Rack1 mediates tyrosine phosphorylation of Anxa2 by Src and promotes invasion and metastasis in drug-resistant breast cancer cells

The following antibodies, reagents, and drugs were used: DMEM/F12, RPMI 1640, DMEM/High glucose medium and trypsin (Hyclone, Logan, UT, USA); Fetal Bovine Serum (FBS, Gibco, Carlsbad, CA, USA); Src inhibitor KX2-391 (Selleckchem, Houston, TX, USA); transwell inserts (Corning Inc., Corning, NY, USA); EGF (Cat No. 236-EG, R&D, USA); Matrigel (BD Biosciences, San Jose, CA, USA); Protein A/G agarose beads (Invitrogen, Carlsbad, CA, USA); Lipofectamine RNAiMax and Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA); antibodies against Rack1 (sc-17754), Anxa2 (sc-28385), p-Anxa2 (sc-135753), GFP (sc-9996, Santa Cruz Biotechnology, Santa Cruz, CA, USA); antibodies against Src (# 2123) and p-Src (# 6943, Cell Signaling Technology, Beverly, MA, USA); antibodies against Flag (F1804) and β-actin (A1978, Sigma, St. Louis, MO, USA); and Rack1, Anxa2, and Src small interfering RNAs (siRNAs, Invitrogen, Carlsbad, CA, USA, detailed information is listed in Table 1).

Human triple-negative breast cancer (TNBC) MDA-MB-468 cells were obtained from American Type Culture Collection (ATCC). The drug-resistant variant cell line MDA-MB-468/EPR was established by our group using a stepwise exposure to an increasing concentration of epirubicin. The human luminal-type breast cancer MCF-7 cells and its drug-resistant cell line MCF-7/ADR was kindly provided by Dr. Zizheng Hou (Henry Ford Hospital, Detroit, MI, USA). Anxa2 stably silenced MCF-7/ADR and control cells have been established in our previous study [18]. HEK-293T cells were obtained from ATCC. The following media supplemented with 10% FBS were used for cell culture: DMEM/F12 for MDA-MB-468/EPR, modified RPMI-1640 for MCF-7/ADR, and DMEM/High glucose for HEK-293T. For siRNA transfection, cells were seeded into six-well plates and cultured to 30–40% confluence. Afterwards, control, Rack1-, Anxa2-, and Src-specific siRNAs were transfected using Lipofectamine RNAiMAX reagent according the manufacturer’s instructions.

Flag-tagged, full-length, and wild-type Rack1 (Rack1^WT) was generated by polymerase chain reaction (PCR) amplification with the following primers: upper: 5′-GAGAGCTAGCATGACTGAGCAGATGACCCT-3′, lower: 5′- GAGAGCGGCCGCCTACTTGTCGTCATCGTCTTTGTAGTCGCGTGTGCCAATGG-3′. Afterwards, the Rack1-coding region was cloned into a lentiviral vector pCDH-hygromycin in the Nhe I and Not I cloning sites. The Rack1^Y246F mutant was generated by mutating the codon TAC (Y) of its amino acid 246 to TTC (F), where tyrosine 246 was replaced by phenylalanine. For Rack1 stable knockdown, a validated siRNA sequence targeting the Rack1 non-coding region (5′-TGGCACACGCTAGAAGTTTATGG-3′) was cloned into the lentiviral vector pLKO.1-hygromycin of the BamH I and Age I cloning sites. The lentiviral plasmids pCDH-Anxa2^WT-GFP, pCDH-Anxa2^Y23A-GFP, and pCDH-Anxa2^Y23D-GFP were obtained in our previous study. In brief, wild-type Anxa2 (Anxa2^WT) was generated by PCR amplification with the following primers: upper: 5′-CGGCTCGAGATGTCTACTGTTCACGAAATCCTG-3′, lower: 5′-CGTGGATCCGTCATCTCCACCACACAG-3′. Afterwards, the Anxa2-coding region was cloned into pEGFP-N3 vector in the Xho I and BamH I cloning sites. Anxa2^Y23A-GFP and Anxa2^Y23D-GFP mutants were generated by mutating the codon TAT (Y) of its amino acid 23 to GCT (A) or GAT (D). Then, the Anxa2-GFP fragments were further subcloned into the lentiviral vector pCDH-hygromycin in the Nhe I and Not I cloning sites. All constructs were confirmed by double-enzyme digestion and DNA sequencing. Lentivirus production was performed as described previously by using a three-plasmid packaging system. In brief, HEK-293T cells were plated in 10-cm dish and cultured to 60% confluence. Afterwards, the cells were co-transfected with the lentiviral plasmid and two packaging plasmids. The virus in the supernatants were collected, concentrated, and used to infect target cells 48 h after transfection. Stable cell lines were selected by using 50 μg/mL of hygromycin B.

Western blot assay was performed as described previously [44]. Briefly, the cells were lysed with 1× SDS lysis buffer; the cell lysates after protein quantification were used for SDS-PAGE separation and subsequently transferred to a PVDF membrane. The membranes were blocked by 5% non-fat milk and probed with corresponding primary antibodies. After washing with 1× TBST, the membrane was incubated with horseradish peroxidase-labeled secondary antibodies, and the signals were subsequently detected using the ECL kit. β-actin was used as a loading control. For EGF-induced Anxa2 phosphorylation assay, control, Rack1, or Src siRNA transfected cells were starved for 8 h. Then, the cells were stimulated with 10 ng/mL of EGF for 0, 5, and 10 min. The cells were lysed, and the total cellular protein was analyzed by using Western blotting analysis. Co-IP assay was performed as described previously [28]. Briefly, MCF-7/ADR and MDA-MB-468/EPR cells were transfected with negative control, Src-, Anxa2-, or Rack1-specific siRNAs for 72 h. Then, the cells were lysed with Tris-Trion X100-based lysis buffer. Afterwards, the quantified cell lysates were pre-cleared with protein G-linked agarose beads, followed by incubation with corresponding antibodies (Anxa2, Rack1, or Src) to enrich immunocomplex. The interacting proteins were captured with Protein A/G conjugated beads and subsequently analyzed by Western blotting with anti-Anxa2, Rack1, or Src antibodies.

Wound healing assay was carried out as described previously [12]. The control and experimental cells were planted in six-well plates and grown to confluence. Afterward, a wound was prepared by scraping the cell monolayer using a 10-μL pipette tip. Afterward, the cells were incubated at 37 °C in a medium containing 0.5% FBS, and the cells were allowed to migrate for 0, 12, 24, 36, 48, and 60 h. The wound areas were photographed under an inverted microscope in indicated times, and the relative migrated distance was calculated with ImageJ. Transwell assays were performed by using Boyden chambers with a 8-μm filter pore. For migration assay, 1 × 10⁵ cells in 200 μL serum-free medium were loaded into the upper chamber. The lower chamber was loaded with 600 μL of 10% FBS-containing medium. After incubation at 37 °C for 16 h, the migrated cells were fixed, stained, and quantified at 200×. For invasion assay, 2.5 × 10⁵ cells were loaded into the upper insert coated with Matrigel, and the incubation time was 24 h. To examine the effect of Src inhibitors on migration and invasion of drug-resistant cells, the cells were pre-treated with Src inhibitors for 2 h. Then, transwell assays were performed in the absence (for migration) or presence (for invasion) of Matrigel.

Six-week-old female SCID mice were purchased from Beijing Charles River. All operations followed the guidelines approved by the Animal Ethical and Welfare Committee of Tianjin Medical University Cancer Institute and Hospital. For metastasis assay, control and Rack1 stably silenced cells, as well as the Rack1 rescued cells, were cultured to log phase. Then, the cells were trypsinized, washed thrice with PBS, resuspended, and adjusted to a concentration of 2 × 10⁷ cells/mL. 2 × 10⁶ cells (n = 8 per group) were injected into SCID mice via tail veins. Three months after injection, the mice were sacrificed; the lungs were dissected, fixed with formalin, embedded in paraffin, and serially sectioned. After hematoxylin and eosin staining, the micrometastasis foci in lungs were visualized and counted under a microscope.

Standard protocols are detailed in Additional file 1: Supplementary methods.

All data were shown as mean ± SD. The statistical differences were analyzed using the GraphPad Prism 7.00 software (GraphPad Software, La Jolla, CA, USA). For multiple sets of comparisons, one-way or two-way ANOVA was performed. P values less than 0.01 (two-tailed) were considered statistically significant. *indicates that the P value is less than 0.01 and ** means P value is less than 0.001.

To determine whether Rack1 is necessary for Anxa2 tyrosine phosphorylation, we silenced the expression of Rack1 in two drug-resistant breast cancer cell lines using two different Rack1-specific siRNAs. As shown in Fig. 1a, Rack1 expression was remarkably downregulated in Rack1 siRNA transfected cells compared with that of the control siRNA transfected group. The level of pY23-Anxa2 was notably decreased in Rack1-silenced cells than in the control cells. Anxa2 tyrosine phosphorylation can be induced by growth factors, such as EGF [13, 15]. We examined the effect of Rack1 knockdown on EGF-induced Anxa2 phosphorylation. As shown in Fig. 1b, Rack1 knockdown attenuated the increase of pY23-Anxa2 induced by EGF in two drug-resistant cells, while the effect of Rack1 silencing on pY23-Anxa2 was evident in MDA-MB-468/EPR cells compared to MCF-7/ADR cells. This variance may be due to the differences in the genetic background between the two cell lines, such as the expression level of endogenous EGFR (Additional file 2: Figure S1), which is higher in MCF-7/ADR cells. Next, we further investigated the function linkage between Rack1 knockdown and cell migration and invasion ability. As shown in Fig. 1c, the knockdown of Rack1 expression in two drug-resistant cells significantly decreased cell migration ability as measured by wound healing assay. Similarly, the results from transwell assay showed that the migration and invasion abilities were significantly inhibited in Rack1-silenced cells compared with control cells (Fig. 1d). To exclude the effect of cell death on migration and invasion, we investigated the effect of Rack1 knockdown on the apoptosis of resistant cells by flow cytometry using Annexin V-FITC/PI double staining method. As shown in Additional file 2: Figure S2, silencing the expression of Rack1 had no significant effect on apoptosis in resistant cells compared to control cells. Therefore, the decrease of cell migration/invasion ability after Rack1 knockdown is not due to the increased incidence of cell death. Collectively, these data demonstrated that Rack1 silencing inhibited Anxa2 tyrosine phosphorylation along with decreased cell migration and invasion abilities.

Src is a well-known upstream kinase of Anxa2 [45‐47]. Therefore, to investigate whether the decreased level of pY23-Anxa2 is associated with the declined cell invasion ability in drug-resistant cells, we blocked Src kinase activity in drug-resistant cells by using Src kinase inhibitor KX2-391. As shown in Fig. 2a, the inhibitor efficiently inhibited the phosphorylation of Src at the tyrosine 416 site, indicating the blockage of this kinase activity. Meanwhile, the level of pY23-Anxa2 was remarkably decreased. Figure 2b shows that the cell invasion ability was significantly suppressed in the Src inhibitor-treated group compared with the control group. Moreover, we silenced the expression of Src in two drug-resistant cells by using two different siRNAs, as shown in Fig. 2c. Src expression was significantly downregulated after transfection of two different siRNAs. Moreover, Src knockdown significantly inhibited EGF-induced tyrosine phosphorylation of Anxa2. While the decrease in the level of pY23-Anxa2 after Src knockdown in MDA-MB-468/EPR cells was not as evident as that in MCF-7/ADR cells, this difference may be due in part to the fact that Src may play other roles in MDA-MB-468/EPR cells in addition to modulating Anxa2 phosphorylation or that a fraction of Anxa2 may also be phosphorylated by other unknown kinases. Moreover, transwell assay showed that the migration and invasion abilities in Src knockdown group were significantly decreased compared with that in control cells (Fig. 2d). Collectively, these findings demonstrated that Src kinase is required for Anxa2 tyrosine phosphorylation, and this result is correlated with the invasiveness of drug-resistant cells.

To determine whether Anxa2 tyrosine phosphorylation is required for the invasiveness of drug-resistant cells, lentivirus expressing GFP-tagged wild-type Anxa2 (Anxa2^WT) as well as its two mutants, Anxa2^Y23A (phospho-deficient mutant) and Anxa2^Y23D (phospho-mimicking mutant), were used to infect Anxa2-silenced MCF-7/ADR cells, which were constructed in our previous study [18]. As shown in Fig. 3a, the expression of Anxa2 was effectively rescued in Anxa2-depleted MCF-7/ADR cells, as verified by Western blot analysis using anti-Anxa2 or anti-GFP-specific antibodies. Consistent with previous reports, transwell assays showed that Anxa2 knockdown significantly inhibited the migration and invasion abilities compared with control cells (Fig. 3b). The re-expression of Anxa2^WT in Anxa2-silenced MCF-7/ADR cells restored cell migration and invasion abilities, whereas the re-expression of Anxa2 ^Y23A mutant or GFP failed to rescue cell motility (Fig. 3b). Moreover, the rescued expression of Anxa2^Y23D mutant even showed an increased migration and invasion abilities compared with that of Anxa2^WT-expressing cells (Fig. 3b). In summary, these data suggested that Anxa2 tyrosine phosphorylation is required for the invasiveness of drug-resistant breast cancer cells.

We previously showed that the scaffold protein Rack1 forms a complex with Src and Anxa2 in drug-resistant cells [28]. We proposed that Rack1 may mediate the binding between Src and Anxa2, given that Rack1 and Src can bind to Anxa2 and modulate its tyrosine phosphorylation. To test this hypothesis, we first examined the effect of Rack1 knockdown on the binding ability of endogenous Src to Anxa2 in two drug-resistant cells through a Co-IP assay using anti-Src or Anxa2 antibodies. As shown in Fig. 4a, Rack1 knockdown conferred no evident effect on the expression levels of Src and Anxa2, whereas the interaction between Src and Anxa2 in two Rack1-silenced cells was notably attenuated compared with that in control cells (Fig. 4a). Consistently, a reciprocal Co-IP assay using anti-Anxa2 antibodies also showed that the binding ability of Anxa2 to Src was decreased in two Rack1-knockdown cells compared with control cells. Therefore, these data suggest that Rack1 is required for the binding of Src to Anxa2. In addition, to investigate whether Src is necessary for the interaction between Rack1 and Anxa2, we silenced Src expression in two drug-resistant cells by using siRNAs and performed Co-IP assay with anti-Rack1 or Anxa2 antibodies. As shown in Fig. 4b, the expression levels of Rack1 and Anxa2 were unchanged after knockdown of Src in the two drug-resistant cells. The amount of Anxa2 co-precipitated by anti-Rack1 antibodies in cell lysates from control cells was similar to that from Src knockdown cells (Fig. 4b). Similarly, a reciprocal Co-IP assay with anti-Anxa2-specific antibodies also confirmed that the binding ability of Anxa2 to Rack1 was not altered in the absence of Src (Fig. 4b). These results suggest that the binding of Anxa2 to Rack1 is unaffected by Src knockdown. Moreover, we further determined whether Anxa2 is necessary for the interaction of Rack1 and Src. As shown in Fig. 4c, Anxa2 knockdown did not affect the expression levels of Src and Rack1. Co-IP assay with anti-Src antibodies showed that the amount of Rack1 bound to Src was the same in the control and Anxa2-silenced cells (Fig. 4c). Consistently, anti-Rack1 antibodies also co-precipitated a similar amount of Src, indicating that the interaction of Rack1 and Src was independent of the presence of Anxa2. These results suggested that Rack1 functions as a scaffold protein and mediates the interaction between Src and Anxa2.

The findings prompted us to further confirm whether Rack1 modulates Anxa2 phosphorylation and cell invasion through its binding to Src. To validate this speculation, Rack1 stable knockdown in two drug-resistant cells was established by using a n shRNA, specifically targeting to its noncoding region. Then, we rescued Rack1 expression in Rack1-silenced cells through infection of lentivirus expressing Flag-tagged Rack1^WT and Src binding-deficient Rack1^Y246F mutant. As shown in Fig. 5a and b, Rack1 expression was effectively rescued in Rack1-silenced cells to levels similar to those in endogenous proteins, as verified by Western blot analysis using anti-Rack1 or anti-Flag antibodies. Re-expression of Rack1^WT in Rack1-silenced cells rescued Anxa2 tyrosine phosphorylation compared with that of control cells. However, rescuing with the Rack1^Y246F mutant failed to recover Anxa2 phosphorylation. Therefore, these results suggested that Anxa2 tyrosine phosphorylation was regulated by Src in a Rack1-dependent manner. Next, we compared the cell migration and invasion ability among these cells. As shown in Additional file 2: Figure S3, wound healing assay showed that re-expression of Rack1^WT significantly increased cell migration ability. On the contrary, re-expressing Rack1^Y246F failed to rescue the cell migration defects caused by Rack1 knockdown. Similarly, transwell-based assay showed that rescuing with Rack1^WT, but not with Rack1^Y246F mutant, recovered cell invasion ability (Fig. 5c, d). These data suggested that Rack1 mediates tyrosine phosphorylation of Anxa2 by Src and promotes invasion in drug-resistant breast cancer cells.

To further determine whether the decreased invasiveness of Rack1-silenced cells occurred mechanically through inhibiting Anxa2 phosphorylation, we introduced GFP-tagged Anxa2^WT or its two mutants into Rack1 stable knockdown cells using lentivirus. As shown in Fig. 6a and b, the expression level of Anxa2 and its mutants were successfully increased in Rack1-silenced cells, as measured by Western blot analysis using anti-Anxa2 or anti-GFP antibodies. The pY23-Anxa2 was slightly higher in Anxa2^WT-expressing MCF-7/ADR cells than in the Rack1 knockdown cells, this is partly due to the total level (endogenous Anxa2 and Anxa2-GFP) of Anxa2 was higher in Anxa2^WT-expressing cells than in the Rack1 knockdown cells, and the presence of less active Src may still phosphorylate a small fraction of Anxa2 in these cells, while the level of pY23-Anxa2 was apparently lower in Anxa2^WT-expressing cells than in control and Anxa2^Y23D-expressing cells. Then, we performed wound healing assay to analyze the cell migration ability. As shown in Additional file 2: Figure S4, the increased expression of Anxa2^WT or Anxa2^Y23D significantly enhanced cell migration ability compared with that of GFP-expressing cells, whereas the increased expression of Anxa2^Y23A mutant failed to rescue the cell migration ability. Consistently, transwell-based assays showed that elevated expression of Anxa2^WT or Anxa2^Y23D in Rack1-knockdown cells notably recovered cell invasion ability. On the contrary, increased expression of Anxa2^Y23A mutant failed to restore the migration and invasion abilities (Fig. 6c, d), although the rescued cell migration/invasion ability in Anxa2^WT- or Anxa2^Y23D-expressing cells cannot reach to that of wild-type cells. Moreover, we further treated Anxa2^WT- and Anxa2^Y23D-expressing MCF-7/ADR cells with Src inhibitor KX2-391. As shown in Fig. 6e, KX2-391 efficiently inhibited the expression of pY23-Anxa2 in Anxa2^WT-expressing cells, while the level of pY23-Anxa2 in Anxa2^Y23D-expressing cells was not affected by Src kinase inhibitor. In addition, transwell-based assay showed that the migration ability of Anxa2^WT-expressing cells can be quenched by Src inhibitor, while Src inhibition has little effect on the migration ability in phospho-mimicking Anxa2^Y23D-expressing cells (Fig. 6f). These results suggested that the decreased cell migration and invasion abilities in Rack1-silenced cells were attributed, at least in part, to the inhibition of Anxa2 tyrosine phosphorylation.

To further investigate the functions of Rack1 on the metastatic potential of drug-resistant breast cancer cells in vivo, we used pulmonary metastasis model in NOD-SCID mice via tail vain injection of MCF-7/ADR cells, control, and the Rack1 stable knockdown cells, as well as the Rack1^WT- and Rack1^Y246F-rescued cells. Three months after injection, apparent reduction of tumor metastatic foci was shown in the lung surface of Rack1-silenced group compared with the control group (Fig. 7a). In addition, the lung surface rescued with the Rack1^WT group displayed more metastatic foci than that in control and Rack1^Y246F-rescued group (Fig. 7a). Moreover, we stained mice lung tissue sections with hematoxylin and eosin to confirm the metastatic foci. As shown in Fig. 7b and c, the size and the number of tumor foci in the lung were significantly decreased in Rack1-silenced group compared with those in the control group. Furthermore, the rescued expression of Rack1^WT, but not Rack1^Y246F mutant, in Rack1-silenced cells increased the number of metastases in the lung compared with that of the control group. Collectively, these data demonstrated that Rack1 is critical for the metastatic potential of drug-resistant breast cancer cells in vivo.