Cancer related circulating and tumor-associated neutrophils subtypes, sources and function

Transcription

1 STATE-OF-THE-ART REVIEW Cancer related circulating and tumor-associated neutrophils subtypes, sources and function Merav E. Shaul and Zvi G. Fridlender Institute of Pulmonary Medicine, Hadassah-Hebrew University Medical Center, Jerusalem, Israel Keywords cancer; immunotherapy; innate immune system; neutrophils; tumor immunology; tumor microenvironment Correspondence Z. G. Fridlender, Institute of Pulmonology, Hadassah-Hebrew University Medical Center, POB 12000, Jerusalem 91120, Israel Fax: Tel: fridlender@hadassah.org.il (Received 18 December 2017, revised 18 April 2018, accepted 29 May 2018) doi: /febs In recent years, the role of neutrophils in cancer biology has been a matter of increasing interest. Many patients with advanced cancer show high levels of neutrophilia, tumor neutrophils are connected to dismal prognosis, and the neutrophil-to-lymphocyte ratio has been introduced as a significant prognostic factor for survival in many types of cancer. Neutrophils constitute an important portion of the infiltrating immune cells in the tumor microenvironment, but controversy has long surrounded the function of these cells in the context of cancer. Multiple evidences have shown that neutrophils recruited to the tumor can acquire either protumor or antitumor function. These findings have led to the identification of multiple and heterogeneous neutrophil subsets in the tumor and circulation. In addition, tumor-associated neutrophils (TANs) were shown to demonstrate functional plasticity, driven by multiple factors present in the tumor microenvironment. In this review, we examine the current knowledge on cancer-related circulating neutrophils, their source and the function of the different subtypes, both mature and immature. We then discuss the pro vs antitumor nature of TANs in cancer, their functional plasticity and the mechanisms that regulate neutrophil recruitment and polarization. Although the vast majority of the knowledge on neutrophils in cancer comes from murine studies, recent work has been done on human cancerrelated neutrophils. In the final paragraphs, we expand on the current knowledge regarding the role of neutrophils in human cancer and examine the question whether cancer-related neutrophils (circulating or intratumoral) could be a new possible target for cancer immunotherapy. Introduction Neutrophils are the predominant circulating leukocyte population in humans, accounting for 50 70% of circulating leukocytes. They are absolutely essential to protect humans and animals from microbial pathogens. Although neutrophils are found in close association with tumor cells and within tumor vasculature [1 3], their exact role in the tumor microenvironment is the subject of controversy [4 6]. Abbreviations Arg1, Arginase1; CCL, chemokine (C-C) ligand; CXCL, chemokine (C-X-C motif) ligand; DC, dendritic cell; G-CSF, granulocyte colonystimulating factor; GM-CSF, granulocyte macrophage colony-stimulating factor; HDN, high-density neutrophil; ICAM, intracellular adhesion molecule; IFN, interferon; IL, interleukin; LDN, low-density neutrophil; LPS, lipopolysaccharide; MDSC, myeloid-derived suppressor cell; MMP, matrix metallopeptidase; MPO, myeloperoxidase; NET, neutrophil extracellular trap; NK, natural killer; NLR, neutrophil-to-lymphocyte ratio; NO, nitric oxide; PBMC, peripheral blood mononuclear cells; PD1, programmed cell death protein 1; PDL1, programmed death ligand 1; ROS, reactive oxygen species; TAM, tumor-associated macrophages; TAN, tumor-associated neutrophil; TGFb, transforming growth factor beta; TLR, toll-like receptor; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor. 1

2 Cancer-related neutrophils subsets and function M. E. Shaul and Z. G. Fridlender The interest in the role of neutrophils in cancer increased during the late 80s and early 90s. However, with the rise of interest in T-cell biology, and since no clear roles for neutrophils in cancer were defined, a gradual decrease in the interest in these major cells of the immune system was noted in the following decade, even as new immunotherapy modalities were developing. Interest has increased during the last years, as recent data suggest more important and significant roles for neutrophils in tumor biology than previously thought [7,8]. Neutrophils play a well-established role in host defense, where they extravagate from the circulation and enter tissues [9,10]. Neutrophils are traditionally considered in the context of their antibacterial functions. It is becoming clear, however, that neutrophils in tumors and their myeloid precursors in the spleen, bone marrow, and blood play an important role in cancer biology [11 15]. Due to their very short life span, it is challenging to isolate and study tumorrelated neutrophils in general, and particularly human tumor-associated neutrophils (TANs). For that reason, most of the data on the characteristics, phenotype, and function of TANs were elucidated using animal models. In a recent review, Eruslanov et al. [16] nicely summarized the current knowledge on neutrophils in human cancer as compared to animal models. Here we provide an up-to-date review on both circulating and intratumoral neutrophils, in animal models of cancer and in human cancer patients. When discussing cancer-related neutrophils, most studies have focused either on circulating neutrophils on one hand, or residing neutrophils in the tumor microenvironment (e.g. TANs) on the other hand. Since circulating neutrophils are the source of intratumoral neutrophils, we will attempt in this summary to connect between TANs and circulating neutrophils and understand the relationships between them. There is clear evidence of the existence of different subpopulations of neutrophils, or at least different states of maturation and activation. However, a major challenge in the characterization of cancerrelated myeloid cells in general, and neutrophils specifically, is the lack of clear membrane-bound markers differentiating between the diverse subpopulations. The problems related to this lack of clear markers will be further discussed below. Finally, we will explore the current knowledge on circulating as well as intratumoral neutrophils in human cancer, their phenotype, their value as prognostic factor, and the possible impact current cancer immunotherapies may have on these cells. Cancer-related circulating neutrophils In tumor-bearing mice, the number of circulating neutrophils increases with tumor development and is associated with disease outcome [6,17]. More recently, the understanding that neutrophils possess a dual role in cancer development has emerged. In our recent work, we found that this dual role previously described in TANs can be similarly noted in circulating neutrophils in tumor-bearing mice as well as advanced human cancer patients [18]. In many patients with advanced cancer, high levels of blood neutrophils have been associated with poor disease outcome [19,20]. In addition, the neutrophilto-lymphocyte ratio (NLR) has been introduced as a significant prognostic factor in many tumor types [21]. In this section, we will review the state of the art data on the characteristics of the different circulating neutrophil subpopulations described in cancer, with special emphasis on animal data. The data on human cancer-related neutrophils will be summarized in a later section. We will discuss the current knowledge on the different subsets of cancer-related circulating neutrophils described, and their differential source, present an overview on the morphology, markers and functional activities of the different subsets, and elucidate our understanding on neutrophils vs polymorphonuclear myeloid-derived suppressor cells (PMN-MDSC), which we and others believe are a subset of circulating neutrophils. Data on cancer-related neutrophils have largely been obtained from the isolation of circulating neutrophils in peripheral blood. Traditionally, circulating neutrophils are isolated on a discontinuous density gradient (Ficoll-Hypaque). Using this technique, neutrophils are found in the high-density (HD) granulocytic fraction, whereas peripheral blood mononuclear cells (PBMCs) are found in the low-density (LD) mononuclear cells fraction [22]. We and others recently found that the distribution of neutrophils after discontinuous density gradient is different from traditionally known. Although most of the mature neutrophils are indeed found in the HD fraction [high-density neutrophils (HDN)], there is also a population of neutrophils in the LD fraction [low-density neutrophils (LDN)]. Moreover, we found that the proportion of this LD population that is negligible in healthy mice and humans, increases with tumor growth, and that LDN consists of two separate neutrophil populations, namely mature and immature neutrophils [18,23 25]. The source of subtypes of circulating neutrophils Multiple studies have demonstrated elevated circulating neutrophil counts and increased neutrophils- 2

3 M. E. Shaul and Z. G. Fridlender Cancer-related neutrophils subsets and function to-lymphocyte ratio in cancer patients [26 29] (see section Neutrophil in human cancer ). High NLR has been associated with increased neutrophil mobilization from the bone marrow and/or the spleen, in response to tumor-secreted cytokines such as IL8, granulocyte colony-stimulating factor (G-CSF), PDGF, and MIP1 [30 33], which circulating levels increase as tumor progresses. A similar increase in circulating neutrophils has also been described in several animal models of cancer, such the 4T1 mammary tumor, Lewis Lung Carcinoma mice model, and the 13762NF mammary adenocarcinoma rat model [18,34]. The finding of the existence of circulating LDN and HDN in cancer has raised new questions regarding the origin for these subsets. Although HDN typically present a mature morphology (segmented nucleus), the LDN fraction contains both mature and immature neutrophils [18]. Although the cause for the presence of mature and immature LDN in cancer is still unclear, the isolation of the neutrophils together with the PBMC fraction could result from (a) different degrees of maturation and (b) different activation and degranulation states. The terminal events in neutrophil maturation include the acquisition of cytoplasmic granules [35,36], which are responsible for increased cell density. Consequently, neutrophils in various stages of maturation may be separated by density gradient centrifugation [37]. The immature fraction of LDNs, could therefore originate from the bone marrow or spleen as immature neutrophils which fail to undergo terminal differentiation (Fig. 1). BrdU pulse-chase experiments aiming to observe newly produced circulating cells suggest that LDN rapidly accumulate in the circulation, whereas HDN appear in the circulation much later (48 h following BrdU administration) [18]. The fact that newly formed neutrophils can be seen in the LD fraction at an earlier time point than the HDN suggests that these are immature cells rapidly leaving the bone marrow. On the other hand, in animal tumor models, LDN can also be derived from the HD fraction as a source of mature neutrophils (Fig. 1). Although circulating mature neutrophils (HDN) have been considered to be terminally differentiated neutrophils, an HDN to LDN transition can occur spontaneously in the circulation of late stage tumor-bearing mice [18]. Importantly, LDN generated from HDN acquired immunosuppressive capacity and impair CD8 + T-cell proliferation [18]. This transition was shown to be partially (but not solely) mediated via transforming growth factor beta (TGF-b), originating from the tumor. Activation of neutrophils from healthy individuals with fmlp or phorbol myristate acetate was also previously found to induce the transition of HDN to the monocytic low-density fraction, possibly due to degranulation [23,24]. Following activation, neutrophils were shown to release to the surrounding granules containing myeloperoxidase (MPO) and GelatinaseB, consequently decreasing their granularity and reducing their density. Another study assessing circulating neutrophils in HIV patients used membranal CD63 as an indirect measure for the release of azurophilic granules and Arginase I. CD63 levels were significantly higher and Arginase I significantly lower in LDN (namely LDGs in the study) as compared to HDN (namely NDGs in the study), suggesting that at least part of the LDN fraction has degranulated and released their arginase [38]. Nevertheless, an exact characterization of the types of granules and their content in the mature HDN vs mature and immature LDN is still missing. Remarkably, we find that LDN also have the capacity (though to a much lesser extent) to switch to HDN [18], possibly simply by maturation. Since LDN contain both mature and immature cells, this transition raises the possibility that in contrast to current assumptions, neutrophil maturation can occur in the circulation, outside of the bone marrow. In that context, it is interesting to mention recent works [39 41] suggesting that neutrophils can return to the circulation after extravasation, and even return to the bone marrow. Morphology and surface markers of tumorrelated circulating neutrophils As suggested above, a major challenge in the characterization of cancer-related myeloid cells in general, and specifically neutrophils, is the lack of clear membrane-bound markers differentiating between the diverse subpopulations. The main separation between HDN and LDN is based on density, precluding the possibility to isolate a specific subpopulation and better characterize it [18]. As mentioned above, circulating neutrophils in cancer are heterogeneous and consist of at least three granulocytes subpopulations, mature segmented HDN, mature LDN and immature LDN. Although maturity markers of neutrophils can be found mostly in the HDN fraction (CD66b, CD11b, CD15, CD16, CD10), they are found at some extent in the LDN subpopulation of neutrophils as well [42]. Murine neutrophils are defined as CD11b + /GR1 + / Ly6G + cells [43]. LDN and HDN, both in mouse and human, display different levels of the important activation marker CD11b + [18,24]. Since there is confusion in the literature between neutrophils and PMN- MDSCs, in some studies markers that were shown as related to MDSC, can be assumed associated with a 3

4 Cancer-related neutrophils subsets and function M. E. Shaul and Z. G. Fridlender Fig. 1. Possible sources of the murine cancer-related neutrophil subsets, in the circulation and the tumor tissue. The source of the different cancer-related neutrophil subtypes is still unclear. The immature subset of the LDN could originate from the bone marrow or spleen as immature neutrophils that fail to undergo terminal differentiation. In contrast, the mature fraction of LDN could be derived from the HD fraction. The finding that TANs in early AB12 murine tumors show stronger cytotoxicity toward tumor cells than TANs in established tumors led to the hypothesis of an N1-to-N2 switch along tumor progression. It is still unknown whether this switch occurs inside the tumor in response to factors secreted in the tumor microenvironment, or reflects a differential recruitment of HDN or LDN to the tumor. fraction of the LDN (at least). One such example is the higher levels of both activation markers CD11b and CD66b in human mature LDN compared to the HDN neutrophils [24]. Interestingly, in HIV patients both LDN and HDN had the same mature segmented morphology, and the expression of CD11b and CD66b was higher in LDN than in HDN [38]. Brandau et al. [25] suggested that MDSCs in the LD fraction of human head and neck, lung, and bladder cancer patient (i.e. LDN) contain large number of different developmental stages of immature neutrophils expressing high levels of CD66b +. Further analysis in this study indicated the existence of three subsets within the neutrophils in the LD fraction, showing high degree of variability among different individuals (CD16 + /CD11b +, CD16 /CD11b +, and CD16 / CD11b cells) [25]. Because CD11b and CD66b are present in the membrane of secretory vesicles, gelatinase granules and/or specific granules, the intensity of these markers could reflect a degree of activation and degranulation. Additional data on surface markers in human circulating cancer-related neutrophils will be presented in a separate section below. Function of circulating neutrophil subsets Studies performed in both animal models and humans show functional differences between HDN and LDN, although the mechanisms leading to these differences are still unknown. Significant functional changes have been described in the circulating neutrophils of cancer patients (e.g. lower ROS production, reduced spontaneous apoptosis, and altered cytokine production). The findings of the existence of both mature and immature LDN in the circulation of cancer patients, together with mature HDN, have raised the question of possible functional differences between these different neutrophils subsets (Table 1). LDN not only accumulate in the context of cancer, but also appear for a short time in the circulation of 4

5 M. E. Shaul and Z. G. Fridlender Cancer-related neutrophils subsets and function Table 1. Schematic phenotypic characteristics described thus far of murine TANs with antitumor (N1) or protumor (N2) function (top panel), and cancer-related circulating neutrophils, HDN, and LDN (bottom panel). N1 TANs TNFa CCL3 ICAM1 ROS Fas Arg1 CCL2 CCL5 VEGF BV8 MMP9 CXCR4 Mature Hypersegmented (?) Tumor cells cytotoxicity CD8+ T cell recruitment ADCC, enhanced apoptosis Proinflammatory cytokines N2 TANs TNFa ICAM1 Arg1 inos CCL2 CCL5 VEGF MMP9 CCL17 CCR5 CCR7 4-1BBL OX40L CD86 CD54 Mature Tumor permissive CD4+ T cells recruitment Promote extravasation Angiogenesis Immune suppression HDN CXCR1 CXCR2 CD62L ROS High density Mature Tumor cell cytotoxicity High phagocytic ability LDN CCL2 CCL3 CXCL1 CXCL2 CXCL10 CXCR2 CD62L CCR5 Low density Mature and immature Reduced phagocytosis Reduced oxidative burst Limit CD8 + T cell proliferation Increased lifespan animals with acute inflammation. In an animal model of peritonitis [18], the number of LDN peaks within two days in the circulation and in the peritoneum of the mice and decreases back as the inflammation is resolved. In contrast, the adoptive transfer of LDN into the inflamed peritoneum was associated with enhanced resolution of inflammation. These findings support the hypothesis that LDN are released and recruited to tissues as part of a process of inflammation resolution. Since tumors represent a chronic state of inflammation, LDN therefore accumulate with tumor progression as a consequence of a state of inflammation that is never resolved. Recent evidences on LDN and HDN functions in cancer suggest that these two subsets have opposite cancer-related functions. Circulating HDN isolated from tumor-bearing mice were shown to possess cytotoxic capacities and to enhance tumor cell death through generation of ROS, therefore inhibiting tumor cell seeding in the premetastatic niche [17]. In contrast, LDN properties appear to promote a supportive microenvironment for tumor development, by displaying reduced cytotoxicity toward tumor cells and strong inhibition CD8 + T cells proliferation. LDN also display reduced inflammatory state, with reduced expression of CXCL1, CXCL2, CXCL10, CCL2, CCL3, CCR5, CXCR2, and CD62L (Table 1). Schmielau et al. demonstrated that the increased presence of granulocytes in the PBMC fraction (therefore LDN) in cancer patients with metastatic adenocarcinomas of the pancreas, colon, and breast correlated with a decrease in interferon c (IFNc) and TNFa production by T cells in these same patients [23]. In this study, the suppression of T-cell function was found to be dependent on the secretion of H 2 O 2 by the neutrophils. A recent study on a multistage mouse model of breast cancer suggested that the prolonged stimulation of hematopoietic stem cells with tumor-secreted G-CSF is responsible for driving the immunosuppressive characteristics of tumor-induced neutrophils [44]. On the other hand, the ability to suppress T-cell function is not restricted to LDN. High Arginase I activity was described in HDN (namely PMN in the study) isolated from both healthy donors as well as renal cell carcinoma (RCC) patients, suggesting these cells retain the ability to suppress T cells [24]. These observations support the idea that not only the immature granulocytes possess the ability to suppress the immune 5

6 Cancer-related neutrophils subsets and function M. E. Shaul and Z. G. Fridlender system but that the mature neutrophils can contribute to immune suppression as well. Our observations in mouse models of breast cancer and mesothelioma further suggest that mature and immature LDN display similar functions, with respect to T-cell suppression and reduced classical effectors functions, in comparison with HDN [18]. We and others have reported that LDN display a reduced rate of apoptosis compared to HDN [18,45]. Neutrophil apoptosis is a fundamental process for their removal and the resolution of inflammation [46,47]. The reason for LDN prolonged life span is still not clear but studies have suggested that it might be the result of factors released from the tumor microenvironment, such as G-CSF [25,45,48,49]. Tumorconditioned media collected from human early lung tumors increased the survival of blood neutrophils compared to neutrophils cultured in regular media [50]. In another study, Rodriguez et al. [24] also found that circulating neutrophils cocultured with RCC cell line show decreased spontaneous apoptosis, although the authors did not find any difference in the apoptosis rate between LDN and HDN. Autocrine secretion of G-CSF from neutrophils was also suggested to promote circulating and intratumoral lifespan [51]. Andzinski et al. showed that IFNb-deficient neutrophils exhibit a significantly lower apoptosis rate, which correlated with increased levels of G-CSF. IFNb therefore could indirectly regulate neutrophil survival. Cancer-related neutrophils vs PMN-MDSC two faces of the same entity? The terminology used to describe the polymorphonuclear population(s) expanding in the context of cancer and affecting tumor biology is currently under active discussion. The different terminologies used by different publications to describe cancer-related polymorphonuclear cells (PMN, PMN-MDSCs, G-MDSCs, and neutrophils) has spread confusion over the nature and the characteristics of these cells in the context of cancer. The expansion of immature myeloid cells in the bone marrow, spleen, and circulation of tumorbearing mice, was described in cancer patients and multiple animal models of cancer almost 20 years ago [52 57]. Due to their suppressive effect especially on T-cell proliferation, these immature cells were named myeloid-derived suppressor cells (MDSCs) [58]. The presence and infiltration of these cells have been shown to be associated with poor prognosis [59 61]. It was later found that the MDSCs population is comprised of at least two subsets granulocytic or polymorphonuclear (G/PMN-MDSC) and monocytic (M-MDSC), which possess different immunosuppressive properties and mechanisms [62 64]. The term MDSC comprises therefore a variety of myeloid cells that are developmentally immature and found at different stages of myelopoiesis. Blood cell maturation involves a series of steps which take place mostly in bone marrow, where the hematopoietic stem cells reside [13]. In tumor-bearing mice and humans with pancreatic or colon cancer (and probably other cancer types), the spleen can become an additional source of neutrophil production [65]. Therefore, MDSCs represent a heterogeneous subset of myeloid cells [11,66] defined by their capacity to suppress T-cell activation and proliferation [53,55,67], and, as neutrophils, were shown to expand and accumulate when tumors are present [54,68]. There is still contrasting evidence on the role of the granulocytic fraction of MDSCs. Some have shown that PMN- MDSC have immunosuppressive properties similar to the monocytic fraction [64,66,69], and others have demonstrated that they are less immunosuppressive [70,71]. Although M-MDSCs are phenotypically clearly defined, the descriptions of PMN-MDSCs added an extra level of complexity to the field as these cells are most closely related to neutrophils and even considered by us and others as a phenotype or subset of neutrophils [72,73]. Although the term PMN-MDSC is by definition meant to describe immature cells, it has been widely used in the literature to describe altogether circulating and tumor-infiltrating polymorphonuclear cells, even when these clearly present typical segmented morphology of mature neutrophils [18,24,74,75]. In addition, as mentioned above, there is a population of circulating neutrophils present in the LD fraction (LDN), which increases with tumor growth and consists of at-least two separate neutrophil populations, mature, and immature [18]. Differencing neutrophils from PMN-MDSC in the circulation is therefore a complex and controversial task, as there is no clear consensus on the differences between these two closely related granulocytic populations. When looking on their described features, there seems to be a significant phenotypic and functional overlap between neutrophils and PMN-MDSCs in the circulation. Murine neutrophils are defined as CD11b + /GR1 + /Ly6G + cells [76]. Murine PMN-MDSCs are also defined as CD11b + /GR1 high /Ly6G +, whereas M-MDSCs coexpress CD11b and Ly6C [63,77,78]. These similar definitions are probably the bases for confusion and overlap between these two populations. In humans, there is no equivalent to the mouse Gr1 marker, and a combination of CD14 /CD15 + /CD66b + /CD16 + defines 6

7 M. E. Shaul and Z. G. Fridlender Cancer-related neutrophils subsets and function mature neutrophils [20]. A more complex panel containing at least six markers is used to evaluate the human MDSCs (CD11b, CD14, CD15, CD66b, HLA- DR, and CD33), and PMN-MDSC are mostly referred to as CD14 /CD15 + /CD66b + /CD16 + /CD11b + / CD33 + /HLA-DR neg [79 81], which is in fact a definition of a subset of neutrophils. In that context, the term PMN-MDSC-like was recently suggested to describe cells that present PMN-MDSCs biochemical or molecular markers but do not have a suppressive activity [63]. To our view, this definition not only spreads more ambiguity around the phenotypic characterization of the granulocyte subsets in cancer, but also diminishes the understanding of the complexity of the immune phenotypic modulation in the context of cancer. We fully agree with the excellent summary and argument recently made on the topic by Coffelt et al. [13], that assigning a name to a cell based on one function (e.g. immunosuppression) implies that G/PMN- MDSCs exist mainly for one purpose, and cannot perform any other activity. In contrast, the dynamic neutrophils are capable to carry out many different and occasionally opposite functions simultaneously [18,50,82 84]. In such a case, neither MDSC nor MDSC-like terminology can define these cells. Hence the same question has been asked in the past over the terminology M-MDSCs and their relation to M1/M2 macrophage terminology [85]. In our view, PMN- MDSCs represent in most studies an immature subset of neutrophils. An important support for that suggestion is that it has been shown that adoptively transferred MDSC can enter tumors and differentiate into a mature phenotype [86]. Based on that, we strongly support the suggestion made by Coffelt et al. to re-evaluate the use of the term G/PMN-MDSC until convincing evidence is generated that distinguishes neutrophils from G/PMN-MDSCs. Tumor-associated neutrophils Neutrophils provide the first line of defense against microbial infections and take a pivotal role in inflammation. Their role in cancer, however, has long been a matter of controversy. Several early studies demonstrated various antitumor properties, including direct cytotoxicity and antibody-dependent cell-mediated cytotoxicity (ADCC) [87]. On the other hand, with the evolution of our understanding of the tumor microenvironment, neutrophils where shown to possess a significantly different set of traits. Neutrophils were shown to directly promote tumor growth by secreting a variety of chemokines and cytokines, as well as actively recruiting other tumor-supporting cells to the tumor microenvironment. TANs were also shown to promote tumor angiogenesis, play a critical role in mediating the angiogenic switch and promote tumor cell motility, migration, and invasion as they secrete enzymes that degrade and modify the extracellular matrix [88]. Finally, we and others [18,82] have shown that neutrophils also take part in immune suppression thereby supporting tumor growth and metastatic progression. This, together with the fact that IFNc stimulated neutrophils may upregulate PD-L1 and acquire the capacity to suppress T-cell proliferation [89] suggest that TANs may serve as a target for antitumor immunotherapies and may also be affected by recent targeting of the PD1/PD-L1 pathway. While the reports regarding neutrophil function in cancer seem conflicting, it is easier to understand them by examining the functional plasticity that neutrophils demonstrate. We have demonstrated in mice that inhibition of TGF-b, available in high concentrations at the primary tumor site, can mediate the functional switch from tumor-promoting neutrophils to antitumor neutrophils. Although these are not exact synonyms, tumor-promoting neutrophils can be mostly termed N2 TANs, whereas antitumor neutrophils can be termed N1 TANs [90]. This finding has led to the hypothesis that neutrophil function in cancer depends on cues received from their microenvironment. For example, in the primary tumor, where TGF-b is abundant, neutrophils act to promote tumor growth. In contrast, in the premetastatic niche where TGF-b levels are low, neutrophils can concomitantly act to limit metastatic seeding via direct cytotoxicity [17,91]. It should be noted that TANs pro or antitumor phenotype is probably not a definite state of activation, but rather a balance of the whole neutrophils properties, resulting from the neutrophils modulation by the milieu and leading to an overall tumor-promoting vs cytotoxic phenotype. We personally believe neutrophils are a heterogeneous population of cells. Neutrophils heterogeneity originates from the fact that these cells are highly responsive to cues (cellular components, chemokines) found in their microenvironment, and may adopt a protumor phenotype in certain conditions and an antitumor phenotype in others. The balance between these phenotypes hence dictates the overall TAN pro vs antitumor properties assessed experimentally in vitro and in vivo. In the current section, we discuss the existing knowledge on TANs in the tumor microenvironment. We discuss the differential recruitment of neutrophils into tumors, and the tumor-supportive and antitumor effects of TANs. We then describe the functional plasticity of TANs, and present the debate of whether the 7

8 Cancer-related neutrophils subsets and function M. E. Shaul and Z. G. Fridlender change from N1 to N2 and vice-versa represents a real polarization or an hyperactivation state. Lastly, we focus on an important and interesting effect of TANs in cancer that is being discovered during the last years, namely the effects of TANs on the immune system. Recruitment of neutrophils into tumors Neutrophils are actively recruited to the tumor microenvironment along a chemotactic gradient involving multiple tumor-secreted factors (Fig. 2) [92]. Neutrophils infiltrate tissues by responding to specific chemokines such as CXCL1 and CXCL2, or cytokines such as TNFa and IFNc [93]. Cell adhesion molecules located on TAN surface (e.g. CD11b) as well as adhesion molecules on the surface of blood vessel endothelial cells (e.g. selectins, ICAM-1 and PECAM-1) play a critical role in the rolling, adhesion, and transmigration of the neutrophils from the bloodstream into the tumor [94]. As a result, the endothelium within tumors was shown to display a distinct adhesion molecule profile compared to the endothelium in healthy tissue [95 97]. The dynamics of neutrophil development and tumor-induced release of granulocytes from the bone marrow to the circulation have been recently summarized [13]. Of the various chemokine receptors expressed on TANs, CXCR1, and CXCR2 have been attributed a major role in TANs recruitment to the tumor [98 100], via binding to the chemokines CXCL1, CXCL2, CXCL5, CXCL6, and CXCL8 secreted from the tumor (Fig. 2). By expressing high levels of these chemokines, tumors interfere with the homeostatic retention of neutrophil in the bone marrow and promote their release to the bloodstream [13], occasionally as early immature forms. In human cancers, IL8/ CXCL8 has been described as the most potent neutrophil chemoattractant to tumors [30,101,102]. In several mouse models of spontaneous and inflammation-driven tumorigenesis, circulating Ly6G + neutrophils were shown to express high levels of CXCR2, and CXCR2 deficiency was demonstrated to significantly attenuate neutrophil recruitment and suppress tumorigenesis [103]. On the other hand, CXCR2 deficiency in a pancreatic cancer model in KPC mice was shown to suppress metastasis, but the authors reported no impact on tumorigenesis [104]. Recently, Yu et al. [105] demonstrated in a 4T1 breast tumor model that TNFaactivated mesenchymal stem cells (MSCs) also promote the recruitment of neutrophils in a CXCR2-dependent manner. CXCR2 has therefore been demonstrated to be a potent protumorigenic chemokine receptor that directs the recruitment of leukocytes in general, and neutrophils in particular, into tissues during tumor development. CXCR2 and CXCR4 were also demonstrated to act cooperatively to balance between the retention and the release of neutrophils from the bone marrow [106]. TAN autocrine secretion of CXCL1 and CXCL2 has been suggested to enhance the infiltration of these cells [107,108], in what seems as a positive feedback loop mechanism in which chemoattractants secreted by neutrophils further Fig. 2. Neutrophils are actively recruited to the tumor microenvironment along a compounded chemotactic gradient. By releasing high levels of chemotactic molecules such as G-CSF, CXCL1, CXCL2, CXCL5, CXCL6, and CXCL8, or exosomal proteins such as S100A8 and S100A9, tumor cells promote the release of neutrophils from the bone marrow and their recruitment into the tumor tissue. In addition, TAN autocrine secretion of CXCL1 and CXCL2 has been suggested to enhance the infiltration of these cells in a positive feedback loop. A G-CSFdependent systemic expansion of neutrophils has also been shown to occur in response to IL-17 secreted from cd T cells and Th17 T cells. 8

9 M. E. Shaul and Z. G. Fridlender Cancer-related neutrophils subsets and function recruit more neutrophils into the tumor. Such a mechanism was previously described in the context of inflammation [109,110]. The two chemokine receptors CCR5 and CCR7 have also been implicated in the migration of neutrophils. Multiple studies have associated CCR5 expression in tumor cells to tumor cell migration, dissemination and tumor progression in models of melanoma, basal breast cancer, and prostate cancer [ ]. Recently, Blattner et al. [114] found a high frequency of CCR5 + neutrophils (described in the study as CD11b + Ly6G + Ly6C low PMN-MDSCs) in melanoma lesions, spleen, and metastatic lymph nodes. CCR5 + neutrophils also showed stronger immunosuppressive phenotype than the CCR5 fraction [115]. Upon activation, neutrophils were also shown to mobilize intracellularly stored CCR7 to their membrane, therefore enhancing their ability to migrate toward CCL19, CCL21, and granulocyte macrophage colony-stimulating factor (GM-CSF) [116]. CCR7 was further demonstrated in vivo to be involved in the migration of neutrophils to the draining lymph nodes, where they may induce and modulate adaptive immune responses [116]. The granulocyte chemotactic protein CXCL6/GCP-2 was shown to promote the accumulation of neutrophils in melanoma tumors, as well as angiogenesis and tumor growth [117]. Beside chemokines, growth factors such as G-CSF and GM-CSF, may also play a role in the chemoattraction of neutrophils into tumors [44,118], in addition to their role in promoting the production of new granulocytic progenitors from the bone marrow. G- CSF was demonstrated to promote neutrophil mobilization in a model of acute inflammation through the activation of the CXCL2-CXCR2 axis [119]. CXCR2- deficient neutrophils become irresponsive to G- CSF-driven mobilization and remain in the bone marrow. In a metastatic breast cancer model, macrophages expressing interleukin 1b (IL-1b) were shown to support a G-CSF-dependent systemic expansion of neutrophils through the activation of IL-17-producing cd T cells [82]. In an orthotopic Kras lung tumor model as well as various models of breast, lung, and lymphoma cell line tumor models, IL-17 produced by Th17 cells induced the expression of G-CSF [120,121] or IL6 [122], leading to immature myeloid cell mobilization and recruitment into the tumor microenvironment. Recently, human cd T cells were also shown to contribute to the accumulation, survival, and proliferation of intratumoral neutrophils (described in the study as CD45 HLADR CD33 + CD11b + CD66b + PMN-MDSCs) in colorectal cancer, through the release of IL-17A, IL8, GM-CSF, and TNFa [123]. Two exosomal proteins, S100A8 and S100A9, have been implicated in the accumulation of granulocytes in a colon carcinoma primary tumor model and the defective myeloid cell differentiation in cancer [124]. These two proteins were also shown to be secreted by neutrophils (described in the study as Gr1-high MDSCs) in a 4T1 breast tumor model and accumulate in the serum of tumor-bearing mice [125,126]. S100A8 and S100A9 promote an inflammation-like state that accelerates the migration of primary tumor cells to the premetastatic niche [127]. By expressing receptors for S100A8/A9, PMN-MDSCs initiate an autocrine loop to increase local levels of the heterodimer [125]. This is hence another example of neutrophils autocrine positive feedback loop that sustains the accumulation of granulocytes in cancer. In addition, the expression of the proto-oncogene cmet on immature precursor cells and TANs was also demonstrated to play a role in neutrophil mobilization and transmigration across the activated endothelium, following G-CSF stimulation [128,129]. We have shown in the past that TGF-b receptor blockade increases the number of neutrophils in tumors. The mechanisms for this increased infiltration are still unclear, but appear to include the modulation of cytokines and chemokines expression, as well as the modulation of adhesion molecules expression on endothelial cells [90,130]. Various immune cell types present in the tumor microenvironment also seem to play a role in the recruitment of neutrophils to the tumor. In the context of inflammation and infectious diseases, macrophages and CD8 + T cells were shown to be important players in the modulation of neutrophil recruitment to the inflamed tissue [131,132]. It may very well be that these cells contribute to the recruitment of neutrophils into the tumor microenvironment as well. The ability of tumor-associated T cells to produce GM-CSF [133], CXCL1, CXCL2 [134] or cytokines such as TNFa and IFNc may promote their ability to recruit and activate neutrophils to the tissue [135]. In line with this hypothesis, CD8 + T cell depletion in a G-CSF-expressing colon carcinoma mouse model was shown to impair TAN infiltration [136]. Conversely, FoxP3 + regulatory T cells, known for their suppressive phenotype, were shown to inhibit neutrophil recruitment to the tumor. This effect was shown to be mediated by decreasing the expression of CXCL1 and CXCL2 in neutrophils [137]. Tumor-supportive roles of TANs An early study by Pekarek et al. [138] nicely demonstrated that depletion of neutrophils can have a robust 9

10 Cancer-related neutrophils subsets and function M. E. Shaul and Z. G. Fridlender inhibitory effect on tumor growth, indicating that this tumor requires granulocytes for rapid growth. This notion was later confirmed in additional tumor models by us and others [ ]. As previously mentioned, we and others have noted that in untreated tumors, TANs appear to mostly develop a protumorigenic phenotype that we have termed N2 TAN [83,90]. Depletion of these protumorigenic neutrophils, therefore, inhibits tumor growth and reduces the level of immunosuppression in the tumor microenvironment [90]. TANs appear to be involved in tumorigenesis and tumor growth through multiple mechanisms including carcinogenesis, angiogenesis, and extravasation (Table 1). Early evidences supporting the concept of neutrophils tumor-promoting effect suggested that ROS production by granulocytes promotes genotoxicity and leads to carcinogenesis. Weitzman and coworkers first demonstrated that phagocytes were mutagenic to mammalian cells and that injection of fibroblasts preincubated with activated neutrophils resulted in tumor development [142]. Number of studies have since reported a link between neutrophils, ROS, mutagenesis, and carcinogenesis, both in vitro and in vivo [143]. Neutrophils have also been demonstrated in multiple studies to enhance the proliferation, migration, and extravasation of tumor cells. Tazzyman et al. [144] showed that neutrophils increase the proliferation of lung carcinoma cells in vitro and that CXCR2 blockade results in the inhibition of neutrophil recruitment and impaired tumor growth in vivo. Neutrophil were further shown to support tumor progression by releasing a wide variety of protumoral factors to the microenvironment, especially proteases, stored in neutrophils granules [145,146]. Neutrophils develop three types of granules as well as secretory vesicles along their maturation. Primary granules appear during the early promyelocyte phase. Secondary granules then form in the immature band cell and tertiary granules finally develop during last stages of maturation and segmentation. The azurophilic (primary) granules contain several serine proteases, such as neutrophil elastase, cathepsins, and proteinase-3, which promote tumor invasion and/or proliferation [ ]. Secretion of neutrophil elastase (NE) enhance proliferation of several cancerous cell lines in vitro [150], and promote the adhesion of neutrophils and tumor cells to the vascular endothelium [151]. The interaction between proteinase-3 and the cell surface receptor RAGE, overexpressed in prostate cancer cell lines, was recently proposed to mediate prostate cancer metastasis to the bone [152]. The secondary granules contain, among others, lactoferrin and neutrophil collagenase (MMP8). The gelatinase (tertiary) granules contain high levels of gelatinase B (also known as MMP9) and ADAM9. By mediating the remodeling of basal membranes and extracellular matrixes, MMP8, MMP9, and ADAM9 were shown to be important players in the infiltration of neutrophils to tumors [145,153,154], but also to the angiogenic process [155,156]. Multiple studies also provide evidences for TANs prometastatic function. The mechanisms by which neutrophils may act to promote metastasis are diverse. The ability of neutrophils to activate the blood vessels endothelium and enhance the adhesion of circulating tumor cells has been suggested [ ]. A recent study by Spiegel et al. [160] showed that neutrophils promote metastasis by impairing the cytotoxic function of natural killer (NK) cells in the primary tumor while facilitating the extravasation of tumor cells to the metastatic niche through modulation of the extracellular matrix. In vitro studies on head and neck squamous cell cancer (HNSCC) lines suggested neutrophils promote metastasis by enhancing tumor cell migration rather than supporting tumor cell proliferation [161]. Recently, a growing number of studies are investigating the possible involvement of neutrophils extracellular traps (NETs) in promoting the migration and extravasation of cancer cells. NETs are fibers of decondensed DNA released from activated neutrophils to trap microbes in the context of bacterial infection [162,163]. A number of studies have recently proposed that NETs might act to capture circulating tumor cells and promote their migration to new sites [164,165]. Tohme et al. [166] showed that the NET-induced secretion of HMGB1 (from the neutrophils) and subsequent activation of the TLR9 pathway in cancer cells promote tumor cell migration, adhesion, and invasion. In addition, TANs ability to modulate other tumorinfiltrating immune cells, such as T cells, macrophages, and NK cells, has been demonstrated to play an important role in promoting a tumor-supportive microenvironment. This aspect of TAN function will be discussed in the section on TANs and the immune system, below. Antitumor effects of TANs Despite the broad literature on the protumor effects of TANs, evidence from multiple studies suggest that cancer-related neutrophils retain the ability to mediate the killing of tumor cells [ ]. In many studies the antitumor roles of neutrophils were assessed with engineered tumor cell lines, or following specific therapies [12]. Interpretation of these studies in the light of the idea of differential neutrophil activation status within tumors is instructive. We found that neutrophils 10

11 M. E. Shaul and Z. G. Fridlender Cancer-related neutrophils subsets and function can assume a more tumor-cytotoxic phenotype (Table 1), for example, during TGFb inhibition [90], IFNb treatment [83], or after immunologic or cytokine activation, where they have the potential to kill tumor cells and inhibit growth [173,174]. Depletion of TANs in animal models during such interventions thus either augments tumor growth and/or blunts the antitumor effects of immunologic treatments [175]. Neutrophils typically kill their target by releasing a combination of reactive oxygen species (ROS), proteases, membrane-perforating agents, and soluble cytokines such as TNFa, IFNs, and IL1b. Several mechanisms for this mediated killing have been proposed in the context of cancer, and include the induction of ROS production and death ligands such as TRAIL [176] or the production of H 2 O 2 by NADPH oxidase [167]. Neutrophils require direct contact to kill tumor cells. Saito et al. [177] reported that neutrophils activate an oxidative process at the contact site with tumor cells and have the capacity to limit metastatic spread. Recently, Gershkovitz et al. [178] described in a 4T1 breast tumor model that neutrophil cytotoxicity toward tumor cells is mediated via the activation of TRPM2, an H 2 O 2 -dependent Ca 2+ channel, in tumor cells. The authors showed that TRPM2-deficiency reduces mammary tumor cells (human and mouse) susceptibility to neutrophil cytotoxicity. A number of studies have also suggested neutrophils actually display antimetastatic properties. This characteristic was mostly attributed to circulating neutrophils and not to TANs. Granot et al. [17] showed that depletion of Ly6G + cells in a 4T1 breast tumor mice model led to a significant increase in lung metastases whereas primary tumor growth was not affected. Accordingly, Lopez-Lago et al. [91] also showed that poorly metastatic carcinoma cells, such as SN12C, activate the cytotoxic and migratory properties of neutrophils and that this activation leads to accumulation of neutrophils with antimetastatic properties in the premetastatic niche. The authors further suggested that chemokines secreted from the tumor cells, such as CXCL5 and IL8, act as a critical rate-limiting step during lung metastatic seeding. Functional plasticity N1 vs N2 The conflicting reports regarding neutrophil functions in cancer had first led to confusion as to the reason for this inconsistency. However, recent evidences over the functional plasticity of cancer-related neutrophils have alleviated these conflicts. It is now well accepted that TANs can acquire both pro or antitumor functions. As mentioned above, in mice, antitumor TANs are occasionally termed by us and others as N1 TANs, and protumor TANs as N2 TANs, depending on the signals encountered in the tumor microenvironment [90,179]. The findings that TANs isolated from early AB12 (mesothelioma) primary tumors showed stronger proinflammatory profile and higher cytotoxicity toward tumor cells than TANs in established tumors in this animal model led to the hypothesis of an N1- to-n2 switch along tumor progression (Fig. 1) [88,180]. In addition to a change in activation state, a modulation in TAN infiltration to the AB12 tumors was also found. TANs at early stages of the AB12 tumor development were localized almost exclusively at the periphery of the tumor whereas at later stages, TANs are found scattered among the tumor cells [180]. Multiple factors secreted from the tumor were shown to impact the balance between these two opposite functions. TGF-b secreted at high levels within the tumor microenvironment was shown to induce an N2 phenotype in TANs [90]. Upon TGF-b blockade, murine CD11b + /Ly6G + neutrophils recruited to the tumor show higher cytotoxicity toward tumor cells and expressed higher levels of proinflammatory cytokines [90]. In contrast, depletion of these neutrophils resulted in decreased tumor growth and in the activation of CD8 + T cells in the tumor. IFNb was recently shown to induce an N1 phenotype in TANs [83]. IFNb-deficient TANs show reduced expression of ICAM1 and TNFa, reduced capacity to kill tumor cells and decreased NETosis. Various additional cytokines have been implicated in modulating TANs phenotype. G-CSF and IL6 have been recently shown to be involved in the functional reprogramming of neutrophils in cancer, through the activation of the STAT3 signaling pathway. IL-6 and G-CSF were reported to cooperate and induce protumor, proangiogenic functions in bone marrow neutrophils, by increasing the expression of MMP9 and BV8 and downregulating TRAIL [181]. Recently, IL- 35 was also demonstrated to indirectly induce an N2- polarization of TANs by increasing G-CSF and IL-6 production in the tumor microenvironment [182], further promoting inos expression in neutrophils and suppression of T-cell function. IL-12 has also been suggested to functionally modulate the neutrophils recruited to the primary tumor toward antitumor properties. IL-12 was shown in various studies to activate antitumor immunity in many ways, including the activation of NK cells, modulation of activated CD4 + T lymphocytes to become Th1 cells, and activation of CD8 + cytotoxic T lymphocytes [183,184]. Treatment combining cyclophosphamide and IL-12 in a colon cancer mouse model was reported to 11