Inappropriate chemokine/receptor expression or regulation is linked to many diseases, especially those characterized by an excessive cellular infiltrate, such as rheumatoid arthritis and other inflammatory disorders. There is now overwhelming evidence that chemokines are also involved in the progression of cancer, where they function in several capacities. First, specific chemokine–receptor pairs are involved in tumour metastasis. This is not surprising, in view of their role as chemoattractants in cell migration. Secondly, chemokines help to shape the tumour microenvironment, often in favour of tumour growth and metastasis, by recruitment of leucocytes and activation of pro-inflammatory mediators. Emerging evidence suggests that chemokine receptor signalling also contributes to survival and proliferation, which may be particularly important for metastasized cells to adapt to foreign environments. However, there is considerable diversity and complexity in the chemokine network, both at the chemokine/receptor level and in the downstream signalling pathways they couple into, which may be key to a better understanding of how and why particular chemokines contribute to cancer growth and metastasis. Further investigation into these areas may identify targets that, if inhibited, could render cancer cells more susceptible to chemotherapy.
- chemokine receptors
- tumour microenvironment
Chemokines comprise a superfamily of about 50 human ligands and 20 receptors that are pivotal regulators of cell migration [1,2]. Traditionally, chemokines and their receptors have been divided into four families on the basis of the pattern of cysteine residues in the ligands (CXC, CC, C and CX3C). They have also been functionally classified as being ‘homoeostatic’ or ‘inflammatory’. Homoeostatic chemokines are constitutively expressed and control leucocyte navigation during immune surveillance. Inflammatory chemokines, which constitute the vast majority, are inducible and control cell recruitment to sites of infection and inflammation . Certain chemokines are also involved in developmental processes such as lymphopoiesis, cardiogenesis and development in the central nervous system . For example, in lymphopoiesis, the transition from bone-marrow-resident haematopoietic stem cells through development of T-cell precursors in the thymus, migration into secondary lymphoid organs for immune response initiation and maturation into circulating memory and effector T-cells involves sequentially co-ordinated changes in the profiles of chemokine receptor expression to guide the cells into the appropriate microenvironments .
Chemokines are 70–130 amino acid soluble proteins that contain one to three disulfide bonds, the exceptions being CX3CL1 (fractalkine) and CXCL16, which have a chemokine domain tethered to a membrane-anchored mucin stalk [6,7]. Despite variable levels of sequence homology, they adopt a characteristic fold that consists of an N-terminal unstructured domain that is critical for signalling, a three-stranded β-sheet connected by loops and turns, and a C-terminal helix (Figure 1A). Although they bind their receptors as monomers in the context of cell migration, many chemokines dimerize or form higher-order oligomers that appear to be important for localization to glycosaminoglycans on cell surfaces and the ECM (extracellular matrix)  and possibly for signalling related to other processes separate from migration [9,10]. The chemokine receptors are seven-transmembrane GPCRs (G-protein-coupled receptors). As such, they have been best characterized with respect to signalling through heterotrimeric G-proteins, primarily involving Gi  (Figure 1B). However, there is ample evidence indicating that chemokine receptors also signal through other G-protein subtypes or even through non-G-protein-mediated pathways [12–15]. Furthermore, although the α-subunit of G-proteins has traditionally been regarded as the major signalling subunit, the βγ-subunits are crucial for the activation of many chemokine-induced pathways. Two of the major pathways activated by Gβγ are PI3Kγ (phosphoinositide 3-kinase γ) and PLC (phospholipase C), whereas Gαi proteins mainly inhibit adenylate cyclase and transduce signals through tyrosine kinases such as Src .
Despite their structural homology and shared ability to induce chemotaxis, different chemokines can elicit other distinct cellular responses [16,17] and/or activate different pathways to elicit a particular response [18,19]. The signalling and physiological response downstream of receptor activation can also vary, depending on the chemokine/receptor combination, the cell type and the pathophysiological state [16,17]. Figure 2 summarizes the major signalling cascades and functional outcomes of chemokine/receptor activation. It is important to bear in mind that specific subsets or combinations of these pathways can be used to induce the functional response.
In the last few years, the involvement of chemokines and their receptors in cancer, particularly metastasis, has been firmly established [20–22]. The association with metastasis is not unsurprising, since it is not a random process of cell migration. On the contrary, it has been known since the early 1900s that cancer cells have a propensity to metastasize to specific organs . Furthermore, metastasis has many features in common with normal cell migration. However, key differences include abnormal chemokine receptor expression, regulation or utilization, often on cells that typically do not migrate. Chemokines then provide a physical address for the secondary destination of the tumour cells.
The process by which tumours grow and metastasize is complex , with many steps required for primary tumour development and establishment of clinically significant secondary tumours (Figure 3). These steps include:
survival and growth of the primary tumour
detachment of tumour cells from the primary lesion
invasion into vascular or lymphatic vessels
homing and adherence to the destination organs
Since alternative environments such as bone marrow and lymph node are not naturally compatible with cells from the breast, for example, cancer cells must both derive and provide signals to favourably shape the tumour microenvironment to become conducive to survival and growth [26–28]. The role of chemokines and their receptors in cancer can thus be divided into three broad categories which contribute to one or more of the above processes:
providing directional cues for migration/metastasis
shaping the tumour microenvironment
providing survival and/or growth signals
In the present review we describe the role of chemokines and chemokine receptors in each of these processes. Table 1 summarizes the chemokine receptors and respective ligands involved in cancer and their general mechanism of tumorigenesis. Although their involvement in these three categories is fairly well established, the exact mechanisms of action are not well understood, and the underlying complexity of the chemokine network makes it difficult to characterize them definitively. Consideration of some of these complexities, discussed at the end of the present review, may be crucial to elucidating more precise mechanisms and thus enable the development of better cancer therapeutics.
CHEMOKINES IN MIGRATION/METASTASIS OF CANCER CELLS
The leading cause of death in cancer patients is from metastasis, namely the formation of secondary tumours in organs distant from the original tumour. It is not a random process, but rather shows bias for particular tissues, and is ordered, specific and molecularly directed [20,24]. Breast cancer, for example, has a tendency to metastasize to the lymph nodes, bone marrow, lung and liver. When Müller et al.  highlighted the role for chemokines in directing organ-specific metastasis, it became clear that chemokine receptor expression patterns on cancer cells and the localization of the corresponding ligands could provide clues for understanding directional metastasis. It has now been established that several chemokines and their receptors play a role in the metastatic process by directing the migration of receptor-bearing tumour cells to sites of metastases where the ligands are expressed. These findings make sense because of the parallels one can draw between lymphocyte trafficking and tumour migration.
The general mechanisms involved in normal cell migration and metastasis are similar. Chemokines cause cell movement by inducing changes in cytoskeletal structure and dynamics. Actin polymerization leads to formation of protrusions (pseudopods) which, with the help of integrins, form focal adhesions with the ECM to help propel the cell forward . Although multiple pathways contribute to chemokine-induced cell migration, PI3K, FAK (focal adhesion kinase) and the Rho family of GTPases (Rho, Rac, Cdc42) are particularly important [30–33]. ERK (extracellular-signal-regulated kinase) and PKC (protein kinase C) signalling, independently or in conjunction with PI3K, may also be involved [34–37]. More detail regarding signalling involved in migration is shown in Figure 2.
Altered chemokine receptor expression on cancer cells
What then distinguishes normal cells from metastatic cancer cells, enabling the cancer cells to migrate when they normally would not? Although there are many contributing factors, in numerous types of cancer the malignant cells exhibit increased or aberrant expression of particular chemokine receptors relative to their normal counterparts, notably CXCR4, CCR7 and CCR10 [22,25,38,39]. In a study of breast cancer, it was demonstrated that CXCR4 was strongly expressed on cancer cells compared with normal breast epithelial tissue, which does not express any CXCR4, and that antibodies against CXCR4 blocked metastasis in a mouse model of breast cancer . Since that seminal publication, CXCR4 and its ligand CXCL12 [SDF-1α (stromal-cell-derived factor-1α)] have been implicated in about 23 different types of cancer . At least two other chemokine systems also play a direct role in metastatic homing of cancer cells: CCR10 in metastasis to skin, where CCL27 (CTACK) is expressed (e.g. melanoma), and CCR7 in lymph node metastasis, where CCL21 (SLC) is expressed [20,25,39,40].
Many reasons for altered chemokine/receptor expression have been identified. Chemokine receptor expression is regulated both at the transcriptional level and post-transcriptionally through RNA stability, translation and receptor desensitization and internalization [41–47]. The tumour microenvironment, and mutant proteins or altered signalling in the cancer cell itself, can also affect chemokine receptor levels. Conditions present within a tumour, such as hypoxia and a rich cytokine environment, including IL-2 (interleukin-2), can induce the transcription of certain chemokine receptors [41,44,45]. For example, hypoxia induces up-regulation of CXCR4 transcription via HIF-1 (hypoxia-inducible factor 1) and through transcript stabilization [42,44]. HIF-1 was also recently found to promote transcription of CCR7, CXCR1 and CXCR2 [46,47]. In renal carcinoma cells it was demonstrated that mutation of pVHL (von Hippel–Lindau tumour suppressor protein), normally responsible for targeting HIF-1 for cell degradation, results in constitutive activation of HIF-1 target genes, including CXCR4 . NF-κB (nuclear factor κB) is part of a key signalling pathway that is often activated in cancer cells and can contribute to the transcription of chemokines [e.g. CXCL1 (Gro-α), CXCL8 (IL-8) and CXCL12] and chemokine receptors such as CXCR1, CXCR2 and CXCR4 [39,48]. At the post-transcriptional level, changes in receptor translation and desensitization by internalization and degradation provide other mechanisms for regulating chemokine-receptor expression. For example, enhanced CXCR4 translation in breast-cancer cells was shown to be associated with the oncogene HER2, which may help to protect CXCR4 from ligand-induced ubiquitination and degradation .
THE TUMOUR MICROENVIRONMENT
A link between inflammation and cancer was observed over 150 years ago when Rudolf Virchow (cited in ) noted that cancers tend to occur at sites of chronic inflammation. Although the relationship between cancer and inflammation is complex, epidemiological studies indicate that inflammatory and infectious diseases are often associated with an increased risk of cancer . In many ways, the microenvironment of tumours mimics that of tissues during the height of an inflammatory response to injury . For example, they both contain a large number of cells from both the innate and adaptive immune system, recruited and activated by a complex profile of chemokines, cytokines, growth factors and proteases. However, unlike the organized morphology of normal tissue, and the ultimate resolution of the inflammation that occurs during healing, tumours exist in a state of chronic inflammation characterized by the presence of malignant cells, development of an aberrant vascular network and the persistence of inflammatory mediators. Within the tumour microenvironment, chemokines and their receptors play roles in modulating angiogenesis, cell recruitment, tumour survival and proliferation and, through these processes, help to define the progression of the cancer. Although the present review focuses primarily on the pro-tumorigenic roles of chemokines, it should be noted that many chemokines/receptors have anti-tumorigenic effects, which would be expected, given their physiological role in protecting the host .
Recruitment of leucocytes: tumour-associated macrophages and DCs (dendritic cells)
One important link between cancer and inflammation is the recruitment of cells, including neutrophils, macrophages, DCs, eosinophils, mast cells and lymphocytes. Of these, TAMs (tumour-associated macrophages)  represent an important component of solid tumours and may account for up to half of the tumour mass. They were first observed in tumours in the late 1970s and therefore represent one of the first specific links between the immune system and cancer.
Macrophages are very heterogeneous and play many roles in the progression of cancer, depending upon the nature of their maturation/polarization. Although an oversimplification of their states, designating the two ends of the spectrum as M1 and M2 represents a convenient way to classify macrophage polarization . M1 (classical) macrophage activation by IL-2/interferon-γ and IL-12 is characterized by high levels of antigen presentation, IL-12/IL-23 production and development of a polarized type I response, leading to tumour cell cytotoxicity and necrosis. By contrast, M2 polarization is characterized by defective IL-12 production , leading to an IL-12low phenotype and is associated with the induction of angiogenic factors, cytokines and proteases. These factors serve to suppress the adaptive immune response and promote angiogenesis, matrix remodelling and tumour growth. A number of excellent reviews have presented evidence that TAMs function as M2-polarized macrophages [28,53,54] that promote a permissive environment for tumour growth. The in vivo significance of macrophage recruitment to tumour growth has been studied in a number of models and in clinical studies, and generally high numbers of TAMs correlate with enhanced vascularization and growth [51,56].
Early studies indicated a role for the pro-inflammatory MCP (monocyte chemoattractant protein) family of proteins, termed ‘tumour-derived chemotactic factors’ in macrophage recruitment to tumour sites . Over the years, the role of CCL2 (MCP-1) in macrophage recruitment and activation via one of its receptors, CCR2, has been extensively studied . CCL2 is produced by many types of tumours, including breast, pancreas, lung, cervix, ovary, melanomas, sarcomas and glial cell tumours, and by fibroblasts, ECs (endothelial cells) and macrophages at the tumour site . CCL2 expression levels correlate with the extent of macrophage recruitment  and, for a number of cancers, expression levels of CCL2 are also related to prognosis [59,60]. The relationship between chemokines, macrophages and prognosis is complex and dependent upon the cancer: this has been termed the ‘macrophage balance hypothesis’. For example, the effect of CCL2 upon tumour progression in melanoma cell lines is concentration-dependent. Whereas melanoma cell lines transfected with high levels of CCL2 promote tumour rejection, those transfected with lower levels support tumour growth . A recent model for melanoma has directly correlated low levels of CCL2 with the presence of M2 macrophages and increased angiogenesis and tumour growth compared with macrophage-depleted melanomas or melanomas expressing no CCL2 at all . Similarly, pancreatic cancer patients with higher circulating levels of CCL2 had higher survival rates than those with lower circulating levels . However, higher levels of CCL2 are associated with increased malignancy in models of mammary adenocarcinomas [64,65].
CCL5 [RANTES (regulated upon activation, normal T-cell expressed and secreted)] also recruits macrophages to tumours, and its increased presence correlates with poor prognosis in breast- and cervical-cancer patients . A direct link between macrophage recruitment by CCL5 and breast tumour growth was observed using a CCL5 variant that contains an extra methionine residue at its N-terminus and functions as a receptor antagonist. Mice treated with the antagonist showed a decrease in tumour size and vascularization compared with control-treated mice . The ability of CCL5 to induce the monocyte expression of CCL2 may also play an important role in monocyte recruitment and shaping the tumour microenvironment .
In total, about 20 chemokines have been detected in neoplastic tissues of various cancers (Table 1). TAMs themselves express chemokine receptors and selected chemokines [e.g. CCL2, CCL17 (TARC), CCL18 (PARC), CCL20 (MIP-3α) and CCL22 (MDC)], suggesting both autocrine and paracrine roles for these proteins in the microenvironment. Interestingly, on TAMs, the expression levels of chemokine receptors such as CCR2 is very low  and may be due to the transition of the cells from blood monocytes to tissue macrophages and/or may prevent these cells from migrating out of the tumour microenvironment once they arrive .
DCs are attracted to the tumour microenvironment by chemokines [71–74], where they are often found at low levels . Like macrophages, DCs comprise a diverse population of cells and are likely to play multiple roles. Although a number of tumours contain mature plasmacytoid DCs capable of priming T-cells, those observed in ovarian cancers are thought to function as pro-angiogenic factors, acting through tumour necrosis factor-α and CXCL8 . Thus DCs are likely to have both anti-tumorigenic (e.g. T-cell priming) and pro-tumorigenic (e.g. angiogenesis, immune tolerance) roles in cancer.
A number of tumours, including primary cutaneous melanomas, ovarian tumours and breast carcinomas, also contain immature DCs [71,73,74]. This suggests that factors in the microenvironment may suppress DC maturation to dampen their anti-tumorigenic effects. Factors such as IL-6 and mononuclear phagocyte colony-stimulating factor present in the tumour environment may also prevent DC maturation by switching the differentiation from DC cells to macrophages . Interestingly, differential localization of immature and mature DCs, most likely recruited by CCL20, has been observed in breast adenocarcinomas . Immature cells incapable of T-cell priming are present within the tumour, whereas mature cells are located in the peritumoural areas, interacting in some cases with T-cells. Since the DC–T cell interaction usually only occurs in the lymph, it suggests that these cells may be mounting a tumour-specific response.
Pro-tumorigenic effects of other stromal cells in the microenvironment
The complex interplay between cells in the microenvironment has previously been highlighted by the discovery of tumour-promoting fibroblasts in epithelial cancers [77,78]. Fibroblasts and myofibroblasts often represent a significant portion of stromal cells in carcinomas. Their direct effects upon carcinoma growth were highlighted in 1999, when it was demonstrated that stromal fibroblasts from human prostate carcinomas had tumour-stimulating properties . In that study, fibroblasts from cancerous or non-cancerous tissues were isolated, mixed with immortal but non-tumorigenic prostate epithelial cells, and co-cultured or injected into immunodeficient mice. Only fibroblasts isolated from the carcinomas were able to stimulate tumour growth. These fibroblasts, termed CAFs (carcinoma-associated fibroblasts), include large populations of myofibroblasts that usually play crucial roles in wound repair, and this observation serves as another example of the association between inflammation and cancer.
The pro-tumorigenic properties of CAFs in mammary carcinomas are partly mediated by CXCL12 . CXCL12 is expressed at high levels by mammary carcinoma CAFs and plays two main roles. First, it directly stimulates tumour growth by binding to, and signalling through, CXCR4 on tumour cells. Secondly, it recruits ECs into tumours, facilitating angiogenesis (see below). However, CAFs exhibit complex and heterogeneous expression profiles, and it is not surprising that other CAF-derived chemokines (e.g. CCL2, CCL5 and CXCL8) also play roles in tumour growth [80–82].
Angiogenesis: recruitment of EC precursors
Angiogenesis is the process of forming new blood vessels and it requires the production of new ECs to form the vessel walls. It is necessary for tumour growth, and cells in the tumour microenvironment release a plethora of small molecules and proteins to promote this process [83–87]. Like leucocyte migration, blood vessel formation is directional and oriented towards the tumour to increase vascularization and enhance growth . Chemoattractants contribute to this process both by recruiting precursor ECs and by inducing their proliferation.
However, the role of chemokines in angiogenesis is quite variable: some are angiogenic, some have no effect on angiogenesis, and others are angiostatic (inhibit angiogenesis). This variability is especially evident within the CXC class of chemokines, as the presence (+) or absence (−) of the ELR (Glu-Leu-Arg) motif near their N-termini correlates with angiogenic or angiostatic characteristics [89–91]. Furthermore, introduction of the ELR motif into the ELR− chemokine CXCL9 [MIG (monokine-induced by interferon γ)] rendered it angiogenic, whereas mutating the ELR motif out of CXCL8 rendered it angiostatic . CXCL12 is the one exception to the ELR correlation within the CXC chemokines, since it lacks the ELR motif, but is nevertherless angiogenic. However, it is believed to mediate its angiogenic effects in part through the induction of VEGF (vascular endothelial growth factor) [16,78].
Non-CXC chemokines can also have angiogenic and angiostatic properties that are not explained by the presence of an ELR motif. For example, CCL1 (I-309), CCL2, CCL11 (Eotaxin) and CX3CL1 can all induce angiogenesis, sometimes directly and sometimes through recruitment of other cells such as TAMs, which in turn release growth and angiogenic factors (e.g. VEGF and basic fibroblast growth factor) [92–94]. Other CC chemokines can inhibit angiogenesis, as has been demonstrated for CCL21 .
All of the ELR+ chemokines bind to CXCR2, which is expressed on ECs . Of these, the TAM-derived CXCR2 ligand CXCL8 is thought to play a particularly important role in angiogenesis, where it is believed to induce the migration and proliferation of ECs . The signalling pathways involved in angiogenesis are essentially the same pathways as those which induce cell migration, survival and proliferation of leucocytes (just specifically in ECs), as discussed elsewhere in the present review. Conversely, the angiostatic properties of many of the ELR− chemokines are thought to be due to activation of CXCR3B on ECs, which results in the inhibition of migration and proliferation [87,97]. CXCR3B is a splice variant of CXCR3 that is found on ECs, whereas the CXCR3A splice variant is expressed on mononuclear cells. Interestingly, CXCR3B contains 52 extra amino acids at its N-terminus relative to CXCR3A, which may modulate its chemokine binding and signalling properties. In contrast with CXCR3B, activation of CXCR3A results in activation of survival and proliferation pathways . This cell-type-specific expression of the CXCR3 splice variants provides an explanation for how ELR− chemokines can activate the standard chemotaxis and survival/proliferation pathways in mononuclear cells while antagonizing these events in ECs in order to inhibit angiogenesis.
Other contributors to the microenvironment: MMPs (matrix metalloproteases)
MMPs present in the tumour microenvironment have been classically associated with processes such as angiogenesis. However, it has become clear that the roles of MMPs are complex, and TAM-derived MMPs play both tumour-promoting and tumour-suppressing roles [98–100]. Chemokines and MMPs have an interesting interdependent relationship, since they control the function of one another. For example, CCL2 induces the expression of MMP-12 by monocytes/macrophages , CXCL8 induces MMP-2 and MMP-9 in ECs , CCL5 and CXCL12 induce MMP-9 expression in monocytes  and CCL2 and CXCL8 induce MT1-MMP . In melanoma cells, the up-regulation of MMP-2 by CXCL8 is associated with increased tumour growth and metastasis , whereas the up-regulation of MMP-9 by CCL5 is thought to contribute to the progression of breast cancer . Conversely, proteolysis of chemokines within their N-terminal signalling domains by MMPs can modulate their activity; examples include the MCPs (CCL2/7/8), CXCL11 (I-TAC), CXCL8 and CXCL12 [7,107]. In the case of the MCPs, the cleavage products have impaired activity and in fact can function as antagonists, whereas the cleavage of the six or seven N-terminal residues of CXCL8 by MMP-9 increases the activity of this chemokine. Truncated CXCL11 failed to attract lymphocytes and was much less potent in inhibiting microvascular EC migration, leading to the concept that the processed form(s) may lead to decreased numbers of tumour-infiltrating lymphocytes and a more angiogenic environment . Overall, the ability of MMPs to dramatically alter the function of a given chemokine makes the full significance of their presence in the microenvironment difficult to define.
Decoy receptors: chemokine receptors that dampen the pro-tumorigenic activities of chemokines?
Recently it has been suggested that the non-signalling ‘decoy chemokine receptors’ DARC (Duffy antigen receptor for chemokines) and D6 may play a role in the tumour microenvironment by binding and sequestering chemokines and thereby inhibiting tumour growth . D6 has an altered DRYLAIV motif in the second intracellular loop (DKYLEIV instead of DRYLAIV) and DARC lacks this motif entirely. Consequently, these receptors fail to couple with G-proteins and cannot induce cell migration, despite their promiscuous chemokine binding profiles [109,110]. Instead, they efficiently internalize their ligands, thus dampening the immune response [110,111].
DARC binds to pro-inflammatory chemokines of both the CC and CXC families, including CCL2 and the angiogenic chemokines CXCL1 and CXCL8. Studies on in vivo models of mice transplanted with non-small cell lung cancer and breast cancer cell lines overexpressing DARC indicated that increased levels of DARC are associated with decreased rates of tumour growth, increased necrosis and decreased metastasis [112,113]. In breast-cancer cells, DARC was thought to be acting, at least in part, by decreasing CCL2 and MMP-9 expression levels, and low levels of DARC in human breast cancer samples were also correlated with a poor prognosis. Transgenic DARC-deficient mice in models of prostate cancer also indicated a role for DARC in reducing tumorigenesis and angiogenesis, by removing prostrate-derived angiogenic chemokines from the circulation. The authors made the link between low levels of DARC expression on erythrocytes and increased mortality rates from prostate cancer .
D6 binds promiscuously, and with high affinity, to a number of CC chemokines . It is thought to play a role in regulating and resolving inflammatory reactions by acting as a scavenger receptor . D6-deficient mice show exaggerated inflammatory responses compared with wild-type mice in models of skin inflammation, with higher levels of inflammatory chemokines in the draining lymph nodes. With respect to cancer, D6 suppresses the development of chemically induced skin tumours, whereas D6-deficient mice have increased susceptibility to tumour development . Interestingly, D6 binds major TAM-recruiting chemokines (CCL2/5/7/8), as well as chemokines postulated to skew inflammatory processes, both towards cell-mediated (e.g. CCL3/4) and antibody-mediated (CCL2/11/17/22) responses. Although further studies are needed to explore the role of decoy receptors in anti-tumour strategies, the anti-tumorigenic properties of both D6 and DARC again highlight the contribution of chemokine-mediated inflammation to cancer.
In summary, chemokines and their receptors play multiple roles in shaping the tumour microenvironment by their ability to attract leucocytes, particularly TAMs, which promote angiogenesis and activate other pro-tumorigenic enzymes and cytokines. However, in order for tumours to thrive, they must also circumvent apoptosis and find ways to promote survival and proliferation. It has recently become clear that tumours utilize chemokine-mediated signalling pathways in order to aid their survival, growth and proliferation, as discussed in the following section.
SIGNALLING: SURVIVAL AND PROLIFERATION
Expression of chemokine receptors on cancer cells may provide the cells with more than a mechanism for migration from the primary tumour to a metastatic site. Receptor signalling may also provide a survival advantage. Chemokine signalling contributes to cancer cell survival in the primary tumour, but it is likely to be of particular importance to metastasized cells. What allows the metastasized cells to ‘feel at home’ in a foreign environment? How are they protected from the immune system? Although these issues involve a multitude of factors, chemokines are likely to be significant contributors. Furthermore, their role in survival signalling is not limited to metastatic cancers. Chemokines that serve as migratory signals to control the homing of lymphocytes to protective niches in the bone marrow and lymph nodes may also promote survival of leukaemia cells.
The molecular strategies for survival and growth are often the result of utilizing, and sometimes reprogramming, existing physiological pathways . In some cases the survival signals are likely to be related to the role of specific chemokine receptor pairs such as CXCL12:CXCR4 and CXCL12:CXCR7 in survival and growth during normal development [118–122]. In other cases, survival and proliferation signalling may result from redirecting existing migration pathways (Figure 2).
Chemokines in survival
Although cancer cells generally have a strong propensity to survive and resist apoptotic stimuli, extracellular survival signals can aide, or even be necessary for, the survival of some cancer cells. For example, CLLs (chronic-lymphocytic-leukaemia cells) rapidly die when cultured in vitro unless they are co-cultured with stromal cells termed ‘Nurselike cells’ . One of the factors secreted by Nurselike cells that contributes to the survival of CLLs in vitro, and presumably in vivo, is CXCL12. CXCL12 has been shown to protect CLLs from spontaneous and drug-induced apoptosis , whereas small-molecule CXCR4 antagonists sensitize the cells to fludarabine-induced apoptosis .
In addition to leukaemia cells , addition of CXCL12 to in vitro cultures of numerous types of CXCR4-expressing cancer cells, including pancreatic adenocarcinoma , glioma  and breast cancer cells , results in their prolonged survival and protection from apoptosis when cultured under suboptimal conditions. CXCL12 also promotes in vivo survival of numerous cancer cells. Administration of antagonists of CXCR4 synergized with chemotherapy in cell killing and tumour regression of glioblastoma multiforme-derived tumour cells  and in a B16 murine model of melanoma . Knockdown of CXCR4 expression through RNAi (RNA interference) or its pharmacological inhibition via AMD3100 (a bicyclam antagonist of the chemokine receptor CXCR4) in a murine 4T1 breast cancer model also decreased the formation of primary tumours and was found to delay and reduce the early growth and/or survival of the 4T1 cells in the lung . These data clearly suggest a role for CXCL12:CXCR4 in survival and/or proliferation of both primary and metastasized cells.
For many years, it was believed that CXCL12:CXCR4 functioned as an exclusive non-redundant pair, but CXCL12 is now known to bind to another chemokine receptor, namely CXCR7 (formerly the orphan GPCR RDC-1) [130,131]. Like CXCR4, CXCR7 plays a vital role in response to CXCL12 in various aspects of embryonic development [122,132]. Recent studies revealed that targeted deletion of CXCR7 results in postnatal death in >95% of the mice, and a different phenotype is observed compared with the CXCR4-knockout mouse, consisting of heart valve defects but normal haematopoietic and neural development . Furthermore, overexpression of CXCR7 is observed in several types of cancers and has been shown to contribute to cell survival and tumour development independently of CXCR4 [130,133]. Small-molecule antagonists of CXCR7 interfere with tumour growth in mouse models of several different cancers. Overexpression of CXCR7 in the MDA-MB 435 breast-cancer cell line, which normally has undetectable levels of CXCR7, resulted in a growth advantage, owing to increased cell survival under suboptimal growth conditions . Implantation of CXCR7-overexpressing MDA-MB 435 cells induced formation of larger tumours in SCID (severe combined immunodeficiency) mice than vector-control transfected cells, despite the absence of CXCR4. RNAi silencing of CXCR7 in the 4T1 breast-cancer cells also resulted in decreased tumour size compared with the wild-type or control RNAi cells. Finally, similar effects of CXCR7 on cell growth/survival were observed in lung-cancer cell lines .
Interestingly, in contrast with the activation of CXCR4 by CXCL12, CXCL12 does not induce calcium flux or migration upon engaging CXCR7, indicating that CXCR7 does not signal in the classic chemokine fashion [122,130]. Many theories have been proposed to explain this phenomenon, including CXCR7 signalling through different pathways, CXCR7 heterodimerization with other receptors or its function as a non-signalling decoy receptor [122,134]. It should be noted that CXCR7 also binds CXCL11, but with lower affinity. Furthermore, CXCL11 is not necessary for development and is naturally absent in some mouse strains, including the C57BL/6 mouse strain which was used as the background CXCR7-knockout mouse mentioned above . The possibility of CXCR7 functioning as a decoy receptor is not unprecedented, since D6 and DARC serve such a function. However, this hypothesis seems unlikely, owing to the importance of CXCR7 in development and the specificity of CXCR7 for CXCL12 and CXCL11 compared with D6 and DARC, which bind multiple CC and CXC chemokines [108,110]. Given the distinct roles of CXCR4 and CXCR7 in developmental processes and their non-identical CXCL12-binding domains , it will be interesting to determine the differences between these receptors in terms of their roles in cancer and whether they function independently and/or synergistically.
Since CXCL12:CXCR4 and CXCL12:CXCR7 are important survival factors in development, it is understandable that they are also prominently involved in cancer cell survival. Yet several other chemokine-receptor pairs also contribute to cancer cell survival (Table 1). For example, the CCL27:CCR10 system is known to attract melanoma cells to the skin, but recent studies suggest that these proteins also promote tumour cell survival by helping to circumvent anti-tumour processes and by providing protection against apoptosis .
Chemokines in cell proliferation and tumour growth
Cancer cells frequently have growth and proliferative advantages over their normal cellular counterparts. As previously discussed, EC proliferation is important for the formation of new blood vessels in order to vascularize tumours and provide routes for metastasis. In addition to the recruitment of TAMs that secrete factors to promote cell proliferation and tumour growth, chemokines directly activate growth and proliferation pathways in the cancer cells themselves. Aberrant protein expression within the cancer cells, due to oncogenes or mutations in tumour suppressors, may then enhance the amplitude and/or duration of the chemokine-activated pathways.
The ability of some chemokines to induce cell growth and proliferation in the context of cancer is well-established. CXCL1, CXCL2 and CXCL3 were originally named Gro/MGSA-α, -β and -γ respectively for ‘growth-related oncogene’ or ‘melanoma growth stimulatory activity/growth regulated protein’ . These closely related chemokines were found to be expressed in approx. 70% of melanomas and function as oncogenes [135,136]. All three chemokines bind to CXCR2 and cause activation of ERK, PI3K and tyrosine kinases to mediate cell proliferation and migration effects [90,91,136–139]. However, induction of cell proliferation is not limited to CXCR2 agonists; other chemokines, including CXCL12, also frequently signal growth and proliferation [27,26,127].
Downstream signalling in survival, growth and proliferation
There is a fair amount of overlap between the survival and proliferation signalling pathways, and this is understandable given that the two processes often work hand-in-hand. It has been demonstrated that stimulation of numerous cancer cells with CXCL12 and other chemokines activates the PI3K/Akt [protein kinase B (PKB)] pathway [11,18,27,46,123,140] which is well known to promote survival effects . Although not all chemokines that promote Akt activation enhance the survival of cells (e.g. under low serum conditions), many do, and this pathway seems to be exploited by a variety of cancer cells [141–143].
Numerous downstream effectors and transcription factors of Akt, ERK1/2, and tyrosine kinase signalling can promote cell survival and proliferation (Figure 2). Chemokine signalling often activates NF-κB, which is commonly downstream of Akt, but can be activated through other pathways, such as the PKC one . NF-κB dimerizes and translocates to the nucleus on activation, where it promotes transcription of various apoptosis inhibitors and cell-cycle-promoting genes . Other downstream targets of Akt include procaspase-9 and the pro-apoptotic Bcl-2 family member, BAD (Bcl-2/Bcl-XL-antagonist, causing cell death), both of which are inhibited upon phosphorylation. The FKHR (forkhead in rhabdomyosarcoma) family of transcription factors, which induce transcription of numerous apoptotic genes, are also inhibited by Akt . Akt-induced activation of Mdm2/Hdm2 (murine double minute 2/human double minute 2), leading to p53 degradation and inhibition of GSK-3β (glycogen synthase kinase-3β), leading to stabilization of β-catenin, also results in downstream inhibition of negative regulators of cell cycle and activation of cell-cycle-promoting genes . Furthermore, via inhibition of TSC2 (tuberous sclerosis complex 2), Akt leads to mTOR (mammalian target of rapamycin) activation, resulting in activation of p70S6K (p70 S6 kinase) and thus enhanced protein translation of numerous cell-growth regulators [137,148]. ERK1/2 signalling may also contribute to survival through some of these pathways, for example via phosphorylation and inhibition of procaspase-9 and BAD [149,150]. Furthermore, ERK1/2 [a MAPK (mitogen-activated protein kinase)] can itself localize to the nucleus and activate transcription factors involved in cell-cycle regulation and differentiation, thereby promoting cell proliferation . Other MAPKs, including JNK (c-Jun N-terminal kinase), have also been implicated in chemokine-induced proliferation signalling . Thus chemokine receptor signalling, resulting in activation of transcription factors involved in anti-apoptotic mechanisms, cell-cycle regulation, and growth-factor production, are yet other mechanisms whereby cancer cells exploit downstream chemokine signalling pathways. These pro-tumorigenic pathways are likely to be particularly important for the ability of metastatic tumour cells to thrive in foreign environments.
COMPLEXITY OF THE CHEMOKINE SYSTEM
As has been discussed above, it is clear that certain chemokine/receptor pairs contribute to cancer progression and metastasis, whereas others are not implicated or may even mediate anti-tumorigenic effects . Particular chemokine/receptor pairs such as CXCL12:CXCR4 have a dominant role in metastasis , likely related to their role in homoeostasis, development and the fact that the ligand is constitutively expressed. However, they cannot always be categorized as ‘pro-tumorigenic’ and ‘antitumorigenic’; although a particular chemokine may exhibit anti-tumorigenic properties in one context, it may still contribute to malignancy in other types of cancers. For example, one of the CCR7 ligands, CCL21, was shown to mediate anti-tumour effects by inhibiting angiogenesis, but it also provides important directional cues for metastasis of cancer cells to the lymph nodes [20,95]. Additionally, low levels of some chemokines, such as CCL2, may induce pro-tumorigenic properties, whereas higher levels inhibit tumorigenesis. Although it is desirable to compartmentalize the roles of chemokines in cancer in fairly defined ways, there are many complexities that should be appreciated. In this section we discuss some of the many possible mechanisms that can complicate the picture.
Diversity of chemokine/receptor responses
Owing to their structural homology and common chemoattractant-related functions (e.g. migration), the diversity of chemokine signalling may be greatly underappreciated. The sheer number of chemokines and receptors, along with the fact that many chemokines bind the same receptor and many receptors engage multiple chemokines, offers the possibility of many outputs. Although such promiscuous partnering gives the appearance of redundancy, emerging evidence suggests that cross-reactivity among ligands and receptors can result in quantitative and qualitative differences in the cellular response [154,155]. For example, it is clear that ligands of the same receptor can elicit different responses, even when their binding affinities are not too dissimilar. Ultimately this must be due to differences in the ligand-induced conformational states and dynamics of the receptors and how they couple into downstream pathways. In an elegant comparative study of CCL17 and CCL22, D'Ambrosio and co-workers showed quantitative differences in CCR4-mediated signalling . CCL22 was much more effective than CCL17 in the induction of integrin-dependent T-cell adhesion, receptor desensitization and internalization. Furthermore, the authors showed that, although CCL22 is the higher-affinity ligand (but only by 2–3-fold), it dissociates more rapidly than CCL17, and they proposed the intriguing hypothesis that the frequency of association/dissociation may be a critical parameter in the activation of certain intracellular signalling pathways. Similarly, CCR7 binds both CCL19 and CCL21 with comparable affinities and demonstrates a similar efficacy in inducing chemotaxis and calcium mobilization. However, CCL19, but not CCL21, led to phosphorylation of CCR7 and subsequent β-arrestin-dependent desensitization in the H9 human T-lymphocyte cell line . The potency and intensity of CCL19-mediated ERK1/2 activation was also higher than that of CCL21-mediated activation .
Distinct responses of a particular receptor to different ligands is also evident in the context of cancer, as the CXCR2 ligand CXCL1, but not CXCL8, was able to activate proliferation and tumour growth in prostate cancer cell lines . Similarly, it is clear that the same chemokine binding to different receptors can also elicit different functional responses, as has already been discussed for CXCL12 binding to CXCR4 versus CXCR7. More quantitative and qualitative analysis of the responses of different chemokine/receptor pairs will be needed to appreciate fully how subtle shifts in binding affinity, kinetics and, ultimately, the induced receptor conformation, can lead to different outputs, which, in turn, will be influenced by cell type. In terms of cancer, this means that it may not always be straightforward to assign a role for a particular chemokine or receptor associated with a cancer. It is important to know what the relevant receptor is for an identified ligand and, vice versa, the relevant ligand for the receptor. Additionally, although the identification of chemokine/receptor mRNA transcripts is often used as evidence for the roles of particular proteins in cancer, the data may be misleading, as protein levels do not always track with mRNA levels. Furthermore, chemokines can be agonists of some receptors and antagonists of others, so what then is their true function? As an example of this concept, the ligands CXCL11, CXCL9 and CXCL10 (IP-10) are agonists of CXCR3, but antagonists of CCR3 , whereas CXCR3 may act as a decoy receptor of CCL11 . N-terminal proteolytic processing may also activate or de-activate chemokines or change their specificity, and the question is: what is the predominant state in the tumour milieu? Other cell-dependent factors, discussed below, can also alter the response.
Complexities of intracellular signalling pathways
Signalling downstream of chemokine receptor activation is also complex, and many factors can influence the functional outcome. Although complexity contributes to fine-tuning of normal chemokine functions, it may also facilitate the ability of cancer cells to adapt various pathways for purposes not normally used in a particular cell type. As discussed above, there is significant overlap between the pathways that are operative in normal chemokine function and those that contribute to cancer. Migration, for example, is critical both to classical chemokine function as well as tumour metastasis. PLC, Akt, ERK1/2 and tyrosine kinase pathways (independently and sometimes in conjunction) have all been implicated in migrational responses in normal cells [18,35,146,159,160]. However, these same signalling molecules also contribute to survival, growth and proliferation in cancer. Furthermore, they can be modulated in many ways in favour of the cancer cell. Paracrine signals from cells in the microenvironment, or autocrine signals, can influence receptor activation and/or regulation. This concept is a major reason why it is so difficult to recapitulate in vivo situations obtaining in in vitro systems. Other complicating factors include G-protein specificity and isoform availability, receptor dimerization, receptor cross-talk and altered signalling and regulation.
Receptor cross-talk, in particular, has been demonstrated to modulate chemokine receptor signalling. For example, EGF (epidermal growth factor) and platelet-derived growth factor are established mediators in ovarian cancer growth and metastasis, and there is now clear evidence for a role of CXCL12:CXCR4 [161,162]. Stimulation with CXCL12 in several ovarian-cancer cell lines resulted in cell proliferation through CXCR4 and biphasic activation of ERK1/2 and Akt, which decreased upon addition of an EGFR (EGF receptor)-specific inhibitor, suggesting cross-talk between CXCR4 and EGFR. In addition, CXCL8 stimulation of CXCR1 and CXCR2 expressing ovarian-cancer cells activates ERK1/2 through interactions with EGFR and c-Src .
Dimer and higher-order oligomerization of GPCRs is also recognized as an important event in the activation and function of many GPCRs [164,165]. Although the relevance of chemokine-receptor homo- and hetero-dimer formation is still under study , it could affect ligand–receptor specificity, the activation of downstream signalling pathways and the duration of the signal in normal or malignant cells. To date, there is evidence suggesting homodimerization (CCR2, CCR5, CXCR1, CXCR2, CXCR4 and CXCR7) and heterodimerization (CCR2/CCR5, CCR2/CXCR4, CXCR1/CXCR2 and CXCR7/CXCR4) of several receptors. Since most of these receptors are implicated in cancer, dimerization could affect the cancer phenotype.
Although cancer cells use much of the same machinery and signalling pathways as do normal cells, they do have altered characteristics, such as the expression of oncogenes or mutations in tumour suppressors that can change or exaggerate the response to chemokines. As has already been mentioned, the mutant pVHL tumour suppressor and the HER2 oncogene contribute to aberrant expression levels of chemokine receptors on cancer cells, thus altering how the cells would normally respond to chemokine signals. In addition, such oncogenes or mutant tumour suppressors could potentially have a dramatic effect on chemokine receptor signalling, leading to prolonged or enhanced pathway activation, or even activation of unique pathways. For example, mutation of PTEN (phosphatase and tensin homologue deleted on chromosome ten), a phosphatase that contributes to inactivation of Akt, could prolong chemokine-induced Akt activation, thus leading to aberrant activity .
Additionally, although inappropriate expression of particular chemokine receptors is certainly relevant to a variety of cancers, increased receptor levels do not always correlate with enhanced signalling . For example, CXCR4 up-regulation was observed in both metastatic and non-metastatic breast-cancer cell lines; however, only the metastatic lines expressed functional CXCR4 . Thus an increase in receptor expression may not always translate into enhanced activity. Furthermore, there may be cancers in which receptor expression is unaltered, yet there are significant changes in the functional response and downstream signalling. Chemokine receptor mutations that cause constitutive activation or that impair desensitization could potentially contribute to tumorigenicity [169–171]. Although there are presently no known endogenously expressed chemokine receptors that exhibit these characteristics, the KSHV (Kaposi's sarcoma herpes virus) GPCR is a CXCR2-like receptor that is constitutively active and contributes to the pathogenesis of Kaposi's lesions . Similarly, point mutations yielding constitutive activation of CXCR2 in mouse embryonic fibroblast NIH 3T3 cells resulted in cell transformation and induced proliferation .
Further adding to the complexity of the situation, many types of cancer cells express multiple chemokine receptors and/or chemokine ligands. CXCR1, CXCR2, CXCR3, CXCR4, CCR7 and CCR10 can all be expressed on melanoma cells and potentially contribute to malignancy . Whether these different receptors contribute independently, redundantly and/or in a co-ordinated manner to the disease remains to be determined. In a tumour microenvironment there is a milieu of growth factors, cytokines and chemokines that most likely function in concert to shape the growth, survival and spread of cancer cells. Although it is critically important to gain a solid understanding of the individual contributions of each factor to the progression of cancer, they do not function in isolation, and it will also be necessary to consider the global picture in the context of complexities such as receptor cross-talk, altered signalling and interactions with other cell types.
The well-established properties of chemokines in controlling cell migration have made them clear candidates for involvement in cancer-cell metastasis. However, the contribution of chemokines to other aspects of cancer, such as growth, proliferation, angiogenesis and cell survival, are also becoming areas of extensive investigation. Many important questions arise from such multifaceted effects of chemokines:
which pathways are activated by chemokines to elicit different responses?
why are only select chemokines involved in cancer?
why do some chemokines have anti-tumorigenic functions while others clearly contribute to malignancy?
how does the same chemokine mediate different effects in normal cells and in different types of cancer cells?
are any of the receptors good drug targets for cancer?
Deciphering signalling pathways activated by chemokines in various cancer cells will be critical to understanding how chemokines influence disease progression and may reveal potential downstream therapeutic targets and consequences of therapeutic intervention.
An interesting paradigm that is emerging in the chemokine field, and may become more relevant in the cancer field in the future, is that of receptor cross-talk, both between other chemokine receptors (e.g. homo-/hetero-dimerization) and between different types of receptors at the cell surface. A number of chemokine receptors have been shown to form homodimers and/or heterodimers, and the complexes often show functional differences compared with their respective monomers . In one recent study, the CXCR4 small-molecule antagonist AMD3100 was used to demonstrate heterodimerization of CCR2/CXCR4 and trans-inhibition of CCR2 . In addition, TAK779, an antagonist of CCR2, CCR5 and CXCR3, was able to inhibit CXCL12 binding to CXCR4 in the context of the heterodimer. Antagonist trans-inhibition presents a unique example of the functional consequences of heterodimer formation, with substantial implications in drug development. From the standpoint of drug development, it also illustrates the importance of understanding the complex network of interactions associated with chemokines and cancer. In doing so, one of the major challenges will be to find ways to recapitulate the influence of the microenvironment in experimental set-ups.
This work was funded by a UARP (University of California AIDS Research Program) Fellowship to S. J. A. (TF06-SD-501), an NIH (National Institutes of Health) Training Grant in Cellular and Molecular Pharmacology (GM007752) to M. O., a Ruth L. Kirschstein NIGMS MARC (National Institute of General Medical Sciences Minority Access to Research Careers) Predoctoral Fellowship (F31) to C. L. S., and awards from the NIH (RO1-AI37113), the Department of Defense (BC060331), UARP (1D06-SD-206) and the Lymphoma Research Foundation to T. M. H.
Abbreviations: Akt, PKB (protein kinase B); CAF, carcinoma-associated fibroblast; CLL, chronic lymphocytic leukaemia cell; DARC, Duffy antigen receptor for chemokines; DC, dendritic cell; EC, endothelial cell; ECM, extracellular matrix; EGF(R), epidermal growth factor (receptor); ELR, Glu-Leu-Arg; ERK, extracellular-signal-regulated kinase; FAK, focal-adhesion kinase; GPCR, G-protein-coupled receptor; GRK, G-protein receptor kinase; HIF-1, hypoxia-inducible factor 1; IL-2, interleukin-2; JNK, c-Jun N-terminal kinase; MAPK, mitogen-activated protein kinase; MCP, monocyte chemoattractant protein; MMP, matrix metalloprotease; NF-κB, nuclear factor κB; PI3Kγ, phosphoinositide 3-kinase γ; PKC, protein kinase C; PLC, phospholipase C; pVHL, von Hippel–Lindau tumour suppressor protein; RNAi, RNA interference; TAM, tumour-associated macrophage; VEGF, vascular endothelial growth factor. Chemokine nomenclature: chemokines are initially referred to as ‘new nomenclature (old nomenclature)’, for example ‘CX3CL1 (fractalkine)’ (where new=CCL#, CXCL#, CL# or CX3CL#; # represents a number and ‘L’ means ‘ligand’); subsequently new nomenclature alone (e.g. CX3CL1) is employed; the nomenclature for the receptors is CCR#, CXCR#, CR# or CX3CR#, where ‘R’ means ‘receptor’; chemoattractant receptor systems are referred to as ‘ligand:receptor’; for example, the chemoattractant CXCL12 and its receptor CXCR4 are referred to as ‘CXCL12:CXCR4’
- © The Authors Journal compilation © 2008 Biochemical Society