There is a long‐standing association between wound healing and cancer, with cancer often described as a “wound that does not heal”. However, little is known about how wounding, such as following surgery, biopsy collection or ulceration, might impact on cancer progression. Here, we use a translucent zebrafish larval model of RasG12V‐driven neoplasia to image the interactions between inflammatory cells drawn to a wound, and to adjacent pre‐neoplastic cells. We show that neutrophils are rapidly diverted from a wound to pre‐neoplastic cells and these interactions lead to increased proliferation of the pre‐neoplastic cells. One of the wound‐inflammation‐induced trophic signals is prostaglandin E2 (PGE2). In an adult model of chronic wounding in zebrafish, we show that repeated wounding with subsequent inflammation leads to a greater incidence of local melanoma formation. Our zebrafish studies led us to investigate the innate immune cell associations in ulcerated melanomas in human patients. We find a strong correlation between neutrophil presence at sites of melanoma ulceration and cell proliferation at these sites, which is associated with poor prognostic outcome.
This study reveals how innate immune cells, in particular neutrophils, that are initially drawn to a wound can subsequently be attracted away to nearby early‐ and late‐stage cancer cells and drive their proliferation.
Both chronic and acute wounds exacerbate cancer growth.
Tissue damage in larval zebrafish, or cancer surgery in adults, draws in neutrophils and macrophages.
Neutrophils are recruited from wounds to nearby pre‐neoplastic cells and deliver trophic signals.
Neutrophil presence correlates with tumour cell proliferative index and indicates poor prognosis in ulcerated human melanoma.
Chronic, persistent inflammation has been shown to damage affected organs and may predispose to many diseases ranging from diabetes to cancer (O'Byrne & Dalgleish, 2001). Epidemiological studies indicate that at least 20% of all cancers begin as a direct consequence of chronic inflammatory disease in different tissues and organs (Grivennikov et al, 2010), and inflammation is considered to be one of the ten “hallmarks of cancer” (Hanahan & Weinberg, 2011). Inflammation‐driven cancers include those associated with chronic viral infections such as hepatitis and hepatocellular carcinoma (Lin et al, 2015) or bacterial infections such as Helicobacter pylori, which accounts for almost all stomach cancers (Mantovani & Sica, 2010). Similarly, local chronic inflammation is a consequence of attack by the parasitic worm Schistosoma hematobium causing bladder cancer in some parts of the world (Condeelis & Pollard, 2006). Local chronic tissue inflammation also often leads to malignant transformation (Werner & Schafer, 2008), as for example in Barrett's oesophagus (Colleypriest et al, 2009). Moreover, studies have shown that patients receiving long‐term therapy (>5 years) with anti‐inflammatory drugs, such as aspirin, have fewer relapses or appearances of new tumours, particularly colon cancer, providing further evidence for the pro‐cancer influence of long‐term inflammation (Rothwell et al, 2011).
Surgery is a key cancer therapy and is still the most effective means to treat human solid cancers, which have not yet metastasised (Ceelen et al, 2013). However, tissue damage is implicated as a possible trigger in the development of various cancers (Combemale et al, 2007; Lee et al, 2009; Kasper et al, 2011; Senet et al, 2012) and may provide a favourable niche for tumour reoccurrence (Hofer et al, 1998), as well as facilitating the growth of pre‐existing micro‐metastases (Bogden et al, 1997), suggesting that surgery may have clinical consequences beyond simply removing the primary cancer (Kuraishy et al, 2011; O'Leary et al, 2013). Classic studies have shown that wounding can lead to tumourigenesis (Hennings & Boutwell, 1970; Clark‐Lewis & Murray, 1978; Leder et al, 1990); for example, wounding Rous sarcoma virus‐infected chickens led to 100% tumour formation at the injured site (Sieweke et al, 1990), and transgenic mice carrying the v‐jun oncogene developed dermal fibrosarcomas after full thickness wounding, whereas identical wounds in non‐transgenic mice healed without tumour formation (Schuh et al, 1990). A more recent retrospective analysis suggested that the use of anti‐inflammatories such as ketorolac (a non‐steroidal anti‐inflammatory drug) given to patients before and after mastectomy led to a lower reoccurrence of their breast cancer, implicating surgery‐mediated inflammation as a key initiator of wound‐induced cancer growth (Forget et al, 2010; Retsky et al, 2013).
Under normal acute inflammatory situations, such as after tissue damage, or an infection, the inflammatory response is self‐limiting and immune cells resolve by apoptosis or returning to the circulation (Martin & Shaw, 2009). In malignant tissues, however, pro‐inflammatory signals continue to intensify to support the needs of the tumour. Hence, the inflammatory response never resolves, and tumours have been likened to “wounds that do not heal” (Dvorak, 1986; Chang et al, 2004; Werner & Schafer, 2008; Troester et al, 2009). Tumour‐associated macrophages (TAMs) and neutrophils (TANs) can constitute a large proportion of the tumour mass (Condeelis & Pollard, 2006) and are associated with poor prognosis in human patients, particularly when tumour‐derived cytokines induce macrophage differentiation from a tumouricidal M1 phenotype to an M2 phenotype, which favours growth and tissue remodelling (Biswas et al, 2006).
In this study, we have utilised zebrafish as a model organism to visualise the relationship between wound‐associated inflammation and adjacent cancers as they develop in vivo. The translucency of zebrafish larvae enables us to live image these interactions from the earliest stages when a pre‐neoplastic cell first arises in otherwise healthy tissue. We express oncogenic RasG12V in specific cell types: melanocytes (to model melanoma) and goblet cells (modelling rare carcinoid tumours). We have previously shown that neutrophils and macrophages interact with the pre‐neoplastic cells, even at a single‐cell stage, before these cells divide to form clones (Feng et al, 2010), and that these interactions are beneficial for pre‐neoplastic cell growth, in part due to release of trophic factors such as PGE2 from myeloid cells (Feng et al, 2012). Here, we show that immune cells recruited to a wound are rapidly drawn out from the wound by competing signals from pre‐neoplastic cells. The amplified exposure to innate immune cells is associated with increased proliferation of pre‐neoplastic cells that we show is dependent upon innate immune cells. We go on to extrapolate these mechanistic studies in zebrafish to human, clinical samples of melanomas with superimposed wounds (ulceration), and we find a close correlation of neutrophil influx with proliferative index of cancer cells in these ulcerated melanomas. We have previously seen that the presence of tumour infiltrating neutrophils and macrophages in primary melanomas was correlated with poor survival (Jensen et al, 2009, 2012). We now show that neutrophil number correlates with proliferation from non‐ulcerated melanomas through to moderate ulceration, and this proliferative microenvironment may help to explain why wound‐induced inflammation may be detrimental to patient survival.
Adult zebrafish tumours arise at sites of repeated tissue wounding
In our previous studies, we observed that zebrafish expressing a mutant RasG12V oncogene in their skin tend to develop invasive melanomas at sites that are prone to friction and other damage, for example on the ventral fin and tail fin (Feng et al, 2010) (Supplementary Fig S1D and E). We wanted to test whether this correlation reflected a true causal association, and so we repeatedly wounded the tail fin of one group of juvenile RasG12V fish fortnightly for 3 months and compared tumour outcome with equivalent siblings that had not been wounded. By the second round of fortnightly wounds, we saw a clear increase in pigment intensity in the wounded tail fins (Fig 1A and B), reflecting recruitment and/or local proliferation of pre‐neoplastic melanocytes. This increase in pigmentation, as measured by threshold analysis (Fig 1C and D), appeared to reach a plateau after five rounds of wounding, and we left the fish unwounded after six wounds. By 18 months after the initial wounding, we observed that 43% of all wounded fish had developed melanoma at the site of wounding while none of the unwounded fish exhibited fin tip melanomas (Fig 1A, B and E), suggesting that chronic wounding can indeed increase the likelihood of cancer growth in tissues that already have a predisposition to developing melanoma. This chronic wounding study was repeated in adult 3‐ to 6‐month‐old fish and showed the same result (Supplementary Fig S2A–D).
Cancer surgery triggers a wound inflammatory response and subsequent influx of neutrophils and macrophages to regions of remaining cancer
Since cancer surgery is an instance whereby a wound is inflicted in the vicinity of a growing cancer, we wondered how a single surgery might impact on these cancer cells. In particular, we wanted to examine how the acute wound inflammatory response which is triggered at any site of tissue injury might draw immune cells to the cancer cells and what the consequences of this might be. The time course of recruitment of inflammatory cells to a tail wound in control fish, without cancer, begins with a first influx of neutrophils (LysC+ cells) within an hour of wounding, with numbers peaking at 24 h post‐injury; while resident macrophage (mpeg+ cells) numbers only begin to increase as neutrophil numbers are diminishing at 6 or 7 days post‐wounding, precisely as previously reported for adult fin regeneration studies (Petrie et al, 2014) (Supplementary Fig S3H). To investigate adult cancer surgery, we used two models. In one of these, we cut a segment from large melanomas on the tail fin of adult fish and subsequently examined the in situ remaining portion of tumour 3 days later (Fig 1F–J). Immunostaining of the initially removed cancer reveals the presence of low levels of neutrophils (Fig 1G), and staining for phospho‐histone H3 shows an associated low level of cell proliferation (Fig 1H). At 3 days post‐surgery, the remaining region of cancer appears heavily populated with neutrophils (Fig 1I), and sections of this region show an associated increase in phospho‐histone H3 staining (Fig 1J), suggesting that local tissue proliferation might be triggered at any site of surgery because of the associated inflammatory influx. To more clearly image neutrophil influx post‐wounding in adult tissues, we selected smaller, flatter melanomas and made punch biopsies in these to include both tumour and healthy tissue (Fig 1K). Whole‐mount imaging of the initially removed biopsy reveals some neutrophils throughout the melanoma right up to the interface between cancer and healthy tissue (Fig 1K'), which reflects previously documented histopathological observations of surgically removed human cancers (Galdiero et al, 2013). At 24 h post‐biopsy, we see significant recruitment of neutrophils to both healthy and tumour wound edge (Fig 1L); however, after 3 days, neutrophils have resolved away from the healthy wound edge but remain at high levels throughout the wounded tumour tissue (Fig 1M).
Live imaging in translucent larvae reveals neutrophils resolving from wounds and recruited to adjacent pre‐neoplastic cells
To gain a more dynamic impression of how wounding may influence the behaviour of innate immune cells in the vicinity of cancer cells, we made a series of laser wounds adjacent to clones of pre‐neoplastic cells on the flanks of zebrafish larvae (Fig 2A and A') which are amenable to live imaging because of their translucency. Our previous studies have shown how fluorescently labelled neutrophils and macrophages are recruited to pre‐neoplastic goblet cells expressing mutant RasG12V and GFP (Feng et al, 2010). These innate immune cells are recruited by stochastic pulses of hydrogen peroxide (Feng et al, 2010), the same signal that has been shown to draw neutrophils to wounds (Niethammer et al, 2009), and they remain at one pre‐neoplastic clone for brief periods before moving on to visit adjacent clones (Fig 2B, Supplementary Movie S1). Tissue wounding triggers an acute, rapid recruitment of large numbers of neutrophils to the wound, but rather than remaining predominately within the wound site for up to 3 or 4 h, as in control wounded fish with no burden of pre‐neoplastic cells (Fig 2C, Supplementary Movie S2), many of these immune cells are distracted from the wound and “visit” the nearby pre‐neoplastic cells (Fig 2D, Supplementary Movie S3); this is most clearly visualised by “footprints” of neutrophil tracks from 90 min to three hours post‐wounding which extend well beyond the wound site in fish carrying a pre‐neoplastic cell burden, whereas they remain in the vicinity of the wound in fish without pre‐neoplastic cells (Fig 2C'' and D''). If wound‐triggered hydrogen peroxide release is blocked by treatment with DPI, then few, if any, neutrophils are drawn to the wound (Fig 2E), and consequently, many fewer visits to nearby pre‐neoplastic cells are seen (Supplementary Movie S4). Because wounding standardly draws many more neutrophils to the flank than are normally present (Fig 2E), this leads to considerably more opportunity for contacts with pre‐neoplastic cells. Indeed, in this period following wounding, we observe 64.2% of Ras+ cells adjacent to a wound receive neutrophil contacts compared to only 26% of pre‐neoplastic cells in a comparable region of an unwounded larvae (Fig 2F). Contacts ranged from less than one minute to more than 90 min, and we have used them as a proxy for neutrophil/pre‐neoplastic cell interactions although we have no evidence that physical contacts between these two lineages are necessary for one cell to influence the other. To follow neutrophil and macrophage recruitment over a substantially longer period, we fixed Ras+ larvae and their WT siblings at various time points up to 5 days post‐wounding (Fig 2G–J). To distinguish macrophages, we immunostained larvae with L‐plastin, which is known to be a pan‐leucocyte marker in zebrafish larvae, and considered cells which were L‐plastin+ but LysC− to be macrophages (Feng et al, 2010; Jones et al, 2013). Over the 5 days post‐wounding, the number of neutrophils recruited to flank pre‐neoplastic cells in wounded Ras+ larvae was significantly higher (P = 0.007) than in unwounded Ras+ larvae (Fig 2P). In contrast, despite the large numbers of macrophages recruited to the wound by 24 h post‐wounding, there appears to be no significant increase in macrophage recruitment to nearby pre‐neoplastic cells when comparing Ras+ unwounded and Ras+ wounded larvae over the total time course (Fig 2Q; P = 0.1019). However, we do observe increased recruitment of macrophages to wounds in Ras+ larvae compared to their WT siblings (Fig 2Q; P = 0.0493), and once recruited, they appear to persist in the area around the wound for longer than in WT siblings due to the presence of the pre‐neoplastic cells (Fig 2M and N). From the earliest time points examined in fixed larvae—3 h after wounding—we observe a significant increase in the number of pre‐neoplastic cells receiving contacts from immune cells (P = 0.0052 for 5 days post‐wound); the number of these contacts peaks by 24 h post‐injury and is maintained for at least 3 days post‐wounding (Fig 2R).
Wounding and the associated increase in inflammatory response appear to trigger increased proliferation by pre‐neoplastic cells
To determine what effect this increased inflammation might have on pre‐neoplastic cell proliferation, we counted the number of pre‐neoplastic cells 3 days post‐wounding in a series of domains extending up to 250 μm away from the wound centre and corresponding approximately to the region of increased inflammation. We see a significant increase in pre‐neoplastic cell number in comparison with equivalent flank regions of unwounded larvae (Fig 3A, B and E) in bands extending 50–150 μm and 150–250 μm from the wound centre (Supplementary Fig S4), and this is corroborated by a significant increase in EdU‐positive pre‐neoplastic cells in 2‐days post‐wounding larvae (Fig 3C and D, Supplementary Fig S5A), where the thymidine analogue EdU has been incorporated into the DNA of proliferating cells. This effect is local, rather than systemic, because we see no significant increase in numbers of Ras+ cells in the epithelium beyond 250 μm from the wound centre (Supplementary Fig S4), or overlying the yolk or in the head (Supplementary Fig S5). Moreover, only wounds that are sufficiently large to trigger a wound inflammatory response over 48 h appear to trigger a pre‐neoplastic proliferative response; small wounds show no increase above background, unwounded, levels (Supplementary Fig S5C). To further test whether this increase is a consequence of the wound inflammatory response, we transiently delayed innate immune cell development by injection of PU.1 and GCSF morpholinos at the one‐cell‐stage embryo. This led to an almost complete depletion of neutrophils and macrophages until 4 days post‐fertilisation (dpf) (Feng et al, 2012) and, as a consequence, we observed a significant decrease in pre‐neoplastic cell numbers, suggesting reduced proliferation both in unwounded control fish, as previously reported (Feng et al, 2010), and in 2‐days post‐wounded fish, by comparison with fish with a normal wound inflammatory response (Fig 3L).
Wound‐associated inflammation drives clonal growth rather than initiation of new clones
The larger number of pre‐neoplastic cells in the vicinity of a wound could be due to increased initiation of new clones or increased cell division within clones, or contributions from both of these. However, it is clear that the numbers of clones counted at 2 days after wounding are generally unaltered by comparison with unwounded fish; rather, the increase in neoplastic cell number is almost entirely due to increase in the size of individual clones (Fig 3M). Time‐lapse studies of individual fish over a period of several days confirm this and indicate that the biggest relative change in clonal size at the wound occurs at about 2 days post‐wounding (Supplementary Fig S6). Such studies also reveal some movement of clones towards the wound during the repair process, likely a passive behaviour reflecting that of their healthy epithelial neighbours as the wound closes.
Depleting innate immune cells by morpholino knockdown of PU.1 and GCSF leads to similar numbers of clones—usually around 10 clones within the wound region—but many of these remain as single cells even after wounding. This strongly suggests that innate immune cells are indeed delivering trophic signals to clones of pre‐neoplastic cells and encouraging proliferation, rather than delivering factors that initiate new clones.
Wound‐associated neutrophils are responsible for driving increased proliferation
We were able to further dissect the role of neutrophils versus macrophages in this wound inflammatory‐driven pre‐neoplastic cell proliferation using morpholinos against GCSF, to delay the development of neutrophils (Liongue et al, 2009) (Fig 3J), and irf8 which blocks the development of macrophages (Li et al, 2011) (Fig 3K). Delaying the development of neutrophils significantly depleted pre‐neoplastic cell numbers, and there was no observed increase in pre‐neoplastic cell number after wounding (Fig 3N). However, when macrophages were depleted (with a compensatory increase in neutrophils), the numbers of pre‐neoplastic cells were only partially reduced, and there was an observed increase after wounding (Fig 3O).
PGE2 is part of the signal that drives wound‐inflammation‐mediated pre‐neoplastic cell proliferation
Since we have previously shown that PGE2 is released by neutrophils and macrophages when recruited to pre‐neoplastic cells in unwounded larvae (Feng et al, 2012), we wondered whether this factor might also be, in part, responsible for the wound‐inflammation‐triggered local proliferation of these cells. To test this, larvae were immersed in either 0.5% DMSO or 10 μM Cox‐2 inhibitor NS398 with 0.5% DMSO and the number of pre‐neoplastic cells was analysed 2 days post‐laser wounding (Fig 4A–C). Although the addition of NS398 does not completely block the increase in pre‐neoplastic cell number post‐wounding, we now see a significant reduction in numbers, such that there is no significant difference between the unwounded DMSO‐treated larvae and the wounded NS398‐treated larvae, suggesting that the addition of a Cox‐2 inhibitor could partially negate the trophic impact of a cancer surgery (Fig 4C). 20 μM of synthetic prostaglandin E2 (dmPGE2) was added at the point of wounding and largely rescued the proliferative response in immune‐depleted zebrafish larvae (Figs 3L and 4G), suggesting that PGE2 derived from immune cells is one contributor towards the trophic factors responsible for the wound‐induced increase in proliferation.
Infiltration of neutrophils correlates both with extent of ulceration and with tumour cell proliferation in human melanoma
While there is considerable anecdotal evidence for wound‐exacerbated cancer progression (Hofer et al, 1998), our zebrafish studies prompted us to investigate whether this association was linked to innate immune cell influx at the wound site. It is already established that ulceration of melanoma is a bad prognostic indicator (Balch et al, 2011), and so we examined whether this might be due to wound inflammation. Haematoxylin and eosin staining of sections allows us to delineate the extent of the epidermal wound and to categorise melanomas as having no ulceration, minimal/moderate ulceration (<70% of total tumour length) or excessive ulceration (>70% of tumour length) (In ‘t Hout et al, 2012) (Fig 5). Our co‐staining for CD66b+ neutrophils and CD163+ macrophages and automated software for quantification of leucocyte numbers reveal a 15 times increase in neutrophils from non‐ulcerated to minimal/moderate ulcerated lesions and 100 times increase from non‐ulcerated to excessively ulcerated lesions (Fig 5A–C and G), whereas there is no such correlation with macrophage numbers and ulceration (Fig 5D–F and H, Supplementary Table S1). We observe that recruited neutrophils can be located throughout the melanoma, but are often associated with the superficial wound site in ulcerated lesions (Fig 5B and C). We see also a significant correlation between tumour cell proliferation [as assessed by Ki67/MelanA double staining (Nielsen et al, 2013)] and extent of neutrophil influx (Fig 6A', B', C' and D). This association appears strongest only up to minimal/moderate ulcers, suggesting that melanoma may reach a proliferation plateau soon after initial ulceration when they first become infiltrated by neutrophils, although we have no means to determine at which time point this might have occurred relative to our biopsies. We see no significant correlation between macrophage numbers and tumour cell proliferation (Fig 6E and Supplementary Table S2).
Infiltration of neutrophils further refines the prognostic indication for melanoma in ulcerated lesions
The thicknesses of the primary melanoma, the presence of ulceration and sentinel node status are all well‐established prognostic factors in melanoma (Balch et al, 2009, 2011). These factors each have independent prognostic impact in our study cohort (Bønnelykke‐Behrndtz et al, 2014) and have been adjusted for in our multivariate analysis. However, in the current study, we find that the density of CD66b+ neutrophil infiltration is an additional, independent prognostic marker of poor melanoma‐specific survival. We see a significant interaction with the extent of ulceration, with neutrophil density associating with a poorer prognosis in non‐ulcerated and excessively ulcerated tumours compared with minimal/moderate ulcerated tumours (Supplementary Table S3). Interestingly, Kaplan–Meier survival analysis showed that an increase in the fraction of infiltrating neutrophils further refines the prognostic impact of ulceration and separates both the non‐ulcerated and ulcerated group of patients (Fig 6F). Previous studies have found ambiguous association between macrophage influx and melanoma prognosis, and our study too suggests no significant link between CD163+ macrophage numbers and prognostic outcome (Supplementary Table S3).
To study the impact of wounding and, in particular, of the wound inflammatory response, upon cancer initiation and progression, we have turned to a genetically tractable and translucent model, the zebrafish, to live image interactions between innate immune cells and pre‐neoplastic cells at sites of tissue damage. We show that continual, chronic wounding of tissues with a genetic predisposition to melanoma leads to an increase in tumour incidence and that even a single acute wound can lead to significantly increased immune cell/pre‐neoplastic cell interactions in the local environment of the wound. These interactions trigger an increase in proliferative index in pre‐neoplastic cells in the proximity of the wound due to increased inflammation, and we show that this trophic effect is, in part, dependent on PGE2 because enhanced growth of pre‐neoplastic cells can be partially restored after immune cell depletion by administering PGE2 to the medium. In line with the results from our zebrafish experiments, we observe that neutrophil influx associates with increased tumour cell proliferation in human melanoma and also serves as a negative prognostic marker for patient survival.
Chronic versus acute, and big versus small wounds
That chronic tissue damage and its associated chronic inflammatory response might be fundamental contributors to cancer progression is now well established (Arwert et al, 2010) and, indeed, inflammation is now considered one of the ten hallmarks of cancer. Many pathologies of organs involving chronic inflammation are associated with malignancy (Hanahan & Weinberg, 2011), such as the progression of chronic skin wounds to squamous cell carcinoma, otherwise known as Marjolin's ulcer (Kerr‐Valentic et al, 2009). Less clear is how acute, rather than chronic, inflammatory episodes might impact on cancer. Our studies indicate that if an acute wound causes sufficient tissue damage to trigger a significant inflammatory response, then those recruited immune cells will likely be drawn to adjacent pre‐ or later stage neoplastic cells and influence their subsequent fate. In the pre‐neoplastic cells that we investigate in zebrafish larvae, this led to an increased proliferative index with a clear increase in clone size, rather than an increase in number of clones, and this is reflected in our clinical studies of ulcerated melanoma where we see an association between extent of wound inflammation and cancer cell proliferation.
The size of the wound is clearly important; this has been previously shown in Rous sarcoma‐infected chickens where a large wound, but not a needle wound, was a sufficient trigger for tumour formation (Sieweke et al, 1989), and more recently in a mouse model predisposed to develop papillomas where there was a linear correlation between wound size and tumour incidence (Hoste et al, 2015). Our studies also show that wounding, per se, does not lead to cancer initiation/progression because when wounds are too small to trigger a significant inflammatory response, or in larger wounds where the immune cells have been deleted, we see no subsequent increase in pre‐neoplastic cell proliferation. The size of the ulceration in human melanoma has also been shown to have an important impact on survival, with larger ulcerations associated with poorer melanoma‐specific survival (In ‘t Hout et al, 2012; Bønnelykke‐Behrndtz et al, 2014). We hypothesise that at least in part, this might be due to the influx of neutrophils as we show in this study.
A role for neutrophils in guiding cancer cell growth
To date, macrophages have generally been considered the lead players of all innate immune cells during the later stages of cancer progression and onwards to metastasis (Condeelis & Pollard, 2006). While not excluding a role for macrophages, our zebrafish data and associated clinical melanoma studies suggest that, where there is a tissue damage‐associated inflammatory cell influence on adjacent cancer cells, the key innate immune cell might be the neutrophil; only when we delete this lineage, do we see a reduction in pre‐neoplastic cell proliferation around the wound, whereas reducing macrophages exerts no significant reduction in pre‐neoplastic cell growth, although our macrophage depletion experiments did lead to increased numbers of neutrophils which could be compensatory. These observations of neutrophil influx post‐tissue damage are reflected in our clinical studies of “wounded” melanomas, where we see no significant association between degree of macrophage infiltration and proliferative index, whereas neutrophil numbers significantly correlate with increased proliferative index at the melanoma ulcer site and with long‐term prognostic outcome. Moreover, there is evidence in other cancers that neutrophil influx may reflect a marker for poor outcome (Donskov, 2013).
Neutrophils are known to infiltrate tumours and can form a significant component of the inflammatory tumour microenvironment (Sionov et al, 2014). Tumour‐associated neutrophils (TANs) have been shown to promote angiotropism (Bald et al, 2014) and to suppress the anti‐tumour immune response from T lymphocytes. Indeed, depletion of TANs can lead to tumour regression (Pekarek et al, 1995). A recent study shows a direct association between neutrophil presence and melanoma metastasis (Bald et al, 2014). In contrast to N1‐type circulating neutrophils, which are known for their capacity to defend skin wounds from foreign pathogens, TANs are thought to exist in an alternatively activated N2 state. When compared to unstimulated neutrophils or myeloid‐derived suppressor cells (MDSCs), TANs show increased expression of chemokines and cytokines (Fridlender et al, 2012) and are likely to be producing many more pro‐tumour factors than simply PGE2 that we have investigated here.
The TAN N2 phenotype is known to be influenced by the TGFβ pathway (Fridlender et al, 2009), which is released at sites of skin damage by platelets and keratinocytes and other cells (Yang et al, 1999), which might, in part, explain the neutrophil response we observe following wounding. Moreover, N2 neutrophils can be switched back to a tumour‐cytotoxic N1 polarity through inhibition of TGFβ (Fridlender et al, 2009), or following immunologic or cytokine activation (Kim et al, 2008), re‐enabling their potential to limit tumour growth (Kim et al, 2008), which is likely to be of significant therapeutic relevance. Clearly, it will be important to investigate further both macrophage and neutrophil phenotypic switches and how these are influenced by their exposure to wounds and then to pre‐neoplastic cells, although currently this is difficult to analyse in zebrafish due to the lack of M1/M2 and N1/N2 markers.
Implications for the clinic
Our studies in zebrafish confirm an indication from many anecdotal clinical observations that cancer surgery and biopsy may influence patient outcomes in ways beyond simple removal of cancer tissue (Naumov et al, 2009). Although we must be cautious in extrapolating observations in zebrafish larvae to cancer surgery in human patients, it is clear that particularly for surgery that may not remove all the cancer, there must be concerns how the wound inflammatory response might influence remaining cancer cells.
Surgery is still the treatment of choice in localised melanoma, while immunotherapy in the form of the anti CTLA‐4 antibody ipilimumab (Hodi et al, 2010) has emerged as a standard treatment in the metastatic setting and there are exciting new treatment options with the PD‐1 antibodies nivolumab and pembrolizumab in development (Topalian et al, 2012; Wolchok et al, 2013; Robert et al, 2014). A primary melanoma is generally excised in a 2‐step procedure with margins, up to 2 cm, depending on the initial thickness. Sentinel lymph node biopsy, for staging and evaluation of micro‐metastasis in the regional lymph node basin, is recommended for patients with melanoma thickness over 1 mm or with the presence of ulceration or signs of excessive proliferation (Jensen et al, 2012). The value of sentinel lymph node biopsy and subsequent removal of regional lymph nodes when positive for micro‐metastasis is questionable, however, as neither confers a survival benefit (Morton et al, 2014). In addition, a high false‐negative rate, defined as the development of macro‐metastasis in the same regional lymph node region after a negative sentinel lymph node biopsy, adds complexity to the discussion. Pathological, surgical and biological factors have been associated with the detection rate of micro‐metastasis in the sentinel node (Riber‐Hansen et al, 2012). However, wound‐associated inflammation may have some impact on non‐sentinel remnant cells, offering them a favourable niche within the lymph node region, which might support further tumour growth. Sentinel lymph node biopsy is currently accepted as an important tool for proper staging and selection of patients for additional targeted or immune‐modulating therapy.
Ulceration of primary melanomas has long been known to reflect a poor prognosis for the patient, but recently it has also been shown that patients with ulceration and microscopic lymph node metastases respond better to interferon than those without ulceration and with macroscopic lymph node metastases (Eggermont et al, 2012). This might suggest that while a chronic inflammatory response initially favours the tumour cell, some further treatments may have the potential to switch the chronic inflammatory response to one that drives an anti‐tumoural response.
At a fundamental level, our studies suggest a layered approach for developing therapeutics that might counter the effects of the damage‐associated acute inflammatory response. A reductionist approach would be to simply delete, or at least dampen, the wound inflammatory response; alternatively, and more subtle, it would be to inhibit the trophic signals delivered to cancer cells by the inflammatory cells. A third strategy would be to take advantage of the enhanced recruitment of inflammatory cells to nearby cancer cells, and encourage the inflammatory cells to kill rather than nurture the cancer cells. Since the acute wound inflammatory response is transient, persisting only until the resolution phase, there will be a limited window post‐wounding when wound‐attracted immune cells can deliver trophic signals to local cancer cells. Our studies in zebrafish larvae show that blocking one of these trophic signals at the time of wounding can, in part, inhibit the inflammation‐dependent proliferative effect. At these stages and for this pre‐neoplastic lineage, we know that PGE2 is a component of the trophic signal, and consequently, prostaglandin synthesis inhibitors are effective. Clinically, long‐term use of prostaglandin inhibitors, especially aspirin, is associated with reduced risk of developing several types of cancer (Rothwell et al, 2012). They may also have an adjuvant role in cancer surgery, although their effects on platelet aggregation must also be taken into consideration.
More importantly, this observation highlights how important it is for us to gain a better understanding of what are the key immune cell‐derived trophic signals for various cancers, and these may be different if the innate immune cells have been drawn to the cancer, and primed in some way, after previously being part of an acute inflammatory wound response. Prostaglandins are likely to be one of the players, for some cancer cell types, but certainly not the only one, and more research should focus on determining the trophic signals originating from wounds and wound‐recruited immune cells which may impact on tumour growth. We need to know considerably more about precisely how immune cells behave in the presence of pre‐neoplastic and cancer cells with various genetic lesions (beyond RasV12 as we have investigated in our zebrafish studies), particularly after they have previously responded to wound cues, before considering how to modulate these behaviours in ways that might benefit the patient.
The overall implication of this study is not that surgery or biopsy collection should be avoided, but rather that the cellular processes that occur as a consequence of local wounding, particularly the wound inflammatory response, require improved understanding as surgical wounds may have unintended consequences. In this regard, it is likely that the move towards more minimal surgery will be beneficial to patients by avoiding unnecessary inflammatory stimulation. A dispersed, infiltrative cancer will still require aggressive surgery to ensure macroscopic clearance, whereas minimal surgery for a contained tumour may reduce inflammatory influx and the negative downstream consequences that we describe in this paper.
The translucency of zebrafish larval tissues has offered us a unique opportunity to live image how the wound inflammatory response impacts on adjacent pre‐neoplastic or cancer cells and reveals some interesting insights. These studies show definitive evidence that innate immune cells drawn to wounds can rapidly move on to competing attractants released by cancer cells, and this observation, particularly, the key role played by neutrophils, has guided our investigations in human studies where the presence of these cells may serve as “warning markers” for cancer‐promoting influences by acute and chronic wound episodes in and around patient tumours. Moreover, a better understanding of these immune cell/cancer cell interactions at sites adjacent to tissue damage will guide potential perioperative therapeutic intervention where cancer cells might remain after surgery.
Materials and Methods
Zebrafish lines and maintenance
Adult zebrafish (Danio rerio) were maintained and crossed as previously described (Westerfield, 2007). In brief, adult zebrafish were reared at a constant temperature of 28°C, mated and fertilised eggs were collected, bleached, maintained and staged according to standard protocols (Westerfield, 2007). Strains included: Et(kita:GalTA4,UAS:mCherry)hzm1 (Et30) (Santoriello et al, 2010), Tg(5XUAS:eGFP‐H‐RASV12)io6 (Santoriello et al, 2010), Tg(LysC:dsRed)nz (Hall et al, 2007) Tg(mpeg1:FRET) (a kind gift from Nikolay Ogryzko and Stephen Renshaw at the University of Sheffield) and tp53M214K (Berghmans et al, 2005). All experiments were conducted with local ethical approval from the University of Bristol and in accordance with UK Home Office regulations.
All the morpholinos were obtained from (GeneTools LLC) and 100‐μm drops (1 nl volume) were injected into one‐cell‐stage embryos. Morpholinos include: pu.1 5′‐GATATACTGATACTCCATTGGTGGT‐3′ (Rhodes et al, 2005); gcsfr 5′‐GAAGCACAAGCGAGACGGATGCCAT‐3′ (Liongue et al, 2009); and irf8 5′‐AATGTTTCGCTTACTTTGAAAATGG‐3′ (Li et al, 2011).
Whole‐mount immunostaining was performed as previously described (Feng et al, 2010). Primary antibodies used in this study include rabbit polyclonal anti‐L‐plastin antibody (1:500) (Feng et al, 2010), chicken polyclonal anti‐L‐plastin antibody (1:200) (Feng et al, 2010), mouse polyclonal anti‐GFP (1:100) (Abcam, ab13970), rabbit polyclonal anti‐RFP (1:200) (MBL), rabbit monoclonal anti‐phospho‐histone H3 (Ser10) (1:200) (Cell Signalling, 3377) and rabbit anti‐Cox2 (1:200) (Cayman Chemicals) (Feng et al, 2012). AlexaFluor‐488 IgG, AlexaFluor‐546 IgG or AlexaFluor‐647 secondary IgGs (all from Invitrogen and used at 1:500) were used to reveal primary antibody localisation. L‐plastin antibody marks all leucocytes at larval stages, and so to determine macrophage numbers in fixed whole‐mount preparation, we counted L‐plastin+; LysC− cells, as previously described (Feng et al, 2010; Jones et al, 2013). To measure larval proliferation, a Click‐iT EdU imaging kit (Invitrogen) was used. In brief, larvae were treated with 400 μM EdU immediately after wounding at 3 dpf and maintained for 2 days post‐wounding when they were fixed, the Click‐iT reaction was performed and the larvae were immunostained according to the manufacturer's instructions. Larvae were then mounted in Citifluor (Agar Scientific) and imaged on a Leica SP5‐II AOBS confocal laser scanning microscope attached to a Leica DM I6000 inverted epifluorescence microscope.
Wounding zebrafish embryos
Larvae were first sorted into Ras+ and WT sibling groups and anesthetised in Danieau's solution containing 0.1 mg/ml tricaine (Sigma). Laser wounds were made as previously described (Redd et al, 2006) using a UV‐nitrogen laser microbeam coupled to a Zeiss Axioplan 2 microscope (Micropoint Laser System, Photonic Instruments). Typical laser wounds required a 3‐s pulse from the Coumarin 440‐nm dye cell laser focused through a 40× Achroplan water immersion objective. Larvae were kept for the required post‐wounding period, when they were fixed and immunostained as described above. The pre‐neoplastic cell numbers in larvae were counted manually (always in the same region above the cloaca) with a 20× objective and 1.5× zoom.
Larvae were treated with 30 μM NS398 or 20 μM dmPGE2 (both from Cayman Chemicals) immediately after laser wounding at 2 or 3 dpf in Danieau's solution containing 0.5% DMSO. After treatment, larvae were fixed in 4% PFA overnight, immunostained and imaged as described above. For DPI treatment, larvae were incubated in 100 μM DPI (Sigma) in Danieau's solution containing 1% DMSO for 45 min prior to wounding and throughout the period of imaging.
Live imaging of zebrafish larvae
Wounded or control larvae were mounted on their sides in 1.5% low‐melting agarose (Sigma), in a glass‐bottomed dish, filled with Danieau's solution containing 0.01 mg/ml tricaine. The climate chamber covering the microscope stage was set at 28°C. Images were collected using a Leica SP5‐II AOBS confocal laser scanning microscope attached to a Leica DM I6000 inverted epifluorescence microscope with a 63× glycerol lens. Movies were recorded at either 1 or 2 frames/min and were exported from Volocity as QuickTime movies using the Sorenson3 video compressor to play at 6 frames/s.
Adult zebrafish surgery and live imaging
Adult zebrafish (either Et30:GalTA4; UAS:eGFP‐H‐RASV12 alone or crossed to tp53M214K, to increase melanoma incidence) were anesthetised in tank system water containing 0.1 mg/ml tricaine (Sigma). Tumours were excised, or the tip of the tail fin was resected, with a microsurgical knife (World Precision Instruments) on a 2%‐agarose plate. Punch biopsies were taken with a 1‐mm sterile disposable biopsy punch (Kai Medical). Images were taken using a Leica camera (DFC320) attached to a Leica MZFLIII dissecting microscope. Live confocal imaging was performed on anaesthetised, punch‐biopsied fish with their tails mounted in 1.5% low‐melting agarose (Sigma) using a Leica SP8 AOBS laser scanning confocal attached to a Leica DM6000 upright microscope with a 10× water immersion lens.
Adult zebrafish immunohistochemistry
Adult zebrafish tissue was fixed in 4% PFA for 2 h at room temperature or overnight at 4°C, washed in PBS and transferred to PBS plus 30% sucrose at least overnight. Tissues were embedded in Tissue‐Tek O.C.T. and frozen in isopentane cooled by liquid nitrogen and 14‐μm section cut by a Bright OTS cryostat onto Superfrost Plus microscope slides (VWR). Frozen sections were washed in PBS with 0.1% Triton X‐100, blocked and incubated overnight with primary antibody (as above) at 4°C. Slides were subsequently washed extensively with PBS with 1% Triton X‐100, re‐blocked briefly and secondary antibody added for 2 h at room temperature, before washing in PBS with 0.1% Triton X‐100 overnight. Slides were mounted in Mowial or ProLong Gold antifade reagent (Invitrogen) and imaged using a Leica SP5‐II AOBS confocal laser scanning microscope.
Post‐image acquisition analysis
The number of pre‐neoplastic cell clones, immune cells recruited and the number of pre‐neoplastic cell contacts were counted manually. Distances from wounds to pre‐neoplastic cell clones were measured using the measure function in Volocity software (Perkin Elmer Improvision). All time‐lapse movie quantification and tracking analysis were performed using Volocity. Individual LysC:DsRed+ cells were identified using automatic ‘‘find objects use intensity'' and ‘‘colour the object'' functions in Volocity for all the time points to generate a ‘‘footprint map''. Adult zebrafish tails were analysed using the threshold function on ImageJ where images were first changed to an 8‐bit image, threshold applied to 130 and the number of particles automatically counted. The area fraction was used to determine the pigmentation of the tail fins, or the accumulation of LysC+, mpeg+ or L‐plastin+ immune cells. Montages of adult zebrafish were made using the photomerge function in Adobe Photoshop. In some instances, the notochord was cropped from the data set using Imaris software (Bitplane) for improved visualisation of overlying RasGV12eGFP goblet cells.
Human tissue collection
Danish patients diagnosed with primary cutaneous melanoma from 2001 to 2008 were retrospectively located from pathology files. Melanomas with verified ulceration (n = 207) were matched consecutively with a reference group of non‐ulcerated melanoma, matched according to Breslow thickness and age. In total, 385 patients with superficial spreading, nodular‐ and lentigo‐malignant melanomas were included. Data on pathological parameters and follow‐up information pertaining to the included patients were collected from electronic patient files, pathology files and the Danish Melanoma Group database. Patients with more than one melanoma, missing follow‐up information or re‐evaluated as metastatic, satellite or benign lesions were excluded to avoid follow‐up bias (n = 17). For proper evaluation and computer estimation, tumours with no sufficient tumour tissue left, largely pigmented melanoma or MelanA‐negative melanomas were excluded. For the neutrophil, macrophage and proliferation analyses, 180, 179 and 164 patients with verified ulcerated melanoma and 201, 203 and 181 patients with non‐ulcerated melanoma were included in the analyses, respectively. The Regional Committee for Health Research Ethics in the Region Middle of Denmark approved the study.
Formalin‐fixed and paraffin‐embedded sections were serial sectioned at 3 μm.
Commercially available antibodies were used for the immunohistochemistry, with a control section for every batch of 20 sections, and evaluated blinded by a senior pathologist (TS). Ultraview fast red was used as detection system, to minimise the influence of the pigmentation of melanophages or melanoma cells. The sections were stained for cd163+ macrophages (Ab Serotec, EDHu‐1, 1:100, cytoplasm) and cd66b+ granulocytes (BD Bioscience, G10F5, 1:200, cytoplasm). For tumour cell proliferation, we used double stains of MelanA (Cell Marque, M2‐7C10, 1:50, cytoplasm) to detect melanoma cells and Ki67 (Spring, sp‐6, 1:100, nuclei) as a marker of proliferating nuclei.
Quantification of melanoma inflammation and proliferation
Objective estimates of the area fraction (%) of infiltrating neutrophils (area intratumoural or intravascular cd66b+ neutrophils/area region of interest × 100), macrophages (area intra‐tumoural or intra‐stromal cd163+ macrophages/area region of interest × 100) and tumour cell proliferation (area Ki67+ MelanA+ cells/area Ki67+ MelanA+ and area of Ki67− MelanA+) were acquired by computer‐assisted image analysis and software from Visiopharm A/S, Denmark (Carus et al, 2013; Nielsen et al, 2013).
Zebrafish data were analysed using Prism 6 software (GraphPad), unpaired two‐tailed Student's t‐test for comparisons between two groups (Prism 6, GraphPad Software) and one‐way ANOVA with appropriate post‐test adjustment for multiple group comparisons (following D'Agnostino‐Pearson omnibus normality tests). Graphs display mean ± SEM unless otherwise indicated. Statistical significance is indicated on graphs using standard conventions, as follows: ns: P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. All experiments were repeated three times (non‐randomised or blinded) and were performed on at least 15 larval or 3 adult fish. Replicates are indicated in relevant figure legends.
Human patient data were grouped by ulceration status (absence vs. presence), and if present dichotomised by the extent (relative length minimal/moderate <70% vs. excessive >70% (In ‘t Hout et al, 2012)). Data were log‐transformed and analysed with linear regression, tested for interactions of ulceration status, ulceration extent, Breslow thickness and sentinel node status. In order to allow log transformation and analysis of sections with no infiltrating neutrophils, zero values were replaced with a defined low value of 0.0005. The assumption of normal distribution was assessed using the residuals and found adequate. Statistical analysis of melanoma‐specific survival was performed using the Cox proportional hazards using a robust sandwich covariance matrix estimate to account for the case control design. Breslow thickness, sentinel node status and ulceration, which are all main factors defining the stage, all demonstrated independent prognostic impact in our study cohort and were adjusted for in the multivariate analysis.
NA, YF and PM conceived and designed the zebrafish experiments. MLB‐B, TS and HS conceived and designed the human patient experiments. NA, LCW and JC performed the zebrafish experiments. MLB‐B performed the experiments in Figs 5 and 6. IJC performed the statistical analysis in Figs 5 and 6. NA, LCW, JC, MLB‐B, HS, YF and PM wrote the paper.
Conflict of interest
The authors declare that they have no conflict of interest.
Supplementary Figure S1
Supplementary Figure S2
Supplementary Figure S3
Supplementary Figure S4
Supplementary Figure S5
Supplementary Figure S6
Supplementary Movie S1
Supplementary Movie S2
Supplementary Movie S3
Supplementary Movie S4
Supplementary Figure and Movie Legends
This work was supported by a Wellcome Trust studentship to NA; a Wellcome Trust Sir Henry Dale Fellowship to YF; a Wellcome Trust Senior Investigator Award and Cancer Research UK programme and BBSRC project grants to PM. MLBB/TS and HS were funded by the Danish Cancer Society, EORTC melanoma group, Korning‐, Arvid Nielssons‐ and Wedell‐Wedellborgs foundation. We would like to acknowledge Jeanetter Bæhr Georgsen for her great technical support, and Allan Vestergaard and Patricia Switten Nielsen for their great help with the automatised methods and protocols. We also thank Lucy MacCarthy‐Morrogh, and Oddbjørn Straume for comments on the manuscript, as well as technical support from the Wolfson Bioimaging facility at the University of Bristol.
FundingWellcome Trust studentship
This is an open access article under the terms of the Creative Commons Attribution 4.0 License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
- © 2015 The Authors. Published under the terms of the CC BY 4.0 license