Reduced annexin A6 expression promotes the degradation of activated epidermal growth factor receptor and sensitizes invasive breast cancer cells to EGFR-targeted tyrosine kinase inhibitors

It has been amply demonstrated that AnxA6 [28, 29] and EGFR [33‐35] are components of lipid raft containing membrane microdomains. It has also been shown that activation of EGFR is independent of AnxA6 expression [23], and that intact lipid rafts were required for the activation of the receptor [35]. Together, this led us to speculate that AnxA6 expression is required for sustained cell surface localization of activated EGFR in BCCs. To test this we first sought to compare the activation and activity of EGFR in the invasive AnxA6 high BT-549 cells with that of the non-invasive AnxA6-low HCC1806 as well as MDA-MB-468 cells (with amplified EGFR expression). We show that the expression of AnxA6 is barely detectable in HCC1806 and MDA-MB-468 cells compared to BT-549 cells (Figure 1A). Meanwhile, BT-549 and HCC1806 expressed relatively similar levels of total EGFR while the expression of EGFR was at least 3-fold higher in MDA-MB-468 (Figure 1A and B). Interestingly, treatment of these cells with EGF stimulated to varying extents, the autophosphorylation of the receptor on Y1068 (p-EGFR Y1068). Analysis of the time course for the activation of EGFR revealed that the receptor remained strongly activated even after 90 min in MDA-MB-468 cells. The activated receptor levels however, decreased with time in both HCC1806 and BT-549 cells (Figure 1C). Figure 1A also shows that in the AnxA6-high BT-549 cells, the activation of EGFR led to a sustained activation of MAP kinase ERK1/2 (p-ERK1/2). Paradoxically, in the AnxA6-low HCC1806 and MDA-MB-468 cells and compared to BT-549 cells, EGFR activation led to relatively reduced activation of ERK1/2 (Figure 1A).

To explain this paradox, we examined the localization of the activated receptor in the three cell lines by immunofluorescence. As shown in Figure 1D, there was a robust EGF-stimulated activation of EGFR in the AnxA6-low MDA-MB-468 cells. Interestingly, the activated EGFR in these cells was essentially localized to the perinuclear/cytoplasmic regions (red arrow heads) and in some cells, sequestered into the nucleus (white arrow head). Similarly, and consistent with Figure 1A, in the AnxA6-low HCC1806 cells, plasma membrane-localized activated EGFR (red arrow head) was also barely detectable (Figure 1D). On the contrary, in the AnxA6-high BT-549 cells, the activated EGFR was predominantly localized to the plasma membrane (red arrow heads). These data suggest that AnxA6 enhances the localization of activated EGFR on the cell surface and that this is accompanied by sustained activation of down-stream effectors such as ERK1/2.

To further ascertain the requirement of AnxA6 in the sustained localization of activated EGFR at the cell surface, and whether this is required for the invasiveness of these cells, we used RNA interference to down-regulate AnxA6 in the AnxA6-high BT-549 cells. Two AnxA6-depleted cell lines designated BT-A6sh2 (clone A1) and BT-A6sh5 (clone A3) were selected from 10 different clones (Additional file 1: Figure S1) and respectively showed ~30% and >80% AnxA6 depletion by Western blotting (Figure 2A). The AnxA6-depleted cells grew more efficiently than the control cells (Figure 2B) and as previously shown [5], their motility was significantly inhibited (Figure 2C). As shown in Figure 2D AnxA6 depletion also induced a transformation from invasive stellate colony morphology with long invasive projections in control BT-549 cells to the non-invasive acinar-like colony morphology in BT-A6sh5 cells. Invasive projections in BT-A6sh5 cells if discernible were much shorter than those in control cells, suggesting loss of invasiveness. This change in colony morphology is dependent on AnxA6 expression level because BT-A6sh2 cells showed intermediate colony morphologies. Based on these data, the BT-A6sh5 cell line with the most AnxA6 depletion was used as the AnxA6-depleted cell line in most of the following experiments. Interestingly, a similar extent of AnxA6 down regulation in MDA-MB-231 cells did not substantially lead to altered colony morphology and if anything, the cells tended to be more motile than the control cells (Figure 2E-H).

To substantiate the different outcomes of AnxA6 depletion in BT-549 and MDA-MB-231 TNBC cell lines on cell growth and motility, we examined the expression of AnxA6 and EGFR genes in these and other breast cell lines by qRT-PCR. This analysis revealed that MDA-MB-231 cells express relatively reduced levels of both AnxA6 and EGFR compared to BT-549 cells (Additional file 2: Figure S2A and B). We also demonstrate that the response of MDA-MB-231 to EGF treatment is also different from that in BT-549 cells in that while PI3 kinase/Akt and MAP kinase ERK1/2 are strongly activated in BT-549 cells, Akt and to a lesser extent ERK1/2 activation in MDA-MB-231 cells are relatively attenuated (Additional file 2: Figure S2B). As previously reported in other AnxA6-deficient tumor cells [20], over expression of AnxA6 in HCC1806 cells on the other hand was associated with reduced activation of the receptor and ERK1/2 (Additional file 3: Figure S3A-D). AnxA6 expression in HCC1806 cells also inhibited their growth in 3D cultures (Additional file 3: Figure S3E). These data suggest that in triple negative breast cancer cells, the modulation of EGFR activation and/or activity by AnxA6 is not only dependent on the AnxA6 expression levels but is also cell type specific.

The desensitization of ligand activated EGFR like most cell surface receptors predominantly occur by rapid internalization of receptor-ligand complexes and degradation in lysosomes. Given the strong cell surface expression of activated EGFR in AnxA6 expressing BT-549 cells, we speculated that the virtually absent expression and attenuated activity of the receptor in the AnxA6-low HCC1806 and MDA-MB-468 cells could be attributed to the fate of the activated receptor. To verify this, BT-549 control or AnxA6-depleted cells were treated with or without EGF for 0–90 min, surface biotinylated and the fate of EGFR examined by western blotting. Assessment of the residual levels of biotinylated surface-associated total and pY1068-EGFR in control (AnxA6-high) and AnxA6-depleted (AnxA6-low) BT-549 cells revealed that EGFR activation per se was indeed unaffected by AnxA6 depletion (Figure 3A). As expected, the levels of remaining ligand-activated EGFR decreased with time in both cell lines. However, the residual cell surface-associated activated EGFR decreased more rapidly in AnxA6-depleted cells (BT-A6sh5) compared to that in control cells (Figure 3A and B). By 90 min 60% of the activated EGFR in control cells was still cell surface-associated compared to only 20% in AnxA6-depleted cells. Similarly, the decrease in total cell surface (biotinylated) EGFR in the control cells was initially more rapid but this continued more slowly thereafter. On the contrary, there was a transient delay in the down-regulation of biotin-labeled EGFR that was followed by a more rapid decrease in the cell surface EGFR levels (Figure 3A and C). Within 90 min of EGF treatment, the receptor in AnxA6-depleted cells decreased to about 10% compared to about 40% in control cells. Taken together and consistent with data in Figure 1D, these data reveal that AnxA6 depletion in invasive breast cancer cells was accompanied by a rapid decrease in the total and activated cell surface EGFR levels. Furthermore, the transient delay in AnxA6-depleted cells versus a rapid initial decrease in the levels in control cells suggests a role of AnxA6 in the internalization and/or trafficking of the activated receptor.

We next sought to determine whether the rapid decrease in the activated cell surface EGFR in AnxA6-depleted cells and/or the relatively minimal activation of ERK1/2 in either HCC1806 or MDA-MB-468 cells could also be attributed to the lack of or relatively low levels of AnxA6. To do this we examined the residual levels of total EGFR in the AnxA6-depleted and control BT-549 cells. This analysis revealed that the EGF-activated (Figure 4A and B) and the total cellular receptor levels (Figure 4A and C) in control cells remained relatively constant while the receptor levels in AnxA6-depleted cells were not only lower (by at least 3-fold after 90 min), but also decreased more rapidly with time. Densitometric analysis of EGF-stimulated activation of ERK1/2 (Figure 4D) and Akt (Figure 4E) also reveal that these downstream targets were strongly inhibited in the AnxA6-depleted BT-549 cells compared to control cells. Together with data in Figure 3, this suggests that AnxA6 is necessary for the stabilization of the receptor on the cell surface and correspondingly, sustained signaling to downstream effectors (ERK1/2 and Akt).

To demonstrate that reduced AnxA6 expression enhanced EGFR degradation, control and AnxA6-depleted BT-549 cells were serum-starved overnight in the presence or absence of chloroquine (CLQ). The cells were then treated with or without EGF and the residual total and activated EGFR were examined by western blotting. As shown in Figure 5A-D, although CLQ treatment did not abolish the degradation of the activated receptor, the total cellular levels of receptor in the control and AnxA6-depleted BT-549 cells were stabilized by CLQ treatment. Interestingly, the levels of the receptor (t-EGFR) in AnxA6-depleted cells were restored to those in the control cell by 90 min (Figure 5A and C). To verify whether there were discernible differences in the degradation and recycling of the activated receptor in the control and AnxA6-depleted BT-549 cells, we examined the co-localization of EGF-activated EGFR with either LAMP1 (a late endosomal marker) or Rab11 (a recycling endosomal marker). As depicted in Figure 5E, within 5 min of EGF treatment, cell surface activated EGFR was clearly discernible in the control cells but the cell surface expression was lost by 90 min (arrows). On the contrary, even within 5 min of EGF treatment, most of the activated EGFR was intracellular in AnxA6-depleted cells (Figure 5E, arrows). We also observed a greater extent of co-localization of activated EGFR with LAMP1 in the AnxA6-depleted BT-549 cells compared to the control cells. Meanwhile, the activated receptor co-localized with Rab11 in both the control and AnxA6-depeleted cells. Together these data suggest that activated EGFR is actively recycled in these cells and that EGFR degradation is enhanced in the AnxA6-depeleted cells.

Given that down regulation of AnxA6 in invasive breast cancer cells was accompanied by a decrease in the total and activated EGFR in invasive breast cancer cells, we speculated that AnxA6-depletion in these cells might affect their response to EGFR-targeted TKIs. To explore this further, AnxA6-depleted and control BT-549 cells were treated with various concentrations (0–10 μM) of EGFR-targeted TKIs for 72 h. Figure 6A shows that gefitinib (Iressa) was the least potent of the four compounds while canertinib (CI-1033) was the most potent. In a concentration-dependent manner, AnxA6-depleted BT-549 cells were more sensitive to lapatinib (Tykerb) and PD153035 (Tyrphostin, AG 1517) compared to control cells. As shown in Table 1, the IC50s for inhibition of cell growth for lapatinib and PD153035 were significantly lower in AnxA6-depleted cells compared to control cells. This suggests that reduced AnxA6 sensitizes invasive BCCs to some EGFR-targeted TKIs.

To validate the above data, we compared the response to these compounds of AnxA6-depleted MDA-MB-231 and BT-549 cells as well as AnxA6-over expressing HCC1806 cells with their respective controls. As depicted in Figure 6B, treatment of these cells with these compounds (5 μM) or DMSO control confirmed that while AnxA6 depletion in the invasive BT-549 cells significantly sensitized the cells to lapatinib, PD153035 and canertinib, AnxA6-depletion in MDA-MB-231 cells did not significantly alter their sensitivity to these compounds. Meanwhile, over expression of AnxA6 in HCC1806 did not alter their response to these EGFR-targeted TKIs. Together these data confirm the variable response of breast cancer cells to EGFR targeted therapies suggest that reduced AnxA6 expression may be more relevant in breast cancer subtypes such as EGFR expressing invasive basal-like breast cancer.

To support the above conclusion, we examined whether AnxA6 expression status also influences the survival of breast cancer patients with varied clinical disease. To do this we used the KM plotter, a recently reported publicly available online survival analysis tool [36] that has been used extensively to analyze the expression of 22,277 genes on the survival of 2,977 breast cancer patients [37‐39]. This analysis revealed that, AnxA6 expression status is not associated with the overall (OS), relapse-free (RFS) or distant metastasis-free (DMFS) survival of all breast cancer patients combined (Figure 7A). AnxA6 expression status also is not associated with the survival of patients with luminal breast cancer or those with different HER2, estrogen or progesterone receptor status. However, AnxA6 expression status is significantly associated with the survival of patients with basal-like breast cancer. As shown in Figure 7B, low AnxA6 expression is associated with a significantly higher RFS for patients with basal-like breast cancer (p = 0.023). High AnxA6 levels were on the other hand, are associated with higher OS (p = 0.0024) and DMFS (p = 0.019) for patients with this breast cancer subtype. A previous evaluation of multiple studies with more than 20,000 patients also showed that high EGFR expression is associated with lower RFS for patients with head and neck, ovarian, cervical, bladder and oesophageal cancers [15]. However, the prognostic value of EGFR expression in other cancers including breast cancer was found to be modest [15].

The following breast cancer cell lines BT-549, MDA-MB-231 (MDA-MB-231), HCC1806 were purchased from American Type Culture Collection (ATCC). MDA-MB-468 (MDA-MB-468) cells were kindly provided by Dr Ann Richmond, Vanderbilt Medical Center (Nashville, TN). These cell lines were cultured in DMEM/F12 containing 10% FBS, penicillin (100 units/ml), and streptomycin (50 units/ml). Cells were maintained at 37°C in a humidified CO₂ incubator. Where indicated, serum-starvation of cells was achieved by culturing the cells in DMEM/F12 containing 0.5% FBS for 24 to 48 h.

Antibodies against EGFR, phospho-EGFR (pY1068), phospho-extracellular signal-regulated kinase 1/2 (pERK1/2, T202/Y204) and phospho-Akt (pS473) were purchased from Cell Signaling Technology (Beverly, MA, USA). Antibodies against ERK2, Akt1, GAPDH and AnxA6 were purchased from Santa Cruz Biotechnology, Inc. Antibodies specific for the flag-epitope (anti-flag M2), β-tubulin and β-actin were obtained from Sigma. Except otherwise indicated, secondary anti-mouse, anti-goat and anti-rabbit horseradish peroxidase-conjugated antibodies were purchased from Santa Cruz Biotechnology, Inc or Sigma. EGFR Tyrosine Kinase Inhibitor Set including canertinib, erlotinib hydrochloride, gefitinib, lapatinib ditosylate and PD153035 hydrochloride was purchased from BioVision Inc. Sulfo-NHS-biotin and protease inhibitor cocktail were products from Sigma.

Small hairpin RNAs (shRNAs) targeting the coding sequence of AnxA6 in pGIPZ lentiviral vector, a non-silencing shRNA control or the empty vector (Open Biosystems Inc.), were used to transfect BT-549 and MDA-MB-231 BCCs using Lipofectamine 2000 (Invitrogen). The cells were selected with puromycin 48 h post-transfection for up to three weeks. Cells expressing high levels of GFP were then sorted by flow cytometry, cloned by limiting dilution, and the isolated clones were expanded in selection medium containing puromycin (2 μg /ml). The following shRNA sequences were used to target AnxA6 in BT-549 and MDA-MB-231 cells: A6sh2, 5'- TTCAGCATTGGTCCGAGTG-3'; A6sh5, 5'-TGTGTCTTCGTCAGTCCCG-3'. For experiments involving over expression of AnxA6, the coding sequence of AnxA6 variant 1 (accession No. NM_001155.4) was amplified from plasmid pCMV-Sport6-AnxA6 (Open Biosystems Inc.). The fragment was cloned into Hind III and Xho I linearized pCMV-3Tag-8 (Clontech), and the construct used to transfect the AnxA6-low HCC1806 BCCs using Fugene6 transfection reagent (Promega). The transfected cells herein designated 1806-Anx6 and 1806-EV were selected with hygromycin, cloned as above and expanded in continued hygromycin (100 μg/ml) selection.

Cells were plated sparsely on glass cover slips and allowed to grow until they formed colonies of a few cells. Cells were then serum-starved overnight and treated with or without EGF for 5 min. Indirect immunofluorescence staining was performed as previously described [5] using pEGFR (Y1068) antibodies and FITC-conjugated secondary antibodies. Cover-slips were mounted with ProLong Gold antifade containing DAPI (Invitrogen). Images were captured using a Nikon A1R confocal microscope with 60× oil-immersion objectives and analyzed using the NIS software.

Cells were grown in complete medium until they were 70% confluent, then serum-starved for 24 h and treated with or without EGF for the indicated time points. Following treatment, cells were washed twice with ice-cold phosphate-buffered saline (PBS) pH 7.4, with 0.5 mM Ca²⁺ and 1 mM Mg²⁺) and then incubated with Sulfo-NHS-biotin (0.2 mg/ml in cold PBS) for 30 min at 4°C. Unreacted biotin was quenched with ice-cold 100 mM glycine in PBS for 15 min at 4°C. Whole cell extracts were prepared in TNE lysis buffer (150 mm NaCl, 10 mm Tris–HCl pH 8.0, 1 mm EDTA, and 1% Nonidet P-40 with protease and phosphatase inhibitors). Biotinylated (cell surface) proteins isolated using Streptavidin-agarose beads and whole cell extracts were used for the detection of cell surface and total cellular EGFR respectively by Western blotting as previously described [5]. Immuno-reactive bands were visualized by enhanced chemiluminescence (ECL; Pierce) and quantified using NIH Image J software.

Breast epithelial and BCCs were cultured until they were 70% confluent then serum-starved overnight and treated with 50 ng/ml EGF (EMD Biosciences) in Hank’s balanced salt solution (HBSS) for the indicated time periods. The EGF treated cells were scraped in ice-cold PBS and total cell lysates prepared as described previously [5]. EGFR activation was detected by immunoblotting with anti-EGFR (pY1068) and antibodies to total EGFR. Activation of downstream signaling cascades was determined by Western blotting using ant-pErk1/2 (pT202/pY204) and anti-pAkt (S473). Immunoblotting with antibodies to either anti-Erk2, anti GAPDH or anti-β-tubulin were used as the loading controls. Immuno-reactive bands were revealed by ECL, scanned and quantified using NIH Image J software. Activation levels were determined as the ratios of phosphoprotein to the total protein or loading controls.

The effects of AnxA6 depletion and TKIs on cell growth were performed in 24-well plates in triplicates using 1 x 10⁴ cells/well, as previously described [5]. The proliferation and viability of the cells were determined by incubating the cells in 1:10 diluted PrestoBlue reagent (Invitrogen) in HBSS supplemented with 1 mM Ca²⁺ and 0.5 mM Mg²⁺ for 2–4 h. Cell proliferation was determined by fluorescence measurement following excitation at 540 nm and emission at 600 nm.

Clonogenic and motility assays were performed in duplicate as previously described [5]. Digital images of the 3D cultures were captured at x10 magnification using DCM200 digital camera and Scopephoto software. For motility assays, cells were counted from at least 5 separate fields per insert.

The online KM plotter was used to compare the impact of AnxA6 expression on the survival of 2,977 breast cancer patients according to the set parameters [36]. In order to analyze the prognostic value of a particular gene, the cohorts are divided into two groups according to the median (or upper/lower quartile) expression of the gene. A survival curve is displayed, and the hazard ratio with 95% confidence intervals and logrank P value are calculated and displayed [36]. We tested the effect of high or low AnxA6 expression on the overall, distant metastasis-free and recurrence-free survival of either all patients or patients with various breast cancer molecular subtypes.

Data were analyzed using Microsoft Excel 2007. Except otherwise indicated data were presented as mean ± SD. Data were analyzed using Student’s t-test; a p-value < 0.05 was considered statistically significant.