, they are generally viewed as passive immunotherapeutics. Moreover, several effectors of innate and adaptive immunity limit the efficacy of oncolytic therapy because they can neutralize viral particles before they reach neoplastic lesions [131,132,151]. This is particularly true for the mononuclear phagocytic system of the liver and spleen, which is able to sequester large amounts of oncolytic viruses upon injection [152,153]; the complement system, to which oncolytic viruses are particularly sensitive [154,155]; and neutralizing antibodies, which can exist in patients prior to oncolytic virotherapy owing to their exposure to naturally occurring variants of the viral strains commonly employed for this purpose [156,157]. This being said, accumulating preclinical and clinical evidence indicates that the therapeutic activity of oncolytic viruses stems, for the most part, from their ability to elicit tumor-targeting immune responses as they promote the release of TAAs in an immunostimulatory context. In support of this notion, oncolytic viruses engineered to drive the expression of co-stimulatory receptors [158][159][160] or immunostimulatory cytokines/chemokines [161][162][163][164][165] reportedly mediate superior antineoplastic effects as compared to their unmodified counterparts [131,132]. Thus, conventional oncolytic viruses also appear to be active, rather than passive, immunotherapeutics.
Although DCs isolated from cancer patients have been shown to exert cytotoxic activity against malignant cells [226], DC-based immunotherapies mediate antineoplastic effects mainly because they engage the host immune system against malignant lesions [227,228]. Thus, all forms of DC-based anticancer interventions constitute paradigms of active immunotherapy.

Peptide-and DNA-based anticancer vaccines
DCs and other antigen-presenting cells (APCs) are also targeted by peptide-and DNA-based anticancer vaccines [83,84,[229][230][231]. In the former scenario, full-length recombinant TAAs or peptides thereof are administered to cancer patients, most often via the intramuscular, subcutaneous or intradermal route, together with one or more immunostimulatory agents commonly known as adjuvants (which potently promote DC maturation) [232][233][234][235][236][237]. The rationale behind this approach is that resident DCs (or other APCs) acquire the ability to present the TAA-derived epitopes while maturing, hence priming a robust TAA-specific immune response [32, 238,239]. The mechanisms underlying the priming of anticancer immune responses by peptide-based vaccines, and hence their efficacy, depend (at least in part) on their size [38]. Thus, while short peptides (8-12 amino acids) are conceived to directly bind to MHC molecules expressed on the surface of APCs, synthetic long peptides (25-30 residues) must be taken up, processed and presented by APCs for eliciting an immune response [38]. Normally, the therapeutic activity of synthetic long peptides is superior to that of their short counterparts, especially when they include epitopes recognized by both cytotoxic and helper T cells or when conjugated to efficient adjuvants [38, 240,241]. This said, some commonly used immunostimulants such as the so-called incomplete Freund's adjuvant (IFA) have recently been shown to limit the efficacy of peptidebased anticancer vaccination [242], calling for the use of alternative immunostimulants. A peculiar type of peptidebased vaccines is constituted by autologous tumor lysates complexed with immunostimulatory chaperones, most often members of the heat-shock protein (HSP) family [243]. This approach is advantageous in that it does not rely on a single TAA but (at least hypothetically) on all TAAs that bind to HSPs (including patient-specific neo-TAAs) [243]. However, generating anticancer vaccines on a personalized basis is associated with considerable costs [243].
DNA-based anticancer vaccines rely on TAAcoding constructs, be them naked or vectored (by viral particles, non-pathogenic bacteria or yeast cells) [32, [244][245][246]. DNA-based vaccines either become a source of such TAA (as it is the case for bacterial and yeast vectors) or transform APCs or muscular cells to do so (as it is the case for naked constructs and viral vectors) [32, [244][245][246][247]. Theoretically, and especially in the presence of adequate adjuvants, this prompts resident DCs or other APCs to prime a TAA-targeting immune response [32, 183,248,249]. A particularly interesting approach in this context is represented by so-called "oncolytic vaccines", i.e., oncolytic viruses genetically altered to code for a TAA [250][251][252]. Promising results have also been obtained with DNA-based vaccines administered per os [253][254][255][256]. In this setting, live-attenuated bacteria expressing a fulllength TAA are taken up by APCs in the intestinal mucosa, resulting in the priming of a robust, TAA-specific immune response in the so-called "mucosa-associated lymphoid tissue" [253][254][255][256].
Both peptide-and DNA-based vaccines have been associated with clinical activity in patients affected by various neoplasms [83,84,[229][230][231]257]. For instance, a peptide-based vaccine targeting the human papillomavirus type 16 (HPV-16) proteins E6 and E7 have been shown to promote complete, long-lasting responses in a significant fraction of patients with vulvar intraepithelial neoplasia [258]. Along similar lines, the administration of a multipeptide vaccine after single-dose cyclophosphamide (an immunogenic alkylating agent, see below) has been shown to prolong overall survival in a cohort of RCC patients [259]. No peptide-or DNA-based anticancer vaccine is currently approved by the US FDA and EMA for use in humans (sources http://www.fda.gov and http://www.ema.europa.eu/ema/). However, vitespen (Oncophage ® ), a heat shock protein 90kDa beta (Grp94), member 1 (HSP90B1)-based anticancer vaccine, has been approved in Russia for the treatment of RCC patients with intermediate risk of recurrence as early as in 2008 [257]. Moreover, three DNA-based anticancer vaccines have been licensed for veterinary use [260][261][262][263], one of which relies on a human TAA (i.e., tyrosinase) [263].
Similar to DC-based interventions, both peptide-and DNA-based anticancer vaccines mediate antineoplastic effects as they (re-)activate the host immune system against malignant cells, hence constituting active forms of anticancer immunotherapy.

Immunostimulatory cytokines
Taken as a family, cytokines regulate (via autocrine, paracrine or endocrine circuits) virtually all biological functions [264][265][266][267]. It is therefore not surprising that various attempts have been made to harness the biological potency of specific cytokines to elicit novel or reinvigorate pre-existent tumor-targeting immune responses [268][269][270][271]. The administration of most immunostimulatory cytokines to cancer patients as standalone therapeutic interventions, however, is generally associated with little, if any, clinical activity [272][273][274][275]. Thus, immunostimulatory cytokines are generally employed as adjuvants for other anticancer (immuno)therapeutics, either as recombinant molecules or encoded within expression vectors [276][277][278][279][280][281][282][283][284]. Notable exceptions include interferon (IFN)-α2b (also known as Intron A ® ), and IL-2 (also known as aldesleukin and Proleukin ® ), which mediate single agent therapeutic activity in patients affected by melanoma, a tumor type particularly sensitive to immunotherapy [274,284]. IFN-α2b is currently approved by the US FDA and EMA for the therapy of hairy cell leukemia (HCL), AIDSrelated Kaposi's sarcoma, follicular lymphoma, multiple myeloma, melanoma, external genital/perianal warts (condylomata acuminata) and cervical intraepithelial neoplasms (both as a recombinant, unmodified protein, and as a pegylated variant), while IL-2 is licensed for the treatment of metastatic forms of melanoma and RCC. Moreover, IFN-α2a (also known as Roferon-A ® ) is approved for use in subjects with HCL and chronic phase, Philadelphia chromosome-positive chronic myeloid leukemia, upon minimal pretreatment (within 1 year of diagnosis). In Europe, IFN-α2a is also licensed for the treatment of melanoma. Of note, GM-CSF (also known as molgramostim, sargramostim, Leukomax ® , Mielogen ® or Leukine ® ) and granulocyte colony-stimulating factor (G-CSF, also known as filgrastim, lenograstim or Neupogen ® ) are approved by the US FDA and EMA for use in humans, but not as part of anticancer regimens [285][286][287][288]. Nonetheless, GM-CSF has been shown to potentiate the clinical activity of several immunotherapeutics, including (but not limited to) peptide-based vaccines and immunomodulatory mAbs [259,289]. Recombinant tumor necrosis factor α (TNFα) is also licensed by several regulatory agencies worldwide (but not by the US FDA), for the treatment of limb-threatening soft tissue sarcoma and melanoma [290][291][292]. However, in this setting TNFα is not employed as an immunostimulatory agent but administered in combination with melphalan (an alkylating agent) to increment the local concentration of the drug (and hence boost its cytotoxicity), and to promote the selective destruction of the tumor vasculature [293].
The antineoplastic activity of immunostimulatory cytokines is expected to depend on the host immune system, implying that they underlie a bona fide paradigm of active immunotherapy. However, the actual mode of action of immunostimulatory cytokines has not yet been fully explored. Moreover, some of these agents may promote a cytokine cascade with unwarranted, potentially lethal effects, and hence should be employed with caution.
The first of these approaches, which is commonly referred to as "checkpoint blockade", has been shown to induce robust and durable responses in cohorts of patients with a variety of solid tumors [39, 300, 321-327]. As it stands, no less than three checkpoint-blocking mAbs are currently approved by international regulatory agencies for use in humans (source http://www.fda.gov): (1) the anti-CTLA4 mAb ipilimumab (Yervoy™), which was licensed by the US FDA for use in individuals with unresectable or metastatic melanoma on 2011, March 25 th [328][329][330][331][332]; the anti-PD-1 mAb pembrolizumab (Keytruda™), which received accelerated approval by the US FDA for the treatment of advanced or unresectable melanoma patients who fail to respond to other therapies on 2014, September 4 th [333][334][335][336][337][338]; and nivolumab (Opvido™), another PD-1targeting mAb licensed by the Japanese Ministry of Health and Welfare for use in humans on 2014, July 07 th [339]. Based on the results of a recently completed Phase III clinical trial demonstrating that nivolumab significantly improves the progression-free and overall survival of patients with BRAF WT melanoma [340], the approval of this mAb by the US FDA is expected within the next few months. The safety and efficacy of ipilimumab, pembrolizumab, nivolumab and other checkpoint-blocking mAbs are being demonstrated in a steadily expanding panel of oncological indications [45,46,341,342]. Of note, some co-stimulatory mAbs including urelumab and PF-0582566 (both of which target CD137) are also under clinical development, with promising results [46,341]. Preclinical data suggest that combining checkpoint blockers with co-stimulatory mAb mediates superior antineoplastic effects [294, 343,344]. At least in part, this reflects the ability of co-stimulatory mAbs to promote NK cell functions [58, 345,346]. In line with this notion, a few clinical trials testing checkpoint blockers in combination with urelumab or lirilumab (a KIR-inhibiting mAb) have just been initiated (source http://www.clinicaltrials.gov).
Designed to (re-)activate the host immune system against malignant cells, immunomodulatory mAbs constitute an established and clinically promising paradigm of active immunotherapy. Interestingly, despite their non-specific mechanism of action, the clinical efficacy of immunomodulatory mAbs (and in particular checkpoint blockers) may be profoundly influenced by the panel of (neo-)TAAs specific to each neoplasm [347].
Extracellular ATP mediates robust immunostimulatory functions as it recruits and activates APCs via purinergic receptor P2Y, G-protein coupled, 2 (P2RY2) and purinergic receptor P2X, ligand-gated ion channel, 7 (P2RX7), respectively [383][384][385][386]. On the contrary, the degradation products of ATP (notably AMP and adenosine), have a pronounced immunosuppressive activity upon binding to adenosine A2a receptor (ADORA2A) and ADORA2B [387][388][389]. Two enzymes operates sequentially to degrade extracellular ATP, ectonucleoside triphosphate diphosphohydrolase 1 (ENTPD1, best known as CD39), which converts ATP into ADP and AMP [390][391][392], and 5'-nucleotidase, ecto (NT5E, best known as CD73), which transforms AMP into adenosine [393,394]. Some human neoplasms express increased amounts of CD39 and/or CD73, reflecting the evolutionary advantage conferred to cancer cells by the stimulation of adenosine receptors [395,396]. Efforts have therefore been dedicated to the development of agents that would limit the extracellular availability of adenosine or inhibit adenosine receptors [392,397]. Preclinical evidence indicates that CD39-or CD79-targeting agents (mostly mAbs) mediate antineoplastic effects as standalone interventions and improve the efficacy of other anticancer agents [397]. The clinical development of these agents, however, has not yet been initiated. Conversely, ADORA2A antagonists are currently being tested in latestage clinical trials, but as a therapeutic option against Parkinsonism [397]. It will be interesting to determine the safety and efficacy of inhibitors of adenosine generation or signaling in cancer patients.
Although it remains unclear whether these agents truly operate by altering the microenvironmental availability of Trp and Kyn [398], the antineoplastic effects of IDO inhibitors critically rely on the host immune system, implying that this constitutes an instance of active anticancer immunotherapy [399]. This also applies to strategies aimed at limiting the extracellular availability of adenosine.
Thus, PRR agonists have spurred interest not only as adjuvants for conventional vaccines [423,424], but also as immunotherapeutic interventions that may mediate antineoplastic effects per se or boost the therapeutic activity of other anticancer agents [34,48,425]. Three TLR agonists are approved by the US FDA for use in cancer patients: (1) the bacillus Calmette-Guérin (BCG), an attenuated variant of Mycobacterium bovis that presumably operates as a mixed TLR2/ TLR4 agonist, which is currently used as a standalone immunotherapeutic agent in subjects with non-invasive transitional cell carcinoma of the bladder [426]; (2) monophosphoryl lipid A (MPL), a TLR2/TLR4-activating derivative of Salmonella minnesota LPS currently utilized as adjuvant in Cervarix ® , a vaccine for the prevention of HPV-16 and -18 infection [427]; and (3) imiquimod, an imidazoquinoline derivative that triggers TLR7 signaling, currently employed for the treatment of actinic keratosis, superficial basal cell carcinoma and condylomata acuminata [422,426]. Of note, picibanil (a lyophilized preparation of Streptococcus pyogenes that operates as a TLR2/TLR4 agonist has been licensed for use in cancer patients by the Japanese Ministry of Health and Welfare (but not by the US FDA) as early as in 1975 [428,429]; while mifamurtide (a synthetic lipophilic glycopeptide that activates NOD2) has been approved by the EMA for the treatment of osteosarcoma in 2009 [430][431][432]. Moreover, the safety and efficacy of several other PRR agonists are currently being evaluated in clinical trials [433][434][435]. These molecules include agatolimod (CpG-7909, PF-3512676, Promune ® ), an unmethylated CpG oligodeoxynucleotide that activates TLR9 [436]; polyriboinosinic polyribocytidylic acid (polyI:C, Ampligen™, Rintatolimod), a synthetic doublestrand RNA that signals via TLR3 [437]; and Hiltonol™, a particular formulation of polyI:C that involves carboxymethylcellulose and poly-L-lysine [48,438].
Some malignant cells express PRRs [439][440][441][442][443][444][445], implying that PRR agonists may not be completely devoid of intrinsic tumor-modulating functions. Nonetheless, a large body of preclinical and clinical literature indicates that the antineoplastic effects of PRR agonists stem from their ability to engage the host immune system. Thus, PRR agonists constitute active immunotherapeutics.

Immunogenic cell death inducers
Some conventional chemotherapeutics, often employed at metronomic doses [446,447], as well as some forms of radiation therapy, can kill malignant cells while stimulating them to release specific DAMPs in a spatiotemporally coordinated manner [15,420,448]. Such DAMPs bind to receptors expressed on the surface of APCs (including TLR4), and not only boost their ability to engulf particulate material (including TAAs and cancer cell debris) but also trigger their maturation/ activation [15,418,448,449]. As a result, APCs acquire the ability to elicit a cancer-specific immune response that (at least in mice) is associated with the development of immunological memory [15,450]. We have dubbed such a functionally atypical form of apoptosis "immunogenic cell death" (ICD) [15]. Importantly, ICD inducers exert optimal antineoplastic effects in immunocompetent, but not in immunodeficient, mice [15,[451][452][453][454]. However, the ability of a specific stimulus to trigger ICD can be properly assessed only by means of vaccination experiments involving immunocompetent mice and syngeneic tumor models [15,455]. As it stands, a few FDA-approved therapies have been shown to constitute bona fide ICD inducers, including: doxorubicin, mitoxantrone and epirubicin (three anthracyclines currently employed against various carcinomas) [186, 449], bleomycin (a glycopeptide antibiotic endowed with antineoplastic properties) [456], oxaliplatin (a platinum derivative generally used for the therapy of colorectal carcinoma) [453,457], cyclophosphamide (an alkylating agent employed against neoplastic and autoimmune conditions) [458][459][460], specific forms of radiation therapy [419,[461][462][463][464][465][466], photodynamic therapy (an intervention that relies on the administration of a photosensitizing agent coupled to light irradiation) [448,467,468], and bortezomib (a proteasomal inhibitor used for the treatment of multiple myeloma) [469,470].
These and other (hitherto experimental) ICD inducers have been viewed as conventional forms of anticancer therapy, exerting antineoplastic effects via cytostatic or cytotoxic mechanisms. However, accumulating evidence indicates that the full-blown therapeutic potential of these molecules relies on the host immune system [15,471]. Thus, we propose to classify ICD inducers as a form of active anticancer immunotherapy.

Others
Other anticancer immunotherapies are approved by regulatory agencies worldwide for use in cancer patients or are currently being investigated for safety and efficacy in preclinical or clinical settings.
Lenalidomide (Revlimid ® , also known as CC-5013) and pomalidomide (Pomalyst ® , also known as CC-4047) are two derivatives of thalidomide (Thalomid ® ) originally developed in the 1990s to achieve improved potency in the absence of significant side effects [472]. Thalidomide was indeed marketed as an over-thecounter sedative, tranquilizer, and antiemetic for morning sickness in various countries in the late 1950s, but was rapidly withdrawn following a peak of infants born with malformation of the limbs [473]. In spite of its pronounced teratogenic activity, thalidomide raised renewed interest as an inhibitor of TNFα secretion in the 1990s [474], and was approved by the US FDA (under a strictly controlled distribution program) for the therapy of erythema nodosum leprosum (a complication of leprosy etiologically linked to TNFα) in 1998 [475]. The combination of thalidomide with dexamethasone (a glucocorticoid) rapidly turned out to mediate therapeutic effects in patients with hematological malignancies, eventually resulting in the approval by the US FDA of this regimen for the treatment of newly diagnosed multiple myeloma [476]. Alongside, lenalidomide (which retains some degree of teratogenicity) was licensed for use in patients with multiple myeloma (also in combination with dexamethasone) and low or intermediate-1 risk myelodysplastic syndromes that harbor 5q cytogenetic abnormalities (as a standalone intervention) [477][478][479][480]. Conversely, pomalidomide (which is devoid of teratogenic activity) has been approved for use in multiple myeloma patients only in 2013, when the approval of lenalidomide has been extended to mantle cell lymphoma (MCL) [481][482][483]. Although the effects of thalidomide, lenalidomide and pomalidomide, which are collectively referred to as "immunomodulatory drugs" (IMiDs), on the immune system have been characterized with increasing precision throughout the past two decades [484], the underlying molecular mechanisms remained obscure [485]. Recent findings indicate that the therapeutic activity of IMiDs depend, at least in part, on their ability to bind the E3 ubiquitin ligase cereblon (CRBN) and hence boost the proteasomal degradation of the B cellspecific transcription factors IKAROS family zinc finger 1 (IKZF1) and IKZF3 [486,487]. Of note, CRBN, which is also involved in the teratogenic effects of thalidomide and lenalidomide [488], regulates the abundance of interferon regulatory factor 4, perhaps accounting for the immunomodulatory functions of IMiDs [489]. Although endowed with intrinsic antineoplastic activity, IMiDs should be considered active immunotherapeutics.
As they progress and respond to treatment, neoplastic lesions are infiltrated by a significant amount of lymphoid and myeloid cells, including CD8 + T lymphocytes, Tregs, tumor-associated macrophages (TAMs) and immunosuppressive B-cell populations [122-124, 490,491]. Robust tumor infiltration by CD8 + T lymphocytes is generally associated with a good prognosis, especially when the intratumoral levels of Tregs are limited [124,492]. Along similar lines, high intratumoral levels of TAMs with a "classically-activated" M1 phenotype (which exert tumoricidal functions, stimulate NK cells and secrete T H 1-polarizing cytokines) generally correlate with improved disease outcome [491,493]. The contrary holds true when the myeloid tumor infiltrate contains high levels of "alternatively-activated" M2 TAMs or specific B-cell subsets, which can secrete not only immunosuppressive cytokines like IL-10 and TGFβ1, but also angiogenic mediators such as VEGFA and enzymes that remodel the extracellular matrix [491,493]. These observations prompted the development of immunotherapeutic regimens based on the depletion/ inhibition of Tregs or B lymphocytes, as well as on the conversion of M2 TAMs to their M1 counterparts.
Denileukin diftitox (also known as Ontak ® ) is a recombinant variant of IL-2 fused to the diphtheria toxin [494]. Owing to its selective cytotoxicity for cells expressing IL-2 receptor α (IL2RA, best known as CD25), denileukin diftitox has been approved by the US www.impactjournals.com/oncotarget FDA and EMA for the treatment of CD25 + cutaneous T-cell lymphoma in the early 2000s [494]. More recently, denileukin diftitox has been tested for its ability to improve the efficacy of various immunotherapies by efficiently depleting Tregs (which also express CD25) in patients affected by various neoplasms [495][496][497]. In some (but not all) these clinical settings, denileukin diftitox enhanced the efficacy of immunotherapy as it provoked a sizeable Treg depletion [496,497]. However, denileukin diftitox has recently been ascribed with a number of immunosuppressive effects [498,499]. This may explain why in some cases denileukin diftitox had no clinical activity [495], and casts doubts on the possibility to use such Treg-depleting agent as a routine anticancer immunotherapeutic. This said, several conventional antineoplastic agents commonly used in the clinic appear to deplete or inhibit Treg, which presumably contributes to their therapeutic activity (see below) [420,421]. Along similar lines, at least part of the clinical activity of ibrutinib (PCI-32765), a small molecule inhibitor of bruton tyrosine kinase (BTK) recently approved by the US FDA for use in patients with MCL and CLL [500][501][502], may stem from its ability to target tumor-infiltrating B lymphocytes or myeloid cells [503]. A clinical trial testing this possibility in pancreatic cancer patients will soon be initiated (LC, personal communication).
Several immunotherapeutic agents exert antineoplastic effects by altering the relative proportion between M2 and M1 TAMs in favor of the latter [491]. These include: (1) tasquinimod, a second-generation orally active quinoline-3-carboxamide analog initially developed as an antiangiogenic agent [504,505]; trabectedin (Yondelis ® ), a marine antineoplastic agent currently approved in Europe, Russia and South Korea for the treatment of soft tissue sarcoma and ovarian carcinoma [506,507]; (3) inhibitors of chemokine (C-C motif) ligand 2/chemokine (C-C motif) receptor 2 (CCL2/ CCR2) signaling [508]; (3) mAbs specific for chemokine (C-X-C motif) receptor 4 (CXCR4) [509]; and (4) small molecule inhibitors and mAbs that suppress colony stimulating factor 1/colony stimulating factor 1 receptor (CSF1/CSFR1) signaling [510][511][512]. With the single exception of trabectedin (which was not developed as an immunotherapeutic agent), none of these strategies is currently approved by the US FDA or EMA for use in humans (sources http://www.fda.gov and http://www. ema.europa.eu/ema/). However, several Phase II-III clinical trials are currently ongoing to establish the safety and efficacy of these active immunotherapeutic agents in patients with various solid tumors (source http://www. clinicaltrials.gov).
Additional, hitherto experimental immunotherapeutic regimens act by stimulating the host immune system to mount a novel (or unleash an existing) immune response against malignant cells. These include: (1) strategies for the depletion of circulating myeloid-derived suppressor cells (MDSCs), a bloodborne population of immature, immunosuppressive myeloid cells that generally accumulate in the course of tumor progression [513][514][515][516]; (2) mAbs that block CD47, one of the major antiphagocytic receptor expressed by malignant cells [517][518][519]; and (3) vaccines relying on the administration of cancer cell lines expressing immunostimulatory molecules (e.g., GM-CSF) upon inactivation or lysis [520].

CONCLUDING REMARKS
During the past three decades, immunotherapy has become a clinical reality [35, 78, 521], and an ever-increasing number of cancer patients are expected to receive, at some stage of their disease, an immunotherapeutic intervention [522,523]. The observations presented above suggest that various immunotherapies previously classified as passive, including several (if not all) tumor-targeting mAbs, ACT and oncolytic viruses, may de facto constitute active forms of immunotherapy. Moreover, accumulating preclinical and clinical evidence indicates that therapeutically relevant anticancer immune responses invariably exhibit some degree of epitope spreading, i.e., they eventually target several TAAs even when they were initially directed against a single one [524,525]. This is not surprising considering that malignant cells exhibit a high degree of genetic/genomic instability and hence are relatively prone to generate so-called "antigen loss variants" that would render TAA-specific immunotherapies completely ineffective with time [526][527][528]. Thus, even if immunotherapies that truly generate an anticancer response with a unique antigen specificity existed [529,530], they presumably would not mediate clinically relevant, long-term immune responses. In turn, this casts some doubts on the practical utility of classifying immunotherapies into "antigen-specific" or "non-specific".
Recently, great attention has been given to the immunostimulatory effects of conventional chemotherapeutics [420,421,531,532]. Indeed, several compounds that have been successfully used in the clinic, including the nucleoside analogs gemcitabine (which is approved by the US FDA for the treatment of pancreatic, ovarian, breast and non-small cell carcinoma) [533,534] and 5-fluorouracil (which is licensed for use in patients affected by various neoplasms) [535,536] have offtarget immunostimulatory effects, in particular when administered as low doses and according to metronomic schedules (while, similar to radiation therapy, they are generally immunosuppressive when given at high doses) [537,538]. It is therefore tempting to speculate that most (if not all) anticancer agents that are truly beneficial to patients operate as active immunotherapeutics, stimulating the host immune system to mount an antigenically broad (and hence insensitive to antigen loss) response against malignant cells. In support of this notion, an ever increasing number of combinatorial immuno(chemo) therapeutic regimens is being designed and tested in clinical trials, with promising results [34]. This being said, only the adequate implementation of protocols to monitor immune system-related parameters among patients participating in clinical trials (immunomonitoring) will provide insights into this possibility [539][540][541][542][543]. Such protocols are inherently complex, calling for international efforts toward standardization [544]. Harmonized immunomonitoring procedures will undoubtedly guide the development of new (immuno)therapies, and facilitate the identification of novel prognostic or predictive biomarkers [544]. We are positive that the next clinical success of anticancer immunotherapy is just behind the door.