Prolactin receptors (PRLr) expressed in a majority of breast cancer are activated by prolactin and growth hormone. The PRLr is commonly stabilized in human breast cancer due to decreased phosphorylation of residue Ser349, which, when phosphorylated, recruits the βTrcp E3 ubiquitin ligase and facilitates PRLr degradation. Here, we show that constitutive oncogenic signaling downstream of ErbB2 and Ras stabilizes PRLr via inhibitory phosphorylation of glycogen synthase kinase-3β (GSK3β) on Ser9. Importantly, inactivation of GSK3β correlates with elevated levels of PRLr protein in clinical human breast cancer specimens. Additional studies using pharmacologic, biochemical, and genetic approaches reveal that GSK3β is a bona fide PRLr kinase that phosphorylates PRLr on Ser349 and is required for the recognition of PRLr by βTrcp, as well as for PRLr ubiquitination and degradation. [Cancer Res 2008;68(5):1354–61]
Malignant transformation of cells and development of tumors result from a number of key events that include stimulation of cell proliferation and inhibition of cell death (1). The pituitary hormone prolactin (PRL), which is also secreted by mammary epithelia, plays a central role in mammary gland development and function. In addition, several lines of evidence strongly implicate PRL in breast tumorigenesis (reviewed in ref. 2). First, PRL promotes growth of human breast cancer cells acting as a survival agent and as a mitogen (3, 4), and up to 95% of primary human breast cancers are positive for PRL and its receptors (5, 6). Second, transgenic mice locally expressing PRL within mammary epithelia develop tumors (7, 8), whereas genetic ablation of PRL receptors severely delays the development of SV40 large T antigen–induced breast carcinomas (9). Finally, epidemiologic studies link elevated levels of circulating PRL with increased risk of breast cancer (10, 11) and its metastases (12), as well as with decreased taxane therapeutic efficacy (13–15) that could be reversed by pharmacologic suppression of PRL levels (14).
Prolactin acts via cell surface receptors that exist as long/ΔS1 isoforms (hereafter called PRLr) and several shorter alternatively spliced variants that often exert dominant-negative effects on signaling via the PRLr. PRL activates the PRLr-associated Jak2 tyrosine kinase and a series of downstream signaling pathways, including signal transducers and activators of transcription (Stat), Erk1/2, phosphoinositide 3-kinase (PI3K)–Akt, and others. Because a high proportion of human breast cancer cells secrete their own PRL, the autocrine effects of PRL may account for the limited success of inhibitors of pituitary PRL synthesis/release against human breast cancers (reviewed in ref. 2). Antagonists of PRLr kill human breast cancer cells in vitro and abrogate the tumorigenesis in the xenograft models demonstrating that persistent signaling induced by locally secreted PRL is essential for growth and survival of these cells (16, 17).
However, PRL also induces proteolytic degradation of PRLr via receptor ubiquitination facilitated by the SCFβTrcp E3 ubiquitin ligase that is recruited to the substrate in a manner that requires phosphorylation of Ser349 within the phosphorylated degron (18). Given that this ligand-induced PRLr down-regulation limits the extent of PRL signaling (2), it is not clear how PRL maintains the survival of breast cancer cells. Previous studies showed that the levels of PRLr are decreased in the breast cancer intratumoral stromal compartment compared with the stroma from benign tissue (6). This decrease was attributed to the down-regulation of PRLr in response to its ligand secreted by tumor cells. However, the levels of PRLr in tumor cells are not decreased in comparison with benign mammary cells (6, 19), suggesting a possibility that down-regulation and degradation of PRLr in tumor cells might be impaired. Indeed, we have reported that phosphorylation of PRLr on Ser349 within its phosphorylated degron is impaired in breast cancer cells and tissues that exhibit increased stability of PRLr and ensuing high levels of its expression (20).
Here, we investigated the mechanisms and outcomes of PRLr stabilization in breast cancer. We found that glycogen synthase kinase 3β (GSK3β) mediates the recruitment of βTrcp and receptor ubiquitination and degradation through phosphorylation of PRLr on Ser349. However, constitutive oncogenic signaling downstream of the Ras pathway inactivates GSK3β through phosphorylation of GSK3β on Ser9. Inhibition of GSK3β activity prevents phosphorylation of PRLr on Ser349 and PRLr ubiquitination, ultimately leading to PRLr stabilization.
Materials and Methods
Detailed experimental procedures are provided in the supplementary data.
DNA constructs, chemicals, and antibodies. PCDNA3-based vectors for mammalian expression of βTrcp2 (21) and Flag-tagged PRLr and pGEX-2T–based vectors for bacterial expression of glutathione S-transferase (GST)–PRLrWT and GST-PRLrS349A were described previously (18). Short haipin RNA constructs for knocking down GSK3α or GSK3β were constructed in a modified version of pSuper-retro vector. Control short hairpin RNA (shRNA) vector contained the sequence against GFP (22). Human recombinant prolactin (PRL) was kindly provided by Dr. A.F. Parlow (National Hormone and Peptide Program,). Antibody against βTrcp (23) and phosphorylated Ser349-PRLr (20) were previously described. Other chemicals and antibodies were purchased.
Cell lines, gene expression, and immunotechniques. Cells were cultured as previously described (24). MCF10A cells stably expressing H-Ras were obtained by cotransfecting H-Ras constructs with pBABE-puro vector, followed by selection in medium containing puromycin (1 μg/mL). Immunoprecipitation and immunoblotting procedures were carried out as described elsewhere (20).
In situ detection and quantification of protein expression. Immunohistochemistry was performed on breast cancer tissue arrays as described (25). Tissue array slides were scanned, and images of each breast cancer specimen were captured at multiple immunofluorescence wavelengths. Automated Quantitative Analysis (AQUA) software (HistoRx) was then used to identify cancer cell masks based on cytokeratin-positive cells. AQUA scores for both PRLr and phosphorylated Ser9-GSK3β were then calculated that correspond to the average signal intensity within the cancer cell compartment normalized per cytokeratin signal. Levels of phosphorylated Ser9-GSK3β and PRLr in cancer cells were assessed in two separate tissue arrays of archival and deidentified invasive breast carcinoma tissues from 170 patients After excluding tissue spots with insufficient tumor tissue left after antigen retrieval for either phosphorylated Ser9-GSK3β or PRLr determination, conclusive data were obtained for 44 and 50 specimens from cohorts A and B, respectively.
Binding, ubiquitination, and degradation assays. In vitro phosphorylation-binding assay, in vivo ubiquitination, and pulse chase assays were carried out as previously described (18, 20). For cycloheximide chase analysis, cells were grown in 100-mm dishes, starved overnight in serum-free DMEM medium, and treated with cyclohexomide (50 μg/mL) and PRL (20 ng/mL) for indicated time before harvesting cells from one-third of each dish. Harvested cells were lysed, and Flag tag PRLr proteins were immunoprecipitated with anti-Flag antibody, separated on SDS-PAGE, and analyzed by immunoblotting.
A constitutive activation of the Ras pathway (26), often reflecting upstream alterations in Neu/ErbB2, as well as amplification of c-Myc (27), and overexpression of cyclin D1 (28) are common in breast cancers. We investigated whether these molecular alterations in breast cancers affect stability of PRLr in human embryo kidney 293T cells that exhibit efficient degradation of PRLr (18). Degradation of PRLr was significantly delayed in cells that received activated H-Ras (Fig. 1A) and exhibited increased Erk activation (Supplementary Fig. S1). In contrast, Ras failed to stabilize another cytokine receptor that is also a substrate for βTrcp, IFNAR1 (data not shown). A similar PRLr stabilizing effect was observed in cells cotransfected with ErbB2/Neu. However, neither c-Myc nor stable and hyperactive cyclin D1T286A mutant (29) was capable of significantly inhibiting the rate of PRLr turnover (Fig. 1A) despite being expressed at detectable levels (Supplementary Fig. S1). These results show that activation of growth factor receptor–dependent and Ras-dependent signaling pathways leads to stabilization of PRLr.
Furthermore, expression of activated Ras in near normal human mammary epithelial MCF10A cells increased the levels of PRLr in mass cultures (Supplementary Fig. S2). Individual clones harboring activated Ras also exhibited increased levels of PRLr concurrent with decreased PRLr phosphorylation on Ser349 (Fig. 1B). Given that Ser349 phosphorylation is required for PRLr ubiquitination and degradation (18), these data imply that PRLr might be stabilized in cells that harbor activated Ras. Indeed, proteolysis of endogenous PRLr was impaired in Ras-expressing MCF10A cells relative to the parental ones (Fig. 1C and Supplementary Fig. S3). Consistent with the notion that Ras might perpetuate PRLr signaling, we observed a more robust and longer lasting activation of Stat5 by a pulse of PRL treatment of cells that harbor activated Ras compared with parental cells (Fig. 1D). Together, these data suggest that activation of Ras pathway in MCF10A cells leads to stabilization of PRLr (most likely due to a decreased phosphorylation on Ser349) and to an increase in the magnitude and duration of PRLr signaling.
Constitutive activation of Ras-dependent pathways [including activation of mitogen-activated protein kinase (MAPK)–dependent and PI3K-Akt–dependent cascades] is responsible for the effects of Ras on cell transformation, growth, and survival (30). Expression of either constitutively active catalytic subunit of PI3K (p110*; ref. 31) or constitutively active MAP/ERK kinase 1 (MEK1; MEKEL; ref. 32) that respectively stimulated either Akt or Erk (Supplementary Fig. S4) led to a delay in PRLr turnover (Fig. 2A). A lesser effect of p110* at 6 h (that was seen in cycloheximide but not in pulse chase analysis; results not shown) could be attributed to a rapid degradation of activated Akt in cycloheximide-treated cells.5
A. Plotnikov and S. Y. Fuchs, unpublished data.
We next investigated whether these pathways may affect PRLr phosphorylation on Ser349 (which is critical for PRLr degradation) in human breast cancer cells. We used T47D cells that exhibited relatively low levels of Ras activation (33) and relatively efficient degradation of PRLr when compared with other breast cancer cell lines tested (20). To modify the activity of Ras-dependent pathways, we treated T47D cells with pharmacologic inhibitors against either PI3K (Wortmannin) or MEK1 (PD098059). The latter inhibitor dramatically decreased the levels of PRLr (data not shown), thereby confounding the interpretation of phosphorylated specific immunoblots. When comparable levels of PRLr were loaded (Fig. 2B), a noticeable increase in the extent of Ser349 phosphorylation of PRLr was observed in response to either of these inhibitors. These data suggest that constitutively active MAPK and PI3K pathways impede PRLr phosphorylation within its phosphorylated degron.
Our previously published data indicated that the activity of a kinase that phosphorylates Ser349 in vivo might be impaired in malignant cells (20). Given these results, we were searching for a PRLr-Ser349 kinase whose activity would be (a) capable of facilitating protein ubiquitination and degradation of its substrates, (b) inhibited by both PI3K-Akt and MAPK pathways, and (c) decreased in human breast cancer cells. GSK3β, which is inactivated in some breast cancers (34) via inhibitory phosphorylation on Ser9 triggered by constitutive activation of PI3K or/and MAPK pathways (29, 35), was considered a candidate kinase. Indeed, we observed that, concurrent with an increase in phosphorylation of PRLr on Ser349, phosphorylation of GSK3β on Ser9 was decreased in T47D cells treated with either MEK1 or PI3K inhibitors (Fig. 2B).
We next tested whether GSK3β might play a role in stabilization of PRLr by Ras in 293T cells by coexpressing either wild-type GSK3β or kinase-deficient GSK3βK85,86MA mutant (GSK3βKD) or GSK3S9A mutant that cannot be phosphorylated in response to Ras activation. The GSK3S9A mutant was expressed to an extent similar to other constructs; this mutant did not impair Erk activation by Ras (Supplementary Fig. S5). However, stabilization of PRLr by Ras was not observed in cells expressing GSK3S9A. Whereas coexpression of wild-type kinase with Ras slightly improved PRLr degradation, catalytically inactive GSK3βKD mutant further stabilized PRLr (Fig. 2C) despite its negative effect on Ras-induced Erk phosphorylation (Supplementary Fig. S5). Furthermore, this kinase-deficient mutant was capable of stabilizing PRLr even when expressed without Ras (data not shown). These results implicate GSK3β activity in regulation of PRLr turnover and suggest that inhibitory phosphorylation of GSK3β on Ser9 might be involved in mediating the stabilizing effects of Ras activation on PRLr.
GSK3β phosphorylates the regulators of transcription and apoptosis (β-catenin, Snail, and Mcl-1) leading to their recognition by βTrcp E3 ubiquitin ligase followed by efficient ubiquitination and degradation of these proteins (34, 36, 37). Remarkably, the levels of Snail were increased in human breast cancers that exhibited high levels of inhibitory phosphorylation of GSK3β on Ser9 (34). We sought to investigate whether primary human mammary tumors display a positive correlation between phosphorylation of GSK3β on Ser9 and expression of PRLr.
Levels of phosphorylated Ser9-GSK3β and PRLr in human breast cancer tissues were determined using AQUA of immunofluorescence signals in cytokeratin-positive cells in breast cancer tissue arrays. Data from a first set of 44 primary invasive breast carcinomas analyzed in parallel revealed that levels of phosphorylated Ser9-GSK3β and PRLr varied markedly between breast cancer specimens and were highly positively correlated (Spearman's rho = 0.50, P = 0.001; Fig. 3A). Importantly, this positive correlation held up in a second independent clinical cohort of 50 primary invasive breast carcinomas (Spearman's rho = 0.33, P = 0.019). These results suggest that GSK3β is involved in regulation of PRLr levels and that inhibitory phosphorylation of GSK3β on Ser9 might contribute to PRLr stabilization and accumulation in human breast cancers.
Correlations between GSK3β activity and PRLr phosphorylation on Ser349 and PRLr levels implicate GSK3β in phosphorylation-dependent recruitment of βTrcp to PRLr followed by its ubiquitination and degradation. We observed that endogenous GSK3β and PRLr were coimmunoprecipitated from cell lysates and that PRL treatment did not affect the extent of this binding (Supplementary Fig. S6). Incubation of recombinant GSK3β with bacterially produced GST-PRLr protein (a fusion of GST with intracellular tail of PRLr; ref. 18) as a substrate in an in vitro kinase assay yielded a signal recognized by anti-Ser349 phosphorylated-specific antibody in an ATP-dependent and Ser349-dependent manner (Fig. 3B). This result indicates that GSK3β is capable of directly phosphorylating PRLr on Ser349.
Pharmacologic inhibition of GSK3β using the SB216763 inhibitor decreased Ser349 phosphorylation of PRLr immunopurified from three different cell lines (Supplementary Fig. S7). We further used shRNA to knockdown either GSK3β or related GSK3α. The efficiencies of these constructs were verified by analyses of protein levels and of phosphorylation of a known GSK3β substrate, β-catenin (Supplementary Fig. S8). Knockdown of GSK3β alone decreased the extent of PRLr phosphorylation on Ser349. This decrease was concurrent with increased levels of PRLr and was further augmented by a combined knockdown of GSK3β and GSK3α (Fig. 3C). Furthermore, expression of kinase-deficient GSK3βKD mutant in T47D human breast cancer cells also decreased Ser349 phosphorylation and increased total PRLr levels, whereas expression of GSK3βS9A active kinase led to the opposite result (Fig. 3D). In all, these data suggest that GSKβ is a bona fide PRLr kinase that is capable of phosphorylating Ser349 and regulating PRLr protein levels.
Given that phosphorylation of PRLr on Ser349 is critical for recruitment of βTrcp E3 ubiquitin ligase (18), we investigated whether GSK3β directly regulates PRLr-βTrcp interaction. To this end, we used an in vitro phosphorylation-binding assay (18), in which bacterially produced GST-PRLr is phosphorylated and then analyzed for efficiency of binding of radiolabeled βTrcp. Under these conditions, βTrcp interacted with the substrate in a manner that depended on the presence of GSK3β and ATP and on integrity of Ser349 (Fig. 4A). In addition, when lysates from 293T cells were used as a source of kinase activity, pretreatment of cells with GSK3β inhibitors, such as LiCl and Kenpaullone, decreased the efficiency of subsequent binding of βTrcp to GST-PRLr (Fig. 4A,, right). Furthermore, treatment of 293T cells with these inhibitors decreased the levels of endogenous βTrcp that was coimmunoprecipitated with endogenous PRLr (Fig. 4B). Together, these data indicate that GSK3β regulates recruitment of βTrcp E3 ubiquitin ligase to PRLr via phosphorylation on Ser349.
Phosphorylation-dependent recruitment of βTrcp to its substrates is a rate-limiting event in regulation of ubiquitination and degradation of these substrates (36). When compared with wild-type GSK3β, the expression of catalytically inactive GSK3βKD noticeably decreased the levels of ubiquitinated PRLr; an opposite result was observed in cells that were transfected with active GSK3βS9A (Fig. 4C). In addition, knockdown of GSK3β alone dramatically impaired PRLr ubiquitination, which could be further decreased by a combined knockdown of GSK3β and GSK3α (Fig. 4D). These data suggest that GSK3β is required for efficient ubiquitination of PRLr.
Given that PRLr levels are increased upon inhibition of GSK3β, we next investigated whether this kinase is involved in regulating PRLr stability and signaling. Treatment of cells with three different GSK3 inhibitors delayed PRLr degradation measured either by pulse chase (Fig. 5A) or cycloheximide chase (Supplementary Fig. S9) analyses. Furthermore, knockdown of GSK3β resulted in an inhibition of PRLr turnover that was further impaired when shRNA against GSK3α were also added (Fig. 5B). In addition, whereas expression of active GSKβS9A decreased the efficiency of activation of Erk triggered by a pulse treatment with PRL expressed to a lesser extent, kinase-deficient GSK3βKD mutant robustly extended the duration of this activation (Fig. 5C). It is likely that stimulating effect of GSK3βKD on Erk activation was specific for PRL-induced events because expression of this GSK3β mutant did not increase Erk activation in Ras-transfected cells (Supplementary Fig. S5). Furthermore, knockdown of both GSK3α and GSK3β noticeably augmented the induction of Erk phosphorylation in response to PRL (Supplementary Fig. S10). In all, these results suggest that GSK3β contributes to an efficient degradation of PRLr and to negative regulation of PRL signaling.
Besides activating diverse signaling pathways, PRL also stimulates down-regulation of its own receptors; this down-regulation is impaired in some human breast cancers and tissues, at least in part via impaired phosphorylation of PRLr on Ser349. Here, we present observations that suggest a mechanism that might underlie PRLr stabilization in malignant tissues. Our data indicate that signaling induced by oncogenic proteins activating Ras and its downstream effectors (MAPK and PI3K-Akt pathways) results in stabilization of PRLr in a manner that requires the integrity of Ser9 of GSKβ. Analysis of primary human breast cancers shows that the inhibitory phosphorylation of GSK3β on Ser9 correlates with increased levels of PRLr. Data obtained from biochemical, pharmacologic, and genetic studies characterize GSK3β as a bona fide PRLr kinase that phosphorylates Ser349, enables the recruitment of βTrcp, and facilitates ubiquitination and degradation of PRLr. Together, evidence presented here delineates a putative mechanism of PRLr stabilization in breast cancers.
Studies aimed at delineation of the mechanisms by which phosphorylation of PRLr on Ser349 is impaired (thus, leading to PRLr stabilization in breast cancers) implicated constitutive activation of Ras downstream pathways (PI3K-Akt and MAPK) in these processes. The effects of activated Ras were no longer observed in cells expressing active GSK3βS9A mutant, suggesting that phosphorylation of the Ser9 residue of GSK3β is important for PRLr stabilization by Ras (Fig. 2). Numerous reports show that activation of either MAPK (33, 35) or PI3K-Akt (38) pathways decreases the activity of GSK3β via an inhibitory phosphorylation on Ser9 (39). Furthermore, inhibition of GSK3β via its Ser9 phosphorylation has been reported in breast cancers (34, 40). Our data implicate GSK3β inhibition in stabilization, accumulation, and ensuing hyperactivation of PRLr similar to various other protooncogenic proteins whose phosphorylation by GSK3β within specific phosphorylated degrons triggers their ubiquitination and degradation, including β-catenin (41), cyclin D1 (29), c-Myc (42), c-Jun (43), Mcl-1 (37), and Snail (34). Although phosphorylation of these proteins by GSK3β recruits diverse F-box proteins within SCF E3 ubiquitin ligases (βTrcp for PRLr, Mcl-1, Snail, and β-catenin; Fbx4 for cyclin D1, ref. 44; and Fbw7 for c-Myc and c-Jun), stabilization of protooncogenic GSK3β substrate proteins that are conducive to cellular transformation seems to represent a common mechanism in the pathogenesis of cancer.
The role of MAPK and PI3K signaling pathways in limiting PRLr degradation is somewhat counterintuitive, given that PRL induces these very pathways (2) while promoting Ser349 phosphorylation of PRLr (20). It is plausible that, in cancer cells, sustained activation of MAPK/PI3K pathways by ErbB2 and/or Ras results in a dramatic GSK3β inhibition and PRLr stabilization, which, in turn, self-perpetuates cellular responses to PRL, whereas a transient activation of MAPK/PI3K in benign cells is insufficient for this process. Future delineation of ligand-induced events and their temporal relationships should be important to understand the negative regulation of PRLr signaling in physiologic processes, such as lactation and mammary gland involution.
Besides Ras-activated signaling, a role for other oncogenic pathways capable of inhibiting GSK3β, such as Wnt pathway that affects degradation of β-catenin (45) and cyclin D1 (46), in stabilization of PRLr cannot be ruled out and is under investigation. Future studies will also determine the specific role of PRLr stabilization and associated molecular alterations in the development and progression of various subtypes of human breast cancer and the significance of these mechanisms for prognosis and drug responsiveness of this lethal disease.
Evidence from biochemical, pharmacologic, and genetic studies presented here identify GSK3β as a bona fide Ser349 kinase for PRLr (Fig. 3). Phosphorylation of a specific Ser349-encompassing phosphorylated degron of PRLr by GSK3β enables the recognition of PRLr by βTrcp that recruits other components of the SCFβTrcp E3 ubiquitin ligase to facilitate PRLr ubiquitination and degradation and, hence, negatively regulates PRL signaling (this study and ref. 18). Consistent with data on other GSK3β substrates (e.g., β-catenin) obtained in cells from mice in which GSK3β was genetically ablated (47), our studies using RNA interference approach suggest an additional role of a related kinase, GSK3α, in regulating PRLr Ser349 phosphorylation, as well as PRLr ubiquitination and degradation (Figs. 4 and 5).
An intriguing question is whether phosphorylation of PRL by GSK3 might be regulated by the ligand. Although the current paradigm suggests that GSK3β is constitutively active (48), several reports indicate that GSK3β can be further induced in response to several environmental stimuli, including UV light (49) and insulin (50). Alternatively, phosphorylation of PRLr by GSK3 could be regulated via priming phosphorylation. The PRLr contains several Ser/Thr residues distal to its phosphorylated degron, but the positions of these residues do not correspond to typical canonic GSK3β substrates priming sites at +3 position (45). Nevertheless, mutation of one of these residues, Ser361, led to a decrease in Ser349 phosphorylation in the rate of PRLr degradation.6
A. Plotnikov, Y. Li, and S. Y. Fuchs, unpublished data.
Whereas numerous epidemiologic and experimental data support important roles of PRL signaling in human breast cancers, the mechanisms that lead to constitutive activation of PRLr signaling that occurs in primary human mammary tumors are poorly understood. Data presented here delineate a GSK3β-dependent mechanism by which PRLr might be stabilized in breast cancers as a result of constitutive activation of Ras-dependent oncogenic pathways. Whereas our data also indicate that decreasing the activity of GSK3β augments cellular responses to PRL (Fig. 5C), it remains to be seen whether this regulation affects all branches of hormone-induced signaling or selectively affects specific pathways.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Susan G. Komen Breast Cancer Foundation grant BCTR0504447 (S.Y. Fuchs) and National Cancer Institute grants CA115281 (S.Y. Fuchs) and CA101841 (H. Rui).
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 Z. Ronai, J.A. Diehl, C.V. Clevenger, and the members of his laboratory for their critical comments and discussions and Z. Ronai, A. Eastman, J.W. Harper, D. Bohmann, J. Downward, M. Cole, M. Lemmon, P. Klein, S. Anderson, A.M. Chan, and E. Krebs for the reagents.