The scaffolding protein NHERF1 regulates the stability and activity of the tyrosine kinase HER2

We examined whether the scaffolding protein sodium-hydrogen exchanger regulatory factor 1 (NHERF1) interacts with the calcium pump PMCA2 and the tyrosine kinase receptor ErbB2/HER2 in normal mammary epithelial cells and breast cancer cells. NHERF1 interacts with the PDZ-binding motif in PMCA2 in both normal and malignant breast cells. NHERF1 expression is increased in HER2-positive breast cancers and correlates with HER2-positive status in human ductal carcinoma in situ (DCIS) lesions and invasive breast cancers as well as with increased mortality in patients. NHERF1 is part of a multiprotein complex that includes PMCA2, HSP90, and HER2 within specific actin-rich and lipid raft-rich membrane signaling domains. Knocking down NHERF1 reduces PMCA2 and HER2 expression, inhibits HER2 signaling, dissociates HER2 from HSP90, and causes the internalization, ubiquitination, and degradation of HER2. These results demonstrate that NHERF1 acts with PMCA2 to regulate HER2 signaling and membrane retention in breast cancers.

interactions with NHERF2 contributed to the apical retention of PMCA2 (38,39). Therefore, we reasoned that similar interactions with NHERFs might anchor PMCA2 and HER2 at the cell surface in breast cancer cells. To explore this hypothesis, we first examined the expression of NHERF1 and NHERF2 mRNA in mammary tumors in rats (40). As shown in Fig. 1A, NHERF1 expression increased in N-methyl-N-nitrosourea-induced mammary tumors compared with normal mammary glands, whereas NHERF2 expression decreased in tumors relative to normal glands. As a result, the ratio of NHERF1 to NHERF2 expression increased in tumors. Therefore, we concentrated further studies on potential interactions between PMCA2, HER2, and NHERF1.
We saw a clear up-regulation of NHERF1 immunofluorescence in hyperplastic regions and in tumors in MMTV-Neu mice (Fig. 1, B and C) compared with the relatively weak NHERF1 immunofluorescence at the apical surface of normal luminal epithelial cells. Furthermore, NHERF1 was not localized at the apical surface of the cells but was expressed uniformly within the plasma membrane, where it co-localized with both HER2 and PMCA2 in hyperplastic regions and/or tumors (Fig. 1, B and C). We also examined immunofluorescence stain-ing for NHERF1 in a series of 16 HER2-positive and four HER2negative human DCIS samples. NHERF1 levels were clearly increased in all HER2-positive DCIS samples compared with HER2-negative DCIS samples (Fig. 1D). Furthermore, NHERF1 co-localized with HER2 and PMCA2 at the membrane of the malignant cells. As with mouse tumors, NHERF1 was uniformly expressed along the plasma membrane in DCIS lesions and was not limited to the apical surfaces of the cells.
We assessed NHERF1 immunofluorescence in invasive breast cancers by utilizing a previously described tumor microarray (YTMA-49) consisting of 652 invasive breast cancers annotated to clinical outcome data covering a mean of 8.9 years (30,41,42). NHERF1 fluorescence staining intensity was measured as a continuous variable using the semiautomated AQUA system (41). We found that NHERF1 expression correlated with clinical pathology HER2-positive status (p ϭ 0.03) and positive nodal status (p ϭ 0.02) (Fig. 1E). Compared with the previous analysis for PMCA2, we found a significant positive correlation between NHERF1 and PMCA2 AQUA scores (p Ͻ 0.001, Fig. 1F). When NHERF1 expression was considered as a continuous variable, there was a trend toward higher NHERF1 levels predicting shorter survival, but the association Figure 1. A, NHERF1 mRNA levels, NHERF2 mRNA levels, and NHERF1/NHERF2 mRNA ratios from normal mammary glands versus mammary tumors harvested from N-methyl-N-nitrosourea-treated rats. B, immunofluorescence staining for HER2 (green) and NHERF1 (red) in mammary ducts from normal mice or hyperplastic regions in the mammary glands from MMTV-Neu mice. The right panels show co-staining with DAPI. Scale bars ϭ 10 m. C, immunofluorescence staining for HER2 (green) and NHERF1 (red) (top row) and for PMCA2 (green) and NHERF1 (red) (bottom row) in tumors from MMTV-Neu mice. Panels on the right show co-staining with DAPI. Scale bars ϭ 10 m. D, top row, immunofluorescence staining for HER2 (green) or NHERF1 (red) or merge in human HER2-positive DCIS. The right panels show a magnified view of the boxed area in the left panel. Center row, staining for PMCA2 (green) or NHERF1 (red) or merge in human HER2-positive DCIS. The right panels show a magnified view of the boxed area in the left panel. Bottom row, immunofluorescence staining for HER2 (green) or NHERF1 (red) or merge in human HER2-negative DCIS. The right panels show a magnified view of the boxed area in the left panel. Scale bars ϭ 10 m. E, correlations between NHERF1 expression and clinical pathological tumor characteristics. F, correlation between PMCA2 and NHERF1 AQUA scores in the YTMA49 tissue microarray. G, survival curves for tumors with NHERF1 AQUA scores above (circles) versus below (triangles) the optimal cut point as defined by the X-tile bioinformatics tool.
was not statistically significant (Cox hazard, p ϭ 0.094). However, when the X-tile bioinformatics tool (43) was used to define an optimal cut point between high and low NHERF1 levels, NHERF1 AQUA levels above this threshold were associated with a statistically significant decreased length of survival (Fig.  1G, Cox hazard, p ϭ 0.015). We examined the interactions between NHERF1, PMCA2, and HER2 and found that the relationships between NHERF1 AQUA scores and survival were lost when either PMCA2 or HER2 was included in a multivariate analysis. These results suggest that the ability of NHERF1 to predict mortality in this cohort is related to its associations with HER2 status and/or PMCA2 levels.

NHERF1 interacts with PMCA2 and HER2 in breast cancer cells
Next we examined whether NHERF1 interacted directly with PMCA2 and/or HER2. First, we examined NHERF1 mRNA levels and immunofluorescence staining in immortalized MCF10A human mammary epithelial cells and the HER2-positive human breast cancer cell line SKBR3. As expected, HER2 mRNA expression was much higher in SKBR3 cells than in MCF10A cells, as were both NHERF1 and PMCA2 mRNA levels ( Fig. 2A). We then performed confocal immunofluorescence staining for NHERF1, HER2, and PMCA2 in SKBR3 cells. As shown in Fig. 2B, NHERF1 co-localized with PMCA2 and HER2 within punctate areas of the plasma membrane that protruded from the apical aspect of the cells. NHERF1 also co-localized with actin (Fig. 2, B and C) as well as with the HER2 signaling partners EGFR and HER3 in similarly appearing membrane domains (Fig. 2, D and E). Interestingly, in serumstarved cells, HER2 co-localized with NHERF1 and actin more broadly at the apical membrane of the cells, but acute pharmacologic activation of HER2 signaling with either EGF or Nrg1 treatment caused both NHERF1 and HER2 to coalesce within discrete actin-enriched membrane protrusions, whereas EGFR and HER3 became internalized, as described previously (26,28,30) (Fig. 2, C-E). To determine directly whether NHERF1, PMCA2, and HER2 co-localized within the same membrane domains, we transiently expressed GFP-tagged PMCA2 and FLAG-tagged NHERF1 in MCF10A cells that constitutively overexpressed HER2. As with SKBR3 cells, HER2, PMCA2 (GFP), and NHERF1 (FLAG tag) co-localized within membrane protrusions at the apical surface of MCF10A cells (Fig. 2F).
Next, we transiently expressed His-tagged NHERF1 with GFP-tagged PMCA2 in CHO cells and MCF10A cells. As shown in Fig. 2G, when we immunoprecipitated NHERF1 using anti-His tag antibody, we also pulled down PMCA2. Expression of a GFP-tagged, mutant PMCA2 lacking the six C-terminal amino acids encompassing the PDZ recognition sequence (⌬6PMCA2) disrupted the ability to co-IP PMCA2 (GFP) with NHERF1 (His) in both cell lines, demonstrating that interactions between PMCA2 and NHERF1 depend on the PDZ recognition domain of PMCA2. We also introduced the same constructs into MCF10A cells overexpressing HER2. At baseline, immunoprecipitation of NHERF1 from MCF10A cells did not pull down HER2 (Fig. 2H). However, given that activation of HER2 appeared to increase immunofluorescence co-localization of NHERF1 with HER2 (Fig. 2C), we also treated the cells with EGF, which led to the ability to co-IP HER2 with NHERF1 (Fig. 2H).
To determine whether endogenous NHERF1, PMCA2, and HER2 were contained within the same protein complex, we performed co-immunoprecipitation experiments in SKBR3 cells and in tumors harvested from MMTV-Neu mice. As shown in Fig. 2I, immunoprecipitation of PMCA2 pulled down NHERF1 from SKBR3 cells and from MMTV-Neu tumors. Despite their co-localization by immunofluorescence, immunoprecipitation of HER2 did not pull down NHERF1 from either SKBR3 cells at baseline or from MMTV-Neu tumors (Fig. 2J). Nevertheless, NHERF1 appears to be necessary for stable interactions between PMCA2 and HER2 in SKBR3 cells. As reported previously, in control cells, immunoprecipitation of PMCA2 pulled down HER2 (Fig. 2K) (30), but, in SKBR3 cells in which we stably knocked down expression of NHERF1, immunoprecipitation of PMCA2 no longer pulled down appreciable amounts of HER2 (Fig. 2K). Furthermore, although GFPlabeled wild-type PMCA2 co-localized with HER2 at the apical cell membrane (Fig. 2F), membrane co-localization was diminished in MCF10A-HER2 cells transfected with GFP-labeled ⌬6PMCA2, and staining for GFP and HER2 was found in membrane invaginations and in intracellular vesicles (Fig. 2L). These results suggest that PDZ-mediated interactions between PMCA2 and NHERF1 are necessary for PMCA2 to stably interact with HER2 within the protruding membrane domains. Taken together, these data suggest that, in SKBR3 and MCF10A cells, NHERF1 is contained with a complex that also contains PMCA2 and HER2. Furthermore, activation of the EGFR and/or HER2 appears to increase the amount of NHERF1 within this complex, where it is necessary to stabilize interactions between HER2 and PMCA2.

NHERF1 is necessary for HER2 signaling in breast cancer cells
We next examined the effects of knocking down NHERF1 on HER2 signaling in SKBR3 and BT474 cells, two HER2-positive breast cancer cell lines. As shown in Fig. 3A, we achieved over 90% stable reductions in NHERF1 levels in NHERF1 knockdown (NHERF1KD) SKBR3 cells and NHERF1KD BT474 cells. Control cells expressed nonspecific shRNAs. In both cell lines, loss of NHERF1 significantly reduced total HER2 and pHER2 levels (Fig. 3A). There was a reduction in total EGFR expression in BT474 cells but not in SKBR3 cells. However, in both cell lines, knocking down NHERF1 dramatically reduced phospho-EGFR levels. NHERF1KD cells also showed reductions in total HER3 and pHER3 expression. Knocking down NHERF1 expression reduced AKT phosphorylation, although total AKT levels were unchanged. We also used immunofluorescence staining to assess HER2 activation in SKBR3 NHERF1KD cells. As shown in Fig. 3, B and C, in control cells, activated HER2 (pHER2 staining) was most prominent in actin-rich membrane protrusions. However, pHER2 staining was practically absent in NHERF1KD cells. Similarly, pAkt staining was located in prominent membrane protrusions in control cells and reduced in the NHERF1KD cells (Fig. 3D). To assess AKT bioactivity, we examined the localization of FOXO1 by immunofluorescence (Fig. 3E) and measured the expression of a FOXO1-luciferase reporter gene (Fig. 3F) (30). In control cells, FOXO1 was expressed primarily in the cytoplasm, whereas NHERF1KD-SKBR3 cells showed prominent nuclear FOXO1 staining.
Knockdown cells also showed increased FOXO1-luciferase gene expression. Given that AKT activation excludes FOXO1 from the nucleus and inhibits reporter gene expression, these results document reduced AKT bioactivity in response to reduced NHERF1 expression (44). Taken together, these findings show that suppressing NHERF1 expression reduces HER2 levels, impairs HER2 activation, and decreases downstream AKT signaling. As might be expected from diminished HER2/ AKT signaling, knocking down NHERF1 also led to a reduction in cell growth, as assessed by cell accumulation and BrdU incorporation (Fig. 3G).

Loss of NHERF1 and HER2 result in similar changes in global gene expression
To test whether NHERF1 regulated HER2 function more systematically, we compared changes in global gene expression NHERF1 and breast cancer caused by knocking down NHERF1 with those caused by knocking down HER2. We performed oligonucleotide microarray analyses to define changes in baseline gene expression between control SKBR3 cells and SKBR3 cells with stable knockdown of either NHERF1 or HER2 expression (NHERF1KD and HER2KD). We identified 826 transcripts that were significantly altered more than 2-fold in either direction in NHERF1KD cells and 840 transcripts in HER2KD cells. The changes in gene expression in each knockdown cell line were remarkably similar; 82% of the transcripts altered in HER2KD cells also changed in NHERF1KD cells (Fig. 4A), and a heat map of the changes in 1058 total genes altered in either cell line or both (Fig. 4D) underscores the remarkable concordance in global gene expression caused by knocking down NHERF1 versus HER2. We validated changes in the expression of six separate genes altered in the microarray by QPCR. As shown in Fig.  4E, the expression of each of these genes was reduced in both NHERF1KD and HER2KD cells, as predicted by the microarray results. Functional annotation of the changes in gene expression demonstrated a strong correlation with HER2/ ERBB2 signaling, and the altered genes were enriched for cancer-associated transcripts (Fig. 4B). For example, when we compared changes in the expression of the 74 genes in the "advanced malignant tumor" category, we found them to be virtually identical between NHERF1KD and HER2KD cells (Fig.  4C). These results demonstrate that knocking down either NHERF1 or HER2 expression results in fundamentally similar changes in global gene expression, supporting the notion that NHERF1 functions within a HER2-regulated biologic pathway.

Loss of NHERF1 Increases intracellular calcium levels and alters HER2-containing membrane protrusions
Given that knocking down NHERF1 or PMCA2 produces similar alterations in HER2 signaling (30), we next asked whether loss of NHERF1 alters PMCA2 levels. As shown in Fig.  5A, knocking down NHERF1 reduced total cellular PMCA2 levels, increased intracellular calcium concentrations approximately 4-fold, and increased the sensitivity of the cells to calci-um-induced apoptosis (Fig. 5, B and C). Reductions in NHERF1 also caused a profound change in membrane structure. At baseline, control SKBR3 cells displayed actin-rich protrusions from their apical aspect that contained HER2, EGFR, and HER3 ( Fig.  5, D-F). Knockdown of NHERF1 resulted in the effacement of these structures, as shown by confocal immunofluorescence staining for actin (phalloidin) as well as by scanning and transmission electron microscopy (Fig. 5, D and G). At baseline, HER2 and EGFR as well as HER2 and HER3 co-localized within punctate membrane protrusions (Fig. 5, E and F). However, knocking down NHERF1 led to an overall reduction and more uniform distribution of staining for EGFR and HER2 as well as an increase in their intracellular localization (Fig. 5E). Knocking down NHERF1 extensively reduced membrane staining for HER3, which, instead, accumulated in an intracellular compartment where it no longer co-localized with HER2 (Fig. 5F). HER2 localizes to lipid raft regions within the plasma membrane (45)(46)(47), and the HER2-containing punctate membrane domains also stained for cholera toxin B, a histologic marker of lipid rafts (Fig. 5H). As before, co-localization between cholera toxin B and HER2 was particularly distinct after acute activation of HER2 signaling with either EGF or NRG1 (Fig. 5I). Strikingly, knocking down NHERF1 expression caused a large reduction in staining intensity for cholera toxin B and loss of co-localization with HER2 both at baseline and after treatment with EGF and NRG (Fig. 5, H and I). We confirmed these findings by examining flotillin 1 and HER2 expression in Triton X-100-insoluble membrane fractions from control and NHERF1KD SKBR3 cells. As shown in Fig. 5J, HER2, PMCA2, and NHERF1 were all contained within flotillin 1-positive, Triton X-100-insoluble fractions from control cells. By contrast, in NHERF1KD cells, there were reductions in the expression of all three proteins in Triton X-100-insoluble fractions and also a reduction in flotillin 1 expression itself within the Triton X-100-insoluble fractions as well as in total cellular lysates (Fig. 5, J and K). These data demonstrate that NHERF1 is important for establishing and/or maintaining actin-and lipid raft rich show staining for HER2 (red), phalloidin (green), and merged staining. The right three columns show staining for HER2 (green), NHERF1 (red), and merged staining. Arrows point to the more prominent apical membrane protrusions that are actin-rich and contain HER2 and NHERF1 that form after treatment with growth factors. D, confocal images of co-immunofluorescence for NHERF1 (green) and EGFR in SKBR3 cells in growth medium (top row), in serum-free medium (center row), and following acute treatment with EGF. E, confocal images of co-immunofluorescence for NHERF1 (green) and HER3 (red) in SKBR3 cells in growth medium (top row), in serum-free medium (center row), and following acute treatment with NRG1. The top and right panels of each image depict optical sections through the cells in two different orientations. Scale Bars ϭ 10 m. F, confocal images of immunofluorescence for HER2 (aqua), PMCA2 (green), or NHERF1 (red) in MCF10A cells constitutively overexpressing HER2 and transfected with GFP-labeled, WT PMCA2, and FLAG-tagged NHERF1. The bottom right image shows merged images for all three stains. The top and right panels of each image demonstrate optical sections in different orientations. Boxed areas are magnified at the right. G, co-immunoprecipitation experiments in CHO or MCF10A cells transiently transfected with GFP-tagged PMCA2 and His-tagged NHERF1. All cells were transfected with either WT GFP-PMCA2 (lanes 1 and 3) or ⌬6-mutant GFP-PMCA2 lacking the C-terminal six amino acids that include the PDZ interaction motif (lanes 2 and 4). Lanes

NHERF1 and breast cancer
membrane protrusions that contain critical components of the HER2 signaling pathway.

NHERF1 is required for the retention of HER2 at the cell surface
As shown in Fig. 6A, treating control cells with EGF led to the internalization of EGFR but not HER2. In contrast, treating NHERF1KD cells with EGF caused internalization of HER2 with EGFR, and both continued to co-localize within the cyto-plasm. Likewise, treatment of control cells with Nrg1 led to the internalization of HER3, but HER2 remained on the cell surface within membrane protrusions (Fig. 6B). However, in NHERF1KD cells, NRG1 caused internalization of HER2 with HER3, and both co-localized in an intracellular compartment. The internalized HER2 in NHERF1KD cells was localized within an endocytic compartment, as demonstrated by co-localization between HER2 and rab5 in intracellular vesicles (Fig.  6C). By contrast, rab5 never co-localized with HER2 in control cells. Knocking down NHERF1 led to the recruitment of the ubiquitin ligase c-cbl and polyubiquitination and degradation of HER2. As can be appreciated, HER2 did not co-localize with c-cbl in control cells but did so in NHERF1KD cells after treatment with EGF (Fig. 6D). In addition, we co-stained control and NHERF1KD cells for HER2 and with the antibody FK2, which recognizes mono-and polyubiquitinated proteins. FK2 staining co-localized with HER2 and phosphorylated HER2 after EGF activation (Fig. 6, E and F) only in knockdown cells. In contrast, FK2 staining co-localized with EGFR or HER3 in control and NHERF1KD cells after treatment with either EGF or NRG1, respectively (Fig. 6, G and H). Co-immunoprecipitation of polyubiquitin and HER2 demonstrated an increase in HER2 ubiquitination in NHERF1KD cells both at baseline and after treatment with EGF or NRG1 (Fig. 6I). Finally, HSP90 has been shown to be important for stabilizing HER2 at the cell surface, and, as expected, we observed co-localization of HER2 and HSP90 in membrane protrusions in SKBR3 cells at baseline or after treatment with EGF ( Fig. 6J) (27,30,31). However, in NHERF1KD cells, HER2 and HSP90 no longer co-localized, and HER2 was internalized. Likewise, immunoprecipitation of HSP90 pulled down HER2 in control cells, but the ability to co-IP HER2 with HSP90 was much reduced in NHERF1 KD cells (Fig. 5K). Together, these data suggest that loss of NHERF1 disrupts interactions between HER2 and HSP90 at the plasma membrane and leads to the ubiquitination and internalization of HER2 upon its activation (Fig. 7).

Discussion
Breast cells express the PMCA2w/b splice variant, which contains a C-terminal PDZ interaction motif and traffics to the apical membrane (34,36,39). During lactation, PMCA2 transports calcium into milk and prevents calcium-induced involution of secretory epithelial cells (35-37, 42, 48). In breast cancer cells, PMCA2 interacts with HER2 within specific membrane domains to maintain a low intracellular calcium concentration that supports active HER2 signaling and prevents HER2 internalization (30). Therefore, the mechanisms that regulate the membrane localization of PMCA2 and allow it to interact with HER2 are of considerable importance to normal lactation and breast cancer. The presence of a C-terminal PDZ-binding motif within PMCA2 suggests that a member of the NHERF scaffolding family might interact with PMCA2 and influence its membrane localization. We found that NHERF1 is prominently expressed in HER2-positive breast cancers, and, in CHO cells and MCF10A cells, we found that PMCA2 interacts with NHERF1 through its PDZ-binding motif. In breast cancer cells, we demonstrate that NHERF1 interacts with a complex that includes PMCA2, HSP90, and HER2 within lipid raft-and actin-rich membrane domains (Fig. 7). NHERF1 stabilizes interactions between PMCA2 and HER2 and between HER2 and HSP90 and maintains the structure and lipid raft content of membrane signaling domains in which activated HER2 resides after treatment with either EGF or NRG1. These data suggest that NHERF1 has a dominant role in determining the membrane localization of both PMCA2 and HER2.
We found that PMCA2 co-immunoprecipitates with both HER2 and NHERF1 and that knocking down NHERF1 disrupts the ability to co-IP PMCA2 and HER2. However, we could only detect co-immunoprecipitation of NHERF1 and HER2 after acute stimulation of HER2 signaling in MCF10A cells. Likewise, immunofluorescence imaging of PMCA2, HER2, NHERF1, and actin demonstrated increased localization to more prominent membrane protrusions after treatment with EGF or NRG1. These findings suggest that activation of HER2 may increase the recruitment of NHERF1 to a complex that includes PMCA2 and HER2 and that the addition of NHERF1 then helps to stabilize and/or augment interactions between PMCA2 and HER2 as well as the structure of the lipid raft-rich membrane protrusions (Fig. 7). Our prior work showed that PMCA2 regulates HER2 cell surface localization and activation as well as the structure of the membrane protrusions in part by maintaining low intracellular calcium concentrations (30). Given that knocking down NHERF1 decreases PMCA2 levels, increases intracellular calcium levels, and mimics the effects of loss of PMCA2 on HER2 signaling, we think it most likely that the ability of NHERF1 to regulate HER2 signaling is mediated directly by its capacity to bind to the C terminus of PMCA2 and retain both PMCA2 and HER2 within the overall signaling complex. However, NHERF1 has also been reported to regulate the assembly of microvilli and rho-mediated cytoskeletal remodeling (8,49), and it is possible that NHERF1 stabilizes interactions between PMCA2 and HER2 indirectly by regulating the cytoskeletal structure of the membrane signaling domains rather than directly by scaffolding PMCA2 and/or HER2. More work will be needed to completely decipher the complex interactions between PMCA2, HER2, NHERF1, lipid rafts, and the actin cytoskeleton.
Knocking down NHERF1 expression in breast cancer cells inhibits HER2/Akt activation, inhibits cell proliferation, sensitizes cells to apoptosis, and results in changes in global gene expression that closely mimic those caused by knocking down HER2 itself. These data suggest that NHERF1 may contribute to the genesis or progression of HER2-positive breast cancers. Consistent with this idea, we found that NHERF1 expression correlated with HER2 status and PMCA2 expression and with metastatic spread to regional lymph nodes in a cohort of 652 breast cancers. In addition, high levels of NHERF1 predicted poorer survival in this same cohort, although, based on multivariate analysis, this may be a function of its correlation with PMCA2 and HER2 expression. These findings agree with several previous studies that documented positive correlations between HER2 and NHERF1 levels in human breast cancers and at least one report showing that NHERF1 correlates with a worse outcome in HER2-positive luminal B breast cancers (7,12,13,50). However, other studies suggested that NHERF1 inhibits breast cancer growth by scaffolding the phosphate and tensin homolog (PTEN) protein and inhibiting PI3K signaling downstream of the PDGF receptor (11,20). By contrast, we found that NHERF1 is required to support AKT activation downstream of HER2. Given that NHERFs can interact with a variety of different signaling molecules, it may be that NHERF1 exerts diverse effects on breast cancers in different signaling contexts. Larger clinical studies with the power to examine associations between NHERF1 and outcome in different molecular subtypes of breast cancer will be needed to more fully understand these contrasting results.
Receptor internalization followed by recycling or degradation is important for regulating signaling from receptor tyrosine kinases (26,28). For reasons that remain incompletely understood, HER2 normally resists internalization and degradation upon activation and tends to remain at the cell surface, where it can continue to signal for prolonged periods, a property thought to be important for its transforming activity (26,29). Interactions between HER2 and HSP90 as well as the localization of HER2 within membrane protrusions or microvillus- Figure 6. A, immunofluorescence for HER2 (green) and EGFR (red) in control (top row) and NHERF1KD (bottom row) SKBR3 cells after treatment with EGF. B, immunofluorescence for HER2 (green) and HER3 (red) in control (top row) and NHERF1KD (bottom row) cells after treatment with NRG1. C, immunofluorescence for HER2 (green) and rab5 (red) in control (top row) and NHERF1KD cells (bottom row) after treatment with EGF. D, immunofluorescence for HER2 (green) and cbl (red) in control (top row) and NHERF1KD (bottom row) cells after treatment with EGF. E, immunofluorescence for HER2 (green) and polyubiquitin (FK2, red) in control (top row) and NHERF1KD (bottom row) cells after treatment with EGF. F, immunofluorescence for pHER2 (green) and polyubiquitin (FK2, red) in control (top row) and NHERF1KD (bottom row) after treatment with EGF. In A-F, the third panel in each row shows merged staining with DAPI (blue), and the fourth panel in each row represents a magnified view of the boxed region of the third panels. G, immunofluorescence for polyubiquitin (FK2, green) and EGFR (red) after acute treatment of control or NHERF1KD cells with EGF. The images on the right show merged staining for both antibodies and DAPI (blue) to mark nuclei. H, immunofluorescence for polyubiquitin (FK2, green) and HER3 (red) after acute treatment of control or NHERF1KD cells with NRG1. The images on the right show merged staining for both antibodies and DAPI (blue) to mark nuclei. Scale bars ϭ 10 m. I, co-IP for polyubiquitin complexes and HER2. IP for ubiquitinated proteins using FK2 antibody pulled down more HER2 from NHERF1KD cells than from control cells with or without treatment with EGF or NRG1. J, immunofluorescence for HER2 (green) and HSP90 (red) in control (first and third rows) and NHERF1KD (second and fourth rows) cells without (first and second rows) or with (third and fourth rows) treatment with EGF. The arrow points to internalized HER2, dissociated from HSP90, after EGF treatment of NHERF1KD cells. Scale bars ϭ 10 m. K, co-immunoprecipitation of HER2 and HSP90 from control and NHERF1KD cells. Less HER2 is pulled down with HSP90 in NHERF1 KD cells. IB, immunoblot. . Model for the interaction of NHERF1 with PMCA2 and HER2 in breast cancer cells. a, at baseline, NHERF1 helps maintain PMCA2, HSP90, and HER2 in small, lipid raft-rich protrusions from the membrane. b, acute activation of HER2 leads to enlargement of the membrane protrusions with recruitment of NHERF1 to support interactions between PMCA2, HSP90, and HER2 that allow HER2 to remain within the protrusions, whereas EGFR and HER3 are internalized into the cell. c, in the absence of NHERF1, there is a reduction in PMCA2 expression and the lipid raft content of the membranes as well as an effacement of the membrane protrusions. In addition, upon stimulation, HER2 no longer remains at the cell surface but is internalized with EGFR and HER3.

NHERF1 and breast cancer
like structures have been reported previously to contribute to its resistance to internalization (27,30,31,(51)(52)(53). Our previous studies established that interactions between PMCA2 and HER2 are critical for HER2 to both interact with HSP90 and remain associated with membrane protrusions, and to avoid internalization, ubiquitination and degradation after its activation (30). We now demonstrate that NHERF1 is a key component of the same complex that stabilizes the membrane protrusions and HER2 surface localization and activity. These results suggest that it may be possible to devise new therapies for breast cancer by targeting NHERF1 to promote the internalization of HER2 and terminate its signaling.

Knockdown cell lines
Stable cell lines expressing shRNA directed against NHERF1 and ErbB2 (HER2) were generated by transducing cells with commercially prepared lentiviruses encoding shRNAs targeting NHERF1 (sc-63330, Santa Cruz Biotechnology) or HER2 (318 -328, Amsbio). Cells were infected by adding the various shRNA lentiviral particles to the culture for 48 h. Clones expressing the specific shRNAs were selected using 5 g/ml of puromycin (Gibco/Life Technologies, NHERF1) or 5 g/ml blasticidin S HCl (Gibco/Life Technologies, HER2).

Immunoblotting
Protein extracts were prepared using standard methods (30), subjected to SDS-PAGE, and transferred to a nitrocellulose membrane by wet Western-blotting transfer (Bio-Rad). The membrane was blocked in TBST buffer (TBS and 1% Tween) containing 5% milk for 1 h at room temperature. The blocked membranes were incubated overnight at 4°C with primary antibodies in Odyssey blocking buffer (927-40000), washed three times with TBST buffer, and then incubated with secondary antibodies provided by LI-COR for 2 h at room temperature. After three washes with TBST buffer, the membranes were analyzed using the Odyssey infrared imaging system (LI-COR). Triton X-100-insoluble fractions were isolated using the caveolae/rafts isolation kit (CS0750) from Sigma. Samples were then analyzed for HER2 and flotillin1 by Western-blotting analysis as above. All immunoblot experiments were performed at least three times, and representative blots are shown in the figures.

Cell proliferation and apoptosis
Cell proliferation was assessed by measuring BrdU incorporation using the cell proliferation ELISA kit (11647229001) NHERF1 and breast cancer from Roche. Apoptosis was measured by TUNEL assay using the cell death detection ELISA kit (11544675001) from Roche (Genentech Inc.). Cell viability and cell numbers were quantified using the XTT cell viability assay (9095) from Cell Signaling Technology (Danvers, MA).

Tissue microarray
The breast carcinoma tissue microarray (YTMA49) consisted of 652 primary breast cancer specimens retrospectively obtained from 1953 to 1983. Cases were evenly divided between lymph node-positive and -negative, with a median follow-up of 8.9 years. Clinicopathologic data were extracted from Yale and Connecticut Tumor Registries, including disease-specific survival. Tissue microarray slides were stained as described previously (42). The primary antibodies were anti-NHERF1 (Affinity BioReagents) and a mouse anti-cytokeratin (Dako, Carpinteria, CA) to distinguish the tumor from stroma. The slides were counterstained with DAPI. Details of the automated image acquisition and analysis using AQUA have been described previously (41).

Gene array
Total RNA was prepared from control SKBR3 cells as well as NHERF1KD and HER2KD cells using TRIzol reagent (Invitrogen), and oligonucleotide microarray analysis was performed by the Yale Center for Genomic Analysis. The isolated RNA was purified using the RNeasy cleanup kit (Qiagen, Valencia, CA). RNA was reverse-transcribed and hybridized to the HumanHT-12 v3 Expression BeadChip (Illumina, San Diego, CA) and then scanned using the Illumina BeadArray reader. The images were analyzed by Beadstudio software. The instructions for quality control and data analysis were provided by Illumina.

Intracellular calcium measurements
Ratiometric intracellular calcium imaging was performed using 5 M Fura-2/AM (Life Technologies) as described previously (30,42). Cells were loaded with 5 M Fura-2/AM (Life Technologies) for 30 min at 37°C and then imaged at a frequency of 1 Hz on a Zeiss Axiovert 100 microscope. Intracellular calcium concentrations were calculated from the background-subtracted fluorescent ratio (R) at 340 and 380 nm of Fura-2/AM-loaded cells using the formula K d ϫ (R Ϫ R min )/ (R max Ϫ R) ϫ (F f 380/F b 380), where K d is the dissociation constant of Fura 2 for calcium (225 nM), R min and R max are the empirically determined minimum and maximum fluorescent ratios, and F f 380/F b 380 is the fluorescence intensity at 380 nm in calcium-free conditions divided by the fluorescence intensity at 380 nm in saturating calcium concentrations (bound).

Statistics
Analyses were performed with Prism 6.0 (GraphPad Software, La Jolla, CA). Error bars represent standard error. Significance for all comparisons between two conditions was calculated using paired t tests, and significance for multiple comparisons was determined using one-way analysis of variance with Turkey post-test corrections. Significance for Kaplan-Meier analyses were calculated using the log-rank test, and correlations between PMCA2 and NHERF1 were calculated using Pearson's correlation coefficient.
Author contributions-J. J., J. N. V., W. K., and P. D. participated in the design, collection, and analysis of vital experiments. J. J. was responsible for the day-to-day planning and execution of the majority of the experiments. J. C. contributed to the analysis and interpretation of the microarray data. W. B. S. and P. A. F. contributed to the design and analysis of experiments examining the interactions between HER2 and NHERF1. J. J. W. was responsible for the oversight and analysis of all experiments and initiated the hypotheses that led to these experiments. J. J., P. A. F., and J. J. W. contributed to the writing of the manuscript.