Positive Feedback Activation of Estrogen Receptors by the CXCL12-CXCR4 Pathway

Estrogens play a pivotal role in reproductive physiology but are also oncogenic in breast cancers. Regulation of target gene expression by estrogens is mediated through direct interaction with the estrogen receptors ERα and ERβ, which belong to the nuclear hormone receptor family of ligand-activated transcription factors (1). Transcription by ERs involves not only interaction with their cognate estrogen response element (ERE) within target promoters but also tethered interactions with AP-1 transcription factors (2). The activation function AF-2 located in the COOH-terminal region of ERs responds to hormone to initiate a combinatorial recruitment of coactivators that facilitate chromatin remodeling and transcription (3, 4). However, the cellular mechanisms involved in ER AF-1 activation are poorly understood. It is known that in response to many growth factors, such as epidermal growth factor (EGF), insulin-like growth factor I, and heregulins, the activation of ERs is associated with phosphorylation of the AF-1 domain (5–7).

CXCR4 is a seven-transmembrane-domain G-protein–coupled receptor, which is activated by the chemokine stromal cell–derived factor 1 (SDF-1; also referred to as CXCL12). The receptor is widely expressed in different tissues and is a prominent chemokine receptor on cancer cells. The chemotactic properties of SDF-1 have been widely associated with metastasis of several epithelial and hematopoietic cancers including breast, prostate, ovary, and lung cancers (8, 9). Initial findings that breast cancer cells, as opposed to surrounding healthy tissue, express CXCR4 have revealed that CXCR4-mediated chemotaxis of cancer cells was mainly involved in the dissemination of metastases to SDF-1–producing target tissues, such as lung, liver, and bone marrow (10). However, although these studies clearly showed the deleterious consequences of CXCR4 expression by cancer cells, the process by which the expression of CXCR4 is selected before metastasis is not explained. It was generally believed that CXCR4 signaling does not promote cell growth. Nonetheless, several findings suggest that CXCR4 confers tumor growth advantages before metastasis dissemination, namely, as a survival factor, by inducing the expression of integrins and matrix metalloproteases associated with invasive phenotypes (11, 12) and neovascularization (13).

Studies of hormone action in human breast cancer cells have been widely informative on the mitogenic actions of estrogens in ER-positive tumors. The identification of several estrogen-induced genes from ERα-positive tumor cells has provided valuable prognostic tools for predicting ER responsiveness to endocrine treatment (14). SDF-1 has been identified as an estrogen-regulated gene in ERα-positive ovarian and breast cancer cells, suggesting a direct pathway by which estrogen may induce SDF-1 production through ERα (15).

Here we show an autologous regulation of SDF-1 through increased ERα and ERβ transcriptional activity in response to SDF-1, which confers growth potential to breast cancer cells. We further show that the CXCR4/SDF-1 pathway promotes ERβ AF-1 phosphorylation at Ser-87 and enables ERβ responsiveness at AP-1 sites despite the presence of tamoxifen. Our findings provide a mechanism by which the CXCR4/SDF-1 and ER signaling pathways mutually contribute to a positive autocrine/paracrine feedback loop in breast cancer.

Plasmids. Expression pCMX plasmids coding for human and mouse ERα and ERβ and ERβ serine to alanine mutants have been described (16–18). The estrogen-responsive EREtkLuc reporter (18) and the AP-1coll-luciferase reporter (19), which contains a fragment (pos. −73 to +63) of the collagenase promoter with an AP-1 element, were described. Plasmids coding for CXCL12/SDF-1α (Invivogen) and for human CXCR4 and its constitutive D132N (NRY) and defective D84N and N119K mutant forms were described (20, 21).

Cell culture. Human breast cancer MCF-7 and human embryonic kidney 293 cells were routinely maintained in DMEM (Sigma) supplemented with 10% and 5% fetal bovine serum, respectively. MCF-7 cells that stably express a ERE-luciferase reporter (17) were grown in the same conditions as parental MCF-7 cells. ERβ-expressing stable Hs578t breast cancer cells (Hs-ERβ) have been described (22).

Transfection and luciferase assay. Cells were seeded in phenol red–free DMEM supplemented with charcoal dextran–treated serum before transfections, and luciferase assays done as described (17, 18). Cells were treated for 16 h with 10 nmol/L 17β-estradiol (E₂), 2.5 to 25 nmol/L recombinant SDF-1, 10 nmol/L ICI-182780, 5 μmol/L 4-hydroxytamoxifen (OHT), 50 μmol/L PD98059 inhibitor (BioMOL Research Labs), and CXCR4 antagonists AMD3100 and T140 (also known as 4F-benzoyl-TN14003). Luciferase values were determined from at least three independent experiments done in duplicate.

RNA isolation and quantitative PCR. Complementary DNA was prepared from MCF-7 cells and Hs578t stable clones as described (23), and PCR amplification was done in a volume of 20 μL with 0.5 to 1 μL of reverse transcription reaction for 25 to 35 cycles. PCR products were analyzed on a MX3000P (Stratagene) and on gel (Alpha Innotech). Sequences of PCR primers are available on request. Values are derived from three to five separate experiments and normalized to glyceraldehyde-3-phosphate dehydrogenase expression.

Cell proliferation assay. Cell proliferation was measured by using the [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl]tetrazolium bromide (MTT) assay essentially as described (17). MCF-7 cells were seeded at low density in phenol red–free DMEM supplemented with dextran charcoal–treated serum. Treatments with 10 nmol/L E₂ in the presence or absence of 1 μmol/L AMD3100 inhibitor were started the following day and maintained in fresh medium every subsequent day. All samples were assayed in triplicate from three to four independent experiments.

Chromatin immunoprecipitation assay. Chromatin immunoprecipitation (ChIP) assays were done as previously described (17, 24). MCF-7 cells were treated with E₂, tamoxifen, and SDF-1 for 40 min. Primer pairs were designed to encompass proximal/distal ERE and AP-1 sites of estrogen-responsive promoters (17, 25, 26).

Cell lysates, immunoprecipitation, and immunoblotting. Site-specific phosphorylation of ERβ was determined by immunoprecipitation and Western blot analysis. MCF-7 cells were serum depleted for 24 h and treated with 25 nmol/L SDF-1 before lysis in TBS containing 0.1% Triton X-100, 1 mmol/L sodium orthovanadate, 1 mmol/L sodium fluoride, 0.1 mmol/L phenylmethylsulfonyl fluoride, and protease inhibitor cocktail (Roche). Immunoprecipitation was done using an anti-ERβ antibody (Santa Cruz) and Western blot analysis was done using a phospho-specific Ser-87 ERβ antibody (Santa-Cruz). Phosphorylated extracellular signal–regulated kinase (Erk) was determined by Western blot analysis using an anti–phospho Erk1/2 antibody (Cell Signaling).

RNA interference. To silence CXCR4 expression, small hairpin RNA (shRNA) duplexes targeting the sequence TGGAGGGGATCAGTATATACA of human CXCR4 (shCXCR4) were inserted into the pLVTH lentiviral vector for small interfering RNA production. Viral particles were produced in 293T cells as described (27) and used to infect MCF-7 cells. CXCR4 efficient knockdown was monitored by Western blot analysis (data not shown).

CXCR4 is required for estrogen-dependent gene expression and growth of human breast cancer cells. To assess the role of the CXCR4/SDF-1 axis on ER-mediated transcription, ERα- and ERβ-positive human breast cancer MCF-7 cells stably transfected with a ERE-luciferase reporter (17) were used. We observed that the increase in ER activity following E₂ treatment was further augmented by exogenous addition of SDF-1 to cells, suggesting that SDF-1 can potentiate the ER response to estrogen (Fig. 1A). The role of CXCR4 in these effects was supported by addition of T140 and AMD3100, two antagonists of CXCR4, which both reduced, in a dose-dependent manner, ER activation by E₂ (Fig. 1B). Similarly, MCF-7 cells transfected with an inactive D84N CXCR4 mutant also exhibited a decrease in ER activity. These findings indicate that estrogen-dependent ER response can be positively modulated through CXCR4 activation by SDF-1.

In line with a previous report that the SDF-1 gene is regulated by estrogen (15), we found a 5.5-fold increase in SDF-1 expression in response to E₂ compared with untreated cells as determined by real-time PCR (Fig. 1C). Interestingly, the increase in SDF-1 expression was reduced by 55% in the presence of the T140 antagonist (Fig. 1C), which suggests that CXCR4 is involved in the regulation of SDF-1 gene. Other known ER-regulated genes, such as progesterone receptor (PR), pS2/TFF1, and EB-1, as well as cyclin D1 and c-Myc, which contain AP-1 sites that can mediate estrogen responsiveness, were all regulated similarly (Fig. 1C). The contribution of CXCR4 was further assessed using the MTT proliferation assay, in which the growth dependency of MCF-7 cells on prolonged treatment with E₂ was significantly reduced with the use of AMD3100 (Fig. 1D). Taken together, these results point toward an important role of CXCR4/SDF-1 to participate in the estrogenic response of MCF-7 cells and the ability of SDF-1 to potentiate these effects.

SDF-1 up-regulates ER activity and target gene expression in the absence of estrogen. To further substantiate that SDF-1 can promote ER function, we observed that ER activity was increased by SDF-1 independently of estrogen in MCF-7 cells (Fig. 2A). Overexpression of wild-type (wt) CXCR4 or the constitutively active D132N mutant also increased basal ER activity in MCF-7 cells (Fig. 2B), suggesting a CXCR4-ER regulatory pathway. In line with our results on the requirement of CXCR4 in maintaining estrogen-induced SDF-1 expression (Fig. 1C), a 2-fold increase in SDF-1 expression was observed in MCF-7 cells treated with SDF-1, supporting the autologous regulation of SDF-1 gene (Fig. 2C). Such effect was abrogated by the antiestrogen ICI-182780, further establishing an essential role of ER in mediating the response to SDF-1. Other ER target genes were similarly up-regulated by SDF-1 (Fig. 2C), which also enhanced MCF-7 cell growth in the absence of estrogen (Supplementary Fig. S1). Silencing CXCR4 expression in MCF-7 cells strongly impaired the response of EB-1 and PR gene expression to SDF-1 (Fig. 2D), in line with an essential role of CXCR4 in mediating responsiveness to SDF-1 in MCF-7 cells.

Activation of ERα and ERβ by the SDF-1/CXCR4 axis. As MCF-7 cells express both ER isoforms, therefore precluding the exact role of each receptor in the cellular response to SDF-1, we performed ERE-driven luciferase assays using human 293 cells transfected with either receptor. In such condition, both ERα and ERβ were activated by SDF-1 with or without estrogen, and these responses were dependent on the presence and the integrity of CXCR4, as shown with the use of CXCR4 antagonist and mutant forms (Fig. 3). We also determined that SDF-1 produced by cells can signal ERα and ERβ in a paracrine fashion, using a coculture system in which CXCR4- and ERα- or ERβ-expressing ERE-Luc reporter 293 cells were incubated with cells transfected with a SDF-1 encoding plasmid (Supplementary Fig. S2). These results indicate that CXCR4 can promote the activity of ERα and ERβ isoforms.

SDF-1 induces ERβ-dependent gene expression in breast cancer cells. To further address the ability of ERβ to transduce the effects of SDF-1 in a breast cancer cell context, we used Hs578t cells stably expressing ERβ (22) and which are CXCR4 positive (data not shown). As shown in Fig. 4A, stable expression of ERβ conferred estrogen responsiveness in terms of target gene expression, as opposed to stable mock control cells (Hs-empty). Addition of SDF-1 to Hs-ERβ cells also increased the expression of PR, EB-1, and SDF-1, providing additional evidence that SDF-1 can positively modulate its own expression through ERβ in breast cancer cells.

SDF-1 promotes ERβ assembly to proximal and distal estrogen-responsive promoters. To address the specific role of endogenously expressed ERβ in the response to SDF-1, we performed ChIP on the estrogen response regions of ER-regulated promoters. In parallel with E₂, treatment of MCF-7 cells with increasing doses of SDF-1 induced a strong association of ERβ with the proximal ERE region of the SDF-1, PR, and pS2 genes (Fig. 4B). Interestingly, such increased occupancy of the SDF-1 promoter is consistent with the autologous regulation of the SDF-1 gene and further shows that the SDF-1 gene is a target of ERβ. Recent genome-wide location analyses of ERα binding sites based on ChIP-on-chip approaches have identified, along with proximal EREs, more distal cis-regulatory enhancer elements that, for certain genes, were reported to confer estrogen responsiveness (26, 28). Given the strong incidence of ERβ to up-regulate PR in response to SDF-1 (Fig. 4A), we then tested whether ERβ could associate with the distal enhancer region located ∼100 kb downstream of the transcriptional start site of the PR gene (28, 29). Indeed, we found that ERβ was even more recruited to the distal PR enhancer region in response to E₂ and SDF-1, compared with the proximal promoter (Fig. 4B). These findings provide evidence that ERβ can share the use of proximal and distal promoter elements originally identified with ERα, and that ERβ activating signals, such as with SDF-1, facilitate such recruitment to up-regulate estrogen-responsive genes in breast cancer cells.

SDF-1 enhances tamoxifen-induced transcription of AP-1 site by ERβ. Whereas antiestrogens, including tamoxifen, are known to efficiently inhibit the expression of ERE-regulated genes, transcription via an AP-1 element by ERα and ERβ can be enhanced by tamoxifen, possibly explaining how certain antiestrogens exhibit estrogen-like effects on cell growth (2, 30). We then tested whether SDF-1 could modulate ERβ activity on a AP-1 luciferase reporter. As expected, and also observed by others (2), E₂ alone had minimal effect on AP-1 response to ERβ, whereas tamoxifen led to a 4.5-fold increase in AP-1 transcription (Fig. 4C). However, a near 2-fold increase in AP-1 activity was observed in both cases following SDF-1 treatment, suggesting that SDF-1 is by itself able to regulate ERβ activity on an AP-1 site. These results were correlated in Hs-ERβ cells regardless of the presence of tamoxifen, with a 2.1- and 2.0-fold increase in the expression of cyclin D1 and c-Myc, respectively (Fig. 4D), two estrogen-regulated genes that contain AP-1 sites (30, 31). To ensure a potential role of AP-1 site, we performed ChIP on the AP-1 site located in the proximal promoter of CcnD1 reported to tether ERα in response to E₂ (25). This region was found potent in recruiting ERβ in response to tamoxifen, which was further increased by addition of SDF-1 (Fig. 4E). These data suggest that SDF-1 can transduce a promiscuous ERβ/AP-1 relationship at the CcnD1 promoter in MCF-7 cells.

The CXCR4-mediated activation of ERβ at ERE and AP-1 sites is AF-1 dependent. Given the ability of the CXCR4/SDF-1 pathway to exert activation of ERβ at both ERE and AP-1 sites regardless of the presence of ER ligands, we evaluated the role of ERβ AF-1, and in particular Ser-106 and Ser-124, identified to mediate mouse ERβ activation in response to ras and growth factors (5, 7, 16). We observed that disruption of Ser-106, either alone or in combination with Ser-94 and/or Ser-124, abolished the activation of ERβ by SDF-1 on an ERE reporter, identifying Ser-106 as critical for ERβ modulation by CXCR4/SDF-1 (Fig. 5A). Similarly, when tested on an AP-1 reporter, the S106A mutant was unable to respond to CXCR4/SDF-1 modulation in the presence of tamoxifen (Fig. 5B). These results highlight the role of Ser-106 in mediating ERβ responsiveness to CXCR4 signaling at both ERE and AP-1 sites.

SDF-1 induces phosphorylation of ERβ at Ser-87 in MCF-7 cells. Mouse ERβ Ser-106 is part of a consensus site for mitogen-activated protein kinase (MAPK)–mediated phosphorylation that is highly conserved among vertebrates and which perfectly matches with Ser-87 of the human ERβ, the difference in numbering being attributable to a 19-amino-acid shorter form of the human versus mouse isoform. As previously reported (32), we found that CXCR4 activation by SDF-1 efficiently promoted Erk activation in transfected 293 cells, and this effect was severely impaired by addition of the CXCR4 antagonists T140 and AMD3100 (Supplementary Fig. S3B). Such activation of Erk through CXCR4 was found essential in providing transcriptional response of ERβ to SDF-1 (Fig. 5C) and correlated with an increased phosphorylation of mouse ERβ at Ser-106 (Supplementary Fig. S3C) and an accelerated ERβ degradation through this site (Supplementary Fig. S4). This identifies Ser-106 as a crucial determinant that couples ERβ transcriptional competence and turnover in response to the CXCR4/SDF-1 pathway.

To confirm these observations in human breast cancer cells, we observed increased Erk activation levels in MCF-7 cells treated with SDF-1 (Fig. 5D), paralleled with an increased human ERβ phosphorylation at Ser-87 (Fig. 5E). Silencing CXCR4 contributed to severely impaired Erk activation and ERβ Ser-87 phosphorylation by SDF-1 (Fig. 5D and E). These findings identify ERβ Ser-87 as a regulatory site targeted by the SDF-1/CXCR4 pathway in human breast cancer cells.

Both ERs and CXCR4/SDF-1 play pivotal roles in breast cancer. Previous data showing that estrogen promoted expression of the CXCR4 ligand SDF-1 have linked the two pathways. In the current study, we show that not only do ERs activate the CXCR4/SDF-1 pathway but, conversely, CXCR4 signaling also promotes ER transcriptional activation, thereby establishing a complete autocrine loop for breast cancer cell growth. Our conclusions are supported by results showing that CXCR4 activation, by either the agonist SDF-1 or activating mutations in the CXCR4 receptor, increases ER-dependent transcription of either ERE- or AP-1–driven reporters and target genes in cancer cells. CXCR4-mediated ER transcriptional activation occurred both in the presence and absence of ER ligands (estradiol or tamoxifen) and affected both ERα and ERβ. The ability of SDF-1 to signal the recruitment of ERβ to the estrogen-responsive promoter region of the SDF-1 gene with subsequent SDF-1 expression emphasizes an autologous positive feedback regulation of SDF-1 in ER-positive breast cancer cells. The phosphorylation of ERβ at Ser-87 by SDF-1, correlated with receptor transcriptional activation and increased turnover, provides a mechanism by which the AF-1 mediates ER responsiveness to CXCR4/SDF-1 and the Erk1/2 pathway in cancer cells.

Whereas the role of the SDF-1/CXCR4 axis in metastasis is well documented (8–10, 33), its potential implication in tumorigenic steps before metastasis remains less clear. Recent reports showing that SDF-1 expression was increased by estradiol, resulting in mitogenic effects in ERα-positive ovarian and breast cancer cells (15), and on the requirement of steroid receptor coactivator SRC-1 (34) suggested that such early steps of tumorigenesis may involve ERs. In the current study, we show a correlation between estrogen-dependent proliferating gene activation by CXCR4 and SDF-1–promoted growth of ER-positive MCF-7 breast cancer cells. Our observations that transcription of ER-regulated genes was induced by SDF-1 but abrogated by antiestrogen ICI-182780 suggest that the mitogenic effect of SDF-1 in these cells requires functional ERs. We show that by targeting both ERα and ERβ, transduction of CXCR4 signaling by SDF-1 allows a direct regulation of ER-responsive genes, including the SDF-1 gene, independently of estrogen stimulation. This interplay between the CXCR4/SDF-1 pathway and ER-mediated transcription defines an autocrine/paracrine feed-forward loop (Fig. 6), providing a possible explanation of how CXCR4 activation participates in the maintenance of continuous cancer cell proliferation. Whether this phenomenon occurs before CXCR4-mediated metastatic dissemination is uncertain, but it raises the intriguing possibility that autologous and continuous CXCR4 activity maintained through ER activation by hormone and/or enhanced Erk signaling may participate in priming metastasis.

Our observation that regulation of ERα and ERβ activity by SDF-1 occurs in the presence as well as in the absence of hormone points toward an important contribution of the AF-1 function in conferring ER responsiveness to SDF-1 signaling. Such active role of AF-1 was further substantiated by the demonstration that SDF-1 leads to AF-1 phosphorylation at Ser-106 of mouse ERβ and the corresponding Ser-87 of human ERβ. Indeed, the AF-1 domain of ERβ contains many putative phosphorylation sites, of which Ser-106 and Ser-124 were described to be directly phosphorylated by Erk1/2, resulting in receptor AF-1 activation in response to EGF or ras and subsequent recruitment of transcriptional coactivators SRC-1 and cyclic AMP–responsive element binding protein-binding protein (7, 16, 35). The critical role of Ser-106 was also important to mediate ERβ response to the CXCR4/SDF-1 axis at both ERE and AP-1 sites. Altogether, our results emphasize a prominent role for the CXCR4/SDF-1 axis to affect ERβ activity through AF-1 phosphorylation to achieve distinct transcriptional events as well as concerted actions with the AF-2.

It has previously been reported that ERβ is able to regulate the activity of the AP-1 transcription factor (2). We here show that the CXCR4/SDF-1 axis increases ERβ recruitment and subsequent transcriptional potential of ERE- and AP-1–regulated genes in the context of breast cancer cells. Importantly, we found an increased expression of two AP-1–regulated genes, cyclin D1 and c-Myc, which are commonly overexpressed in primary breast tumors and recognized markers of early steps in breast tumorigenesis (36). Our findings that the CXCR4/SDF-1 axis stimulates ERβ assembly and activity at AP-1 sites brought potential implications for resistance of cancer cells to antiestrogen treatment. Resistance to prolonged tamoxifen exposure is a major clinical problem, which develops in two thirds of women treated for breast cancer, such that aggressive metastatic dissemination of tumor cells eventually ensues with a poor prognosis (14). The cellular mechanisms by which ER-positive tumor cells overcome antiestrogen effects and exhibit excessive proliferation remain uncertain at present. Our results are consistent with the possibility that the autocrine regulation and coupling between ER-regulated pathways and the CXCR4/SDF-1 axis may function despite the presence of tamoxifen. Indeed, the capacity of ERβ to enhance AP-1 activity and gene expression by SDF-1 and its promiscuity with the AP-1 region of cyclin D1 promoter were both maintained in the presence of tamoxifen, in line with previous reports that AP-1 regulates transcription of cyclin D1 in the presence of antiestrogens (30). Our observations that this process may require Erk activation to phosphorylate ERβ Ser-87 are consistent with altered MAPK transduction pathways and enhanced AP-1 signaling associated with tamoxifen resistance (37). Elevated MAPK activity produced a mitogenic phenotype in MCF-7 cells (38), and enhanced Erk1/2 activity was measured in tumors from ER-positive breast cancer patients maintained on tamoxifen (39). Although tamoxifen efficiently uncouples coactivator SRC-1 potentiation from ER AF-2 function, no such effect was observed on AF-1 activity on Erk1/2 activation (16, 35). It is therefore tempting to speculate that the ability of CXCR4 signaling to target and up-regulate AF-1 activity during tamoxifen exposure might supplant AF-2 as the primary route of ER activation in tamoxifen-resistant breast cancer cells.

Our data show not only that ERβ up-regulates the expression of SDF-1, as also observed for ERα (15) but also that SDF-1 increases ERα and ERβ activity, thereby extending the role of CXCR4 to ER signaling in breast cancer cells (Fig. 6). Although not referred as a perfect ERE (15, 40), the estrogen response region of the SDF-1 gene was highly potent in recruiting ERβ in response to SDF-1, which identifies the SDF-1 gene as a target of ERβ, and further provides a mechanism by which the autologous regulation of the SDF-1 gene can occur in breast cancer cells. Similarly, recruitment of ERβ to other estrogen-responsive promoters, such as PR and pS2, supports the ability of ERβ to increase gene expression in the context of enhanced SDF-1 production. In the case of PR, the apparent propensity of ERβ to bind distal as well as proximal EREs points toward a complex interplay that likely involves cooperating factors to facilitate gene transcription (41). CXCR4 was also reported to be enhanced by estradiol in endometrial adenocarcinoma (42) and breast tumor xenografts (43). Altogether, these data suggest that both components of the CXCR4/SDF-1 axis, the receptor and the ligand, are likely to be regulated in an ER-dependent fashion.

Recent reports have identified the chemokine receptor CXCR7 as an additional receptor for SDF-1, with potential implication in tumor development (44, 45). Our findings do not rule out the implication of CXCR7 in the effect of SDF-1 in MCF-7 cells. However, our observation that CXCR4 silencing impairs the modulation of ER-dependent gene expression and Erk-mediated ERβ phosphorylation suggests that functional CXCR4 is required for transducing these effects by SDF-1 in MCF-7 cells. Notwithstanding its implication in sustaining the growth of breast cancer cells and its potential to affect CXCR4 function (46), the role of CXCR7 in regulating estrogen-dependent functions remains as yet uncharacterized and deserves further investigation.

No potential conflicts of interest were disclosed.

Grant support: Canadian Institutes of Health Research, Cancer Research Society, Inc., and the Canadian Foundation for Innovation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Julie Martin for technical assistance and members of the laboratory for critical reading and useful comments. K. Sauvé is supported by graduate student awards from the Groupe de Recherche sur le Médicament de l'Université de Montréal and from the Fondation de l'Hôpital Ste-Justine; J. Lepage holds an award from the Natural Sciences and Engineering Research Council of Canada; M. Sanchez from the Fondation de l'Hôpital Ste-Justine; and N. Heveker and A. Tremblay are New Investigators of the Canadian Institutes of Health Research.