Introduction

The basal-like subset of breast cancer was first identified as tumors lacking hormone receptor and human epidermal growth factor receptor 2 amplification with a gene expression profile similar to basal or myoepithelial cells of the breast (Perou et al., 2000). Basal-like breast cancers account for 8–37% of all breast cancers (depending on the definition criteria) and is associated with a poor prognosis; increased development of distant metastasis, decreased survival rate and increased mortality (reviewed in (Rakha et al., 2008; Voduc and Nielsen, 2008; Rakha and Ellis, 2009)). As these tumors lack the expression of estrogen and progesterone receptors as well as human epidermal growth factor receptor 2, there are limited options that are largely ineffective in treating patients suffering from basal-like breast cancer.

Large-scale genetic analyses of human tumor samples have generated a wealth of molecular information and have identified potential tumor suppressors, oncogenes and prognostic factors (Perou et al., 2000; Sørlie et al., 2001; Sotiriou and Pusztai, 2009). The challenge now is to study the function of these new genes and understand their mechanisms of action to validate their clinical utility and confirm their potential use as targets of intervention. The arrestin family member ARRDC3 is one such gene that was recently found within a cluster on chromosome 5 deleted in 17% of basal-like breast cancers (compared with 0% in luminal breast-cancers) suggesting a role as a tumor suppressor (Adelaide et al., 2007). ARRDC3 contains structural homology to the arrestin family of proteins, which has an important role in the internalization and subsequent regulation of G protein-coupled receptors (GPCRs). Although ARRDC3 has been classified as an α-arrestin, there is a cluster of acidic and hydrophobic residues that may bind to clathrin, suggesting a role in membrane protein internalization (Alvarez, 2008). This is further supported by the putative structural homology between ARRDC3 and other arrestin family members VPS26 and β-arrestin (Aubry et al., 2009). A role for ARRDC3 connecting the PPARγ signal and endosomal functions has been suggested (Oka et al., 2006).

Integrins (ITGs) are cell surface adhesion molecules which mediate cell–extracellular matrix and cell–cell interactions. Binding of ITGs to their ligands initiates several signaling events that modulate many cellular behaviors, such as adhesion, proliferation, survival, motility, gene expression and differentiation (Dowling et al., 1996; Fuchs et al., 1997; Hynes, 2002; Watt, 2002; Arnaout et al., 2005; Lipscomb et al., 2005; Wilhelmsen et al., 2006; Germain et al., 2009; Vicente-Manzanares et al., 2009). The ITGβ4 subunit was initially identified in cancer as a tumor-related antigen associated with metastasis (Falcioni et al., 1989; Lee et al., 1992) and was later found to promote motility and invasion in carcinoma cells (Rabinovitz and Mercurio, 1996; O’Connor et al., 1998; Nikolopoulos et al., 2004). ITGβ4 signaling increases invasive potential and has been shown to sustain the survival of carcinoma cells in stressful environments (Yoon et al., 2005; Baril et al., 2007; Chen et al., 2009). Recent human studies have found that ITGβ4 expression is correlated to breast cancer size and grade (Diaz et al., 2005), and significantly correlates with basal-like breast cancer (Lu et al., 2008). ITGβ4 expression is also linked to poor patient prognosis in a variety of cancers (Raymond et al., 2007).

Despite its significance in tumor progression, surprisingly little is known about the regulation of ITGβ4 at the protein level. It is phosphorylated during signal transduction and thought to be internalized, trafficked around the cell, and either recycled or degraded (Hemler, 2001; Caswell and Norman, 2006, 2008; Wilhelmsen et al., 2007; Dutta and Shaw, 2008; Germain et al., 2009). It has been hypothesized that after hemidesmosome disrupting signals, newly liberated ITGβ4 is rendered capable for signaling. Activation results in the release of ITGβ4 from interactions with the cytokeratin cytoskeleton and allows for de novo interaction with the actin cytoskeleton and signaling molecules (Rabinovitz and Mercurio, 1996; Yoon et al., 2005; Mitra and Schlaepfer, 2006). However, little is known about the mechanism for which this occurs.

To understand the function of ARRDC3 we used overexpression and short hairpin RNA-mediated downregulation in human breast cancer cells. We showed significant effects on breast cancer cell migration and growth that coincided with dramatic effects on the cell surface protein ITGβ4. In this study we showed that ARRDC3 functions as a novel regulator of tumor progression in breast cancer because of its effects on ITGβ4 internalization and degradation.

Results

ARRDC3 expression is downregulated during tumor progression

To investigate how ARRDC3 expression is altered during carcinogenesis, we used Oncomine to analyze previously published microarray data. Levels of ARRDC3 mRNA are lower in breast cancer tissues when compared with normal tissue (Figure 1a). Furthermore, ARRDC3 mRNA levels decrease on the transformation of human mammary epithelial cells (Figure 1b). Oncomine data also revealed a decrease in ARRDC3 mRNA in metastatic tumors when compared with the primary tumor (Figure 1c). In addition, ARRDC3 levels are lower in tumors of breast cancer patients that relapsed within 5 years of diagnosis when compared with patients that remained disease free for >5 years (Figure 1d). Combined, these data indicate that ARRDC3 repression occurs early in carcinogenesis, decreases throughout tumor progression and suggests that low ARRDC3 may indicate a poor prognosis.

Figure 1
figure 1

Downregulation of ARRDC3 is an early event in carcinogenesis. Oncomine was used to analyze previously published microarray data. (a) Levels of ARRDC3 mRNA are decreased in human breast cancers when compared with normal breast tissue. (b) ARRDC3 mRNA levels decrease in human mammary epithelial cells after transformation with various oncogenes. (c) Levels of ARRDC3 mRNA are decreased in metastatic lesions when compared with the primary tumor in the same patient. (d) Expression of ARRDC3 mRNA is decreased in the initial ER+ tumors from patients that have relapsed when compared with 5-year disease-free patients. All data, including P-values, were calculated from Oncomine.

ARRDC3 affects in vitro cancer cell tumorigenicity

To evaluate the potential role of ARRDC3 as a suppressor of tumor growth, we generated stable cancer cell lines either overexpressing or repressing ARRDC3 in the basal-like breast cancer cell line MDA-MB-231 cells. The MDA-MB-231 line was used because it is considered a basal-like breast cancer cell line that had detectable levels of ARRDC3 and ITGβ4 when compared with other cell lines (Supplementary Figure S1). In Addition, MDA-MB-231 cells depend on ITGβ4 signaling for cell adhesion, survival, migration and invasion (Lipscomb et al., 2003, 2005). Expression of ARRDC3 indicated that there was not a chromosomal 5 deletion therefore, we could manipulate levels to observe effects on tumorigenicity.

We first noticed a change in the comparative growth rates. ARRDC3 overexpression causes a decrease in cell growth rates, whereas the repression of ARRDC3 increases cell growth (Figure 2a). To investigate the effects on invasiveness, sublines were used in a chemoinvasion assay. Overexpression of ARRDC3 caused a 50% reduction in the number of invasive cells, wheras repression of ARRDC3 caused a twofold increase (Figure 2b). Sublines were then used in a wound assay to test whether ARRDC3 affects cancer cell migration. We found a dramatic decrease in cell migration when ARRDC3 was overexpressed (Figure 2c). Sublines with repressed ARRDC3 exhibited a significant increase in migration rate. We next evaluated the importance of ARRDC3 for anchorage independent growth by calculating the colony number and size after 4-weeks growth in soft agar. There was a significant decrease in colony number in the ARRDC3 overexpressing sublines, whereas sublines with repressed ARRDC3 showed a significant increase in colony number (Figure 2d). ARRDC3 expression also affected the size of the colonies; sublines with repressed ARRDC3 had a higher percentage of colonies larger than >200 μm, whereas lines overexpressing ARRDC3 had a lower percentage of colonies >200 μm (Figure 2e).

Figure 2
figure 2

Expression of ARRDC3 affects in vitro tumorigenicity Each line was tested in triplicate and data bars represent mean±s.e.m. Single asterisk represents P<0.05, whereas a double asterisk represents P<0.001 as determined by Student's t-test. (a) Growth curves from the stable lines show that ARRDC3 overexpression suppresses cellular proliferation, whereas repression enhanced proliferation. (b) Matrigel chemo-invasion assay show that ARRDC3 overexpression suppresses invasion, whereas repression enhances invasion. (c) Wound assay of mitomycin-C treated cells shows that overexpression of ARRDC3 leads to a decrease in cell migration, whereas RNA interference-mediated repression of ARRDC3 increases cell migration. (d) ARRDC3 overexpression suppresses anchorage independent growth, whereas repression enhances anchorage independent growth. (e) Both colony number and colony size are affected. Dark gray bars represent total colonies (at least 50 μm), whereas light gray bars represent colonies larger than 200 μm.

ARRDC3 suppresses in vivo tumorigenicity

To assess the role of ARRDC3 in tumor growth in-vivo, the stable subclones of MDA-MB-231 cells were injected into the mammary fat pad of nude mice (n=20 for each subline). All subclones formed tumors efficiently in 85–100% of the mice suggesting that ARRDC3 does not affect in vivo tumor incidence. However, sublines with repressed ARRDC3 formed measurable tumors more quickly (2 weeks) when compared with the control line (3 weeks). In contrast, the formation of tumors from the ARRDC3 overexpressing line was delayed (4 weeks) when compared with the control line (3 weeks). After 7 weeks of injection, mice were killed and tumors were dissected and measured. ARRDC3 significantly suppressed tumor growth in vivo as determined by the final tumor volume measurements (Figures 3a–c). All tumors showed a similar undifferentiated solid tumor morphology (Supplementary Figure S2a). To determine whether there was a difference in the proliferation, sections from the xenograft tumors (n=5 tumors for each cell line) were stained for Ki67. As expected from the in vitro studies, repressing ARRDC3 increases the percent of Ki67 positive cells by almost twofold, whereas overexpressing ARRDC3 decreases the number of Ki67-positive cells by approximately threefold (Figure 3d and Supplementary Figure S2b).

Figure 3
figure 3

ARRDC3 negatively regulates in vivo tumorgenicity. Stable cells lines of MDA-MB-231 cells were injected into the mammary fat pads of immunocompromised mice (n=20 injections for each cell line) and grown for 7 weeks. (ab) Overexpression of ARRDC3 (ARRDC3) led to a decrease in in vivo tumor size compared with vector control (Flag), whereas repression of ARRDC3 (shARRDC3) led to an increase in tumor size compared with control (shCtl). (c) When quantified, the differences in final tumor volume were statistically significant compared with scrambled short hairpin RNA and empty vector controls. (d) ARRDC3 affects the in vivo proliferation of tumor cells as determined by percentage of Ki67-positive cells. Xenograft tumor sections were analyzed for Ki67 using IHC. Over 1500 cells per tumor were counted and scored either Ki67+ or Ki67−. In all, four to five tumor samples per group were analyzed and data bars represent mean±s.e.m. Single asterisk represents P<0.05, whereas a double asterisk represents P<0.001 as determined by Student's t-test.

To assess the contribution of ARRDC3 to in vivo tumor cell survival, sections of xenograft tumors were evaluated for central necrosis. The areas of necrosis and total tumor area were quantified using ImageJ software (n=4 tumors of each cell line). Although larger tumors are generally more necrotic, tumors derived from the shARRDC3 lines contained significantly less necrosis when compared with control tumors (Figures 4a and b). In contrast, tumors from the ARRDC3 overexpressing cell line, although smaller, were highly necrotic (Figures 4a and b). Terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick end labeling staining was also carried out on sections from the xenograft tumors. The expression of ARRDC3 seemed to have no effect on the number of apoptotic cells (Figure 4c and Supplementary Figure S2c), suggesting that ARRDC3 functions independently from the apoptotic pathway.

Figure 4
figure 4

ARRDC3 repression promotes in vivo cell survival. Tumors were cut in half along the longest axis therefore the center-most section of the tumor was used to make slides. Changes in necrosis are more evident when tumors of comparable size are analyzed. (a, b) Repression of ARRDC3 (shARRDC3) leads to a decrease in xenograft tumor necrosis (outlined in black) compared with control (shCtl), whereas overexpression of ARRDC3 (ARRDC3) leads to an increase in xenograft tumor necrosis compared with vector control (Flag). Dashed line represents the point where two images were merged to visualize the entire tumor section. Five tumors from each line were analyzed and the area of necrosis measured. Data bars represent mean±s.e.m. Single asterisk represents P<0.05, whereas a double asterisk represents P<0.001 as determined by Student's t-test. (c) ARRDC3 does not affect the number of apoptotic cells in xenograft tumors. Terminal deoxynucleotidyl transferase-mediated dUTP-biotin nick end labeling (TUNEL) analysis was performed on tumor sections and the number of positive cells in 10 high-powered fields ( × 40) was counted. Five tumors per group were analyzed and data bars represent mean±s.e.m. NS, not significant.

ARRDC3 negatively regulates ITGβ4

Initial studies using adenoviral vectors to overexpress ARRDC3 gave approximately 40-fold increase in protein and caused a complete detachment of MDA-MB-231 cells from the substratum after 1 week in culture suggesting a defect in cell adhesion. As cell adhesion and tumorigenic properties of MDA-MB-231 cells are dependent on ITGβ4, we hypothesized that ARRDC3 was affecting ITGβ4 expression. To determine whether ARRDC3 expression influences ITGβ4 levels in breast cancer cells, the stable lines with approximately fivefold ARRDC3 overexpression and RNA interference knockdown were analyzed by western blot. Overexpression of ARRDC3 caused a significant decrease in ITGβ4 protein levels, whereas cells with repressed ARRDC3 expression show a marked increase in ITGβ4 (Figure 5a). To verify ITGβ4 was only affected in cells with altered ARRDC3, cells were infected with a green fluorescent protein-expressing ARRDC3 adenovirus and evaluated for ITGβ4 levels using immunofluorescence. Cells that overexpress ARRDC3 showed significant downregulation of ITGβ4 surface levels, whereras adjacent uninfected cells showed no change in ITGβ4 protein levels (Figure 5b). Additionally, a partial detachment of ARRDC3 overexpressing cells from the substratum can be seen in culture 24 h after infection (Figure 5b). As expected, ITGβ4 expression and cell attachment were not affected by infection with the control green fluorescent protein adenovirus. Flow cytometric analysis also showed that overexpression of ARRDC3 causes an ablation of ITGβ4 protein from the cell surface (Figure 6c).

Figure 5
figure 5

ARRDC3 negatively regulates ITGβ4. (a) Western blot analysis shows that ITGβ4 protein level is affected by ARRDC3 expression. Stable alterations of ARRDC3 expression in MDA-MB-231, MDA-MB-435 and MDA-MB-435+β4 cell lines. When ARRDC3 expression was repressed using two separate short hairpin RNA constructs (shA or shC), cells showed an increase in ITGβ4 protein when compared with control cells (Un or shCtl). Conversely, when cells were made to overexpress ARRDC3 (ARRDC3), ITGβ4 levels decreased when compared with vector control cells (Flag) or untransfected parental cells (Un). (b) Immunofluorescence shows that when MDA-MB-231 cells overexpress ARRDC3 using a green fluorescent protein (GFP)-expressing adenovirus (adARRDC3), ITGβ4 levels decrease when compared with uninfected cells (no GFP expression) and cells infected with a control GFP-expressing adenovirus (adGFP).

Figure 6
figure 6

In breast cancer cells, ARRDC3 directly interacts with ubiquitinated ITGβ4 and negatively regulates protein levels in a mechanism dependent on the proteosome. (a, b) Cells can maintain high levels of ITGβ4 after ARRDC3 overexpression if treated with proteosome inhibitor Lactacystin for 6 h. (a) Western blot analysis shows that ITGβ4 levels after ARRDC3 overexpression is restored when the proteosome is inhibited. Inhibition of the proteosome is demonstrated by the accumulation of IκBα. (b) Immunofluorescence demonstrates that ITGβ4 levels are retained after ARRDC3 overexpression if cells are treated with proteosome inhibitor. Cells positive for adARRDC3 (green fluorescent protein (GFP+)) are outlined with a dashed line. (c) ARRDC3 causes a complete removal of ITGβ4 from the cell surface that is partially rescued with proteosome inhibition. The cell surface of live cells were then stained with an ITGβ4 antibody and analyzed by flow cytometry. Uninfected GFP-negative cells were gated out before ITGβ4 levels were examined. (d) Pretreatment of MDA-MB-231 cells with proteosome inhibitor lactacystin before the endogenous coimmunoprecipitation with antibodies against ARRDC3 and ITGβ4 enriches ubiquitinated forms of ITGβ4. Bands seen in the ubiquitin immunoblots were identical in size to ITGβ4 bands. Dotted line indicates noncontiguous lanes from the same film.

Interestingly, although ITGβ4 is altered in our stable MDA-231 cell lines (Figure 5a), these cells did not show the loss of adhesion seen in adenoviral infected cells. This is likely because of the higher expression levels that can be achieved with adenovirus.

ARRDC3 regulates ITGβ4 protein levels in a proteosome-dependent manner

We next wanted to determine the mechanism of ITGβ4 degradation in breast cancer cells. Parental MDA-MB-231 cells were infected with adARRDC3 and then treated with the proteosome inhibitor lactacystin for 6 h. The addition of 25 μM lactacystin prevented the ARRDC3-mediated decrease in ITGβ4 protein levels as determined by immunofluorescence and western blots (Figures 6a and b). Lactacystin treatment did not prevent internalization of surface ITGβ4, although a low level of ITGβ4 was still detectable on the surface (Figure 6c), suggesting that undegraded protein can be recycled to the cell surface. To exclude the possibility of lysosomal contribution to the protein degradation, ARRDC3 overexpressing cells were treated with the lysosomal inhibitor, chloroquine at 50 μM for 18 h. Chloroquine treatment did not prevent ITGβ4 reduction after infection with adARRDC3 (Supplementary Figures S3a and b).

As the regulation of ITGβ4 by ARRDC3 seemed to be dependent on the proteosome, and as to other arrestin family members have a role in targeting receptors for degradation (Gurevich and Gurevich, 2006; Bhandari et al., 2007; Kendall and Luttrell, 2009; Zhang et al., 2009), we asked whether ARRDC3 interacts with ITGβ4 after it is targeted for degradation. Untreated MDA-MB-231 cells show an undetectable amount of ubiquitinated ITGβ4, even after immunoprecipitation. However, when cells were pretreated with proteosome inhibitor lactacystin, higher molecular weight bands of ITGβ4 were detected after immunoprecipitation with both ARRDC3 and ITGβ4 antibodies (Figure 6d). Immunoblotting for Ubiquitin revealed that almost equal amounts of ubiquitinated-ITGβ4 is coimmunoprecipitated with the ARRDC3 antibody when compared with the ITGβ4 immunoprecipitation (Figure 6d). Thus, ARRDC3 binds ITGβ4 after ubiquitination, and seems likely to have a role in targeting ITGβ4 for degradation.

ARRDC3 directly interacts with activated ITGβ4

A novel phosphorylation site on ITGβ4, serine-1424, was recently identified as important in hemidesmosome disassembly and was enriched on the trailing edge of migrating cells (Germain et al., 2009). This phosphorylation, along with the phosphorylation of other serines, results in the disassembly of the hemidesmosome and mobilization of ITGβ4 to actin-rich protrusions (Rabinovitz et al., 2004; Wilhelmsen et al., 2007; Germain et al., 2009). To determine whether this site is important in ARRDC3-mediated ITGβ4 internalization, endogenous expression patterns of ITGβ4, ITGβ4-pS1424 and ARRDC3 were examined using confocal microscopy on migrating cancer cells. ARRDC3 colocalized with ITGβ4 only on the lagging edge of the cell, where ITGβ4-pS1424 is enriched (Figure 7a). We determined that ARRDC3 physically interacts with ITGβ4-pS1424 by endogenous coimmunoprecipitation (Figure 7b). Densitometric analysis show approximately 70% of ITGβ4-pS1424 is coimmunoprecipitated by ARRDC3 antibody when compared with ITGβ4 antibody. The densitometric ratio of precipitated pS1424/ITGβ4 is 3.5-fold higher for the ARRDC3 coimmunoprecipitation when compared with the ITGβ4 coimmunoprecipitation suggesting that ARRDC3 preferentially interacts with this activated form of ITGβ4.

Figure 7
figure 7

ARRDC3 preferentially interacts with ITGβ4 when phosphorylated on serine-1494. (a) Confocal images of a migrating MDA-MB-231 cell after wounding. Endogenous expression of ARRDC3, ITGβ4 and ITGβ4-pS1424 was detected. ARRDC3 colocalizes with ITGβ4 on the lagging edge of the cell, where ITGβ4-pS1424 is located. (b) MDA-MB-231 cells were treated with phosphatase inhibitors 1 μM sodium orthovanadate and 1 μM sodium fluoride for 3 h. Cell lysates were used in endogenous coimmunoprecipitation with antibodies against ARRDC3 and ITGβ4. ARRDC3 enriches ITGβ4-pS1424 when compared with total ITGβ4. Bands seen in the phospho-S1424 immunoblots were identical in size to ITGβ4 bands.

ARRDC3 specifically targets ITGβ4

Cancer cells frequently have defects within vesicular trafficking and/or endocytic pathways (Mosesson et al., 2008). To demonstrate that ARRDC3 specifically targets ITGβ4, rather than stimulating generalized endocytosis, we examined several cell surface proteins after cells were infected with either control or ARRDC3 over-expressing adenovirus. Analysis of nonpermeabilized cells by flow cytometry shows that ARRDC3 overexpression did not affect surface levels of ITGβ1, CD44 and EpCam (Figure 8).

Figure 8
figure 8

The regulation of ITGβ4 by ARRDC3 is specific and not a product of generalized increased endocytosis. MDA-MB-231 cells were infected with either a green fluorescent protein (GFP)-expressing adenovirus (+GFP) or an ARRDC3-expressing adenovirus (+ARRDC3). Nonpermeabilized cells were then analyzed by flow cytometry for various cell surface proteins. ARRDC3 overexpression did not affect surface levels of ITGβ1, CD44 or EpCam.

Effects of ARRDC3 on in vitro tumorigenicity is dependent on ITGβ4

To determine whether ITGβ4 is required to mediate the effects of ARRDC3 on cancer cell behavior, we used MDA-MB-435 (a cancer line which does not express ITGβ4) and MDA-MB-435+β4 cells (a daughter cell line engineered to overexpress ITGβ4), to create lines with altered ARRDC3 levels and assayed for tumorigenicity (Figure 5a).

Similar to MDA-MB-231 cells, the MDA-MB-435+β4 cells showed marked changes in proliferation, migration, invasion and anchorage independent growth. Overexpression of ARRDC3 in MDA-MB-435+β4 cells caused decreased proliferation and migration, whereas repression of ARRDC3 lead to increased proliferation and migration (Supplementary Figures S4a and b). Cell invasion assays showed a 50% decrease in the number of invasive cells in ARRDC3 overexpressing cells, whereas repression of ARRDC3 caused a twofold increase (Supplementary Figure S4c). Soft agar assays again showed that overexpression of ARRDC3 inhibits anchorage independent growth, whereas repressed ARRDC3 promotes in vitro tumorigenicity in the MDA-MB-435+β4 cells. In contrast to the ITGβ4-positive cell lines, ARRDC3 had only marginal effects on the parental, ITGβ4-negative MDA-MB-435 cell line (Supplementary Figures S4a–d). Altogether, these data show that ARRDC3 affects in vitro tumorigenicity, principally in an ITGβ4-dependent fashion.

ARRDC3 downregulation and coordinate ITGβ4 upregulation in human breast carcinogenesis and tumor progression

As there is an extensive connection between ITGβ4 and breast carcinogenesis, we examined protein levels of ARRDC3 and ITGβ4 in normal breast tissue and primary human breast cancers. ITGβ4-positive cells were located in the basal layer of normal ducts, whereas ARRDC3 was more highly expressed in luminal cells (in which ITGβ4 expression is negligible) but weakly expressed in basal cells (in which ITGβ4 expression is highest) (Figure 9b). In breast cancers, ARRDC3 expression was inversely correlated to ITGβ4 by immunofluorescence and western blot (Figures 9a and c–f).

Figure 9
figure 9

Expression of ARRDC3 in human breast tumors. (a) Western blot analysis of protein extracts from primary human breast tumors (invasive ductal carcinomas). ARRDC3 and ITGβ4 expression are inversely correlated. (bf) Immunofluorescence of primary breast or breast tumor tissue demonstrated an inverse correlation to ARRDC3 and ITGβ4 protein levels. (b) ARRDC3 is primarily expressed in the luminal cells in normal breast tissue. (c) Grade 1 breast tumors tend to express high levels of ARRDC3 and low levels of ITGβ4. (d) Grade 2 human breast tumors showed varied expression levels of ARRDC3. Expression is always the inverse of ITGβ4 expression. (e) Grade 3 breast tumors tend to express low or undetectable levels of ARRDC3 and high levels of ITGβ4. (f) Basal breast cancers with high ITGβ4 expression express very low or no levels of ARRDC3.

The combination of our data and the current knowledge of ITGβ4 strongly suggested that ARRDC3 downregulation would lead to a more aggressive or metastatic phenotype. To analyze the expression pattern of ARRDC3 in tumors, 52 human breast tumors of varied grade were used for ITGβ4 and ARRDC3 coimmunofluorescence. Expression of ARRDC3 was inversely correlated to tumor grade (Table 1). Within grade 1 tumors samples, five out of six showed high ARRDC3 staining while the outlier exhibited strong ITGβ4 staining. Interestingly, ARRDC3 expression varied greatly in grade 2 tumors, yet levels were inversely correlated to ITGβ4 expression. Grade 3 tumors generally expressed low or undetectable levels of ARRDC3. Although 11 out of 52 tumors samples expressed no or low levels of ARRDC3 and ITGβ4, only 1 out of 52 tumors samples had intense staining for both ARRCD3 and ITGβ4. Overall, the data support the hypothesis that ARRDC3 functions as a regulator of breast cancer progression through its affects on ITGβ4.

Table 1 Expression of ARRDC3 in human breast tumors

Discussion

In the present study, we describe a novel regulatory pathway involving the internalization and degradation of the cell surface protein ITGβ4, which has significant effects on breast cancer cell growth and tumorigenicity. We showed that ARRDC3, previously identified by genetic screening as a potential tumor suppressor (Adelaide et al., 2007), is a novel regulator of breast cancer progression. We show that ARRDC3 directly binds to ITGβ4, specifically when phosphorylated at S1424, ultimately leading to proteosome dependent degradation. The subsequent change in ITGβ4 protein levels significantly affects in vitro tumorigenic properties such as proliferation, migration, invasion and growth in soft agar. In-vivo analyses show that ARRDC3 negatively regulates tumor growth in nude mice. In addition, data from human breast cancer samples show that ARRDC3 expression is inversely correlated to ITGβ4 protein levels and tumor grade.

ITGβ4 is part of a genetic signature correlated to basal-type breast cancer (Lu et al., 2008). It was therefore very interesting to note that ARRDC3 is part of a cluster on chromosome 5 deleted in 17% of the same basal-type breast cancer subset (compared with 0% in luminal breast cancers) (Adelaide et al., 2007). Seven of eight basal-like breast cancer samples we analyzed had little to no expression of ARRDC3 (Table 1), suggesting that ARRDC3 may be downregulated by additional mechanisms other than chromosomal deletion. We found that when ARRDC3 protein was low or absent in human breast cancers regardless of grade or receptor status, these tumors generally had high levels of ITGβ4. We do not know the mechanism of downregulated ARRDC3 within these tumors, however chromosome deletion is likely not the only cause of ARRDC3 defects. Further genetic, epigenetic and mutational analyses of the ARRDC3 locus are needed to more fully investigate the transcriptional regulation of ARRCD3 within tumors.

Interestingly, some tumors we evaluated maintained expression of both ARRDC3 and ITGβ4. It is possible that the effects of ARRDC3 on ITGβ4 are simply dose dependent and moderate expression of ARRDC3 maintains intermediate levels of ITGβ4. Low or moderate levels of ARRDC3 may also allow for the recycling of ITGβ4 similar to the effects of β-arrestin on GPCRs (Luttrell and Lefkowitz, 2002). The presence of increased surface ITGβ4 we observed after treatment with a proteosome inhibitor may be evidence of such a mechanism (Figure 5c). This would facilitate the speed in which filamentous actin protrusions form, thus promoting tumor progression. However, further trafficking studies are necessary to determine whether ITGβ4 is actively being recycled to the cell surface in the absence of degradation.

Until recently, there have been relatively few investigations of the factors involved in ITG endocytosis. Internalization of ITGs α5β1, αvβ6, αvβ3, α6β1 have been associated with clathrin-mediated endocytosis (Caswell and Norman, 2008; Ezratty et al., 2009), whereas ITGs α5β1, αvβ3, α2β1 (Caswell and Norman, 2008; Shi and Sottile, 2008) have been connected to caveolae dependent endocytosis. ITGα6β4 has been shown to be associated with lipid rafts (Gagnoux-Palacios et al., 2003) suggesting that caveolae may have a role in localization. However, not all ITGα6β4 was found associated with lipid rafts. Our data provides additional insight to the area of ITG endocytosis. As ARRDC3 contains clathrin-binding motifs, it is possible that it functions within a clathrin-dependent mechanism although further investigation of the protein complexes within this pathway is needed. Lastly, ARRDC3 does not seem to affect β1IGTs (Figure 8), and because of the unique long cytoplasmic tail of ITGβ4, ARRDC3 is likely specific for this IGT.

It has been hypothesized that during the progression from normal epithelium to invasive carcinoma, the function of ITGβ4 switches from a mechanical adhesive device into a signaling-competent receptor. In this case, ITGβ4 needs to be liberated from hemidesmosome where it can then be trafficked to actin-rich motility structures (filopodia and lamellae) (Mercurio et al., 2001; Santoro et al., 2003; Lipscomb and Mercurio, 2005). Our findings highlight the importance of the ITGβ4-pS1424 site in this process as endogenous ARRDC3 colocalizes with this phosphorylated form of ITGβ4, likely during hemidesmosome disassembly and ITGβ4 internalization. The classic arrestin family member β-arrestin 1 is known to down-regulate G-protein mediated signaling, initiate internalization from the membrane and mediate additional signaling cascades (Claing et al., 2002). β-Arrestin 2 has been shown to have a similar role in controlling the single transmembrane receptor, transforming growth factor-β-R III (Chen et al., 2003). In addition, there is emerging evidence suggesting a role for β-arrestins in cancer (Buchanan et al., 2006; Raghuwanshi et al., 2008; Li et al., 2009). It is clear from our data that ARRDC3 has a role in the regulation of ITGβ4 protein levels and likely contributes to the control of ITGβ4 function during breast cancer progression.

Our data does not exclude the possibility that ARRDC3 has an ITGβ4 independent mechanism of action in cancer cells or in normal cells. The effects of ARRDC3 on cancer cells lacking ITGβ4 had marginal statistical significance and the overall effects were small (Supplementary Figures S2a–d). It is therefore unclear whether these effects are biologically significant. Considering its similarity to β-arrestins, ARRDC3 may also regulate GPCRs. Possible targets could include smoothened and frizzled receptors (class-6 GPCRs) or RhoA/C (class-1 GPCRs); which also have extensive implications in cancer.

In summary, our data identifies ARRDC3 as a novel regulator of breast cancer growth and progression that targets ITGβ4 for internalization and proteosome dependent degradation. We reveal several oncogenic properties affected by changes in ARRDC3 expression and identify a correlation between ARRDC3 expression and human breast tumor grade or aggressiveness. Our data also open several potential avenues of future research to better understand the potential significance of this novel regulatory pathway in carcinogenesis.

Materials and methods

Cells and reagents

MDA-MB-435 cancer cells expressing wild-type ITGβ4 were generated previously (Shaw et al., 1997). MDA-MB-231 breast carcinoma cells were obtained from the Lombardi Breast Cancer Depository (Georgetown University). The following antibodies were used: rabbit pAb 505 (ITGβ4-specific, gift from Dr Arthur Mercurio, UMass Medical School, Worcester, MA, USA); rat mAb 439-9B (ITGβ4-specific, BD Biosciences, San Jose, MA, USA); rabbit pAb S1424 (ITGβ4 phospho-serine 1424 specific, gift from Dr Isaac Rabinovitz, BIDMC, Boston, MA, USA); rabbit pAb ARRDC3 (Abcam, Cambridge, MA, USA); and mouse mAb Arrdc3 generated in our laboratory.

Cloning and generation of stable lines

pSuper vectors (oligoengine) containing control short hairpin RNA (shCtl target sequence: TTCTCCGAACGTGTCACGT) and short hairpin RNAs targeting ARRDC3 (target sequence A:GGCCTTGGCTACTACCAGT; target sequence C: GCGTGGAATATTCACTAAT) were generated in our lab per manufacturers recommended protocol. Full-length human ARRDC3 cDNA and Flag-LacZ fusion cDNA were cloned into the pBABE-puro expression vector (Addgene, Cambridge, MA, USA). Cells were transfected with Fugene HD (Roche Applied Science, Indianapolis, IN, USA) following manufacturer's protocol. Stable transformants were selected using puromycin and pooled.

Adenoviral preparation

Full-length human ARRDC3 cDNA was initially cloned into pLE green fluorescent protein (Clontech, Mountain View, CA, USA). The fragment containing ARRDC3+green fluorescent protein was then cloned into Adeno-X LP CMV using the Adeno-X Expression System 2 kit (Clontech). Adeno-X maxi purification kit (Clontech) was used to purify the adenovirus. For all steps, the manufacturer's protocol was followed. The amount of virus needed for each cell type was determined empirically.

Immunofluorescence

Cells cultured on coverslips or frozen sections of O.C.T.-fixed human breast cancer tissue (obtained from the UMass Cancer Center Tumor Bank with IRB approval #13176) were fixed with acetone and permeabilized by incubating with phosphate-buffered saline+Triton and blocked with 5% goat serum (Gibco, Carlsbad, CA, USA). Slides were incubated with primary antibody (rabbit polyclonal ARRDC3 (Abcam, Cambridge, MA, USA) and/or ITGβ4 439-9b (BD-Pharmingen, San Jose, CA, USA)) for 2 h at room temperature or overnight at 4 °C. After washing, slides were incubated with secondary antibodies conjugated with either fluorescein isothiocyanate or Texas-Red (Vector Laboratories, Burlingame, CA, USA) for 2 h (room temp). Slides were mounted using Vectashield with 4′-6-diamidino-2-phenylindole (Vector Laboratories) and were analyzed using a fluorescence microscope.

Wound assay

In all, 5 × 105 cells were evenly plated in a six-well plate. At 24 h after plating, confluent cells were treated with 15 μg/ml mitomycin C (Roche Applied Science) for 90 min at 37 °C. The monolayer was scratched with a P200 pipette tip and washed three times to remove floating cells. The wound closure was then monitored by digital photography.

Growth in soft agar

In all, 1.0 × 103 cells were suspended in 2 ml of serum-containing medium containing 0.3% agar and overlaid on a 1ml base layer of 0.75% agar in six-well plates. The soft agar was overlaid with complete medium (0.5 ml per well), which was changed every 2 days. After 3–4 weeks of incubation, viable colonies were stained by adding 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide into the covering media. Images were captured using bright-field microscopy and the number and size of the colonies were analyzed using ImageJ software. Only colonies with a diameter of >50 μm were counted.

Invasion

Matrigel invasion assays were carried out as described previously using 6.5-mm Transwell chambers (8-μm pore size, CoStar, St Louis, MO, USA) (Shaw et al., 1997). After 4 h, the cells that had invaded to the lower surface of the filters were fixed in methanol for 10 min. Filters were mounted on glass slides using Vectashield mounting medium containing 4′-6-diamidino-2-phenylindole (Vector Laboratories). Invasion was quantified by counting the number of stained nuclei in five independent fields in each transwell using ImageJ software.

Xenograft mouse studies

Animals’ care was in accordance with guidelines approved by IACUC. Cells were trypsinized, washed with sterile phosphate-buffered saline and resuspended in 35 μl phenol red-free Matrigel immediately before injection. Old female immunocompromised mice (9 weeks) (nu/nu; National Cancer Institute) were anesthetized briefly with isofluorane and cells were injected into the #3 and #8 mammary fat pad (1 × 106 cells per injection, two injection sites per mouse). Tumor volume was determined using the following formula: (4/3)π(1/2 × smaller diameter)2 (1/2 × larger diameter). Final dissected tumor volume was determined using the following formula: (4/3)π(1/2 × length)(1/2 × width)(1/2 × height). Percent tumor necrosis was determined by outlining the necrotic area in the area finder function of ImageJ and dividing it by the total area of the tumor.

Immunoprecipitation

Cells were treated for 3–4 h with 1 mM sodium orthovanadate or 1 mM NaF or 50 μM Lactacystin and solubilized at 4 °C for 15 min in NP-40 Lysis buffer supplemented with 1 mm sodium orthovanadate, 1 mm NaF and protease inhibitors (Complete mini tab; Roche Applied Science). Aliquots of cell extracts containing equivalent protein amounts were incubated overnight at 4 °C with antibodies and protein-A/G-Sepharose (Santa Cruz, Santa Cruz, CA, USA) with constant agitation. Immune complexes and aliquots of cell extracts containing equivalent total protein amounts were then analyzed by western blot.

Western blot analysis

Cell extracts containing equivalent protein amounts were resolved by sodium dodecyl sulfate–polyacrylamide gel and were transferred to nitrocellulose membranes. Membranes were blocked for an hour using a 50 mM Tris buffer (pH 7.5) containing 0.15 mM NaCl and 0.05% Tween 20 and 5% (w/v) Blocker (BioRad, Hercules, CA, USA) in phosphate-buffered saline+ Tween 20. Membranes were incubated overnight at 4 °C in the same buffer containing primary antibodies. Proteins were detected by enhanced chemiluminescence (Pierce, Rockford, IL, USA). For phospho-immunoblots, the blocking buffer for the primary antibodies contained 5% (w/v) bovine serum albumin.

Statistical analysis

All values in the present study were expressed as mean±s.e.m. unless otherwise noted. The significant differences between the groups were analyzed by a Student's t-test and a P-value of <0.05 was considered significant.