Sci. Signal., 24 May 2011
p38α Signaling Induces Anoikis and Lumen Formation During Mammary Morphogenesis
Huei-Chi Wen1,2,3, Alvaro Avivar-Valderas1,2, Maria Soledad Sosa1,2, Nomeda Girnius4, Eduardo F. Farias1, Roger J. Davis4, and Julio A. Aguirre-Ghiso1,2*
1 Department of Medicine, Tisch Cancer Institute at Mount Sinai, Mount Sinai School of Medicine, New York, NY 10029, USA.
Abstract: The stress-activated protein kinase (SAPK) p38 can induce apoptosis, and its inhibition facilitates mammary tumorigenesis. We found that during mammary acinar morphogenesis in MCF-10A cells grown in three-dimensional culture, detachment of luminal cells from the basement membrane stimulated mitogen-activated protein kinase (MAPK) kinases 3 and 6 (MKK3/6) and p38α signaling to promote anoikis. p38α signaling increased transcription of the death-promoting protein BimEL by phosphorylating the activating transcription factor 2 (ATF-2) and increasing c-Jun protein abundance, leading to cell death by anoikis and acinar lumen formation. Inhibition of p38α or ATF-2 caused luminal filling reminiscent of that observed in ductal carcinoma in situ (DCIS). The mammary glands of MKK3/6 knockout mice (MKK3–/–/MKK6+/– ) showed accelerated branching morphogenesis relative to those of wild-type mice, as well as ductal lumen occlusion due to reduced anoikis. This phenotype was recapitulated by systemic pharmacological inhibition of p38α and β (p38α/β) in wild-type mice. Moreover, the development of DCIS-like lesions showing marked ductal occlusion was accelerated in MMTV-Neu transgenic mice treated with inhibitors of p38α and p38β. We conclude that p38α is crucial for the development of hollow ducts during mammary gland development, a function that may be crucial to its ability to suppress breast cancer.
Introduction Back to Top
The creation of ducts and alveoli containing luminal spaces surrounded by polarized epithelial cells attached to a basal lamina is a hallmark of postnatal mammary gland development (1). Studies of mammary epithelial cells (MECs) in three-dimensional (3D) cultures and of mammary gland development in vivo have shown that lumen formation results from the removal of cells that have detached from the extracellular matrix (ECM) in the terminal end bud (TEB) and in elongating mammary ducts (2). Lack of cell-ECM adhesion elicits an acute "stress signaling" response that results in caspase-dependent (anoikis) (3) and caspase-independent (4, 5) cell death. Anchorage-independent cell survival and enhanced proliferation are conferred by proteins encoded by oncogenes, such as human epidermal growth factor receptor 2 (known as HER2/neu). Enhanced survival can also be conferred by overexpression of antiapoptotic proteins, such as B cell lymphoma 2 (Bcl-2). Both proteins promote the accumulation of luminal cells characteristic of breast atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS). Clinically, a reduced response to oncogene- and replication-induced stress that results in senescence predicts subsequent tumor formation and worse outcome in DCIS patients (6). However, the precise stress signaling pathways that might be deregulated to promote resistance to anoikis and thereby lumen filling are not fully understood.
The p38 mitogen-activated protein kinase (MAPK) pathway integrates various types of stress signals and mediates anoikis in normal epithelial cells (7–9). Four mammalian isoforms of p38 [p38α (MAPK14), p38β (MAPK11), p38 (MAPK12), and p38 (MAPK13)] have been identified (10). p38α promotes growth arrest by inhibiting cyclin D1 gene transcription and protein abundance (11) and by activating the p53 to p21 and p16 to Rb pathways (11–14). p38α also regulates the spindle assembly checkpoint (15) and can delay the G2 to M transition of the cell cycle (16). It also inhibits tumor initiation by sensing oncogene-induced oxidative stress (17). Accordingly, inactivation of p38α leads to mouse mammary tumorigenesis in vivo (13, 18), and ~15% of human primary breast carcinomas show amplification of PPM1D, the gene that encodes protein phosphatase Mg2+/Mn2+-dependent 1D, which dephosphorylates p38α, thereby inhibiting its activity (12). However, the location and timing of p38 signaling required for mammary tumor suppression remain unknown. The p38 pathway, which inhibits signaling through the ERK1/2 (extracellular signal–regulated kinase 1 and 2) MAPK pathway, is activated by loss of adhesion in MECs (7) and induces anoikis in colonic epithelial cells (8, 9). We thus hypothesized that p38-dependent inhibition of ERK1/2 signaling and thereby induction of anoikis might be central mechanisms to limit the accumulation of luminal cells during mammary morphogenesis.
Here, we used 3D cultures of immortalized nontumorigenic human MCF-10A MECs to show that p38α-mediated anoikis acts as an early barrier to prevent the inappropriate survival and growth of luminal cells. Upon cell detachment, MAPK kinase 3 and 6 (MKK3/6)–mediated phosphorylation of p38α led to activation of ATF-2 (activating transcription factor 2) and induced c-Jun–dependent transcription of the proapoptotic gene Bim. This led to the induction of caspase-3–dependent anoikis of the detached cells and lumen formation in mammary acini. MECs hypomorphic for p38α were incapable of activating ATF-2 and stimulating c-Jun protein increase and the subsequent increase in BimEL (the extra-long form of Bcl-2–interacting mediator of cell death) mRNA and protein, resulting in resistance to anoikis and luminal filling. This phenotype was recapitulated by ATF-2 knockdown. Further, occlusion by excess luminal cells of the mammary gland ducts and TEBs was apparent in MKK3/6 knockout (MKK3–/–/MKK6+/–) mice and in mice treated with p38α/β inhibitor but not in wild-type controls. Systemic inhibition of p38 had marked effects in transgenic mice in which the Her2/neu oncogene is under the control of the mouse mammary tumor virus promoter (MMTV-Neu), which developed DCIS-like lesions after only 2 weeks of treatment. Our data reveal how p38α signaling shapes normal mammary acinar morphogenesis and inhibits HER2/neu-driven tumorigenesis. Our data also define at what stage of mammary gland development p38α might act to suppress tumorigenesis and how its inhibition could accelerate disease progression.
Results Back to Top
Loss of ECM attachment activates MKK6-p38α signaling and anoikis in MCF-10A cells
Integrin and growth factor signaling become uncoupled in MCF-10A cells grown in suspension, initiating a stress signal that results in cell death (19). Consistent with previous studies (7), immunoblot (IB) analysis indicated that p38 phosphorylation was increased in detached MCF-10A cells relative to that in attached cells (Fig. 1A). We also observed activation of p38 in primary mouse MECs (mMECs) and immortalized mouse embryonic fibroblasts (MEFs) grown in suspension (Fig. 1, A and C). When focusing on MCF-10A cells, we found that increased p38 phosphorylation was accompanied by phosphorylation of its upstream activators MKK3 and 6 (MKK3/6) and of its downstream target, the heat shock protein 27 (HSP27) (Fig. 1A), confirming activation of the p38 signaling pathway. p38 was not activated by centrifugation or trypsinization of cells (fig. S1A). Further, blocking β1-integrin ligand binding in attached cells with the AIIB2 monoclonal antibody (20) increased p38 phosphorylation to a degree comparable to that induced by growth in suspension (Fig. 1B). The phospho-p38 (Thr180/Tyr182)–specific antibody we used to assess p38 phosphorylation detects all isoforms of activated p38. Using antibodies selective for the different p38 isoforms, we found relatively similar amounts of endogenous p38α, p38β, p38, and p38 in MCF-10A cells (fig. S1B). We focused mainly on p38α, and its role in anoikis and mammary morphogenesis.
To identify the upstream MAPK kinase responsible for activating p38 in cells grown in suspension, we assessed p38 phosphorylation in MEFs derived from wild-type, MKK3–/–, MKK6–/–, or MKK3–/–/6–/– mice and grown on fibronectin-coated plates or for 24 hours in suspension culture. In adhered conditions, these cells produced both the p38α and the p38β isoforms (fig. S1C). MKK6 phosphorylates both isoforms, but predominantly p38α, whereas MKK3 phosphorylates only p38α and not p38β (21, 22). Ablation of MKK3 had no obvious inhibitory effect on p38 phosphorylation in suspended cells, but loss of MKK6 or of both MKK3 and MKK6 markedly inhibited it (Fig. 1C); these findings suggest that both MKK3 and MKK6 contribute, albeit to different degrees, to activation of p38α in response to cell detachment. In agreement with the MEF data, treatment of detached MCF-10A cells with SB203580 [which inhibits both the p38α and the p38β isoforms but not p38 or p38 (23, 24)] decreased HSP27 phosphorylation (Fig. 2A). Treatment 24 hours before and throughout the assay with SB203580 also significantly inhibited anoikis after 24 and 48 hours in suspension, although this inhibition was no longer apparent at 72 hours (Fig. 1D). Together, these data demonstrate that detachment from ECM can activate the MKK3/6 to p38α to HSP27 pathway in MCF-10A cells.
Activation of p38α mediates acinar lumen formation by inducing BimEL and anoikis
Activation of p38α signaling can induce anoikis or growth arrest (9, 25). We studied anoikis and growth arrest during mammary acinar morphogenesis using MCF-10A 3D cultures (26). As previously reported (27), MCF-10A cells formed acini in 3D culture and, by day 15 of morphogenesis, when luminal cells are cleared through anoikis, MCF-10A acini formed lumens (Fig. 1E). Treatment with SB203580 or small interfering RNA directed against p38α (siRNAp38α) blocked luminal clearing, resulting in acini filled with cells (Fig. 1E). Proliferation, as measured by Ki67 staining, was not significantly increased (Fig. 1F), but acini were slightly larger (~30%) (Fig. 1G). We conclude that the primary function of p38α during MCF-10A morphogenesis is to promote lumen formation.
BimEL can be transcriptionally activated by p38 after different stress signals (28, 29), and it is required for lumen formation (7, 30). As previously shown (19), cell detachment increased BimEL abundance. We found that increased BimEL abundance correlated with robust phosphorylation of p38α and HSP27 (Fig. 2A). siRNAp38α or pharmacological inhibition of p38α/β with SB203580 or SCIO469 (specific for p38α) (31, 32) almost completely reversed the increase in Bim mRNA and protein in suspended MCF-10A cells (Fig. 2A). Using a Bim-luciferase (Bim-Luc) reporter construct (33), we determined that RNA interference (RNAi) knockdown of p38α in MCF-10A cells or deletion of p38α in MEFs (34) decreased Bim promoter activation in cells grown in suspension (Fig. 2B). Expression of a constitutively active mutant form of p38α (p38αCA) (35, 36) stimulated the Bim promoter to the same extent as did cell detachment (Fig. 2B), indicating that p38 activation is sufficient to activate Bim gene transcription. In addition, the increased luciferase activity in response to p38αCA paralleled increases in endogenous Bim mRNA, indicating that the Bim-Luc assay is a faithful reporter of Bim expression (Fig. 2B). In 3D culture, SB203580-treated acini showed significantly less BimEL than did untreated cells (Fig. 2C). This decrease in BimEL correlated with reduced luminal apoptosis (detected by cleaved caspase-3 staining) in acini formed by SB203580- or siRNAp38α-treated cells at day 8 and day 10 of morphogenesis, respectively, relative to cells treated with empty vehicle or control siRNA (Fig. 2, D and E, and fig. S1D). Together, these data indicate that p38α-regulated expression of Bim is associated with lumen formation during mammary acinar morphogenesis.
ERK1/2 and p38α have opposing effects on BimEL abundance
ERK and p38 have opposing effects on apoptosis (37–39) and, whereas p38 activation increases BimEL abundance, ERK1/2 reduced BimEL protein accumulation (28, 29, 40). We hypothesized that a signaling imbalance favoring p38α over ERK1/2 could increase BimEL abundance in detached luminal cells. Conversely, a high ERK1/2-to-p38α signaling ratio in ECM-attached cells might decrease BimEL induction.
Either treatment with the MEK1/2 (mitogen-activated or extracellular signal–regulated protein kinase kinase 1 and 2) inhibitor U0126 to decrease ERK signaling (Fig. 3A), or activation of p38α signaling by expressing either a constitutively active form of MKK6 [Mkk6b(E)] or p38αCA increased BimEL abundance in adherent MCF-10A cells incubated in full growth media with 5% horse serum (conditions of high basal ERK1/2 activation) (Fig. 3, A and B). SB203580 treatment of adherent cells increased ERK1/2 phosphorylation and prevented induction of BimEL by U0126 so that its abundance did not increase beyond that found under basal conditions (Fig. 3A). Moreover, U0126-mediated induction of cell death was reversed by SB203580 treatment (fig. S1F). These data suggest that, upon inhibition of MEK1/2-ERK signaling, p38 activity is required to induce BimEL production and apoptosis. Furthermore, a p38α siRNA inhibited Mkk6b(E)-dependent BimEL induction (Fig. 3B). This and the fact that p38 phosphorylation was greatly reduced further support the notion that MKK6 activation of primarily p38α (the phospho-p38 antibody detects all four isoforms) leads to increased BimEL abundance (Fig. 3B). The increase in BimEL apparent in p38αCA-transfected MCF-10A cells was enhanced by increasing concentrations of U0126 (Fig. 3C). This additive effect of inhibiting ERK1/2 and activating p38α suggests that the two pathways converge to regulate BimEL abundance.
We next investigated whether mutual regulation of ERK1/2 and p38α activities might be required to regulate BimEL production. Consistent with the effects of SB203580 treatment in MCF-10A cells, we found that U0126 did not increase BimEL protein abundance in p38α–/– MEFs, whereas it did in wild-type cells (Fig. 3D). Loss of BimEL induction in suspended p38α–/– MEFs was rescued by transient expression of a wtp38α vector (Fig. 3E), indicating that the lack of BimEL induction was not an epiphenomenon of p38α deletion. Detachment of wild-type or p38α–/– MEFs (Fig. 3F) or of MCF-10A cells transfected with control siRNA or siRNA targeting p38α (Fig. 2A) confirmed that cells hypomorphic for p38α showed decreased induction of BimEL in response to growth in suspension. SB203580 also inhibited BimEL induction in suspended wild-type MEFs (Fig. 3G), arguing that our pharmacological and genetic approaches specifically target p38α or p38α/β. Together, these data indicate that p38α is essential for the increase in BimEL caused by inhibition of ERK1/2 signaling. We also found sustained ERK1/2 phosphorylation in detached MCF-10A cells treated with SB203580 or transfected with siRNA to MKK6 (Fig. 3H), which suggests that activation of MKK6 and p38 signaling is in part responsible for inhibition of the ERK1/2 pathway. Furthermore, the activation of ERK1/2 observed upon inhibition of p38α in MCF-10A cells (Fig. 3, A and H) was also observed in adhered p38α–/– (Fig. 3F) and in MKK6–/– MEFs (Fig. 3H) where loss of p38α or MKK6 signaling caused enhanced ERK1/2 phosphorylation. However, we did not observe this effect in p38α–/– cells in suspension. This may be due to differences between deleting p38α and inhibiting its kinase activity or to the involvement of other p38 isoforms. Nonetheless, p38α was required for BimEL induction in all cases (Fig. 3, B to F). Inhibition of ERK1/2 signaling may, by enhancing Bim expression (19) and protein stability (40), cooperate with p38α to increase BimEL protein abundance in detached cells.
Activation of ERK1/2 and p38α in MCF-10A cells grown in 3D culture was observed in separate populations of cells. ERK1/2 phosphorylation was confined to the ECM-bound cells; phosphorylated p38 (p-p38) was detected exclusively in matrix-deprived luminal cells at the center of the acinar structures (Fig. 3I). Low-frequency detection (one to two cells per acinus) of cells with p38α phosphorylation was consistently found in the acinar structure throughout days 6 to 10 in culture and lost thereafter (Fig. 3I and fig. S1G). This is most likely due to the transient nature of these phosphorylation events (Fig. 3I and fig. S1G). Thus, whereas ERK1/2 activation is restricted to ECM-attached cells, phosphorylation of p38 specifically occurs in luminal cells and is apparent at all stages of morphogenesis during which apoptosis occurs. Together, these data indicate that a low ratio of ERK1/2 to p38α signaling is required for the increase in BimEL and, consequently, for anoikis of luminal cells and thereby lumen formation during mammary acinar morphogenesis.
ATF-2 and c-Jun mediate p38α-dependent expression of BimEL during mammary acinar morphogenesis
An increase in BimEL abundance relative to that in attached cells was apparent at 0.5 hours of suspension culture. BimEL abundance remained high for 4 hours and decreased slightly at 8 hours (Fig. 4A). These changes in protein abundance were not accompanied by significant changes in mRNA abundance (Fig. 4B). At 8 hours and up to 16 hours, protein abundance increased steadily, as did mRNA (Fig. 4, A and B). At 24 hours, although Bim mRNA abundance was lower than at 16 hours, it was still more than twice that in attached cells, and this was accompanied by increased protein abundance. Changes in the abundance of p-p38, p-ERK, and p-HSP27 were apparent for all three proteins at 4 hours and thereafter (Fig. 4, A and B). These data suggest that loss of attachment triggers a transient increase in BimEL stability, possibly due to posttranslational modifications (41, 42), followed by a sustained transcriptional activation initiated by a decrease in the ERK1/2/p38 activity ratio.
We hypothesized that p38α-dependent activation of specific transcription factors (TFs) might stimulate the increase in Bim expression apparent 8 hours and beyond in suspension. Neither Forkhead box O3a (FOXO3a) nor C/EBP homologous protein (CHOP), two TFs that stimulate Bim expression and apoptosis in response to stress (43, 44), was responsible for Bim mRNA induction in suspension (fig. S1, H and I). Therefore, we focused on ATF-2, a TF that is phosphorylated at Thr71 and Thr69 in response to activation of c-Jun N-terminal kinase (JNK) and p38α, which in turn increases c-Jun transcription (45–47). Both ATF-2 and c-Jun can also heterodimerize and, along with other components of activating protein 1 (AP-1), they stimulate the transcription of genes that promote cell death (48, 49). Mice lacking an ATF-2 allele develop invasive ductal mammary carcinomas that display a solid tubular structure (50). We found that detached MCF-10A and mMEC cells strongly induced ATF-2 phosphorylation (Fig. 4C). When focusing on MCF-10A cells, we found that suspension-induced ATF-2 phosphorylation also correlated with increases in c-Jun and BimEL abundance (Fig. 4C). ATF-2 phosphorylation and the increase in c-Jun abundance were not observed in detached p38α–/– MEFs or in siRNAp38α-transfected MCF-10A cells (Fig. 4, D and E); these findings suggest that p38α played a critical role in activating ATF-2 and inducing c-Jun production in suspended cells.
Both ATF-2 phosphorylation and c-Jun abundance were increased in cells located at the center of the acini that were not attached to the ECM (Fig. 5A), and SB203580 treatment decreased ATF-2 phosphorylation (Fig. 5B); hence, in suspended MCF-10A cells, ATF-2, c-Jun, or both might mediate luminal cell anoikis. To determine whether c-Jun was responsible for BimEL induction, we briefly incubated suspended MCF-10A cells transfected with control or c-Jun expression vector. We detected c-Jun and BimEL as early as 2 to 4 hours in suspension (Fig. 5C), at a time when transcriptional activity is submaximal (Fig. 4B). Next, we determined whether inhibition of c-Jun or ATF-2 could block the increase in BimEL in detached cells. Transfection of a dominant-negative form of c-Jun, TAM-67 (51), inhibited the increase in BimEL (Fig. 5C) and of its mRNA (fig. S1J) and activation of its promoter (Fig. 5C). Similarly, knockdown of ATF-2 inhibited activation of the Bim promoter in suspended cells (Fig. 5D). Thus, ATF-2 and c-Jun appear to mediate increased Bim expression in response to p38α signaling in cells detached from the ECM, and thereby potentially contribute to acinar lumen formation.
To determine their role in lumen formation conclusively, we knocked down ATF-2 during acinar morphogenesis (Fig. 5E) and found that this prevented proper lumen formation in enlarged MCF-10A acinar structures (Fig. 5E). These structures showed significant reduction of BimEL and cleaved caspase-3 staining in cells filling the lumen (Fig. 5E). Our monolayer and 3D studies in MECs show that activation of ERK1/2 and p38α is temporally and spatially regulated so that in luminal cells, p38α activation is mutually exclusive with that of ERK1/2. The increase in p38 activity leads to activation of ATF-2 and production of c-Jun, which, combined with the loss of ERK1/2 activity, induce BimEL and anoikis of luminal cells.
p38α/β inhibition triggers luminal filling in ducts and TEB in normal and MMTV-Neu mammary glands
MMTV-PPM1D mice have been studied in the context of HER2/neu signaling in tumor development studies (52). However, when and how p38 signaling might antagonize the development of HER2/neu-induced tumors was not studied. Furthermore, whether PPM1D overexpression affects specific stages of normal mammary gland morphogenesis in vivo in the absence of HER2/neu through inhibition of p38 signaling was not reported. To determine this, we studied mammary ductal branching and lumen formation in a strain of normal inbred FVB female mice treated with SB203580. We also compared mammary gland development in the C57B strain of wild-type female or MKK3–/–/MKK6+/– C57B mice, where p38 activation is greatly decreased in many tissues (53, 54).
We injected 4-week-old FVB female mice with vehicle or SB203580 (10 mg/kg) intraperitoneally every 48 hours (12, 13) for 2 or 4 weeks. Analysis of stained whole-mount mammary glands showed that SB203580 accelerated ductal tree elongation and branching (Fig. 6, A and B). Close examination revealed thickened ducts and elongated solid TEBs in SB203580-treated glands, in contrast to the well-defined hollow ducts and normal TEB structures found in the control group (Fig. 6, C and D). Histological analysis of hematoxylin and eosin (H&E)–stained sections (Fig. 6, E to J) confirmed the whole-mount analysis. In ducts, the myoepithelial and epithelial cell layers showed increased cellularity after p38 inhibition, and occupancy of lumens by epithelial cells was common (Fig. 6, G to J, and fig. S2A). The same pattern of myoepithelial and epithelial cell layer organization was corroborated with immunofluorescence to detect α smooth muscle actin (α-SMA, as a marker of myoepithelium) and cytokeratins 8 and 18 (CK8/18, as a marker of epithelium) (fig. S2A). Detection and quantification of cleaved caspase-3 revealed predominantly apoptotic luminal cells in normal ducts (Fig. 7O). SB203580-treated mice showed a significant reduction in apoptosis of luminal cells (Fig. 7O). All of the above characteristics of SB203580-treated mice were recapitulated in mice carrying homozygous or heterozygous deletion of the MKK3 and MKK6 alleles in all tissues (53, 54) (Fig. 6, K to T); mice homozygous for deletion of both MKK3 and MKK6 are embryonic lethal (53). MKK3–/–/MKK6+/– mice (5 weeks old) showed reduced ATF-2 phosphorylation in epithelial and stromal cells, confirming reduction of MKK3/6-p38 signaling in these tissues (fig. S2C). These mice also showed a marked acceleration of ductal tree expansion, so that it almost completely filled the mammary fat pad (Fig. 6, K and L). Additional analysis (Fig. 6, O to T) showed that the ducts and TEBs of MKK3–/–/MKK6+/– mammary glands, unlike those of wild-type mice, had filled lumens (Fig. 6, R to T). This correlated with reduced apoptosis in MKK3–/–/MKK6+/– mice as determined by cleaved caspase-3 staining (Fig. 7O). There was a nonsignificant trend toward reduced percentage of small- and medium-sized TEBs in SB203580-treated mice or MKK3–/–/MKK6+/– mice relative to their respective controls and a significant increase in the percentage of larger-size TEBs (fig. S2D). p38α/β inhibition had no effect on cell proliferation, as measured by p-Rb (Ser807/811) or p-H3 (Ser10) staining in situ (fig. S3), in either the pharmacological or the genetic analyses. Thus, using two different strains of mice and two different experimental approaches, we showed that MKK3, MKK6, and p38 signaling regulate mammary branching morphogenesis and restrict growth of epithelial and myoepithelial cells and TEB size. At the time points analyzed, these effects coincided with p38-induced luminal anoikis, but not with inhibition of proliferation.
Next, we analyzed whole-mount sections of mammary glands of uniparous MMTV-Neu (wtp38α) mice (20 to 24 weeks old) treated with vehicle control or SB203580. After only 2 weeks of treatment, the number of ducts had increased by 40% (P = 0.004, Mann-Whitney test) in the SB203580 group (n = 5 per group) (Fig. 7, A, B, and P); this was accompanied by an increased ductal density and an increase in the size and number of side buds and TEBs (Fig. 7, C to F). Histological analysis showed occluded lumens in ductal structures (48 ± 0.6% control versus 72 ± 19% SB203580-treated mice, P < 0.05), with a substantial increase in the thickness of the epithelial and myoepithelial cell layers, which was confirmed by α-SMA and CK8/18 immunofluorescence (Fig. 7, G to J, and fig. S2B). MMTV-Neu control mice showed decreased luminal apoptosis relative to control FVB mice (Fig. 7O). The increased number of occluded ducts correlated with a 58% (P = 0.027) decrease in the percentage of luminal apoptotic cells (determined by cleaved caspase-3 staining) in SB203580- versus control-treated mice (Fig. 7, K, L, and O). As in normal mammary epithelium, p38α/β inhibition with SB203580 did not enhance in situ cell proliferation (fig. S3). We also observed hyperplastic accumulations of intermingled epithelial and myoepithelial cells emerging from ductal walls in SB203580-treated mice (Fig. 7, I and J, and fig. S2B). Many of these structures were solid and highly disorganized, resembling dysplastic foci or early neoplastic lesions (Fig. 7, D, F, H, and J). As indicated earlier, when during mammary gland development p38 inhibits HER2/neu tumor formation has been unclear. Here, we conclude that MKK3/6 and p38α/β are critical regulators of proper branching morphogenesis and that, at least at the time points we tested, p38α/β might restrict HER2/neu-induced tumors at the time of hyperplasia development. This occurs by promoting luminal cell apoptosis, but not by decreasing cell proliferation. Our in vivo data thus confirm the results of the 3D morphogenesis assay and show that p38α limits luminal filling, excessive side branching, and tissue expansion in mammary epithelium expressing the HER2/neu oncogene.
Discussion Back to Top
Here, we show that p38α has a previously unrecognized role in promoting anoikis of luminal cells during acinar morphogenesis, thereby preventing the accumulation of luminal cells. We found that either pharmacological or genetic inhibition of the p38 stress-signaling axis resulted in resistance to apoptosis, leading to disruption of the normal acinar architecture. These data also support a role for both MKK3 and MKK6 in controlling lumen formation and maintaining normal architecture of the mammary epithelium. Consistent with previous studies (27, 30), our data show that anoikis is required to prevent lumen occlusion in ducts and TEBs and that lack of apoptosis—without any obvious increase in cell proliferation—has a marked effect on ductal tree expansion. It is possible that inhibition of p38 signaling affected stem and progenitor cells in the mammary tissue, contributing to the enhanced branching morphogenesis. If these and other more differentiated cells were protected from apoptosis by p38 inhibition, then the net effect would be larger than anticipated, because all surviving cells would contribute to the expanding mammary epithelium.
We found that inhibition of ERK1/2 signaling in ECM-detached cells mediated BimEL induction and anoikis. We also found that the balance between ERK1/2 and p38α signaling, which controls tumor cell entry into quiescence (55), liver development (56), and the expression of genes required for suppressing transformation and tumorigenesis (17, 57), provides an important signal integration point for cells to commit to survival and proliferation or to anoikis. However, whether ERK1/2 inactivation is solely due to lack of growth factor and ECM signaling has been unknown. Our results show that MKK6 and p38α activation in cells grown in suspension is necessary for inhibition of ERK1/2. We further showed that the combined inhibition of ERK1/2 activity and p38α activation contribute to the expression of Bim. Results by Reginato et al. (19) suggest that reduced ERK1/2 activity contributes to Bim mRNA induction. ERK1/2 is thought to inhibit BimEL expression at the transcriptional (58) and also at the posttranslational level (41). p38α phosphorylation of BimEL at Ser65 has been proposed as a mechanism for increasing BimELs apoptotic function, but whether this was only due to enhanced protein stability was not shown (59). In contrast, phosphorylation of BimEL by ERK1/2 at Ser65 increases protein degradation (40, 60). The phosphorylation by ERK1/2 depends on the previous phosphorylation of BimEL by p90 ribosomal S6 kinase (p90RSK), an ERK1/2 but not p38 target. This dual phosphorylation by ERK1/2 and p90 RSK targets BimEL for degradation (61). Here, we show that ECM detachment induces a biphasic regulation of BimEL abundance in MCF-10A cells. The first phase is most likely posttranslational and accounts for the increase in BimEL abundance up to 8 hours in suspension. This could be due to the loss of ERK phosphorylation at Ser65 (Ser69 in human) among other mechanisms. After 8 hours, the much stronger increase in BimEL abundance depends on enhanced transcription. Although posttranslational regulation in suspended cells may still occur at times beyond 8 hours, it seems that both transcriptional and posttranslational mechanisms ensure that the increase in BimEL abundance is sufficient to commit cells to anoikis.
The TFs that control increased Bim expression during anoikis have been unknown. We now show that neither FOXO3a nor CHOP appears to increase Bim transcription in cells cultured in suspension, whereas ATF-2 and c-Jun act as TFs downstream of ERK and p38 to increase Bim transcription. Both steady-state measurements of Bim mRNA and Bim-Luc reporter activity support the notion that p38α increases the transcription of Bim through ATF-2. Mice hypomorphic for ATF-2 develop lumen-occluded invasive ductal mammary carcinomas (50), suggesting a role for this TF in lumen formation. We found that ATF-2 knockdown phenocopied the effects of p38α abrogation. Our results indicate that ATF-2 phosphorylation by p38α is required to increase the abundance of c-Jun in luminal cells destined to undergo apoptosis. Arguably, ATF-2 can also be pro-tumorigenic (62), and the role of c-Jun downstream of p38α depends on the cell tissue context. For example, in squamous carcinoma cells where p38α/β was inhibited (63), or in livers or MEFs of p38α–/– mice, c-Jun abundance and phosphorylation are increased (64). However, in mouse lungs in which p38α is conditionally deleted, c-Jun phosphorylation is enhanced but not its abundance (65). Nevertheless, our data are consistent with previous results (50) indicating a critical role for ATF-2 in regulating acinar lumen formation. Knockdown of ATF-2 caused a more profound effect on acinar enlargement than did loss or inhibition of p38α, suggesting that multiple stress signals might converge on ATF-2. Moreover, our 3D morphogenesis data correlated with the role of p38 in regulating luminal apoptosis in ducts and TEBs in vivo. Surprisingly, genetic inhibition of MKK3/6 or pharmacologic inhibition of p38α/β increased the number of epithelial and myoepithelial cells without any apparent increase in proliferation. It is possible that the pro-proliferative effect of p38α/β inhibition may occur at earlier times during mammary gland development. Although our data analysis focused on the p38α and β isoforms, we cannot eliminate the possibility that p38 and p38, both of which are substrates of MKK3 and MKK6 (66), might also play a role in inducing mammary epithelial apoptosis in vivo.
In cancer, p38 appears to execute various and sometimes opposing programs, which might depend on the state of oncogenic progression (17, 63, 67). Despite these complexities, the tumor-suppressive function of p38 is well established (68). Although p38 does not appear to be mutated in cancer, its tumor-suppressive functions could be abrogated by different means. For example, in breast cancer, amplification of Wip1, a p38 phosphatase, appears to circumvent p38s antitumor effects. Genetic or pharmacologic inhibition of p38α/β isoforms accelerates mammary tumor progression (13, 52). We explored the possibility that p38-mediated anoikis of luminal cells during mammary acinar development could be a crucial point at which p38 might act to curtail breast cancer development. In MMTV-Neu mammary tissue, SB203580 treatment accelerated epithelial tissue growth and the development of hyperplastic ducts with occluded lumens, supporting the notion that p38α/β signaling could limit the development of Her2/neu-induced hyperplasia. The massive expansion of the ductal tree in MMTV-Neu mammary tissue upon p38α/β inhibition indicates that p38α/β acts to restrain HER2/neu signaling. Thus, upon p38α/β inhibition, unrestricted expansion occurs, perhaps through the same mechanisms as in the FVB mice treated with SB203580 or in the MKK3–/–/MKK6+/– mice.
In summary, our data shed light into the morphogenetic and temporal windows (for example, during branching morphogenesis and lumen formation and maintenance) during which p38α/β activation might block ductal filling, and the development of hyperplasia. p38α/β signaling, which mediates apoptosis and growth arrest, increases with aging as do that of tumor-suppressive cell cycle inhibitors that limit unscheduled proliferation (69). Thus, it is tempting to speculate that loss of p38 signaling could facilitate survival and proliferation of immortalized or transformed ductal or alveolar luminal cells. Future studies will determine whether these events indeed precede the development of lesions during early steps of tumorigenesis.
Materials and Methods Back to Top
Reagents and plasmids
SB203580 was purchased from Calbiochem and LC Laboratories. SCIO469 was a gift from A. Verma (Albert Einstein Cancer Center, New York). MEK1 inhibitor U0126 was purchased from Promega.
Lipofectamine RNAiMax was purchased from Invitrogen and FuGENE HD from Roche. The Dual-Luciferase Assay kit was purchased from Promega. The expression plasmid MKK6b(E) was described previously (63). The p38αD176A + F327S construct was a gift from O. Livnah (The Hebrew University of Jerusalem, Givat Ram, Jerusalem). c-Jun and TAM-67 were gifts from M. J. Birrer [Center for Cancer Research, National Cancer Institute (NCI), Bethesda, MD]. The pGL3-BimP-luc reporter plasmid was a gift from L. A. Greene (Columbia University, New York, NY).
p38α MAPK siRNA II was purchased from Cell Signaling Technology, MKK6 from Santa Cruz Biotechnology, ATF-2 from Dharmacon, and CHOP and Silencer negative control from Ambion.
Antibodies used in this study were directed against p38α, p-p38 (T180/Y182), and ERK1 (BD Biosciences); p-ERK1/2 (T202/Y204) and CHOP (Santa Cruz Biotechnology); GAPDH (glyceraldehyde phosphate dehydrogenase) and Bim (Calbiochem); HA (hemagglutinin; Roche); Ki67 (Zymed); p-MKK3/6 (S189/207), p-p38 (T180/Y182), p-HSP27 (S82), p-Rb (S807/811), p-histone H3 (S10), ATF-2, p-ATF-2 (T71), cleaved caspase-3 (N175), p38β, p38, p38, and c-Jun N-terminal–specific (Cell Signaling Technology); and c-Jun DNA binding domain–specific (Santa Cruz Biotechnology). In addition, horseradish peroxidase–conjugated (Vector Laboratories and Chemicon International) and Alexa Fluor–conjugated secondary antibodies (Molecular Probes) and polyclonal antibody to cytokeratins 8 and 18 (Progen Biotechnik) were used. Monoclonal antibody against α-SMA was a gift from A. Bernstein (Mount Sinai School of Medicine, New York, NY).
Forward and reverse primer sequences were as follows: Bim as previously described (21); human p38α, 5'-TCCAGACCATTTCAGTCCAT-3' and 5'-AAAAACGTCCAACAGACCAA-3'; mouse p38α, 5'-CCCCAGAGATCATGCTGAAT-3' and 5'-AGGTCAGGCTCTTCCACTCA-3'; human p38β, 5'-TACTTGGTGACCACCCTGAT-3' and 5'-GCTGGTAAACCAGGAATTGA-3'; mouse p38β, 5'-ATTCTACCGGCAAGAGCTGA-3' and 5'-GTCCTCGTTCACCGCTACAT-3'; human p38, 5'-TGATGAGACCCTGGATGACT-3' and 5'-TCGCCTAGCTTCTCATGTTT-3'; mouse p38, 5'-AGGCAGGCAGACAGTGAGAT-3' and 5'-AGGGTGCGGTCTACATCATC-3'; human p38, 5'-ATCCTCAGCTGGATGCACTA-3' and 5'-CCCCTTGAACAGAGTTTTCC-3'; and mouse p38, 5'-ACAAGACTGCCTGGGAGCTA-3' and 5'-CCCAAAGTCCAGGATCTTCA-3'.
Cells, culture conditions, and 3D morphogenesis assays
MCF-10A and MEF monolayer cell culture was maintained as described (26, 53). MCF-10A cells have been described previously (70). Wild-type and knockout MKK3, MKK6, and MKK3/6 MEFs were previously described (26, 53), and wild-type and knockout p38α MEFs were a gift from A. Porras (Universidad Complutense de Madrid). For suspension assay, MCF-10A cells were pretreated overnight with or without 10 µM SB203580 or SCIO469, or transfected with negative control or specific siRNA for 48 hours in complete growth media. Subsequently, MCF-10A or MEF cells (4 x 105/ml) were incubated in ultra-low attachment plates (Corning), kept in their respective complete growth medium with or without treatment, and harvested at the indicated time points. To quantify percentage of cell viability, detached cells were washed with phosphate-buffered saline, disaggregated, and collected as single-cell suspensions with cell-strained cap (BD Falcon) to be incubated in 1:2 trypan blue stain (BioWhittaker). The number of total and nonviable cells was determined with a counting chamber. MCF-10A 3D morphogenesis assay was carried out as described (26). For treatments and transfections during morphogenesis, 5 µM SB203580 or 20 nM p38α or 60 nM ATF-2 and MKK6 siRNA was supplemented to the assay media every 24 or 48 hours, respectively.
RNAi and complementary DNA transfections and luciferase reporter assay
Transfections of plasmids and siRNA oligonucleotides were done in six-well plates with 2 µg of DNA mixed with 6 µl of FuGENE HD or 3.2 nM (p38α) or 9.6 nM (ATF-2 and MKK6) siRNA oligonucleotides in 8 or 22.5 µl of Lipofectamine RNAiMax per well, respectively, and incubated for 24 and 48 hours, following the manufacturers instructions. For reporter assay, parental or transfected MCF-10A cells (described above) and MEF cells were cotransfected with 0.5 µg of Bim reporter plasmid and 0.1 µg of Renilla vector in 2.5 ml of complete growth media in six-well dishes with FuGENE HD. Twenty-four or 48 hours later, cells were washed, lysed, and harvested with buffer provided in the Promega Luciferase System. Relative luciferase activities were obtained with a TD-20/20 luminometer (Turner Designs) and analyzed by normalizing the luciferase activity to Renilla luciferase activity.
Western blotting, immunofluorescence, and image processing
MCF-10A and MEF cells were lysed, and protein was analyzed by immunoblotting as previously described (70). MCF-10A 3D acinar structures were fixed at day 6, 8, 10, or 15 and processed for size and immunofluorescence microscopy analysis as previously described (70). Detection of phosphoproteins was improved by incubating, fixing, and treating acinar structures with phosphatase inhibitors. Confocal analyses were performed with the Leica SP5 DM confocal microscopy system equipped with four lasers: an ultraviolet (UV) diode (405 nm), an argon laser (458, 476, 488, and 514 nm), a 543-nm HeNe laser, and a 633-nm HeNe laser. Pictures of luminal spaces from the equatorial section (the largest diameter from top to bottom) of mammary spheres were taken with 63x magnification. Quantitative measurements of optical density in IB were performed with ImageJ image processing program by National Institutes of Health. Numbers under the IB bands show fold change for the indicated protein ratio and are expressed as means ± SEM. Values for control conditions were set at 1. When indicated by an asterisk (*), the differences were statistically significant at P < 0.05. All IB results are representative of at least three independent experiments (n = 3). Acinar size was calculated with SPOT software following the equation [(length x width2)/2 = acini volume (mm3)] and plotted with GraphPad Prism.
In vivo experiments, mammary gland whole mounts, and immunohistochemistry
Four-week-old FvB (NCI) female mice were injected intraperitoneally with dimethyl sulfoxide (DMSO) (five mice) or SB203580 (five mice, 10 mg/kg) every 48 hours for 4 weeks. SB203580 compound from two sources (see Reagents and plasmids section) was tested in vivo and in vitro, and no apparent differences in efficacy were observed. The same strategy was followed for 24- to 32-week-old MMTV-Neu (71) female mice. C57 MKK3+/+/MKK6+/+ and MKK3–/–/MKK6+/– mice were previously described (53, 54). Mammary gland whole-mount staining was performed as previously described (72). Briefly, mammary glands were excised and fixed in 10% buffered formalin overnight. Next, the samples were incubated in Carnoys fixative solution for 4 hours followed by serial hydration. Finally, mammary glands were incubated in carmine alum solution overnight, dehydrated, and then left in xylol for 16 hours. Stained whole mounts were preserved in methyl salicylate solution. To measure ductal tree density, four transecting lines (2.4 cm long) were drawn across four areas of the mammary gland proximal and distal to the nipple. The number of ducts intersecting these lines was counted and averaged for each animal and expressed as number of ducts per centimeter (unpaired t test). Immunohistochemistry from embedded paraffin sections was performed as previously described (72). The sections were processed with VectaStain ABC Elite Kit (Vector Laboratories), and the signal was detected with DAB Substrate Kit for peroxidase (Vector Laboratories).
Statistical analysis was performed with MS Excel or GraphPad Prism 5.0 software. P values were calculated with one-way analysis of variance (ANOVA) followed by the Bonferroni multiple comparison post test or the unpaired t test with P < 0.05 considered statistically significant.
Supplementary Materials Back to Top
Fig. S1. p38 isoform expression, TF regulation, cell survival, and Bim expression.
Fig. S2. Epithelial and myoepithelial cell organization, p38 activity, and TEB size distribution in control and p38-inhibited mammary glands.
Fig. S3. Effect of p38 inhibition on p-Rb and p-histone H3 staining intensity in FvB and wild-type versus MKK3–/–/MKK6+/– mammary glands.
References and Notes Back to Top
Citation: H.-C. Wen, A. Avivar-Valderas, M. S. Sosa, N. Girnius, E. F. Farias, R. J. Davis, J. A. Aguirre-Ghiso, p38α Signaling Induces Anoikis and Lumen Formation During Mammary Morphogenesis. Sci. Signal. 4, ra34 (2011).
The editors suggest the following Related Resources on Science sites:
In Science Signaling
THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES:
Science Signaling. ISSN 1937-9145 (online), 1945-0877 (print). Pre-2008: Science's STKE. ISSN 1525-8882