Hypoxia-inducible Factor-1α Induces ErbB4 Signaling in the Differentiating Mammary Gland*

Background: HIF-1α regulates mammary gland development by an unknown mechanism. Results: HIF-1α promotes ErbB4 protein accumulation, activity, and ErbB4-dependent differentiation of mammary epithelial cells. Conclusion: HIF-1α is a regulator of ErbB4 signaling in vitro and in vivo. Significance: A mechanism involved in the regulation of mammary gland development by HIF-1α is described. Conditional knock-out of Hif1a in the mouse mammary gland impairs lobuloalveolar differentiation during lactation. Here, we demonstrate that expression of ErbB4 was reduced in the lobulalveoli of mice with mammary gland-specific deletion of Hif1a. Erbb4 was not, however, a direct target gene for transcriptional regulation by HIF-1α in vitro. HIF-1α overexpression or HIF accumulating prolyl hydroxylase inhibitors reduced ErbB4 endocytosis, promoted transcriptional co-regulatory activity of ErbB4, and stimulated ErbB4-induced differentiation of mammary carcinoma cells. Consistently, RNA interference-mediated down-regulation of HIF-1α resulted in reduced ErbB4 protein amount and reduced mammary carcinoma cell differentiation. These findings indicate that HIF-1α is a physiologically relevant regulator of ErbB4 and that ErbB4 is involved in HIF-regulated differentiation of the mammary gland.

Hypoxia-inducible factor-1␣ (HIF-1␣) 2 is a transcription factor that binds specific hypoxia-response elements in the DNA with its basic helix-loop-helix domain. HIF-1␣ regulates transcription of several genes involved in angiogenesis, glycolysis, and pH regulation (1). HIF-1␣ protein degradation is suppressed in ischemic conditions and in malignancies, and HIF-1␣ is considered a promising drug target for cancer therapy (1).
HIF-1␣ is also necessary for normal development of several tissues, including the heart, the vasculature, and the brain (2). In pregnant mice, Hif1a deficiency results in condensation of the lobuloalveoli of the mammary gland and impaired production of milk during lactation (3).
ErbB4 is a growth factor receptor that mediates the signals of EGF-like ligands, such as the neuregulins (NRG), betacellulin, and heparin binding EGF-like growth factor (4). ErbB4 belongs to the ErbB/HER family of receptor tyrosine kinases, that also includes the EGF receptor (ErbB1, HER1), ErbB2 (HER2), and ErbB3 (HER3) (5). Although EGF receptor and ErbB2 are clinically relevant cancer drug targets, both tumor-promoting and differentiation-inducing roles have been proposed for ErbB4 (6). Erbb4 is alternatively spliced generating functionally unique ErbB4 isoforms (7). These isoforms differ either at the extracellular juxtamembrane domain (isoforms JM-a and JM-b) or the intracellular cytoplasmic domain (isoforms CYT-1 and CYT-2), generating variability in proteolysis-dependent signaling or coupling to intracellular proteins, respectively (7).
In addition to critical roles in regulating cardiovascular (8), neural (8,9), and renal (10) development, ErbB4 has been demonstrated to be necessary for differentiation of the mouse mammary gland (9,11,12). Interestingly, mice deficient of either Erbb4 or Hif1a in their mammary epithelia exhibit a similar failure in the formation and differentiation of the milk-producing lobuloalveoli (3,11,12).
The overlap of the phenotypes of Hif1a-and Erbb4-deficient mice implies that ErbB4 could be a novel HIF-1␣-regulated gene with a function downstream of HIF-1␣ in the developing mammary gland. To test this hypothesis, we analyzed ErbB4 expression in mice deficient of HIF-1␣ in the mammary gland (Hif1a Flox/Flox MMTV-Cre) and addressed the effect of HIF1A gain and loss of function on ErbB4 expression and function in mammary epithelial cells in vitro. Here, we report that HIF-1␣ induces ErbB4 protein accumulation in the mammary epithelial cells both in vivo and in vitro and that this process is necessary for the normal ErbB4-mediated differentiation of mammary gland epithelial cells.

EXPERIMENTAL PROCEDURES
Conditional Knock-out Mice and Immunohistochemistry-Mice with mammary gland specific targeting of Hif1a (Hif1a Flox/Flox MMTV-Cre line A) have been described earlier (3). Paraffin sections of the mammary glands from mice at pregnancy day 18 or lactating day 1 were immunostained with the rabbit polyclonal antibodies anti-ErbB4 (sc-283; Santa Cruz Biotechnology) and anti-GLUT-1 (ab14683; Abcam). Immunohistochemical analysis was performed using HistomouseMax IHC staining kit (Invitrogen) following the manufacturer's protocol.
Cell Culture-T-47D human breast cancer cells were maintained in RPMI supplemented with 10% FCS. HEK293 human embryonic kidney cells and MDA-MB-468 human breast cancer cells were maintained in DMEM supplemented with 10% FCS.
Immunofluorescence Staining-T-47D cells as well as MDA-MB-468 transfectants expressing ErbB4 JM-a CYT-1 were plated on glass coverslips and treated for 20 h in DMEM with or without 500 M DMOG followed by 30 min of treatment with or without 50 ng/ml NRG-1. After fixing with ice-cold methanol, the cells were washed with PBS and placed in blocking solution (5% goat serum and 0.01% Tween 20 in PBS) for 1 h at room temperature followed by overnight staining at 4°C with the anti-ErbB4 E200 diluted in blocking solution. After five washes with PBS, the secondary antibody goat anti-rabbit Alexa 488 (Invitrogen) and 0.5 g/ml DAPI (Sigma-Aldrich) diluted in blocking solution was applied for overnight incubation at 4°C. After five washes with PBS, the samples were mounted with Mowiol (Calbiochem). Samples were imaged with a LSM510 META (Carl Zeiss) confocal microscope using planapochromat NA1.4 63ϫ objective. One-micrometer-thick optical sections were imaged through the middle plane of the cells. The localization of the anti-ErbB4 epitope was analyzed from unprocessed images using FIJI ImageJ software (21). First, the cell was outlined right outside of the plasma membrane using segmented line tool, and fluorescence intensity was measured yielding total fluorescence intensity of the cell (I total ). Then the segmented line was drawn right inside of the plasma membrane, and the intensity was measured again, yielding internal fluorescence intensity (I internal ). The amount of internalized signal was calculated using the following equation.
Antibody Internalization Assay-MDA-MB-468 cells expressing ErbB4 JM-a CYT-1 were plated on glass coverslips and treated for 20 h in DMEM with or without 500 M DMOG. Next, 200 ng/ml of an antibody recognizing the extracellular domain of ErbB4 (clone H4.77.16; Thermo Scientific Pierce) was added and incubated with the cells for 2 h at 37°C. The cells were washed with PBS and fixed in ice-cold methanol. Following a wash with PBS, the cells were incubated for 1 h at room temperature in the presence of the secondary antibody goat anti-mouse Alexa 568 (Invitrogen) and 0.5 g/ml DAPI. After four washes with PBS, the cells were mounted with Mowiol. Two-micrometer-thick optical sections were imaged through the middle plane of the cells. The amount of fluorescence signal reflecting the amount of internalized antibody was quantified as described above.
Cell Surface Biotinylation-MDA-MB-468 cells expressing ErbB4 JM-a CYT-1 were plated at 500,000 cells/60-mm dish. The following day, the medium was replaced with DMEM including 0 or 500 M DMOG for 20 h. The cells were washed three times with ice-cold PBS. The cell surfaces were biotinylated with 2 mM cell-impermeable biotinylation reagent Sulfo-NHS-LC-Biotin (Thermo Scientific Pierce) for 20 min at room temperature. The cells were washed twice with ice-cold PBS and once with ice-cold DMEM. Following biotinylation, the cells were treated with 50 ng/ml of NRG-1 for 0, 10, or 30 min. Subsequently, cell surface proteins were cleaved by incubation in trypsin solution (2.5 g/liter trypsin (Lonza), 0.2 g/liter K-EDTA (BDH Laboratory Supplies), and 1.0 g/liter glucose (Amresco) in PBS) for 20 min on ice and 10 min at room temperature. Trypsinized cells were centrifuged, and the cell pellet was resuspended in 200 l of lysis buffer (1% SDS, 150 mM NaCl, 100 mM Tris-Cl, pH 7.4). The samples were incubated at 95°C for 5 min. Fifty microliters of each sample was diluted with 950 l of TNN buffer (0.5% Nonidet P-40, 5 mM EDTA, 250 mM NaCl, 50 mM Tris-HCl, pH 7.4) and incubated together with streptavidin-agarose beads at room temperature for 2 h. After five 1-ml washes with TNN buffer, the beads were resuspended in sample buffer, incubated at 95°C for 5 min, cooled down, and subjected to Western analysis.
In Vitro Differentiation Assay-MDA-MB-468 cells expressing ErbB4 (isoform JM-a CYT-1) or vector control cells were cultured in Matrigel (BD Biosciences), as previously described (15). Briefly, the cells were trypsinized and resuspended in DMEM ϩ 10% FCS at a density of 500,000 cells/ml. Thirty microliters of the cell suspension was combined with 180 l of ice-cold Matrigel, and 50 l of the mixture was added to 96-well plate wells. The gels were allowed to solidify for 30 min at 37°C, after which 150 l of medium containing 50 ng/ml NRG-1 was added on top. After 7 days in culture, cell clusters were scored under a 20ϫ objective with an Olympus CK40 microscope (Olympus). For Western analyses, the cells were cultured at equivalent conditions on 24-well plates for 5 days. The cells were released from Matrigel using BD cell recovery solution (BD Biosciences), as recommended by the manufacturer.
Luciferase Reporter Assays-To address the activity of the STAT5-sensitive ␤-casein promoter, HEK293 cells were plated on 24-or 96-well plates at a density of 1.5 ϫ 10 4 cells/ cm 2 . The following day, the cells were transfected with 50 ng of empty pcDNA3.1 vector, pcDNA3.1ERBB4JM-aCYT-2-HA, or pcDNA3.1ERBB4JM-aCYT-2K751R-HA together with 50 ng of pME18S-STAT5a, 100 ng of the reporter plasmid pGL3-␤-casein-LUC, and 50 ng of the control plasmid pEGFP-C3 (Clontech) (plasmid quantities given for 24 wells). The day after transfection, cells were subjected to treatment with 0 or 200 M CoCl 2 and/or 0 or 10 M AG 1478 for 24 h. Luciferase activity was measured using Bright-Glo luciferase assay (Promega) following the manufacturer's protocol. The luminescence signal from the reporter plasmid was normalized by the fluorescence signal from the control plasmid.
Chromatin Immunoprecipitation-T-47D cells were starved overnight in the absence of serum and then treated for 24 h with 0 or 500 M DMOG and for 45 min with 0 or 50 ng/ml of NRG-1. Chromatin immunoprecipitation was performed as previously described (22) using rabbit anti-ErbB4 (sc-283; Santa Cruz Biotechnology) to precipitate the intracellular domain of ErbB4 or rabbit IgG (Cell Signaling Technology) as a negative control. Immunoprecipitated chromatin was PCRamplified using PCR SuperMix (Invitrogen) and the primers 5Ј-actgtcctccagtcattgtct-3Ј and 5Ј-tggtccatcagcttctgtgac-3Ј previously shown to amplify a region of the human ␤-casein promoter that binds ErbB4 (23).
Statistical Analyses-In vitro experiments were analyzed with two-sided Student's t test or with one-way analysis of variance associated with Dunnet's t post hoc test calculated with IBM SPSS 20 software. Data about the effect of HIF1A-targeting siRNAs on ErbB4-mediated differentiation (shown in Fig.  6C) was analyzed using repeated measures analysis of variance with Tukey HSD post hoc test to control for baseline variations between experiments.
Loss of Erbb4 in the lactating mammary gland leads to a similar phenotype as loss of Hif1a (3,11,12), implying that HIF-1␣ and ErbB4 reside in the same signaling pathway regulating mammary gland differentiation. To address whether HIF-1␣ regulates the level of ErbB4 protein in vivo, tissue sections from pregnancy day 18 mammary glands of the mice with conditionally targeted Hif1a alleles were analyzed by immunohistochemistry with an antibody against ErbB4. When compared with the wild-type control (Fig. 1A), immunoreactivity for ErbB4 was strongly reduced in the differentiating Hif1a Ϫ/Ϫ mammary epithelial cells (Fig. 1B). As a control, Hif1a knockout was also shown to result in a significant suppression of immunoreactivity for the known (2) HIF-1␣-regulated gene product glucose transporter-1 (GLUT-1) (Fig. 1, C and D).
HIF-1␣ Promotes Accumulation of Endogenous ErbB4 Protein in Vitro-To further test the effect of HIF-1␣ on ErbB4 expression, T-47D human breast cancer cells endogenously expressing ErbB4 (24) were transfected with an empty vector or plasmids encoding wild-type HIF-1␣ or the HIF-1␣ mutant HIF-1␣ P402A/P564A, resistant to prolyl hydroxylation-dependent proteolysis. Western analysis of the transfectants indicated that ectopic expression of both HIF-1␣ constructs increased the ErbB4 protein level ( Fig. 2A). Moreover, HIF-1␣ targeting siRNAs suppressed the ability of the known inducer of endogenous HIF stability, DMOG, to increase endogenous ErbB4 protein level in the T-47D cells (Fig. 2B).
To address the hypothesis that HIF-1␣ promoted ErbB4 protein accumulation as a direct transcriptional activator of the ERBB4 gene, T-47D cells were treated with another chemical inducer of endogenous HIF-1␣ stability, CoCl 2 , and the amount of ERBB4 mRNA was measured by real time RT-PCR. However, no up-regulation of ERBB4 mRNA expression was observed, whereas CoCl 2 treatment significantly induced the expression of the known (2) HIF-1␣-regulated gene GLUT1 (Fig. 2C).

ErbB4 Induction by HIF-1␣
Moreover, HIF overexpression did not stimulate ERBB4 mRNA expression. ERBB4 mRNA level was 0.3% of ␤-actin mRNA level 48 h after transfection of a HIF plasmid as compared with 0.4% after transfection of a vector control. These data suggest that both endogenous and ectopically expressed HIF-1␣ can promote accumulation of the ErbB4 protein at the post-transcriptional level.
HIF-1␣ Promotes Accumulation of Both the Full-length ErbB4 and Its Cleaved Intracellular Domain Fragment-The 180-kDa full-length ErbB4 in T-47D cells is processed to an 80-kDa intracellular domain (ICD) fragment with potential signaling activity (25). However, although considerable amounts of the 80-kDa fragment are produced, it is also rapidly degraded in the T-47D background (26). To address the effect of HIFregulation on the different ErbB4 species, HEK293 cells were transfected to express ErbB4 (cleavable isoform JM-a CYT-2). As expected for regulation taking place at the post-transcriptional level, treatment with CoCl 2 promoted accumulation of ectopically expressed ErbB4 (Fig. 3A, lane 4 versus lane 2). Both the full-length 180-kDa and the carboxyl-terminal 80-kDa ICD fragment of ErbB4 were accumulated in response to CoCl 2 , and co-transfected HIF-1␣ siRNA suppressed the accumulation of both ErbB4 species (Fig. 3A). Consistently, DMOG and overexpressed HIF also accumulated ErbB4 in HEK293 cells (data not shown). The effect of CoCl 2 was not dependent on ErbB4 cleavage or on the type of the ICD of ErbB4 because it promoted accumulation of both cleavable (JM-a) and noncleavable (JM-b) ErbB4 isoforms with two different types of ICDs (CYT-1 or CYT-2) (Fig. 3B).
HIF-1␣ Induction Reduces ErbB4 Endocytosis-The steadystate levels of receptor tyrosine kinases at the cell surface are regulated by the rate of receptor endocytosis and degradation (27). To assess the putative effect of HIF activation on ErbB4 internalization, T-47D cells endogenously expressing ErbB4 were treated with DMOG, and the localization of an ICD epitope of ErbB4 was visualized by immunofluorescence microscopy (Fig. 4A). Indeed, more immunoreactivity for ErbB4 associated with the plasma membrane, as opposed to cytosol, in cells treated with DMOG. DMOG treatment also significantly suppressed ErbB4 internalization in cells in which ligand-induced receptor endocytosis was stimulated by NRG-1 (Fig. 4B). The effect of DMOG on ErbB4 localization was inhibited by siRNAs targeting HIF-1␣, indicating that the DMOG effect was indeed dependent on HIF-1␣ (Fig. 4B). Similar findings were made when MDA-MB-468 breast cancer cell transfectants expressing ErbB4 JM-a CYT-1 were analyzed (data not shown).
Consistent observations of DMOG-suppressed ErbB4 internalization were also made when the MDA-MB-468 transfectants were cultured for 2 h in the presence of an antibody recognizing the extracellular domain of ErbB4 (clone H4.77.16) followed by immunodetection of the subcellular localization of the antibody (Fig. 4, C and D). Finally, DMOG reduced the amount of biotinylated ErbB4 translocating from the cell surface to a trypsin-insensitive intracellular compartment in response to stimulation with the NRG-1 ligand (Fig. 4E). These findings indicate that HIF induction promotes ErbB4 accumulation by reducing ErbB4 endocytosis.
We have previously demonstrated that trafficking of ErbB4 from the cell membrane to degradation occurs via Rab5-positive early endocytic vesicles (28). Interestingly, it was recently reported that Rab GTPase-binding effector protein 1 (RABEP1, rabaptin-5), a critical regulator of Rab5 (29), is involved in HIFpromoted accumulation of EGF receptor (30). To address whether RABEP1 was also involved in the mechanism by which the HIF-activator DMOG suppressed ErbB4 internalization, the effect of RABEP1 down-regulation by RNA interference was assessed. As predicted, the knockdown of RABEP1 inhibited ligand-induced endocytosis of ErbB4 biotinylated at the cell surface (Fig. 4F). RABEP1 was also necessary for the DMOG-induced accumulation of ErbB4 (Fig. 4G), suggesting that ErbB4 accumulation was indeed endocytosis-dependent.
HIF-1␣ Promotes ErbB4 Signaling-To evaluate the activity of the ErbB4 protein accumulated by HIF-1␣, ligand-induced phosphorylation of endogenous ErbB4 in T-47D cells was analyzed by Western blotting using a phospho-specific anti-ErbB4 recognizing phosphorylated Tyr-1248 at the carboxyl terminus of ErbB4. Overexpression of the degradation-resistant HIF-1␣ P402A/P564A mutant increased the duration of ErbB4 phosphorylation from 10 to 30 min after NRG-1 stimulation  AUGUST 8, 2014 • VOLUME 289 • NUMBER 32 (Fig. 5A). However, this prolongation in tyrosine phosphorylation was not associated with significantly altered kinetics of Erk or Akt phosphorylation at the same time points (Fig. 5A). This finding is consistent with previous observations indicating that in cells expressing both ErbB4 and ErbB2 (such as T-47D cells (24)), the Erk and Akt pathways are predominantly activated by the ErbB2 kinase in response to NRG-1 (15).

ErbB4 Induction by HIF-1␣
To more specifically analyze the effect of HIF-1␣ on a signaling output dependent on the ErbB4 kinase, a reporter luciferase assay measuring the transcriptional activity of the ␤-casein pro- The cells were biotinylated using a cell-impermeable reagent followed by treatment with 50 ng/ml NRG-1 for 0, 10, or 30 min. The cells were trypsinized, and internalized biotin-labeled ErbB4 was detected by pulldown with streptavidin-agarose beads followed by Western blotting with anti-ErbB4 (E200). The filter was reblotted with streptavidin-HRP to control loading. RABEP1 (sc-271069) and actin (sc-1616) were analyzed by Western blotting from the cell lysates. F, Western analysis of cell surface biotinylated proteins. MDA-MB-468 cells expressing ErbB4 JM-a CYT-1 were transfected with a negative control siRNA (Ϫ) or RABEP1-targeting siRNAs (#1 and #2) and cultured for 20 h in the presence of 0 or 500 M DMOG. After biotinylation, the cells were treated for 30 min with NRG-1, trypsinized, and analyzed for ErbB4 endocytosis as in E. RABEP1 (sc-271069) and tubulin (sc-9104) were analyzed by Western blotting from parallel lysates. G, Western analysis of ErbB4 (E200) and RABEP1 (sc-271069) expression. MDA-MB-468 cells expressing ErbB4 JM-a CYT-1 were transfected with a negative control siRNA (Ϫ) or RABEP1-targeting siRNAs (#1 and #2) and cultured for 20 h in the presence of 0 or 500 M DMOG. The filter was reblotted with anti-actin (sc-1616) to control loading.

ErbB4 Induction by HIF-1␣
moter was set up in HEK293 cells. This assay serves as a readout of transcriptional co-activator activity of ErbB4 (31)(32)(33) because ErbB4 enhances activation of ␤-casein promoter via STAT5 (23). Overexpression of ErbB4 (JM-a CYT-2) significantly increased the activity of the ␤-casein promoter, and the effect was further enhanced by the addition of CoCl 2 (Fig. 5B). Inhibition of ErbB kinase activity with the chemical inhibitor AG 1478 partially suppressed the increase achieved by ErbB4 overexpression, and introduction of a kinase-inactivating mutation (kinase-dead ErbB4 K751R) totally abolished the capability of CoCl 2 to increase ErbB4-stimulated promoter activity (Fig. 5B).
To assess the potential of HIF-1␣ to induce binding of the released ErbB4 ICD to endogenous ␤-casein promoter, chromatin immunoprecipitation experiments were carried out in the T-47D mammary cell background. Induction of endogenous HIF-1␣ by DMOG increased the NRG-1 ligand-induced binding of ErbB4 to the ␤-casein promoter (Fig. 5C). Taken together, these findings indicate that HIF activity can promote ErbB4 signaling.
HIF-1␣ Promotes ErbB4-dependent Mammary Epithelial Differentiation-The synergistic effect of HIF inducers and ErbB4 on ␤-casein promoter indicated that regulation of ErbB4 by HIF-1␣ may also be relevant for milk protein synthesis and lactogenic differentiation of the mammary epithelial cells. To address the biological role, an in vitro model of mammary epithelial cell differentiation was set up. For these experiments, we chose the MDA-MB-468 mammary carcinoma cells because they had previously been used to study ErbB4-dependent mammary epithelial differentiation (15,34). The MDA-MB-468 cells, lacking endogenous ErbB4 expression (15,35,36), were retrovirally transfected with the empty vector or a construct encoding ErbB4 JM-a CYT-1, an isoform known to promote mammary epithelial differentiation (31,32).
When the infected cell lines were cultured for 7 days in Matrigel, they formed both clusters of randomly associated The columns represent mean luciferase activity normalized with EGFP fluorescence Ϯ standard error from three independent experiments (n ϭ 9; except for ErbB4 JM-a CYT-2 K751R, n ϭ 6). Protein expression was controlled by Western analyses with anti-ErbB4 (E200) and anti-pErbB4 (Tyr-1248, #4757). C, chromatin immunoprecipitation (IP). T-47D cells were starved overnight in 0% FCS and then treated for 24 h with 0 or 500 M DMOG followed by 45 min of treatment with 0 or 50 ng/ml of NRG-1. Samples were immunoprecipitated with anti-ErbB4 (sc-283) or with the negative control IgG. Immunoprecipitated chromatin was analyzed by PCR amplifying the ErbB4 binding site of the human ␤-casein promoter. AUGUST 8, 2014 • VOLUME 289 • NUMBER 32 JOURNAL OF BIOLOGICAL CHEMISTRY 22465 cells (non-acinar colonies) and acinus-like colonies demonstrating signs of epithelial organization into ball-like structures (Fig. 6A). As expected based on earlier reports (31,32), overexpression of ErbB4 JM-a CYT-1 significantly enhanced the percentage of acinar colonies of all the colonies (Fig. 6B), indicating enhanced differentiation. In contrast, simply treating the cells with 100 M of the HIF-1␣-inducing DMOG was not sufficient to promote differentiation of the vector control cells lacking ErbB4 (Fig. 6B, white columns). When DMOG was administered to the cells expressing ErbB4, however, this compound promoted differentiation over the level achieved by ErbB4 overexpression alone (Fig.  6B, gray columns). Interestingly, experimentation with HIF-1␣-targeting siRNAs demonstrated that the effect of ErbB4 expression on the formation of acinar colonies was also dependent on the presence of HIF-1␣ (Fig. 6C). These observations indicate that HIF-1␣ enhances ErbB4-mediated differentiation of mammary epithelial cells in vitro.

DISCUSSION
Here we provide evidence indicating that accumulation of ErbB4 is part of the mechanism by which HIF-1␣ promotes lactogenic differentiation of the mammary epithelium during pregnancy. Deletion of HIF-1␣ was shown to reduce ErbB4 protein level in the lactating mouse mammary gland in vivo and HIF-1␣ expression increased ErbB4 protein in human mam-

ErbB4 Induction by HIF-1␣
mary epithelial cells in vitro. Experiments with an in vitro model of mammary epithelial differentiation indicated that HIF-1␣ activity was both sufficient and necessary for the ability of ErbB4 to induce differentiation. HIF-1␣ induction also promoted the ability of ErbB4 to bind and activate the promoter of the gene for ␤-casein, a major component of milk in the lactating mammary duct (37). This interaction between the HIF-1␣ and ErbB4 signaling pathways is expected to be biologically significant because mice carrying deletions of either Erbb4 or Hif1a exhibit similar phenotypes in their lactating mammary glands (3,11,12).
HIF-1␣ is a transcription factor that is known to stimulate the expression of several genes by direct activation of specific response elements in the target gene promoter (1,38). However, the mechanism by which HIF-1␣ promoted the accumulation of ErbB4 protein did not seem to involve direct transcriptional activation. Treatment with the HIF-inducing CoCl 2 did not increase the amount of ERBB4 transcript, whereas a similar treatment promoted the accumulation of mRNA encoded by GLUT1, a well known HIF target gene (2), by 80-fold. In addition, ErbB4 protein was accumulated when encoded by an expression vector lacking the regulatory sequences of the endogenous ERBB4 locus.
In contrast, our data indicate that HIF-1␣ promotes ErbB4 accumulation as a result of reduced endocytosis. The prolyl hydroxylase inhibitor DMOG reduced endocytosis of ErbB4 in MDA-MB-468 breast cancer cells, and functional endocytosis was needed for the prolyl hydroxylase inhibitor-promoted accumulation of ErbB4. Consistently with these findings, it has been reported that the loss of VHL, a protein recognizing prolyl hydroxylase-modified HIF, reduces endocytosis of EGF receptor and fibroblast growth factor receptor-1 in renal cell carcinoma cells (30,39). Although the exact mechanism by which HIF-1␣ promotes ErbB4 accumulation remains to be elucidated, the accumulation of both the full-length ErbB4 and its 80-kDa carboxyl-terminal fragment was dependent on RABEP1, a regulator of Rab5-mediated endocytosis (29).
Interestingly, HIF-1␣ has recently been shown to physically interact with the ICD of ErbB4 (20), raising a possibility that the direct molecular contact might have a role in regulating the stability of ErbB4. However, the interaction between HIF-1␣ and ErbB4 ICD is expected to be restricted to the nucleus and require proteolytic release of a the soluble ICD (20). Thus, our findings demonstrating that HIF-1␣ also promoted accumulation of the noncleavable JM-b isoforms of ErbB4 are inconsistent with a model in which ErbB4 stability was regulated by a direct HIF-1␣ interaction. On the other hand, the observation that ErbB4 expression alone was not sufficient to promote MDA-MB-468 cell differentiation when HIF-1␣ expression was down-regulated by RNA interference suggested that a direct or indirect interaction of the two proteins may also have a role in mediating the differentiation signaling downstream of ErbB4. Because previous work has demonstrated that the interaction of HIF-1␣ with the ErbB4 ICD leads to enhanced stability of HIF-1␣ (20), it is possible that HIF-1␣ and ErbB4 can reciprocally enhance each other's stability and signaling, generating a positive feedback loop (40). Given the similar mammary gland phenotypes of the Hif1a-and Erbb4-deficient mice, it is tempting to speculate that such a feedback loop of HIF-1␣ and ErbB4 could be involved in the differentiation of the lactating mammary gland.
HIF-2␣ has also recently been reported to regulate ErbB4 signaling in breast cancer cells (41). The mechanism involves increased expression of the ErbB4 ligand amphiregulin and subsequent ErbB4 phosphorylation (41). Interestingly, HIF-2␣, but not HIF-1␣, protein expression has been associated with favorable survival and luminal epithelial differentiation in clinical breast cancer samples (41). However, whereas amphiregulin regulates mammary gland development in mice (42,43), no defects of mammary gland development have been reported for Hif2a/Epas1 null mice. These findings imply that HIF-2␣ and HIF-1␣ may regulate ErbB4 by variable mechanisms in different biological contexts.
The induction of ErbB4 by HIF-1␣ in the HEK293 cells implied that ErbB4 may be induced by HIF-1␣ also outside of the context of the mammary gland. Indeed, in addition to mammary development, Hif1a deletion has been reported to produce similar phenotypes to Erbb4 deletion during trabeculation of the heart (8,44,45), as well as during migration of the cells of the neural crest (45,46). It remains to be elucidated what kind of role, if any, the cross-talk between HIF-1␣ and ErbB4 plays in these processes. In addition to development, HIF-1␣ and ErbB4 may regulate each other in neoplastic tissues. Consistent with a widely distributed signaling role on the same pathway, our previous in silico analysis demonstrated that the expression of ErbB4 significantly associates with HIF-1␣ activity in a number of normal and malignant human tissues (20). In conclusion, our findings indicate that HIF-1␣ may promote differentiation of the lactating mammary gland via a mechanism involving suppressed endocytosis of the ErbB4 receptor tyrosine kinase.