CXCL12 is a pleiotropic chemokine capable of eliciting multiple signal transduction cascades and functions, via interaction with either CXCR4 or CXCR7. Factors that determine CXCL12 receptor preference, intracellular signalling route and biological response are poorly understood but are of central importance in the context of therapeutic intervention of the CXCL12 axis in multiple disease states. We have recently demonstrated that 5T4 oncofoetal glycoprotein facilitates functional CXCR4 expression leading to CXCL12 mediated chemotaxis in mouse embryonic cells. Using wild type (WT) and 5T4 knockout (5T4KO) murine embryonic fibroblasts (MEFs), we now show that CXCL12 binding to CXCR4 activates both the ERK and AKT pathways within minutes, but while these pathways are intact, they are non-functional in 5T4KO cells treated with CXCL12. Importantly, in the absence of 5T4 expression, CXCR7 is upregulated and becomes the predominant receptor for CXCL12, activating a distinct signal transduction pathway with slower kinetics involving transactivation of the epidermal growth factor receptor (EGFR), eliciting proliferation rather than chemotaxis. Thus the surface expression of 5T4 marks the use of the CXCR4 rather than the CXCR7 receptor, with distinct consequences for CXCL12 exposure, relevant to the spread and growth of a tumour. Consistent with this hypothesis, we have identified human small cell lung carcinoma cells with similar 5T4/CXCR7 reciprocity that is predictive of biological response to CXCL12 and determined that 5T4 expression is required for functional chemotaxis in these cells.
The 5T4 trophoblast oncofoetal glycoprotein is expressed by many different carcinomas while showing only low levels in some normal tissues (Hole and Stern, 1988; Southall et al., 1990). 5T4 expression has been shown to influence cell adhesion, cytoskeletal organisation and motility (Awan et al., 2002; Carsberg et al., 1996), properties which might account for its association with poorer clinical outcome in some cancers (Mulder et al., 1997; Starzynska et al., 1994; Starzynska et al., 1998; Wrigley et al., 1995). 5T4 molecules are highly N-glycosylated transmembrane glycoproteins whose extracellular domain contains two regions of leucine-rich repeats (LRR) and associated flanking regions, separated by an intervening hydrophilic sequence (King et al., 1999). LRR are found in proteins with diverse functions and are frequently associated with protein-protein interactions (Kobe and Kajava, 2001).
5T4 oncofoetal antigen is associated with very early embryonic stem (ES) cell differentiation and altered motility (Ward et al., 2003) and is also a part of this coordinated process of epithelial-mesenchymal transition (EMT) (Eastham et al., 2007; Spencer et al., 2007). EMT is associated with key morphogenetic events in embryonic development but are also considered contributory to the metastatic spread of epithelial tumours. We have recently shown that 5T4 glycoprotein facilitates functional CXCR4 expression leading to CXCL12 mediated chemotaxis in differentiating ES cells, MEFs and cancer cells (Castro et al., 2012; Southgate et al., 2010).
CXCL12 is a pleiotropic chemokine regulating cellular survival, proliferation as well as chemotaxis but has also been associated with cancer metastasis (Balkwill, 2004; Vandercappellen et al., 2008). CXCL12 was originally identified as the only monogamous chemokine in the CXC family, binding specifically to the widely expressed cell surface seven transmembrane domain G-protein coupled receptor (GPCR) CXCR4 (Burger and Kipps, 2006; Nagasawa et al., 1996). This exclusivity was challenged by discrepancies of CXCL12 binding in the absence of CXCR4 which led to the identification of the orphan receptor RDC1, subsequently established in the CXC chemokine family as CXCR7 (Burns et al., 2006). In addition to CXCL12, CXCR7 binds with low affinity to the chemokine CXCL11 (I-TAC) (Burns et al., 2006).
Biological responses elicited following chemokine binding to the specific receptors involve G protein coupled signalling. For example, upon CXCL12 binding, CXCR4 undergoes a conformational change that facilitates activation of heterotrimeric G proteins and signalling effectors at the plasma membrane (Marchese, 2006). This initiates signalling cascades along multiple pathways including PI3K and MAPK resulting in downstream phosphorylation of proteins such as AKT and ERK respectively (Ganju et al., 1998; Zhang et al., 2005). However, CXCR7 lacks the consensus DRYLAIV motif of the CXC family on the second intracellular loop of the receptor that is purportedly involved directly in G protein coupling. It also fails to activate G protein mediated GTP hydrolysis or initiate calcium mobilization in response to ligand binding (Burns et al., 2006). These data have supported the view that CXCR7 functions as a decoy receptor. However, CXCR7 significantly increases cell proliferation and elevates cellular adhesion in response to CXCL12 in other conditions (Begley et al., 2007; Burns et al., 2006; Miao et al., 2007; Tripathi et al., 2009). Whether CXCR7 functions like a GPCR in mediating a signal transduction process has been a topic of continued debate (Rajagopal et al., 2010). It has been proposed that CXCR7 acts to scavenge or sequester CXCL12, thereby generating gradients of CXCL12 that lead to differential signalling through CXCR4 (Boldajipour et al., 2008; Dambly-Chaudière et al., 2007). Since CXCR4 and CXCR7 can form heterodimers, at least when transiently overexpressed in transfected cells, the possibility that CXCR7 serves as a co-receptor for CXCR4 leading to enhanced CXCL12 mediated G-protein signalling has been proposed (Levoye et al., 2009; Sierro et al., 2007). These mechanisms all involve ligand binding to CXCR7 requiring crosstalk with CXCR4 to activate intracellular signalling pathways (Hartmann et al., 2008). Importantly it has now been shown that CXCR7 interacts with β-arrestin in a ligand-dependent manner (Kalatskaya et al., 2009; Luker et al., 2009; Rajagopal et al., 2010) providing the first confirmed mechanism of direct CXCR7 mediated signal transduction.
Using WT and 5T4KO MEFs, we have investigated CXCL12 receptor preference, signalling pathways and biological responses. In WT MEFs, CXCL12 binds CXCR4, activating both the ERK and AKT pathways within minutes. However in the absence of 5T4 expression, CXCL12 binds CXCR7 which is upregulated, activating a distinct signal transduction pathway with slower kinetics involving the transactivation of the EGFR and eliciting proliferation rather than chemotaxis. The reciprocal relationship between 5T4 and CXC7 is also described in certain small cell lung carcinoma (SCLC) cell lines in which 5T4 is required for chemotaxis to CXCL12.
CXCL12 stimulation of the MAPK and PI3K pathways are disrupted in 5T4 KO but not WT MEFs
In WT MEFs CXCL12 activated dose-dependent signal transduction cascades along both the PI3K and MAPK pathways, with kinetics characteristic of a G-protein coupled response, leading to phosphorylation of AKT and ERK respectively (Fig. 1A,B). However, in 5T4KO MEFs no such cascade was activated in response to CXCL12 (Fig. 1A,B). Data from three independent experiments were compiled and densitometric analysis confirmed that no significant ERK or AKT activation could be observed in 5T4KO MEFs in response to CXCL12 (Fig. 1C–F).
The pathways of signal transduction in WT MEFs complied with classical PI3K and MAPK routes, as phosphorylation of AKT and ERK were significantly disrupted by inhibition of canonical upstream kinases in each pathway, PI3K (80% inhibition) and MEK, respectively (70% inhibition) (Fig. 2). Uncoupling of G-proteins from the cell surface receptors with pertussis toxin (PTX) also resulted in signal transmission inhibition (78%), confirming that each pathway was activated by a GPCR (Fig. 2). The GPCR in question was confirmed to be CXCR4 with the use AMD3100 (AMD), a specific CXCR4 antagonist, which disrupted both PI3K (79%) and MAPK (78%) signalling in these cells. A specific CXCR7 antagonist, CCX733 (CCX), had no effect (Fig. 2).
Although CXCL12 elicited no signal along the PI3K or MAPK pathways in 5T4KO MEFs, the absence of the 5T4 gene in these cells did not cause a gross defect in either signalling pathway. Both PI3K and MAPK networks were intact in 5T4KO MEFs and were inducible by other well-characterised ligands that act independently of chemokine receptors. MAPK signalling was activated by phorbol 12-myristate 13-acetate (PMA) in both WT and 5T4KO MEFs equally (Fig. 2A), and PI3K signalling was similarly activated by insulin in MEFs from both genotypes (Fig. 2B). These data confirmed that the signalling defects reported in 5T4KO MEFs were attributable to dysfunctional CXCR4 rather than disrupted signalling networks.
CXCL12 induces chemotaxis in WT but not 5T4KO MEFs and proliferation of 5T4KO but not WT MEFs
CXCL12 has a well defined role as a chemo-attractant; however, while both WT and 5T4KO MEFs display similar levels of overall migration in a 10% serum gradient (Fig. 3A), only WT MEFS show significant chemotaxis to CXCL12 (P<0.001). This migration is dependent upon CXCR4 activity and can be inhibited by AMD3100 (P = 0.001) (Fig. 3B). CXCL12 has been demonstrated to elicit multiple cellular functions beyond chemotaxis, including the survival and proliferation of normal and cancerous cells. Accordingly we investigated the influence of CXCL12 on the growth of WT and 5T4KO MEFs. Interestingly we observed that despite an apparent failure to elicit a G protein-coupled signal transduction cascade (Fig. 1), CXCL12 was able to enhance the proliferation of 5T4KO MEFs under limiting growth conditions as assessed by counting cell numbers (Fig. 3C). This proliferative response was specific to the 5T4KO MEFs as their WT counterparts did not display enhanced proliferation in response to the chemokine (Fig. 3C). Proliferation was further assessed over a range of times (Fig. 3D) and CXCL12 concentrations (Fig. 3E) by MTT analysis.
CXCL12 induced 5T4KO MEF proliferation is mediated by upregulated CXCR7 expression
The proliferative response of 5T4KO MEFs to CXCL12 is clearly not associated with the typical time course reported for GPCR activation of signalling cascades, so any influence on the downstream pathways over longer time courses post stimulation was investigated. We found that a delayed transmission of signal along the MAPK pathway in response to CXCL12 existed in both WT and 5T4KO MEFs (Fig. 4A). In WT MEFs, in addition to the classical G protein-coupled early response (Fig. 1), a signal was also observed around 60 minutes. In 5T4KO MEFs responses to CXCL12 along the MAPK axis were principally seen at ∼30 and 120 minutes (Fig. 4A). Specific CXCR antagonists showed that signal transduction was mediated by CXCR4 in WT and CXCR7 in 5T4KO MEFs, respectively (Fig. 4B). Consistent with this, FACs (Fig. 4C) and western blot (Fig. 4D) analyses showed that while CXCR4 cell surface and overall protein levels are similar across the genotypes, CXCR7 is upregulated in 5T4KO compared to WT MEFs and is only present at the cell surface of 5T4KO MEFs. To corroborate these findings, and confirm the specific function of the CXCR7 antagonist CCX733, physical knockdown of CXCR7 expression in 5T4KO MEFs by specific shRNAs was performed. Relative to scrambled control shRNA infected cells, CXCR7 expression was significantly reduced by 69% (Fig. 4E) which correlated with a 46% reduction in CXCL12 induced ERK phosphorylation in these cells (Fig. 4F).
These observations correlated functionally as the CXCL12 induced proliferation in 5T4KO MEFs was blocked by CXCR7 but not CXCR4 antagonism (Fig. 4G). While expression of 5T4 enhances CXCL12/CXCR4 functionality its absence also marks reduced expression of the alternate CXCR7 receptor. Thus 5T4 may act to alter the equilibrium of CXCL12 receptor preference between the higher affinity CXCR7, inducing proliferation, and the pro-migratory CXCR4.
CXCR7 signalling and biological function in 5T4KO MEFs is specific to CXCL12
While CXCR4 binds only CXCL12, CXCR7 can bind both CXCL12 and CXCL11, a chemokine that is shared with the CXCR3 receptor (Qin et al., 1998). We investigated these alternative ligands and receptors in the MEF signalling networks. CXCR3 was not detected at the surface of either WT or 5T4KO MEFs by flow cytometry (Fig. 5A). Since CXCR3 may not be at the surface of these cells, and is primarily expressed on activated leukocytes, a leukaemic cell line (SD1) was added to the analyses as a positive control and whole cell lysates were probed by western blot. CXCR3 was expressed in the control SD1 cell line but displayed limited expression in whole cell lysates from both MEF genotypes (Fig. 5B). These data suggest that any CXCL11 effect on signal transduction or function would have to be delivered via the CXCR7 receptor. However, CXCL11 did not activate the MAPK pathway with early or late kinetics (Fig. 5C) nor enhance the proliferation of WT or 5T4KO MEFs (Fig. 5D). However, CXCL11 has been shown to compete with CXCL12 for CXCR7 binding, and in so doing inhibit CXCL12 driven biological functions, thereby acting as a CXCR7 antagonist (Zabel et al., 2011; Zabel et al., 2009). Accordingly, pre-incubation with CXCL11 blocked CXCL12 proliferation in 5T4KO MEF in a dose dependent manner (Fig. 5E), indicating competitive inhibition of CXCL12 binding to CXCR7. These data confirm that the CXCR7 dependent signal transduction and biological response in 5T4KO MEFs is mediated by CXCL12.
CXCR7 activation of the MAPK pathway involves transactivation of EGFR and depends upon Src kinase
Chemokine receptors have been reported to activate G protein-independent signal transduction mechanisms that rely upon the activation of Src kinases (Daaka et al., 1997; Luttrell et al., 1999) but they can also transactivate various growth factor receptors including the EGFR (Porcile et al., 2004). Here we show that CXCR7 signal transduction in 5T4KO MEFs, manifest as late phosphorylation of ERK, is blocked by both specific Src and EGFR tyrosine kinase inhibitors. CXCR7 signalling along the MAPK pathway is dependent upon EGFR activity in these cells, and can be blocked by the EGFR tyrosine kinase inhibitor Tyrphostin (Fig. 6A). It is well established that tyrosine kinase inhibitors can have off-target effects due to a lack of specificity, we therefore validated our pharmaceutical approach by knockdown of the EGFR in these cells. EGFR expression was significantly reduced (62%) in 5T4KO MEFs relative to scrambled control shRNA infected MEFs (Fig. 6B), which correlated with a 49% reduction in ERK phosphorylation in response to CXCL12 (Fig. 6C). Furthermore, CXCL12 activation of 5T4KO MEFs results in phosphorylation of the EGFR with similar kinetics to those observed for ERK activation (Fig. 6D), which can be prevented by knockdown (69%) or antagonism (97%) of the CXCR7 receptor, confirming that CXCR7 activity is responsible for EGFR activation (Fig. 6E,F).
CXCR7 signal transduction in these cells is also dependent upon Src kinase activity and can be blocked with a pan Src inhibitor PP1 (Fig. 6G). CXCR7 mediated phosphorylation of the EGFR can also be blocked by PP1 suggesting that Src kinase acts downstream of CXCR7 but upstream of the EGFR in this novel signal transduction pathway (Fig. 6G). Finally, this pathway was confirmed to be of functional significance as intervention with Tyrphostin disrupted CXCL12 mediated proliferation in 5T4KO MEFs (Fig. 6H).
5T4 and CXCR7 expression in human small cell lung carcinomas (SCLC)
If there is a reciprocal relationship between 5T4 and CXCR7 then it might be predictive of CXCL12 biological function in human cancer. Small cell lung carcinoma lines (H69, H82, H146, H345, H524, H526, H1048, H1963 and DMS114) were screened for expression of 5T4, CXCR4 and CXCR7 by flow cytometry or western blotting. Only DMS114 and H524 cells were found to be 5T4 negative, while all lines expressed CXCR4 (not shown). These 5T4 negative cell lines were paired with a morphologically similar, 5T4 positive SCLC cell line, H1048 and H146, respectively, and the reciprocal relationship between 5T4 and CXCR7 at the cell surface was investigated by flow cytometry. We discovered that, as in MEFs, CXCR7 was only detected at the surface of the 5T4 negative cells, DMS114 and H524 (Fig. 7A).
H1048 and DMS114 are both adherent and have a similar morphology and behaviour in culture so we used these cells to test whether 5T4 expression also marks different biological responses to CXCL12.
H1048 (Fig. 7B) but not DMS114 (Fig. 7C) cells showed significant CXCL12 chemotaxis which could be blocked by AMD3100 but not CCX733 (P<0.005 data not shown). Under the limiting culture conditions employed, enhanced growth with CXCL12 was not observed however there was a greater CXCL12 dose dependent survival of DMS114 (up to ∼80%) compared to H1048 cells (up to ∼30%) (Fig. 7D). The improved survival of DMS114 cells was significantly inhibited by anti-CXCR7 (9C4) blocking antibody pre-treatment (Hartmann et al., 2008) (Fig. 7D) and by the small molecule CXCR7 antagonist CCX733 but not AMD3100 (Fig. 7E). The limited protection seen in H1048 cells was not influenced by either of these treatments and appears unrelated to CXCR4 or CXCR7 interaction (Fig. 7D,E).
In conjunction with greater CXCL12 induced survival, DMS114 but not H1048 cells were significantly protected from both late and early stages of apoptosis as determined by annexin staining by flow cytometry (Fig. 7F) and detection of cleaved Caspase-3 by western blot (Fig. 7G). Only the CXCR7 positive, 5T4 negative DMS114 cells were protected from Caspase-3 activation by CXCL12. Furthermore, pre-incubation of DMS114 cells with CCX733 precluded the anti-apoptotic effects of CXCL12 in these assays (Fig. 7G), confirming that in all cases the biological responses of CXCL12 were mediated by CXCR7.
In order to confirm a direct role for 5T4 in the biological functions examined in the SCLC cells, we used a panel of shRNAs to knock down the expression of 5T4 in the H1048 cells. Relative to scrambled controls we were able to reduce the expression of 5T4 by fivefold (Fig. 8A) and assessed these cells for the expression of both CXCR7 and EMT markers. Knockdown of 5T4 in H1048 cells had no effect on cell surface levels of CXCR7 or CXCR4; however, we did observe that expression of the mesenchymal marker N-Cadherin was reduced (Fig. 8A). Importantly, while migration across a serum gradient was unaffected, chemotaxis toward CXCL12 was reduced in the 5T4 knock down H1048 cells (Fig. 8B), relative to scrambled controls (Fig. 8C).
We have previously demonstrated that 5T4 oncofoetal glycoprotein facilitates functional CXCR4 expression leading to CXCL12 mediated chemotaxis in mouse embryonic cells (Southgate et al., 2010) and certain human tumours (Castro et al., 2012). Here, we show that in the absence of 5T4 the alternate CXCL12 receptor, CXCR7, is upregulated in MEFs and its activation elicits a different pathway of signal transduction with an altered functional cellular response.
In WT MEFs CXCL12 binds to CXCR4, initiating an immediate G-protein coupled signal via the MAPK and PI3K pathways resulting in cellular chemotaxis. In 5T4KO MEFs activation is propagated along the same pathways but with distinct kinetics atypical of GPCR responses and leads to increased cell proliferation. Using specific receptor antagonists we demonstrated that CXCR7, and not CXCR4, transduces the response to CXCL12 in 5T4KO MEFs. It is worthy of note that the CXCR4 antagonist AMD3100 has been shown in some cases to have partial agonistic activity on CXCR7 (Kalatskaya et al., 2009); however, this observation is not a consensus view, and in our system at least AMD3100 does not appear to promote CXCR7 function. This observation does, however, serve to illustrate the apparent distinction between CXCR4 and CXCR7 in potentiation of cellular responses to CXCL12, namely chemotaxis and proliferation/survival, respectively. A CXCR7 influence on survival and efficient differentiation of B cells into antibody producing cells has been reported (Infantino et al., 2006). A proliferative activity associated with CXCR7 has been shown in fibroblasts with ectopic expression supporting tumour formation in nude mice (Raggo et al., 2005). In experimental acute renal failure there is upregulation of CXCL12 which recruits progenitor cells for tissue repair where CXCR7 is critical for CXCL12-mediated survival, whereas CXCR4 is involved in chemotaxis (Mazzinghi et al., 2008). In addition, CXCR7 is also expressed by a variety of tumour cell lines (Burns et al., 2006; Miao et al., 2007) as well as in primary human tumours (Goldmann et al., 2008; Schutyser et al., 2007; Wang et al., 2008) where the tumour growth and aggressiveness can correlate with receptor expression.
Although some overexpression studies have implied that CXCR7 and CXCR4 may act in concert, via the formation of heterodimeric receptor pairs (Levoye et al., 2009; Sierro et al., 2007), our data in MEFs show independence of signalling and function. Coordinated but discrete roles of the two receptors have been described in primordial cell migration in the formation of the lateral line during Zebrafish development. Here, the spatial distribution of CXCR4 and CXCR7 expression is crucial since disruption of the segregation of the receptors results in primordial cell stalling (Dambly-Chaudière et al., 2007). In B cell development, CXCR7 has an established role and is a marker of mature B cells (Infantino et al., 2006) while CXCR4 is detectable at the cell surface during all stages of development it is only weakly responsive in mature B cells (D'Apuzzo et al., 1997; Honczarenko et al., 1999). In human placenta there is also a reciprocal expression pattern of the CXCL12 receptors. CXCR4 expression is higher in early (8–10 weeks) compared to term placenta (Kumar et al., 2004) conversely CXCR7 expression is higher in term rather than early placenta (Tripathi et al., 2009). In mouse embryonic stem cells there is reciprocity of CXCR7 and CXCR4 membrane expression marked by 5T4. Undifferentiated ES cells are 5T4 negative and express membrane CXCR7 but its biological function here is unknown. Upon differentiation CXCR7 is downregulated and 5T4 upregulated transcriptionally which provides for functional chemotaxis to CXCL12 through CXCR4 (Southgate et al., 2010).
In mouse embryonic cells there is a correlation of 5T4 expression with CXCR4 predominance over CXCR7 although how the latter is regulated is not known. It is likely that the balance of CXCL12 receptor expression will be dynamically controlled to provide the morphogenetic requirements of cells and tissues. There are several factors that are likely to influence the receptor that CXCL12 will bind. Firstly, CXCR7 has more than a ten-fold higher affinity for the chemokine than CXCR4 (Balabanian et al., 2005; Burns et al., 2006). Furthermore, the enzyme peptidylarginine deiminase (PAD) has been demonstrated to differentially reduce the binding affinity of CXCL12 for its cognate receptors. PAD sequentially citrullinates the three N-terminal arginines of CXCL12 to produce CXCL12 isoforms with altered citrullination patterns. Interestingly, mono-citrullination on Arg(8) has no effect on CXCL12 binding to CXCR7 but reduces CXCR4 binding by 30 fold; the triple citrullinated isoform Arg(8, 12 and 20) cannot bind CXCR4 but still has reduced affinity for CXCR7 (Struyf et al., 2009). It is perhaps not surprising that CXCL12 appears to favour binding CXCR7 over CXCR4 given that CXCR4 is so widely expressed. It would be detrimental for all cells to respond chemotactically to CXCL12, for example epithelia would become dispersed and tissue integrity compromised. 5T4 is one factor that acts to bias the CXCL12 response in favour of CXCR4 but this is only manifest in the context of coincident CXCR7 downregulation for which the mechanisms are unknown. It is tempting to speculate that CXCR7 and 5T4 may be co-regulated.
These observations are further complicated by the increasing number of non-cognate CXCL12 receptors, and CXCR4/7 ligands, all of which have the potential to offset the equilibrium of the CXCL12 axis. For example, syndecan has recently been shown to interact with both the CXCR4 receptor and the CXCL12 ligand (Schanz et al., 2011), and macrophage migration inhibitory factor (MIF) can bind to both CXCL12 receptors, acting as a non-cognate ligand to prevent CXCL12 driven metastasis (Tarnowski et al., 2010). Indeed a protein of similar structure to 5T4, LRRC4, has been shown to influence CXCR4 signalling kinetics and function in glioblastoma (Wu et al., 2008). It is our contention that 5T4 is another of these factors that govern the CXCL12 receptor predominance equilibrium.
In this study, CXCL12/CXCR4 elicited a typical GPCR signal along canonical routes. A canonical route of signalling for CXCR7 has not been described and indeed the capacity to activate secondary messenger systems has been questioned. Thus some studies have provided evidence that CXCR7 represents a silent or decoy receptor responsible for either sequestering extracellular CXCL12 (Mahabaleshwar et al., 2008; Tiveron and Cremer, 2008) or modulating CXCR4 signalling by forming heterodimers (Levoye et al., 2009; Sierro et al., 2007). Although virtually all reports have agreed that CXCR7 does not couple to G proteins, functional roles for the receptor that belie a signalling event suggest a non-classical positive signalling role for this receptor. Importantly direct CXCR7 mediated signalling, can occur through β-arrestin. Initial studies indicated an interaction between CXCR7 and β-arrestin in a ligand dependent manner (Luker et al., 2009). Subsequently it was determined that CXCR7 can signal directly through β-arrestin and act as an endogenous β-arrestin-biased receptor (Rajagopal et al., 2010). The concept of β-arrestin biased receptors is well supported and it is believed that a balance exists between G protein- and β-arrestin-mediated pathways. Agonist binding typically results in signalling mediated by G proteins and β-arrestins accompanied by receptor desensitisation and internalisation by β-arrestins. This is in contrast to a system with biased signalling, where signalling is mediated selectively through only one of these two pathways. Interestingly, biased receptors have been genetically engineered from balanced receptors by mutation of key residues involved in G protein coupling, such as mutations of the highly conserved DRY motif of the angiotensin II type 1A receptor (AT1AR) to generate the variant AT1AR(DRY/AAY) (Wei et al., 2003) or mutations of three highly conserved residues in the β2 adrenergic receptor to generate the β-arrestin biased β2AR(TYY) (Shenoy et al., 2007). Surprisingly, similar mutations are observed naturally in decoy receptors including CXCR7, which lacks a typical sequence surrounding the DRY motif (Graham, 2009), suggesting that biased receptors have evolved in the genome, that CXCR7 may be one of them and that other receptors that are currently thought to be orphans or decoys may also signal through non-G-protein-mediated mechanisms (Rajagopal et al., 2010).
These observations suggest that CXCR7 may represent a naturally β-arrestin biased receptor. Recent studies of the β-adrenergic (β2AR) GPCR have demonstrated a mechanism of β-arrestin mediated signalling which is consistent with our observations on CXCR7 signal transduction. This work shows a role of β-arrestins in β2AR directed MAPK activation which varies with cell type, agonist dose and time course. The β2AR can couple to both Gas and Gai proteins, as well as β-arrestin; early studies suggested that both coupling to Gαi and β-arrestin resulted in ERK activation (Daaka et al., 1997; Luttrell et al., 1999). Careful dissection of the β2AR MAPK pathway in HEK293 cells, showed early ERK activation (between 0 and 5 min) was reduced by treatment with pertussis toxin (and was therefore G protein-dependent), while late ERK activation (between 5 and 60 min) was inhibited by knockdown of either β-arrestin-1 or -2 (Shenoy et al., 2007). The phenomenon of late ERK activation is analogous to our findings in 5T4KO MEFs and supports a role for β-arrestin mediated signal transduction. Perhaps most interestingly, the mechanism of late MAPK pathway signal transduction involves activation of Src kinase. The implication is that β-arrestin is able to recruit Src kinase as part of a larger scaffold which, following ligand binding, allows the activation of the kinase by β-arrestin. Furthermore, it has also been suggested that the β2AR can interact directly with Src kinase to facilitate delayed activation of the MAPK pathway (Huang et al., 2004; Sun et al., 2007).
We propose (at least in MEFs) that CXCL12 binding to CXCR7 activates, via β-arrestin, intracellular Src kinase which then transactivates the EGFR leading to signalling along the MAPK pathway. Unfortunately, it is difficult to directly test the role of β-arrestin with no specific inhibitor available. Other approaches such as RNAi knockdown or overexpression would have global effects or generate non physiological interactions likely to make interpreting the results difficult. Our demonstration that CXCR7 activation in 5T4KO cells is able to induce transactivation of the EGFR via activation of Src is not without precedent. There are multiple examples of GPCR and chemokine receptor activation leading to growth factor receptor transactivation. CXCR4 for example has been demonstrated to transactivate both HER2 and the EGFR both via a Src kinase dependent mechanism (Chinni et al., 2008; Porcile et al., 2004); furthermore, CXCL12 has been shown to activate both Src and the EGFR in non-transformed N15C6 prostate epithelial cells (Kasina et al., 2009).
Src transactivates the EGFR by phosphorylation at Tyr845 in the kinase domain, stabilising the activation loop, maintaining the active enzyme state, and providing a binding surface for substrate proteins (Cooper and Howell, 1993; Hubbard et al., 1994). Singh et al. (Singh and Lokeshwar, 2011) have shown in prostate cancer cell lines that CXCR7 can be co-precipitated with and is essential for function of the EGFR, but they did not observe any CXCL12 activated EGFR activity. This may be due to activating mutations of the EGFR which circumvent the requirement for CXCR7 directed phosphorylation of the EGFR by Src. Such mutations have been described in the EGFR, for example the L858R missense mutant which resulted in high levels of basal phosphorylation at Tyr845 (Sordella et al., 2004), and multiple mutations within the EGFR activation loop have been identified in prostate cancer (Cai et al., 2008), a disease that is consistently linked with EGFR hyperactivity.
This study builds upon the role of the 5T4 oncofoetal antigen in the CXCL12 chemokine axis. Our data suggest that 5T4 is not only involved in the positive function of CXCR4, but that also its absence reciprocally marks CXCR7 function. Taken together our data suggest that 5T4 may play a role in breaking the natural CXCR7 receptor bias of the chemokine CXCL12 and provides for a switch in homeostatic function induced by the chemokine. We have also begun to decipher the signal transduction mechanisms utilised by the CXCR7 receptor, clarifying multiple observations of an apparent role for CXCR7 in survival/proliferation. Preliminary studies in SCLC suggest that our observations in MEFs may be of functional relevance to human cancer although more detailed studies of the signalling pathways will need to be performed. One might speculate that in a primary tumour, 5T4 surface expression by cells at the periphery would provide for chemotaxis to CXCL12 secreting endothelial cells and metastatic spread whereas in the heart of the tumour preferential CXCR7 expression could detect lower levels of chemokine and promote cell growth. Therapeutic approaches, targeting selective inhibition of CXCR7 and CXCR4 signalling exploiting 5T4 dependency, might provide for better tumour selectivity.
Materials and Methods
5T4 Knockout (KO) mice
The 5T4KO heterozygote C57BL/6 animals were crossed and the progeny genotyped as previously described (Southgate et al., 2010).
Generation and propagation of MEFs
MEFs were prepared and genotyped 5T4KO or WT from day 13 mouse embryos from mating 5T4KO heterozygote C57BL/6 mice. Each embryo was dispersed and trypsinized for 25 min at 37°C, and the resulting cells were grown for 1 day in T-150 flask. These cells were designated as passage number 0 (P0). Cells were passaged every 2-3 days and experiments were generally performed on passages 4–7. MEFs were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 ug/ml streptomycin at 37°C under 5% CO2.
Human small cell lung carcinoma cell lines
Small cell lung carcinoma (SCLC) cell lines were obtained from ATCC; H69, H82, DMS114, H196, H526, H1048, H1963 and were grown and authenticated as previously described (Dean et al., 2011). The cells were screened for expression of 5T4, CXCR4 and CXCR7 by flow cytometry or western blotting as previously described (Southgate et al., 2010). Cell suspensions were prepared and labelled with mouse monoclonal antibody to human 5T4 (mAb-h5T4), rabbit anti-CXCR7 (Ab72100, Abcam), rabbit anti-CXCR4 (Ab2074, Abcam) or relevant IgG controls (DAKO) and subsequently with secondary reagents, goat anti-mouse IgG1 (A21121, Invitrogen) or AF488 goat anti-rabbit IgG (A11034, Invitrogen).
Chemotaxis was assessed using transwell chambers as previously described for cellular motility assays with variations described below (Spencer et al., 2007). Briefly, the FluoroBlok system (BD Biosciences) was employed to assess the kinetic migration of MEFs. FluoroBlok cell culture inserts are designed with a patented light-tight PET membrane that efficiently blocks the transmission of light within the range of 490–700 nm, allowing fluorescence detection in a simplified and non-destructive manner. Fluorescently labelled cells present in the top chamber of the insert are shielded from bottom-reading fluorescence plate readers and microscopes by the BD FluoroBlok membrane. Once labelled cells migrate through the membrane, they are easily detected by a bottom-reading fluorescence plate reader thereby eliminating cell scraping and manual cell counting to enable analysis of both kinetic and end point migration and invasion assays.
The transwells (8-µm pore size) were immersed in gelatine solution overnight (0.1% in PBS) and rinsed in PBS. Transwells were blocked in BSA for 30 min at 37°C/5% CO2 and washed in PBS. MEFs were cultured as described above, serum starved for 24 hours and labelled with DiIl16 dye (Cambridge Biotech) for 1 hour at 37°C. Excess dye was removed by washing in PBS; cells were then counted, and resuspended in phenol red-free, 1% serum culture medium, and added to the transwell plates onto a preformed chemotactic gradient (CXCL12 at 30 ng/ml or serum containing 10% FCS as a positive control) and incubated for up to 24 hours at 37°C/5% CO2. Fluorescent readings were taken from the bottom of the plate every 30 min at an excitation wavelength of 549 nm and an emission wavelength of 565 nm. In all experiments there was no evidence of differential plating with varying conditions; chemotaxis was presented as fluorescent intensity over the indicated time course. P-values were calculated using unpaired Student's t-test. All chemotactic experiments were performed at least three times with triplicates for each condition.
Proliferation was determined by MTT analysis, enumeration of live cells (as qualified by the exclusion of trypan blue) and brightfield microscopy. WT or 5T4KO MEFs were serum starved for 24 hours prior to chemokine stimulation and commencement of analysis for up to 96 hours. At the indicated time points cells were imaged prior to counting or MTT analysis. MTT solution in PBS was added to each well for 4 hours. After removal of the medium, DMSO was added to each well to dissolve the formazan crystals. The absorbance at 540 nm was determined using a micro plate reader (Molecular Devices). WT and 5T4KO MEFs showed identical metabolism of MTT enabling the use of this assay to assess growth over time and with different chemokine doses. Triplicate wells were assayed for each condition. Inhibition studies were performed in the presence of AMD3100 (Sigma) for CXCR4, 9C4 (Calltag) and CCX733 and 11G8 (ChemoCentryx) for CXCR7, tyrphostin (Sigma) for the EGFR and CXCL11 (R&D) for competitive inhibition studies.
Detection of intracellular signal transduction by SDS-polyacrylamide gel electrophoresis (PAGE) and western blotting
Signal transduction was assessed by the phosphorylation (activation) of canonical intracellular mediators of signal transduction. WT and 5T4 KO MEFs were serum starved for up to 48 hours prior to stimulation with different ligands for different times with reactions terminated by ice cold PBS. The compounds PD98059 (50 µM), LY294002 (50 µM) (both Cell Signaling Technology), CCX733 (1 µM), AMD3100 (10 µM), PP1 or tyrphostin (all Sigma), in order to inhibit MEK1, PI3 kinase, CXCR7, CXCR4, Src kinase or the EGFR respectively, were applied to cells for 1 hour prior to CXCL12, CXCL11 (12.5 nM), phorbol 12-myristate 13-acetate (PMA) (50 nM) (Sigma) or insulin stimulation. Cells were lysed in M-PER supplemented with protease and phosphatase inhibitor cocktails, (Thermo Fisher). Samples were prepared in reducing or non-reducing PAGE loading buffer as appropriate (Thermo Fisher), heated to 100°C for 3 minutes and loaded on to a preformed 10% or 4–15% gradient gel (BioRad) and run in Laemmli buffer (25 mM Tris-base, 192 mM glycine, 0.1% SDS). Western transfer to PVDF membranes used a BioRad Mini-PROTEAN Tetra cell system. For western probing, primary antibody concentrations were selected in accordance with the supplier's instructions, antibodies used were: anti-m5T4 (B3F1 (Southgate et al., 2010)), polyclonal rabbit anti-CXCR4 (Abcam), anti-CXCR7 antibodies from Abcam (Ab72100), Genetex (GTX82935) and Calltag (9C4), anti-CXCR3 (R&D), anti-ERK1/2, phospho-ERK1/2 (Thr202/Tyr204), AKT and phospho-AKT (Thr308), anti-cleaved caspase-3 (Asp175) (all Cell Signaling Technology) and anti-Tubulin (Abcam). Following secondary antibody labelling using appropriate HRP conjugates, (AbSerotec) hybridising bands were detected using SuperSignal West Dura (Thermo Fisher).
CXCR7 expression at the surface of cells was determined by flow cytometry using two layer immunofluorescence with anti CXCR7 antibodies (as described) or an isotype matched control at 4°C for 1 hour followed by rabbit anti-mouse immunoglobulins/RPE F(ab′)2 (Dako) at 4°C for 30 min. The samples were analysed using FACSCalibur Flow cytometer.
In primary murine embryonic fibroblasts the expression of CXCR7 and EGFR was knocked down using lentiviral shRNA vectors from Santa Cruz Biotechnology (CXCR7 sc-142643-V, EGFR sc-29302-V and Control sc-108080). Briefly, 5T4 KO MEFs were incubated with lentiviral particles (MOI = 3), coding shRNA for CXCR7 and EGFR overnight in the presence of polybrene (4 ug/ml) in full DMEM media. The cells were then incubated in fresh full DMEM media for 72 hours and expression of CXCR7 and EGFR was evaluated.
For knockdown of human 5T4, four shRNA plasmids were designed, using an experimentally validated algorithm, to knock down 5T4 by RNA interference in addition to one negative control plasmid. Each vector was tagged with a puromycin-resistance gene. H1048 cells (p:23) were transfected at 70% confluence. DNA from each plasmid was mixed separately with FuGENE6 transfection agent (Roche, 11814443001) in a ratio of 2 µg DNA/6 µl transfection agent for a total 100 µl by using serum-free Opti-MEM medium (Gibco, 51985). The plasmid/transfection agent mixture was left at room temperature for 30 minutes before addition to growth media. 48 hours later growth medium was changed to fresh one supplemented with 0.5 µg/ml puromycin (Sigma, P9620-10). Only successfully transfected cells survived in wells with puromycin. Puromycin dose was determined by a minimum dose-to-kill experiment (puromycin added to 70% confluent cells at increasing dose from 0.1 µg/ml to 10 µg/ml and the lowest concentration which successfully killed all cells after 5 days was used). The cells were allowed to grow in the presence of puromycin and then passaged before 5T4 expression was determined.
For all western blot data, densitometric analyses were performed using ImageJ software on at least three independent blots per experiment. Band density, relative to loading controls were normalised and compared statistically by Student's t-test or ANOVA, significant data are described in the Results sections.
Means and standard deviations are presented. Statistical significance was calculated by either two-tailed unpaired Student's t-test or ANOVA as appropriate.
We thank Jian-Mei Hou for the initial SCLC screen, Mark Penfold at Chemocentryx for the generous provision of CCX733, 11G8 antibody and for useful discussions, and all people involved in the PICR core facilities.
This work was supported by a programme grant from Cancer Research UK [grant number C480/A12328]; and S.S. was supported by the Wigan Cancer Research Fund as a Joseph Starkey Clinical Research Fellow.