Inflammatory Responses during Tumour Initiation: From Zebrafish Transgenic Models of Cancer to Evidence from Mouse and Man.

The zebrafish is now an important model organism for cancer biology studies and provides unique and complementary opportunities in comparison to the mammalian equivalent. The translucency of zebrafish has allowed in vivo live imaging studies of tumour initiation and progression at the cellular level, providing novel insights into our understanding of cancer. Here we summarise the available transgenic zebrafish tumour models and discuss what we have gleaned from them with respect to cancer inflammation. In particular, we focus on the host inflammatory response towards transformed cells during the pre-neoplastic stage of tumour development. We discuss features of tumour-associated macrophages and neutrophils in mammalian models and present evidence that supports the idea that these inflammatory cells promote early stage tumour development and progression. Direct live imaging of tumour initiation in zebrafish models has shown that the intrinsic inflammation induced by pre-neoplastic cells is tumour promoting. Signals mediating leukocyte recruitment to pre-neoplastic cells in zebrafish correspond to the signals that mediate leukocyte recruitment in mammalian tumours. The activation state of macrophages and neutrophils recruited to pre-neoplastic cells in zebrafish appears to be heterogenous, as seen in mammalian models, which provides an opportunity to study the plasticity of innate immune cells during tumour initiation. Although several potential mechanisms are described that might mediate the trophic function of innate immune cells during tumour initiation in zebrafish, there are several unknowns that are yet to be resolved. Rapid advancement of genetic tools and imaging technologies for zebrafish will facilitate research into the mechanisms that modulate leukocyte function during tumour initiation and identify targets for cancer prevention.


Macrophages and Neutrophils are Co-Opted by Cancer to Perform Tumour-Supporting Activities
High levels of macrophages and neutrophils in cancerous lesions correlate with poor prognosis in humans. The pro-tumour activities of these leukocytes are evident in both mammalian and zebrafish models.

Tumour-Associated Macrophages
Tumour-associated macrophages (TAMs) are the most abundant type of leukocyte found within tumours, comprising up to 50% of the tumour mass in epithelial tumours [82,83]. The degree of macrophage infiltration correlates with poor clinical prognosis across a broad range of tumour types [3]. Both tissue-resident macrophages and blood monocyte-derived macrophages infiltrate tumours, where they are influenced by signals derived from cancer cells and the local tumour microenvironment to perform tumour-promoting activities [84]. The activation state of TAMs more closely resembles that of 'alternative' or 'anti-inflammatory' M2 polarization, as opposed to 'classical' or 'pro-inflammatory' M1 polarization, e.g. genes commonly expressed by TAMs include scavenger receptors, arginase-1, matrix metalloproteinases, TGF-β and IL-10, accompanied by downregulation of pro-inflammatory cytokines and MHC Class II [85,86]. Although, it is of note that the concept of M1/M2 polarisation is an oversimplification and TAMs are a highly heterogenous population [3,[87][88][89]. The tumour-supporting roles of TAMs have been studied extensively in the past two decades, revealing mechanisms by which TAMs promote proliferation, angiogenesis, invasion and metastasis [84,[90][91][92]. In zebrafish cancer xenograft models, zebrafish macrophages have also been shown to promote angiogenesis, invasion and metastasis [93,94].

Tumour-Associated Neutrophils
Neutrophils are frequently found within tumours and, following the convention of TAMs, the term 'tumour-associated neutrophil' (TAN) has been adopted. However, TANs are less well characterised in comparison to their TAM counterparts and have been shown to have both pro-and anti-tumour effects [95]. Nonetheless, in numerous studies a high intra-tumour neutrophil density Cells 2020, 9, 1018 6 of 28 has been associated with poor patient prognosis [96,97]. Multiple tumour-promoting functions of TANs have been described, including the encouragement of proliferation, angiogenesis, metastasis and immunosuppression [95,98,99]. In addition, through the secretion of genotoxic nitric oxide and reactive oxygen species (ROS), neutrophils exert a mutagenic effect, and thus provide a further driver for cancer development. On the other hand, higher levels of nitric oxide release from neutrophils can be cytotoxic for cancer cells, contributing to tumour suppression [100,101]. Further anti-tumour roles of neutrophils may include triggering apoptosis via activation of TRAIL [102] and promoting the activity of anti-tumour cytotoxic T lymphocytes [103]. In zebrafish, tumours also exhibit high levels of infiltrating neutrophils, which promote metastasis and express genes involved in angiogenesis and immunosuppression [104][105][106][107]. Anti-tumour neutrophil responses are yet to be described within zebrafish cancer models.
In addition to mature neutrophils, under pathogenic conditions such as chronic inflammatory disease and cancer, immature myeloid cells, known as myeloid-derived suppressor cells (MDSCs), are recruited from the bone marrow prior to terminal differentiation [85,108]. Granulocytic MDSCs (G-MDSCs) are thought to be immature neutrophils since they morphologically and phenotypically resemble neutrophils. G-MDSCs promote tumour cell survival, angiogenesis, invasion, metastasis, and immunosuppression; functions similar to those of mature pro-tumour TANs [109]. In contrast, existing studies have not described anti-tumour functions for G-MDSCs [110]. It is of note that not all cancer studies make a clear distinction between mature TANs and immature G-MDSCs, and there is a lack of clarity as to whether they are independent cell types or whether they originate from the same progenitors [110][111][112]. The existence of MDSCs has not been explored in zebrafish and it is uncertain whether neutrophils in zebrafish cancer studies are recruited in a mature or immature state.

Evidence of Tumour-Promoting Inflammation in Early Tumourigenesis of Mouse and Man
Whilst there is abundant evidence that TAMs and TANs have tumour-promoting roles within established tumours, relatively little is known about the role of macrophages and neutrophils at the earliest stage of tumour initiation. This is largely because the early stages of tumourigenesis are difficult to detect in humans and mammalian models. The exception to this is colorectal cancer, for which distinctive structures known as 'adenomatous polyps' are formed at the early neoplastic stage [113]. These polyps are easy to detect and their removal is a routine procedure. Furthermore, the common initiating mutation, APC, was discovered over 30 years ago by studying families with familial adenomatous polyposis (FAP). FAP is caused by an inherited mutation in the APC gene, leading to abundant polyp formation and, without intervention, malignant disease [114,115]. This discovery was swiftly followed by the creation of APC mutant mouse models which represent not only FAP but also spontaneous colorectal cancers, 80% of which feature mutations in APC [116,117]. The ability to detect and model early neoplastic legions in this way enabled the discovery of factors which promote early tumourigenesis, such as prostaglandin E 2 (PGE 2 ) PGE 2 is a potent inflammatory mediator, synthesised by the COX-2 enzyme. Both PGE 2 and COX-2 are present at high levels in colorectal cancer, adenomatous polyps and even pre-neoplastic microadenoma [118,119]. Chemical inhibition of COX-2, as well as the deletion of the COX-2 gene or various PGE 2 receptors, was found to dramatically decrease intestinal polyp formation in numerous mouse models of colorectal cancer, including APC mutant mice and mice exposed to carcinogens [120][121][122][123][124]. Non-sterile anti-inflammatory drugs (NSAIDs), such as sulindac and COX-2 inhibitor celecoxib, have been tested for the prevention of both FAP and spontaneous colorectal cancers and have proven to effectively reduce polyp formation and prevent cancer [125][126][127]. Epidemiological studies have also shown that the use of COX-2 inhibitor, aspirin, is associated with decreased incidence of colorectal cancer [128]. Unfortunately, the use of COX-2 inhibitors to prevent cancer is limited by the dangerous cardiovascular side effects that are associated with long-term use [129]. PGE 2 is also implicated in the promotion of many other types of cancer, including cancer of the breast [130], liver [131], lung [132], brain [133] and pancreas [134,135]. This suggests that the Cells 2020, 9, 1018 7 of 28 phenomenon by which inflammation fuels early tumour promotion may be a common feature of cancer development. Therefore, the discovery of alternative methods to target this early inflammatory response is imperative for cancer prevention. Inhibiting the recruitment of either macrophages or neutrophils also dramatically reduced polyp formation mouse models of colon cancer and prevented carcinogenesis [136][137][138][139]. This demonstrates the importance of inflammatory leukocytes in promoting early tumourigenesis. As such, both macrophages and neutrophils are promising targets for cancer prevention, but further research is required to understand precisely when and how along the pathway of oncogenesis leukocytes are recruited and co-opted by the cancer cell.
Research in this field has been impeded by a lack of suitable in vivo models. Most mouse models for the study of tumourigenesis rely on read-outs such as tumour size and occurrence, whilst measurements of the early pre-neoplastic stage are largely missing. These models also lack temporal resolution and the means to dissect complex mechanisms involving interactions between multiple cell types. For these reasons, the zebrafish is a uniquely advantageous model for the study of inflammation during early tumourigenesis.

Zebrafish Studies Reveal Leukocyte Recruitment and Trophic Function at the Pre-Neoplastic Stage
Studies utilising transparent zebrafish larvae to model tumour initiation were the first to report a trophic inflammatory response at the pre-neoplastic stage. Feng et al. 2010 first showed that leukocytes are rapidly recruited following the expression of a single oncogene [6]. Subsequent zebrafish studies have shown a similarly rapid recruitment of neutrophils and/or macrophages to the skin, liver and brain in response to transgenic expression of numerous oncogenes, e.g. HRAS G12V , KRAS G12V , NRAS Q61K , Src, Xmrk, Myc and Akt [6,7,[39][40][41]60,68,104,[140][141][142][143]. Live imaging has provided insight into the dynamics of the intrinsic inflammatory response following tumour initiation within these models. Furthermore, the depletion of leukocytes has revealed that both macrophages and neutrophils have a potent tumour-promoting effect as early as the pre-neoplastic stage. Here we highlight the details of these findings within each tissue type:

Skin
In vivo live imaging with fluorescent leukocyte markers in the Tg(kita:Gal4TA, UAS:eGFP-HRAS G12V ) melanoma model showed that leukocytes are recruited to PNCs at the earliest stage of tumour initiation, even prior to the expansion of single clones [6]. Retention of leukocytes in the PNC microenvironment was sustained over time, indicative of chronic non-resolving inflammation. The up-regulation of pro-inflammatory cytokines in PNC-bearing larvae further confirmed an early onset of inflammation in response to tumour initiation. Hydrogen peroxide (H 2 O 2 ), a potent chemoattractant at wounds, was identified as the local chemoattractant guiding neutrophils towards PNCs. This discovery was facilitated by the use of live imaging, which allowed a direct comparison of neutrophil dynamics between tumour initiation and wounding, and the use of a fluorescent H 2 O 2 probe to visualise secretion from PNCs. Although both neutrophils and macrophages were observed to engulf material from PNCs, pan-leukocyte depletion (by Gcsfr + PU.1 morpholino) lead to a greatly decreased number of PNCs demonstrating their pro-tumour function [6,7]. Neutrophil depletion alone (using Gcsfr morpholino) lead to greater reduction of PNCs than macrophage depletion alone (using Irf8 morpholino), suggesting that neutrophils have a greater contribution to trophic inflammation in this model. Furthermore, PGE 2 was found to be one of the trophic factors released by leukocytes, directly promoting PNC proliferation through the EP1 receptor [7]. It is of note, however, that exogenous PGE 2 only partially rescued PNC numbers in pan-leukocyte-depleted larvae, which suggests that leukocytes produce additional, non-redundant trophic signals.
A more precise temporal resolution of the inflammatory response to tumour initiation was observed by tamoxifen-inducible transformation of keratinocytes with GFP-tagged HRAS G12V . Here, neutrophils were recruited within 8 "hours post induction" (hpi) and leukocyte depletion resulted in a 50% reduction of PNC number at 48 hpi [144]. By using correlative light and electron microscopy (CLEM) to study leukocyte recruitment in this model, it was observed that the dominant route by which immune cells gain access to PNCs is through proteolysis-independent breaching of weak spots within the basement membrane. However, proteolysis-dependent damage to the basement membrane was seen underneath areas of PNC clonal expansion, which suggests that PNCs can have invasive properties. Indeed, HRAS G12V -transformed keratinocytes expressed the invasion-related genes Mmp9 and Slug [104]. PNCs also expressed the chemokine, IL-8, which was shown to attract neutrophils via Cxcr2. Interestingly, reduced neutrophil recruitment also partially attenuated the expression of Mmp9 and Slug in PNCs, suggesting that neutrophils encourage the invasive behaviour of PNCs. Transcriptomic analysis of PNC-associated neutrophils in an NRAS Q61K melanoma model, showed upregulation of genes that correspond with growth-and invasion-promoting properties, e.g. FGF-1 and -6, cathepsin-H and galectin1 -and -3 [60].

Liver
Using the Tet-On system to induce KRAS G12V expression in the hepatocytes of zebrafish larvae triggered a trophic inflammatory response in the liver [41]. As seen in response to tumour initiation in the skin, neutrophils were recruited at approximately 8 hpi. Upon entering the liver, neutrophil motility was decreased, indicative of a chronic inflammatory response. Such retention behaviour has also been observed for TANs in mouse tumour models [145]. Following the induction of KRAS G12V , both PNCs and neutrophils expressed high levels of IL-1β and TGF-β. Abrogation of TGF-β signalling reduced liver size, decreased the number of infiltrating neutrophils, and triggered the upregulation of pro-inflammatory genes (IL-1β, TNFα, IL-6, IL-8, IL-12), indicating that TGF-β is a key modulator of neutrophil function in the pre-neoplastic microenvironment of the liver.
The Tet-On inducible system also allowed  to explore the role of both neutrophils and macrophages in response to tumour initiation in the adult fish liver [140]. PNCs expressed both the neutrophil chemoattractant, IL-8, and the macrophage chemoattractant, CSF-1, and an array of pro-inflammatory genes were upregulated in both macrophages and neutrophils (IL-1β, Cxcl1b, Nfkb2, CSF-1, IL-6 and IL-8). Leukocyte depletion completely attenuated the increase in PNC proliferation and survival observed in leukocyte-bearing larvae, restoring the size of the liver to that of wild-type. Depletion of macrophages had a comparable effect compared to the depletion of neutrophils; both resulted in a partial attenuation of liver enlargement and equivalent reduction in PNC proliferation, suggesting that both cell types provide important trophic signals. A cortisol-dependent sex bias in disease severity was observed in male fish consistent with that observed in human hepatocellular carcinoma (HCC). TGF-β expression and the levels of both macrophages and neutrophils in the liver were strongly correlated with this bias, further implicating TGF-β as a mediator of pro-tumour inflammation.
Transformation with other common HCC oncogenes, Xmrk and Myc, also elevated the expression of TGF-β in PNCs and triggered neutrophil and macrophage recruitment to the liver [141]. This demonstrates that oncogene-dependent inflammation is not specific to KRAS G12V . β-catenin overexpression also caused oncogenic transformation in hepatocytes, with associated recruitment of macrophages and neutrophils [78]. Whilst the independent influence of oncogene-dependent inflammation is yet to be determined in this model, a combination of β-catenin expression and high fat diet induced a strong tumour-promoting inflammatory response, which was associated with TNFα positive macrophages. Finally, modelling a rare type of liver cancer (fibromellar carcinoma), by transgenic expression of a DnaJ-PKAc fusion protein, resulted in leukocyte recruitment to the liver [146]. Here too, an increase of TNFα positive macrophages was observed, alongside a liver-wide activation of caspase-1, indicating inflammasome activation and IL-1β release. Chemical inhibition of either TNFα or caspase-1 reduced neutrophil and macrophage numbers to wild-type levels, demonstrating the potency of these cytokines as recruitment factors.

Brain
Glioblastoma can be modelled in the zebrafish by transgenic overexpression of Akt in neural cells, a gene commonly upregulated in human glioblastoma [142]. Oncogenic Akt transformation lead to an overabundance of microglia in the brain. These microglia exhibited reduced mobility, sustained contact with PNCs and a distinctive amoeboid morphology indicative of activation. Live imaging combining both macrophage-and microglia-specific reporters showed that the increase in microglia numbers was due to a recruitment of peripheral macrophages, which subsequently differentiated into microglia upon entry into the brain parenchyma. Treatment with anti-inflammatory drug, dexamethasone, or CSF-1R inhibitor reduced macrophage recruitment and PNC proliferation. Complete abrogation of macrophage recruitment by Irf8 −/− or Cxcr4 −/− resulted in even greater suppression of PNC proliferation. This demonstrates that macrophages/microglia have a potent trophic role during glioblastoma initiation and Cxcr4 is required for their recruitment.
Interestingly, microglia-mediated trophic support to PNCs in the brain might be cell-contact dependent. By live imaging with a transgenic Ca 2+ reporter, Chia et al. (2019) discovered that PNC-microglia interactions are governed by Ca 2+ -dependent ATP release from PNCs and detection of extracellular ATP by the P2y12 receptor on microglia [143]. This mechanism is normally employed under physiological conditions that govern recruitment of microglia to highly active neurons [143,[147][148][149][150]. Reduction of Ca 2+ levels, inhibition of ATP release, or inhibition of the P2y12 receptor all significantly abrogated PNC-microglia interactions and greatly reduced the rate of PNC proliferation, equivalent to that of macrophage/microglia depletion described above [142,143]. Therefore, ATP-P2y12 signalling, and possibly direct PNC-microglia contact, were required for the tumour-promoting effect of Akt-dependent inflammation in the brain.
Whilst Chia et al. (2018) reported an absence of neutrophils in the brain during Akt-dependent tumour initiation, in contrast, neutrophil recruitment into the brain was observed by Powell et al. (2018) following KRAS G12V transformation of astrocytes [68,142]. Although there is a discrepancy between these studies, both peripheral macrophage and neutrophils are found within human brain tumours [151]. In this case, KRAS G12V -expressing astrocytes also expressed high levels of IL-8, and blocking neutrophil recruitment by abrogation of Rac2 or Cxcr1 significantly reduced the proliferation of pre-neoplastic astrocytes [68]. This suggests that neutrophils have a trophic influence upon tumour initiation in the brain, and that IL-8-Cxcr1 signalling is required for neutrophil recruitment.

Mechanisms Governing Intrinsic Pro-Tumour Inflammation: Parallels between Zebrafish, Mouse and Man
Overall, the above zebrafish studies have identified numerous signals involved in the activation and recruitment of inflammatory leukocytes in response to tumour initiation (see Figure 1), in addition to trophic signals that feedback on PNC proliferation. Although equivalent studies of the pre-neoplastic stage do not exist in mammalian models, there is evidence that these signals are present at early stages of mammalian tumourigenesis (see Table 3). Mammalian studies also provide mechanistic clues as to how these signals govern tumour-promoting inflammatory responses. present at early stages of mammalian tumourigenesis (see Table 3). Mammalian studies also provide mechanistic clues as to how these signals govern tumour-promoting inflammatory responses. IL-1β is secreted by PNCs in response to oncogenic transformation, activating local immune cells, which in turn express inflammatory cytokines such as IL-1β and TNFα. H2O2 released from PNCs acts as a local chemoattractant for the recruitment of inflammatory leukocytes. CSF-1 and Cxcl12 contribute to the recruitment of macrophages through interaction with CSF-1R and Cxcr4 respectively, these signals are both PNC and macrophage derived. IL-8, expressed by both PNCs and neutrophils in response to tumour initiation, is responsible for neutrophil recruitment via Cxcr1/2.

Pro-Inflammatory Cytokines are Expressed in Response to Tumour Initiation
Zebrafish skin and liver cancer models showed an upregulation of the pro-inflammatory cytokines IL-1β and TNFα in response to tumour initiation [6,7,68,78,140,146]. Both of these inflammatory cytokines were important for the recruitment of leukocytes to the liver [146].
In mouse models of skin and colon cancer knock-out of IL-1β/IL-1R or downstream signal transducer, MyD88, lead to a dramatic and sustained abrogation of leukocyte recruitment, in addition to a reduced occurrence of colon polyps and skin papillomas [152][153][154]. This demonstrates that IL-1β is required for instigating tumour-promoting inflammation during early tumourigenesis in mice. IL-1β activates and amplifies host inflammatory responses by activation and recruitment of innate immune cells via expression of pro-inflammatory cytokines and chemokines [155,156]. In support of this mechanism in response to tumour initiation, in vitro studies have shown that IL-1β, TNFα and IL-6 expression are induced by transformation with oncogenic forms of Ras in various human and mouse cell types [154,[157][158][159][160][161].
Knock-out of TNFα or TNFR1 also reduced leukocyte infiltration and increased resistance to tumour development in mouse models of skin and liver cancer [162][163][164]. TNFα signalling is wellestablished to upregulate pro-inflammatory cytokines and chemokines, and can itself act as a chemoattractant signal, which may explain the requirement of TNFα for leukocyte recruitment in both zebrafish and mouse studies [165]. Dichotomously, TNFα signalling can also directly induce apoptosis of cancer cells, or promote cell survival through activation of NF-κB signalling [165]. Although the direct effect of TNFα upon PNCs has not been explored in zebrafish models of tumour initiation, the latter mechanism is more likely since no signs of leukocyte-mediated cell death have been recorded and TNFα positive macrophages promote tumourigenesis in zebrafish liver cancer models [140,146].

Pro-Inflammatory Cytokines are Expressed in Response to Tumour Initiation
Zebrafish skin and liver cancer models showed an upregulation of the pro-inflammatory cytokines IL-1β and TNFα in response to tumour initiation [6,7,68,78,140,146]. Both of these inflammatory cytokines were important for the recruitment of leukocytes to the liver [146].
In mouse models of skin and colon cancer knock-out of IL-1β/IL-1R or downstream signal transducer, MyD88, lead to a dramatic and sustained abrogation of leukocyte recruitment, in addition to a reduced occurrence of colon polyps and skin papillomas [152][153][154]. This demonstrates that IL-1β is required for instigating tumour-promoting inflammation during early tumourigenesis in mice. IL-1β activates and amplifies host inflammatory responses by activation and recruitment of innate immune cells via expression of pro-inflammatory cytokines and chemokines [155,156]. In support of this mechanism in response to tumour initiation, in vitro studies have shown that IL-1β, TNFα and IL-6 expression are induced by transformation with oncogenic forms of Ras in various human and mouse cell types [154,[157][158][159][160][161].
Knock-out of TNFα or TNFR1 also reduced leukocyte infiltration and increased resistance to tumour development in mouse models of skin and liver cancer [162][163][164]. TNFα signalling is well-established to upregulate pro-inflammatory cytokines and chemokines, and can itself act as a chemoattractant signal, which may explain the requirement of TNFα for leukocyte recruitment in both zebrafish and mouse studies [165]. Dichotomously, TNFα signalling can also directly induce apoptosis of cancer cells, or promote cell survival through activation of NF-κB signalling [165]. Although the direct effect of TNFα upon PNCs has not been explored in zebrafish models of tumour initiation, the latter mechanism is more likely since no signs of leukocyte-mediated cell death have been recorded and TNFα positive macrophages promote tumourigenesis in zebrafish liver cancer models [140,146].

Chemokines Recruit Leukocytes to PNCs
Zebrafish models of both brain and liver cancer have indicated that CSF-1 contributes to the recruitment of macrophages in response to tumour initiation [140,142]. Likewise, CSF-1 signalling is a well-established macrophage recruitment signal in mammalian cancers [166][167][168][169][170]. With respect to early tumourigenesis in mice, abrogation of CSF-1 signalling reduced tumour formation in mouse models of thyroid cancer and pancreatic neuroendocrine cancer [171,172]. However, blocking CSF-1 signalling in breast cancer models only affected the progression to invasive metastatic carcinoma [173]. Similarly, in the development of mouse colon cancer, lack of CSF-1 lead to a modest reduction in the number of small polyps but a dramatic reduction in the number of large polyps, suggesting a role in promoting later progression [174]. Therefore, the involvement of CSF-1 during tumour initiation might be tissue-specific. In addition to its role in macrophage recruitment, CSF-1 contributes to the control of the M2 phenotype of mammalian TAMs; both pre-clinical and clinical studies have shown that inhibition of CSF-1 signalling does not always deplete TAMs from established tumours, but rather impairs their tumour-promoting functions [170,175]. This suggests that other factors also drive the recruitment and retention of macrophages in cancer. Chia et al. (2018) showed that, whilst CSF-1 played a role in macrophage recruitment, signalling to Cxcr4 was absolutely required for the recruitment of macrophages to the Akt-transformed brain in zebrafish tumour initiation, and that PNC-derived Cxcl12 was the major ligand for this interaction [142]. Cxcr4 signalling has been shown to promote progression of many cancers in mammals including glioblastoma, through both cancer cell autonomous signalling and leukocyte interactions [106,[176][177][178][179][180][181]. However, the pro-tumour effect of Cxcr4 signalling at the preneoplastic stage was macrophage specific [142]. In the mouse APC min model of colorectal cancer, Cxcr4 −/− mice exhibited reduced macrophage infiltration and polyp formation [182]. Furthermore, in the mouse Lewis lung carcinoma model, inhibition of Cxcr4 greatly reduced tumour-associated inflammation and tumour growth [183]. Thus, there is some evidence that Cxcr4 signalling recruits macrophages and has a pro-tumour effect during early stage tumourigenesis in mammalian models. Cxcr4 has also been implicated in the recruitment of neutrophils and G-MDSCs to cancer cells in mouse xenograft models, resulting in tumour-promoting effects [182,[184][185][186].
IL-8 was expressed in zebrafish PNCs in response to oncogenic Ras expression [140,187] and IL-8 signalling through Cxcr1/2 chemokine receptors was shown to drive pro-tumour neutrophil recruitment at the pre-neoplastic stage in zebrafish models of skin and liver cancer, contributing to PNC proliferation and sustained tumour-promoting inflammation [68,104]. Likewise, Ras-induced expression of IL-8 promotes neutrophil recruitment and tumourigenesis in mouse xenograft models [159,188,189]. IL-8 has also been found to provoke a neutrophil-dependent anti-tumour effect in mouse models, consistent with the ability of TANs to perform both pro-and anti-tumour functions [190]. More broadly, Cxcr1/2 signalling has been shown to recruit pro-tumour TANs and/or G-MDSCs in a variety of mammalian cancers [191][192][193][194][195]. With respect to early tumourigenesis, Cxcl1 and Cxcl2 have been shown to drive neutrophil and G-MDSC infiltration in mouse models of colorectal cancer, promoting polyp and tumour formation [137,196,197]. These findings suggest that signalling via the IL-8 -Cxcr1/2 chemokine axis promotes tumour initiation.

Plasticity and Heterogeneity of Leukocytes
The expression of pro-inflammatory cytokines by macrophages and neutrophils in response to tumour initiation in zebrafish studies corresponds with the hypothesis that tumour-associated inflammation begins with a pro-inflammatory response, which becomes progressively more anti-inflammatory throughout the course of tumourigenesis [86,87,[198][199][200][201]. This phenomenon mimics the inflammatory wound response; in both zebrafish and mammals the wound response is characterised by early M1/N1 activation followed by a switch to M2/N2 activation states, the latter of which promotes tissue repair and resolution [51,[202][203][204]. The exception here is that tumours provoke a sustained response, thus creating a positive-feedback cycle of non-resolving inflammation and tissue repair [198]. Interestingly, macrophages and neutrophils within zebrafish studies had trophic effects at the pre-neoplastic stage (as early as 24 h following oncogenic transformation), and in the presence of pro-inflammatory signals [6,7]. This suggests that, in the case of intrinsic inflammation, the M2/N2 switch occurs at the pre-neoplastic stage. The simultaneity of pro-inflammatory and trophic signals may be achieved by an intermediate phenotype between M1/N1 and M2/N2, and/or heterogeneity of leukocyte responses. In support of the latter, leukocyte heterogeneity was observed in zebrafish studies of tumour initiation; only 20% of leukocytes recruited to PNCs in the skin expressed PGE synthase, whilst minor populations of macrophages were positive for either M1 (TNFα) or M2 (arginase-1) markers [6,7]. Furthermore, in response to oncogenic transformation in the liver, only 35% of macrophages were TNFα positive [146].

PGE 2 as Trophic Factor and Immunomodulator
As previously discussed, PGE 2 is a central component of pro-tumour inflammation in humans and mice, and has been identified as a potent leukocyte-derived trophic signal at the pre-neoplastic stage in zebrafish [7]. In support of this finding, PGE 2 is also secreted by macrophages within neoplastic polyps in humans and mice [205][206][207]. PGE 2 signalling has a direct trophic function in mouse models of colon cancer, for example, through activation of β-catenin [128]. Macrophage-derived PGE 2 also increases COX-2 expression in neoplastic cells, creating a positive feedback cycle, which greatly increases the local level of PGE 2 in the colon, driving tumourigenesis [208].

TGF-β Governs Pro-Tumour Neutrophils
In zebrafish models of liver tumour initiation, high levels of TGF-β expression by PNCs are associated with both disease severity and the expression of anti-inflammatory genes by pro-tumour neutrophils [41,141]. The mechanism has also been indicated in mammals by treatment of mouse tumours with a TGFBR inhibitor, which resulted in a reduction of neutrophils and reduced tumour size [226]. These neutrophils showed an upregulation of pro-inflammatory cytokines, chemokines and iNOS, and had a direct cytotoxic effect on tumour cells due to ROS secretion. Although TGF-β signalling can itself affect cancer cell survival, the reduction in tumour size was neutrophil-dependent. Further studies in mouse models of cancer have also implicated TGF-β as a regulator of pro-tumour G-MDSCs. These studies showed that anti-TGFβ antibodies had a therapeutic effect by depleting MDSCs in a breast cancer model [227], and myeloid-specific deletion of TGFBR2 reduced tumour metastasis in breast, lung and melanoma models [228]. More recently, the role of TGF-β in the recruitment of pro-tumour neutrophils has been described in mouse models of hepatocellular carcinoma through Cxcl5, the expression of which corresponds to disease severity in human patients [229]. Whilst these studies do not represent tumour initiation, they do indicate that TGF-β has a pro-tumour effect in mammalian cancer, including liver cancer, via recruitment and modulation of neutrophils.

Extracellular ATP Attracts Pro-Tumour Microglia/Macrophages
Chia et al. (2019) showed that the trophic effect of microglia during glioblastoma tumour initiation was dependent upon Ca 2+ -mediated ATP release from PNCs and signalling through the P2y12 receptor [143]. This represents a repurposing of an endogenous mechanism found in both zebrafish and mice, which governs microglia recruitment to areas of high calcium, such as injury and seizures [147][148][149][150]. Similarly, in a mouse model of glioma, extracellular ATP signalling through the purinergic receptor, P2X7R, promotes the recruitment of macrophages and microglia [230]. Although microglia are quite divergent in comparison to other macrophage types [231], extracellular ATP signalling though purinergic receptors also modulates the function of macrophages in inflammatory responses outside of the brain and promotes M2 polarisation in TAMs [232][233][234]. Therefore, this represents a mechanism for further study with regards to tumour initiation in other tissue types. Table 3. Signals governing the trophic inflammatory response to tumour initiation in zebrafish are conserved in mouse models of early neoplastic development.

Zebrafish (Pre-neoplastic Stage)
Mouse (Early Neoplastic Stage) Released by PNCs and neighbouring cells in the skin [6]. Unknown Promotes PNC proliferation and local leukocyte recruitment [6].

IL-1β
Expressed in response to tumour initiation in the skin, liver and brain [6,41,68,140,146].
Promotes neoplasm formation and leukocyte recruitment in the skin and colon [152][153][154]. Promotes PNC proliferation and leukocyte recruitment in the liver [146].

TNF-α
Expressed in response to tumour initiation in the skin and liver [6,78,140,146]. Promotes neoplasm formation and leukocyte recruitment in the skin and liver [162][163][164]. Promotes PNC proliferation and leukocyte recruitment in the liver [146].

CSF-1
Expressed in response to tumour initiation in the liver [140].
Promotes neoplasm formation and macrophage recruitment in thyroid and pancreas [171,172]. Promotes PNC proliferation and leukocyte recruitment in the brain [142].
Only promotes later stages of breast and colon cancers [173,174].

Cxcl12-Cxcr4
Cxcl12 is expressed in response to tumour initiation in the brain [142]. Promotes neoplasm formation and macrophage recruitment in colon and lung [182,183]. Promotes PNC proliferation by macrophage recruitment [142].
IL-8 is expressed in response to tumour initiation in the skin, liver and brain [6,41,68,104,140].
Promotes neoplasm formation and the recruitment of neutrophils and G-MDSCs in the colon [137,196,197]. Cxcr1/2 signalling promotes PNC proliferation and neutrophil recruitment in the brain and skin respectively [68,104].

PGE 2
Produced by leukocytes in response to tumour initiation in the skin [7].

ATP
Released by PNCs in response to tumour initiation in the brain [143]. Unknown Promotes PNC proliferation and microglia contact via purinergic signalling [143].

Future Perspectives
Zebrafish cancer models have shown that PNC-induced inflammation is an important player with respect to tumour initiation. Hitherto, the role of inflammatory leukocytes at the pre-neoplastic stage was largely unknown. Imaging of transparent zebrafish larvae has demonstrated that neutrophils and macrophages are recruited in response to tumour initiation and both have a trophic effect upon PNCs. The ever-expanding collection of transgenic reporters, combined with efficient methods for genetic manipulation in the zebrafish, provide a toolbox for the future analysis of cellular and signalling responses within the developing PNC niche. As such, the zebrafish model will provide much needed insight to support the development of novel strategies for cancer prevention.
Thus far, zebrafish studies have identified some of the chemoattractants, inflammatory mediators and trophic factors that contribute to PNC-induced inflammation. Whilst these signalling molecules correspond with pro-tumour responses in mouse and man, further research is required to confirm their mechanisms of action at the pre-neoplastic stage. There also remain additional signals that are yet to be discovered, for example, it is apparent that PGE 2 is not the sole leukocyte-derived factor influencing PNC proliferation at this stage [7]. Furthermore, the mechanisms that govern the pro-tumour phenotype of macrophages and neutrophils remain unclear. A starting point to tackle this question would be to study known factors involved in polarising TAMs and TANs in the tumour microenvironment for a role at the PNC niche. Chemical inhibitors for blocking TAM recruitment and/or M2 polarisation are currently in clinical trials for cancer treatment and may also be promising for cancer prevention [235]. Considering the pro-tumour role of neutrophils in zebrafish studies of tumour initiation, targeting neutrophils may also be an effective approach. Moreover, a more potent effect may be achieved by developing strategies to reprogram macrophages and neutrophils to an 'M1' or 'N1' state, wherein they may partake in anti-tumour responses. Some promise has been shown in this area, regarding the reprogramming of TAMs in mouse models [175,220,235,236], but this is yet to be explored in zebrafish. Traditional methods for studying leukocyte polarisation in mammalian immunology involve the analysis of multiple cell surface markers by flow cytometry, a method which is not compatible with zebrafish due to a lack of zebrafish-specific antibodies. However, the recent development of single-cell RNA-sequencing technologies provides a powerful solution. This approach will provide insight into macrophage and neutrophil heterogeneity within the PNC microenvironment and identify marker genes that can be used for the development of novel transgenic reporters. It will also drive the discovery of novel factors that mediate trophic interactions between leukocytes and PNCs.
In addition to promoting PNC proliferation, zebrafish cancer models have also provided evidence that genes involved in angiogenesis and invasion are expressed at the pre-neoplastic stage by both PNCs and inflammatory cells [60,104,140,144]. This is of interest since these processes are traditionally associated with later stages of cancer but are likewise difficult to capture at early stages in mammalian models [2]. This presents angiogenesis and invasion, and their relationship with inflammation, as subjects for further study in zebrafish tumour initiation models. Dissecting complex multicellular mechanisms such as these will be facilitated by the ease of genetic manipulation in the zebrafish. For example, a system for cell-type specific CRISPR-mediated gene depletion has recently been developed [237,238]. This will help to pick apart the functions of specific cell types and the roles of pleiotropic genes.
Alongside innate inflammatory cells, tumour-infiltrating lymphocytes also play key roles in the tumour microenvironment [13][14][15]. Furthermore, recent advances in immunotherapy have revolutionized cancer treatment by promoting the adaptive immune response against cancer cells [239,240]. However, little is known as to when and how adaptive immune cells respond to pre-neoplastic or early neoplastic stage tumour development. Zebrafish may prove a useful model for addressing this question. The developmental origin, anatomical location and genetics underlying the development of adaptive immunity are largely conserved between zebrafish and mammals [241][242][243][244][245][246]. In recent years, studies of zebrafish lymphocytes have also revealed the existence of functionally conserved mature T cell subsets, such as CD4 + helper T cells [247,248], T regulatory cells [247,249,250] and γδ T cells [251]. The characterisation of mature lymphocyte subpopulations in zebrafish remains an ongoing topic of research, with an expanding collection of fluorescent reporter lines, e.g. pan T cell reporter, Tg(lck:eGFP) [244]; CD4 + T cell reporter, Tg(CD4-1:mCherry) [247]; and regulatory T cell reporter, Tg(foxp3a:eGFP) [249,250]. This will allow researchers to explore the role of T cells during early tumour development using in vivo live imaging.
Other common components of the tumour microenvironment include extracellular vesicles (EVs) and stromal cells, which may also play a role at the pre-neoplastic stage. EVs are known to transmit signals that modulate the behaviour of recipient cells. Cancer-derived EVs have been shown to promote M2 polarisation of TAMs in some cases, and the expression of pro-inflammatory cytokines in others [252][253][254]. Recently, both a fluorescent probe [255,256] and a transgenic line [257] have been developed that specifically label EVs in zebrafish, making it possible to track their transit in a live in vivo model. For example, one study showed that tumour-derived EVs activated macrophages, resulting in a macrophage-dependent promotion of metastatic outgrowth at distal sites [255]. These tools will be invaluable in addressing the question of whether PNC-derived EVs modulate host inflammatory cell function. Stromal cells within the PNC microenvironment are also of interest. 'Cancer-associated fibroblasts' have been identified as cancer-promoting agents in mammalian tumours [258], but their role at the pre-neoplastic stage remains largely unexplored. One study has recently shown that leukocytes recruited in response to tumour initiation in the liver release serotonin, which activates mesenchymal stromal cells and encourages them to secrete TGF-β, amplifying pro-tumour inflammation [259]. This demonstrates a tumour-promoting role for mesenchymal stromal cells during liver tumour initiation and suggests a direction for future study in other tissues.
As well as detailed mechanistic studies, the zebrafish model is also amenable to high-throughput screening. The small size of zebrafish embryos allows imaging of the whole organism and several individuals can be screened simultaneously. For example, Precazzini et al. (2019) developed an automated, imaging-based screen with 384-well plates to detect modifiers of melanocyte hyper-proliferation in a zebrafish melanoma model [260]. Taking advantage of existing fluorescently tagged cancer models, it would be possible to perform in vivo imaging-based screening for cancer preventative chemicals. The development of highly efficient methods for CRISPR-mediated gene deletion in first generation larvae, known as 'crispants', also provides the means to carry out high-throughput CRISPR screening for the discovery of novel targets [238]. Thus, the zebrafish is a powerful model with the potential to reveal mechanisms governing pre-neoplastic lesion development, and to test strategies for cancer prevention.