HER2/EGFR–AKT Signaling Switches TGFβ from Inhibiting Cell Proliferation to Promoting Cell Migration in Breast Cancer

TGFβ signaling plays a critical role in breast cancer development. In the early stages of breast cancer formation, TGFβ signaling suppresses tumor initiation by blocking cell proliferation, while at the late stage of breast cancer development, TGFβ signaling can enhance cancer progression and malignancy through promoting epithelial–mesenchymal transition (EMT) and cell migration, sustaining cancer stem cells and modulating tumor microenvironment (1–5). These biological effects of TGFβ are achieved via Smad-mediated transcription regulation (6, 7) or non-Smad pathways (8, 9). Smad2 and Smad3, which are activated by the TGFβ–type I receptor (TβRI)-mediated C-terminal serine phosphorylation, regulate gene expression in the nucleus in cooperation with Smad4 and other factors. Furthermore, phosphorylation of serine/threonine residues in the middle linker region by various kinases has been shown to regulate the stabilization, nuclear localization, and transcription activity of Smads (10, 11).

HER2 (ErbB2, Neu), which is a member of the epidermal growth factor (EGF) receptor family, belonging to the superfamily of tyrosine kinase receptors (RTK), is amplified and overexpressed in about 30% breast cancer, correlated with poor patient prognosis (12). HER2 can form homodimers to activate downstream signaling independent of ligands and cause cell transformation (13). The activity of phosphoinositide 3-kinase (PI3K)/AKT, major downstream mediators of HER2/EGFR signaling, is often amplified in breast cancers (14).

The cross-talk between TGFβ signaling and HER2/EGFR signaling has been studied to a certain extent during breast cancer development (15, 16). In HER2-positive breast cancer cells or HER2-overexpressing breast epithelial cells, the antiproliferative effect of TGFβ is attenuated, while TGFβ–mediated migration and invasion are enhanced (17–19). Moreover, higher metastasis and less apoptosis were observed in the transgenic mice with overexpression of both HER2 and constitutively active TβRI (20, 21). However, the underlying molecular mechanism is largely unclear. Here, we report that HER2/EGFR can switch the function of TGFβ signaling from growth inhibition to migration promotion in breast cancer cells including triple-negative breast cancer cells. HER2/EGFR signaling via AKT stimulates phosphorylation of Ser208 in the linker region of Smad3 and thereby increased its nuclear accumulation, and the nuclear Smad3 activates the expression of the genes related to EMT and cell migration.

The human HER2 and AKT were subcloned into pcDNA-4TO-myc, pCDNA3.1⁺-HA, and pEGFP-N1 plasmids. Human TβRI cDNA was subcloned into pCMV5-HA plasmid. TβRI(GGD) is the constitutively active form with three point mutations (L193G, P194G, and T204D; ref. 22). The shRNAs for Smad2, Smad3, and NUAK1 were obtained from the human shRNA libraries (Sigma), and the sequences are shown in Supplementary Table S1. Antibodies and inhibitors used in this study were shown in Supplementary Tables S2 and S3. Herceptin was obtained from Alpha Diagnostic International. Recombinant human TGFβ1 was purchased from R&D Systems, EGF and FGF from Invitrogen, and insulin was from Sigma. Protein A-Sepharose or Glutathione Sepharose 4B were from GE Healthcare.

Cell lines shown in Supplementary Table S4 from ATCC or China Infrastructure of Cell Line Resource were authenticated using short tandem repeat (STR) profiling. KYSE410 was from SAILY BIO with STR profiling. All the cell lines had been confirmed without Mycoplasma contamination by PCR using the primers: 5′-GGGAGCAAACACGATAGATACCCT-3′ and 5′-TGCACCATCTGTCACTCTGTTAACCTC-3′. NMuMG and MCF-10A cells were used with less than four passages, and other cells were less than 10 passages after thawed.

Transfection of DNA plasmids into HEK293T, NMuMG, MCF-10A, and MDA-MB-231 cells was performed with Lipofectamine 2000 (Invitrogen) or Vigofect (Vigorous Biotechnology). ShRNAs were expressed in MDA-MB-231 cells with a lentiviral system (Sigma), and the stable cell lines were selected with 0.3 μg/mL puromycin (Sigma). Wild-type Smad3 or its mutants were cloned into pENTR1A plasmid under the control of CMV promoter in the p2k7_bsd lentiviral backbone (gift from Dr. Kehkooi Kee, Tsinghua University, Beijing, China). The viral supernatants were used to infect breast cell lines. Blasticidin (3 μg/mL; Sigma) was added into culture medium for 5 days to select cells with stable viral integration.

These experiments were described as previously reported (22). Some of the immunoblots were reprobed after stripping (stripping buffer: 100 mmol/L β-mercaptoethanol, 2% SDS, 62.5 nmol/L Tris–HCl pH6.8). For immunofluorescence, the following primary antibodies were used: anti-Smad2 (1:100), anti-Smad3 (1:100), anti-Flag (1:400), anti-F-actin (1:100), anti-E-cadherin (1:500), and anti-GFP (1:200).

For reporter assay, luciferase activity was measured 36 hours after transfection with the dual luciferase reporter assay system (Promega) following the manufacturer's protocol. The experiments were repeated in triplicate, and the data are represented as the means ± SD after normalization to Renilla activity.

For wound-healing assay, the confluent cell monolayer in 12-well plate was wounded by scraping the cells with a yellow 20 μl pipette tip and then treated with growth factors or inhibitors for indicated time. Quantitation was done by measuring the migrating wound gap width and compared with the start wound from three independent experiments. For Transwell assay, cells were plated in medium with 1% FBS and inhibitors in the upper chamber of a Transwell plate (Corning), while the media containing 10% FBS, inhibitors, or TGFβ were placed in the lower well. After incubation for indicated time, cells were fixed with methanol for 20 minutes, stained with crystal violet dye for 30 minutes, and photographed. Experiments were carried out at least three times.

MDA-MB-231 cells were pretreated with DMSO, 100 pmol/L TGFβ1, 5 μmol/L lapatinib, or TGFβ1 + lapatinib for 4 hours, and then crosslinked with 1% formaldehyde solution before lysis. The DNA/protein complex-containing lysates were then sonicated to obtain DNA fragments ranging 500 to 1,000 bp on average. After centrifuged at 12,000 rpm for 10 minutes, the supernatant was diluted and preabsorbed by 50 μL Protein A beads (Zymed) and then incubated with 10 μg anti-p-Smad3 or anti-Smad2 antibodies overnight at 4°C. The immunocomplexes were collected with 100 μL protein A beads, eluted, and reverse crosslinked by overnight incubation at 65°C. Whole-cell extract DNA (input fraction reserved from the sonication step) was also treated for crosslinking reversal. Immunoprecipitated DNA and total DNA were then purified by treatment with RNase A, proteinase K, and phenol:chloroform:isoamyl alcohol extraction. PCR primers for chromatin immunoprecipitation (ChIP) are shown in Supplementary Table S5.

Total cell RNAs were isolated with TRIzol (Invitrogen), and cDNA was synthesized with Revertra Ace (Toyobo). Real-time PCR was performed with a Roche480 system (Roche). Expression values were normalized to GAPDH expression. The PCR oligo sequences are shown in Supplementary Table S5.

Cells with different treatments were harvested and subjected to the nucleus extraction kit (Applygen Technologies Inc.) for the nuclear and nonnuclear constituent extraction. The proteins in nuclear and nonnuclear fractions were analyzed by immunoblotting.

About 3,000 cells were placed in the 48-well plates and then subjected to different treatments. Every day, the cell numbers were counted with the CCK-8 cell counting kit (Dojindo Molecular Technologies).

Recombinant human AKT1 protein (Abcam; 200 ng) was incubated with 2 μg purified Smad3 protein in a 50 μL-reaction buffer containing 25 mmol/L Tris–HCl (pH 7.5), 150 mmol/L NaCl, 10 mmol/L MgCl₂, 1 mmol/L DTT, and 5 mmol/L ATP for 1 hour at room temperature. The reaction was stopped by adding the SDS loading buffer. The protein phosphorylation was detected by immunoblotting.

GST-Smad3 WT and mutant (S208A and S208E) proteins were prepared from Escherichia colistrain DE3. RanBP3 and importin-β1 proteins were generated using the Quick Coupled transcription/translation system (Promega). Smad3 proteins, RanBP3, and importin-β1 were incubated for 2 hours in the binding buffer (0.5% NP-40, 150 mmol/L NaCl, 50 mmol/L Tris–HCl, 5 mmol/L EDTA) and monitored by immunoblotting.

For lung metastasis formation, 2 × 10⁵ MDA-MB-231 cells stably expressing firefly luciferase pretreated with DMSO, 500 pmol/L TGFβ, 10 μmol/L lapatinib or TGFβ + lapatinib for 24 hours were injected into the lateral tail vein of nude mice (n = 6 per group) in a volume of 0.1 mL. To measure metastasis ability of SKBR3 cells, 1 × 10⁶ cells pretreated with DMSO, 100 pmol/L TGFβ, 0.01 μmol/L lapatinib, or TGFβ + lapatinib for 24 hours were injected into the lateral tail vein of nude mice (n = 6 per group) in a volume of 0.1 mL. For 4T1 cells, 1 × 10⁶ cells stably expressing firefly luciferase pretreated with DMSO, 100 pmol/L TGFβ, 2 μmol/L lapatinib, or TGFβ + lapatinib for 24 hours were injected into the lateral tail vein of BALB/C mice (n = 5 per group) in a volume of 0.15 mL. Bioluminescence images were collected after 21 to 30 days by using the Lumina II imaging system (PerkinElmer). The color scale depicts the photon flux (photons per second) emitted from xenografted mice. After sacrifice, the lungs or livers were excised for metastatic foci analysis, and the number of metastatic nodules of each group was analyzed. *, P < 0.05; **, P < 0.01; ***, P < 0.001. For subcutaneous tumor formation, 3 × 10⁶ MDA-MB-231 cells pretreated with DMSO, 100 pmol/L TGFβ, 5 μmol/L lapatinib, or TGFβ + lapatinib for 24 hours in 0.1 mL of PBS were inoculated in the flank of the mice. After 30 days, the tumors were excised, and fixed in 4% paraformaldehyde and imbedded in paraffin. Paraffin sections were cut at a thickness of 7 μm for IHC analysis. All animal experiment protocols were approved by the Animal Ethical Committee of Tsinghua University. Balb/c nude mice of 5 weeks old were used for subcutaneous tumor formation studies.

All tumors were primary and untreated before surgery. All subjects received radical mastectomy or modified radical mastectomy. The axillary lymph nodes were routinely dissected, and lymph node metastasis was determined based on histologic examination. Tumor size was defined as the maximum tumor diameter measured on the tumor specimens at the time of operation. Histologic types of the total 223 samples were defined according to the WHO classification criteria into grade I (23 cases), grade II (136 cases), and grade III (74 cases). Clinical stage was defined according to the sixth edition of AJCC classification criteria. They were formalin-fixed and paraffin-embedded tumor specimens, including 156 cases of invasive ductal breast cancer and 67 cases of ductal carcinoma in situ at a tissue bank maintained in the Department of Pathology of the First Affiliated Hospital of Nanchang University, Jiangxi, China. The protocol to obtain the clinical samples and use of these samples were approved by the Institutional Review Board at the First Affiliated Hospital of Nanchang University, and the studies were conducted in accordance with the ethical guidelines of the Declaration of Helsinki. Written informed consent was obtained from the patients.

IHC staining was done on 3-μm thick sections. Hematoxylin and eosin (H&E) staining was reviewed to ensure the cancer tissue and normal mammary epithelia adjacent to cancer were in the same section. Tissue slides were deparaffinized with xylene and rehydrated through a graded alcohol series and purified water. The slides were incubated in 3% hydrogen peroxide solution for 10 minutes and autoclaved in 10 mmol/L sodium citrate buffer (pH 6.0) for 2.5 minutes. After washed with PBS, the slides were incubated with the primary antibodies (p-Smad3, 1:50; HER2, 1:200; p-Y1248, 1:200) overnight at 4°C, and then with the second antibody for 30 minutes at 37°C, and then processed according to the DAB Detection Kit (Polymer). All of the slides were observed by at least 2 independent investigators blinded to the histopathologic features and patient data of the samples. The widely accepted German semiquantitative scoring system was used for assessing the staining intensity and area extent. Each specimen was assigned a score according to the intensity of the nucleic or membrane staining (no staining, not detected = 0; weak staining, light yellow = 1; moderate staining, yellowish brown = 2; strong staining, brown = 3) and the extent of stained cells (0% = 0, 1%–24% = 1, 25%–49% = 2, 50%–74% = 3, 75%–100% = 4). The final immunoreactive score was determined by multiplying the intensity score with the extent of score of stained cells, ranging from 0 (the minimum score) to 12 (the maximum score).

The Student t test, Kruskal–Wallis test, χ² test, Mann–Whitney U test, and log-rank test were used as indicated in the figure legends. *, P < 0.05 were considered statistically significant. *, P < 0.05; **, P < 0.01; ***, P < 0.001. For the bar charts, data were plotted as mean ± SD from at least 3 independent experiments. For immunoblot gel quantification, the band intensity of total proteins was normalized to loading control (GAPDH or tubulin) and the band intensity of phosphorylated proteins normalized to the total proteins after the gels were scanned and band intensity was qualified with ImageJ.

Previous reports indicate that HER2 can cooperate with TGFβ to enhance cell migration and invasion while attenuating TGFβ–induced growth inhibition (16–18, 23). To confirm this, we examined the effect of TGFβ on the proliferation of HER2-postive human breast cancer SKBR3 and MDA-MB-453 cells. As shown in Fig. 1A, TGFβ1 alone had no apparent effect on cell proliferation, but dramatically inhibited it in the presence of lapatinib. In contrast, TGFβ lost the ability to promote the migration of both SKBR3 and MDA-MB-453 cells in the presence of lapatinib (Fig. 1B). Similar results were obtained with Herceptin (Supplementary Fig. S1A–S1B), which is a recombinant humanized monoclonal antibody of HER2. We also explored the TGFβ responsiveness of triple-negative breast cancer cell lines MDA-MB-231 and SUM159, which have no immunohistochemically detectable expression of HER2, estrogen receptor, and progesterone receptor (24, 25), but HER2 expression is detectable by immunoblotting as shown in Supplementary Fig. S1C and in other reports (26, 27). TGFβ1 had limited or no apparent effect on cell proliferation, while lapatinib significantly enhanced the cytostatic effect of TGFβ1 (Fig. 1C). Similarly, TGFβ lost the ability to induce migration of these cells in the presence of lapatinib (Fig. 1D and E; Supplementary Fig. S1D–S1F). Similar results were also obtained with the HER2 inhibitor CP724714 or the EGFR inhibitor gefitinib (Supplementary Fig. S1G–S1I). In addition, HER2 overexpression or EGF treatment could further enhance TGFβ or constitutively active TβRI-mutant TβRI(GGD)-induced migration of MDA-MB-231 cells (Fig. 1F; Supplementary Fig. S1J–S1K). Similarly, TGFβ1 inhibited proliferation of MCF-10A cells while HER2 overexpression could attenuate this effect (Fig. 1G). HER2 overexpression could also enhance TGFβ–stimulated migration of MCF-10A cells (Fig. 1H). These results suggest that HER2/EGFR activity is critical for TGFβ to stimulate migration of breast cancer cells, including triple-negative breast cancer cells, while inhibition of HER2/EGFR activity enhances the cytostatic effect of TGFβ.

As reported previously (17, 28), TGFβ induced EMT in MCF-10A, NMuMG, and MDA-MB-231 cells as evidenced by cell morphologic change, stress fiber formation, E-cadherin loss, and upregulation of mesenchymal markers (N-cadherin, Snail, PAI-1, twist, vimentin, and fibronectin), and this effect was blocked by lapatinib (Fig. 1I and J; Supplementary Fig. S1L–S1N).

Then, we examined the role of TGFβ and HER2/EGFR in tumor formation and metastasis in vivo and found that TGFβ significantly inhibited subcutaneous tumor formation, which was further enhanced by lapatinib (Fig. 1K). Ki67 staining of tumor xenografts supported the above observation (Supplementary Fig. S1O). Tail intravenous injection of treated cells (MDA-MB-231, HER2-high SKBR3, or mouse breast cancer 4T1 cells) led to lung or liver metastasis, and TGFβ promoted the metastasis while lapatinib attenuated this effect (Fig. 1L; Supplementary Fig. S1P and S1Q). All these results support the notion that HER2/EGFR is important for the functional switch of TGFβ from growth inhibition to tumor metastasis.

As Smad2 and Smad3 are the downstream transducers of TGFβ signaling, we examined whether HER2/EGFR signaling affects their activities using CAGA-luciferase reporter that is sensitive to Smad3 and ARE-luciferase reporter that responds to Smad2 (29, 30). As shown in Fig. 2A and B, Supplementary Fig. S2A–S2B, lapatinib reduced TGFβ–induced expression of CAGA luciferase while having minimal effects on ARE luciferase in MDA-MB-231, SKBR3, MCF-10A, and NMuMG cells. Similar results were obtained with CP724714 in MDA-MB-231 cells (Supplementary Fig. S2C). In line with these results, HER2 overexpression enhanced CAGA-luciferase reporter activity but not ARE luciferase in MCF-10A and NMuMG cells (Fig. 2C; Supplementary Fig. S2D). These data suggest that HER2/EGFR signaling modulates the activity of Smad3, but not Smad2.

Although lapatinib had effect on neither TGFβ–induced C-terminal tail phosphorylation nor stability of Smad2 and Smad3 (Fig. 2D), the TGFβ–elicited nuclear accumulation of Smad3, but not Smad2, was dramatically reduced by lapatinib or gefitinib (Fig. 2E and F; Supplementary Fig. S3A–S3D). Consistently, HER2 overexpression enhanced Smad3 nuclear localization but had no effect on Smad2 in response to TGFβ (Supplementary Fig. S3E–S3H). Nuclear localization of Smad3 is tightly regulated by both nuclear import and export events (31). In agreement with the previous report that RanBP3 binds to Smad3 and mediates their nuclear export (32), TGFβ1 reduced the Smad3-RanBP3 interaction, but lapatinib blocked this reduction (Fig. 2G; Supplementary Fig. S3I), indicating that HER2/EGFR signaling promotes the nuclear accumulation of Smad3 by reducing Smad3 export. Although TGFβ1 reduced the Smad2–RanBP3 interaction, lapatinib had no effect on Smad2–RanBP3 interaction (Fig. 2H).

Phosphorylation of serine/threonine residues in the linker region of Smad proteins has been shown to regulate their stabilization, nuclear localization, and transcription activity (10, 11). As the above data suggest that HER2/EGFR signaling modulates the nuclear translocation of Smad3 but does not affect the C-tail phosphorylation and stability, we then examined the effect of HER2/EGFR on the linker phosphorylation of Smad3. EGF induced the phosphorylation at T179 and S208 of Smad3 in MDA-MB-231, SKBR3, MCF-10A, and NMuMG cells, which was blocked by lapatinib or gefitinib (Fig. 3A; Supplementary Fig. S4A-D). The similar results were observed in non–small cell lung cancer A549 cells, gastric cancer BGC-823 cells, and esophageal cancer KYSE410 cells (Supplementary Fig. S4E). EGF-stimulated S213 phosphorylation was observed only in MDA-MB-231, and S204 phosphorylation was not inhibited by lapatinib in MCF-10A cells (Fig. 3A; Supplementary Fig. S4C). Furthermore, HER2 overexpression enhanced Smad3 S208 phosphorylation (Fig. 3B; Supplementary Fig. S4F and S4G), and the kinase activity of HER2 was important for this enhancement as the kinase-dead mutant HER2(KM, K753M) did not exhibit the effect (Fig. 3C). All these data together suggest that HER2/EGFR signaling consistently stimulates Smad3 S208 phosphorylation.

HER2/EGFR activates a series of downstream signaling pathways, such as AKT and ERK (extracellular signal-regulated kinase) signaling (12). As shown in Fig. 3D and Supplementary Fig. S5A–S5E, the AKT inhibitor MK2206 attenuated Smad3 S208 phosphorylation, while the ERK inhibitor PD0325901 reduced EGF-mediated Smad3 S208 phosphorylation only in MDA-MB-231 and NMuMG, but not in MCF-10A, SUM159, and SKBR3 cells. Consistently, ectopic expression of wild-type AKT and its active DD mutant (T308D and S473D), but not the kinase-dead (KD, K179A) mutant, enhanced S208 phosphorylation (Supplementary Fig. S5F–S5G). In vitro kinase assay also confirmed that S208 is a direct phosphorylation site of AKT (Fig. 3E). These data indicate that AKT is the major kinase downstream of HER2/EGFR to mediate Smad3 S208 phosphorylation.

As AKT can be activated by various extracellular stimuli (33, 34), we tested whether other factors that were able to activate AKT can induce Smad3 S208 phosphorylation. As shown in Supplementary Fig. S5H, Smad3 S208 phosphorylation was associated with AKT activation, but not ERK activation, further supporting the role of AKT in mediating HER2/EGFR signaling to induce Smad3 S208 phosphorylation.

Phosphorylation of the linker region of Smad proteins modulates their stability and subcellular localization (10, 11). As HER2/EGFR signaling influenced the nuclear localization of Smad3, we assessed whether AKT-mediated S208 phosphorylation plays a role in this process. As shown in Fig. 4A and B and Supplementary Fig. S6A, inhibition of AKT activity by MK2206 or the phosphatidylinositol 3 kinase (PI3K) inhibitor LY294002 attenuated the nuclear accumulation of Smad3, but not Smad2. Consistently, MK2206 reduced the TGFβ–stimulated expression of the Smad3-responsive CAGA-luciferase, but not that of Smad2-responsive ARE luciferase (Fig. 4C; Supplementary Fig. S6B). Moreover, the MEK/ERK inhibitor PD0325901 had no effect on the TGFβ–induced nuclear accumulation of Smad3 (Supplementary Fig. S6C). S208A mutation that prevents phosphorylation abolished the TGFβ–induced nuclear accumulation of Smad3, and S208E that mimics phosphorylation entered the nucleus as wild-type Smad3 (Fig. 4D), while alanine replacement of T179, S204, and S213 had no effect on TGFβ–induced nuclear translocation of Smad3 (Supplementary Fig. S6D). RanBP3 interacted strongly with Smad3(S208A), but weakly with Smad3(S208E), while importin-β1 associated with none of them (Fig. 4E), suggesting that S208 phosphorylation prevents the nuclear export of Smad3. However, mutation of the linker phosphorylation sites of Smad3 or inhibition of AKT activity did not influence the receptor-mediated C-terminal tail phosphorylation (Supplementary Fig. S6E–S6F). Suppression of AKT activity by MK2206 also enhanced TGFβ1–induced antiproliferation and blocked cell migration promoted by HER2 and constitutively active TβRI(GGD) overexpression in MDA-MB-231 cells (Fig. 4F and G). These data together suggest that AKT-mediated S208 phosphorylation is critical for the nuclear accumulation of Smad3.

We then examined the gene expression in MDA-MB-231 cells after treatment with TGFβ, lapatinib, or both (TLap) for 24 hours. The expression of the genes associated with EMT and cell migration (fibronectin, Snail, N-cadherin, MMP9, CTGF, ANGPTL4, NUAK1/ARK5, and PAI-1) was stimulated by TGFβ but attenuated to different extent in the TLap group, while E-cadherin expression was opposite (Fig. 5A). In contrast, TLap treatment upregulated the expression of the G₁ CDK inhibitors p15, p21, and p57 (Fig. 5B). The effect of lapatinib on TGFβ-regulated gene expression was resembled by gefitinib and MK2206 (Fig. 5C; Supplementary Fig. S7A). These results indicate that inhibition of HER2/EGFR or AKT activity alters the TGFβ–regulated gene-expression profile, shifting its function from inhibiting proliferation to enhancing migration.

Knockdown of Smad2 or Smad3 abolished the TGFβ–induced antiproliferation in the presence of lapatinib (Supplementary Fig. S7B–S7C). However, knockdown of Smad3, but not Smad2, blocked TGFβ–enhanced cell migration, which could be rescued by reexpression of WT Smad3 but not Smad3(S208A) mutant in the Smad3 knockdown cells (Fig. 5D and E). Furthermore, Smad3(S208E) reexpression enhanced cell migration of MDA-MB-231 cells even without TGFβ. Consistently, Smad3 depletion strongly attenuated the TGFβ–induced expression of fibronectin, Snail, CTGF and PAI-1 and abolished the TGFβ–mediated suppression of E-cadherin, and lapatinib had no further effect (Fig. 5F). In contrast, Smad2 depletion had no effect (Fig. 5G). Accordingly, lapatinib reduced the TGFβ–induced binding of Smad3, but not Smad2, to the promoters of PAI-1, CTGF, and NUAK1 (Fig. 5H). These results further indicate that HER2/EGFR modulates TGFβ–regulated expression of these EMT and cell migration genes via Smad3, while TGFβ–induced antiproliferative effect needs both Smad2 and Smad3.

Smad3 activation has been shown to be correlated to breast cancer progression (15, 16, 35). To study whether HER2/EGFR signaling regulates Smad3 activation during breast cancer development, we examined the levels of C-terminally phosphorylated Smad3 and Y1248 phosphorylation of HER2 in human breast cancer samples. A total of 223 breast cancer specimens (78 Luminal A: ER⁺PR⁺HER2⁻, 62 Luminal B: ER⁺PR⁺HER2⁺, 52 HER2⁺: ER⁻PR⁻HER2⁺, and 31 triple-negative breast cancer (TNBC): ER⁻PR⁻HER2⁻), mostly from invasive breast carcinomas, were analyzed by IHC for p-Smad3(CT), HER2, and p-Y1248 levels. Compared with adjacent normal tissues, p-Smad3(CT) staining was significantly higher in cancer tissues (Fig. 6A and B). p-Smad3(CT) staining was categorized into three levels (low, scores 0–4; medium, scores 5–8; and high, scores 9–12) by the German semiquantitative scoring system that takes into consideration both staining intensity and area. Of 223 samples, 49.8% showed medium levels and 33.2% expressed high levels of p-Smad3(CT; Table 1). Significantly, p-Smad3(CT) staining had a positive correlation with pathologic grades and clinical stage (Fig. 6C–E and Table 1). Moreover, p-Smad3(CT) levels were correlated with HER2 and p-Y1248 levels (Fig. 6A, C, F; Table 1). These results demonstrate a positive correlation of HER2 activation and Smad3 activation in malignant breast cancers. The levels of p-Smad3(S208) were also correlated with cancer progression and with the HER2 p-Y1248 and p-AKT levels in cancer samples (Fig. 6G and H; Supplementary Fig. S7D). These data further support the conclusion that HER2-induced activation of AKT is important for Smad3 S208 phosphorylation and thus its activation in breast cancer development.

TGFβ signaling suppresses tumor formation via its antiproliferative effect, but promotes the invasion and metastasis of breast cancers synergistically with HER2 (17, 18, 20, 21, 23). Several mechanisms have been proposed as to how HER2 modulates TGFβ signaling. ERK and p38 MAPK and PI3K/AKT have been suggested to mediate the TGFβ–induced migration and invasion of breast epithelial cells when HER2 is overexpressed (17, 18), although the underlying mechanism is unclear. HER2 impairs the antiproliferative effect of TGFβ, probably by modulating the TGFβ–mediated expression of tumor suppressor genes or the oncogene Myc (36–38). Recent studies showed that TGFβ and HER2/EGFR pathways synergized to promote EMT by regulating the Smad3-PCBP1 interaction and CD44 splicing (39). In the current study, we present the evidence that HER2/EGFR signaling switches the antiproliferative function of TGFβ to the proinvasion and metastasis function by controlling the nuclear localization of Smad3 via AKT (Fig. 6I).

We found that HER2 contributed to the functional switch of TGFβ in both HER2-high and HER2-low breast cancer cells. Our data further showed that inhibition of HER2 activity with lapatinib could enhance TGFβ–induced growth inhibition and block cell migration and invasion. It has been reported that lapatinib can increase migration/invasion of lapatinib-resistant MDA-MB-231 cells, but it has minimal effect on normal MDA-MB-231 cells (40), which is in line with our results. Therefore, lapatinib probably has different effects in TNBC treatment when it is combined with different therapy methods. Our findings suggest that lapatinib in combination with activation of TGFβ signaling may be a promising therapy method to treat breast cancers.

AKT has been shown to inhibit Smad3 function. It mediates the inhibitory effect of IGF-1 on the TGFβ–triggered activation of Smad3 in rat prostatic epithelial NRP-152 cells (41). In hepatic carcinoma Hep3B cells, AKT was reported to interact with Smad3 and block its phosphorylation and nuclear localization induced by TGFβ, resulting in attenuation of the TGFβ–elicited apoptosis (42, 43). The inhibitory effect of AKT on Smad3 was shown to be independent of its kinase activity. However, it has also been reported that AKT can enhance the DNA binding capacity of Smad3 and TGFβ–induced EMT and migration of lung cancer A549 cells by directly phosphorylating Smad3 at T179 (44). Consistently, we found that AKT-mediated HER2/EGFR signaling enhances TGFβ–induced migration of breast cancer cells by promoting the nuclear accumulation of Smad3. These observations indicate that AKT regulates TGFβ response in a cell type–dependent manner.

Several kinases have been shown to phosphorylate the linker region of Smad3 and regulate its stability, subcellular localization, and transcription activity (reviewed in refs. 10 and 11). Various kinases, such as MAPKs, GSK3β, CDKs, and ROCK, can phosphorylate the Smad3 linker at T179, S204, S208, or S213. Our data indicated that AKT was critical for EGF-induced S208 phosphorylation, and inhibition of AKT activity impaired the nuclear accumulation of Smad3, but not Smad2. As the phosphor-mimicking mutant S208E had a lower binding affinity to the nuclear exporter RanBP3, it is possible that AKT-mediated phosphorylation of S208 impairs Smad3 interaction with RanBP3 and thereby the nuclear export of Smad3. The prolonged nuclear retention of Smad3 could result in its stronger transcription activity. Phosphorylation at S208 has been shown to either promote or inhibit Smad3 activity, dependent on different signal inputs or/and various combinations of phosphorylated sites in the linker. For instance, in response to TGFβ and HGF, JNK can phosphorylate S208 and other sites and enhance TGFβ–induced EMT and invasive capacity (45, 46); phosphorylation at S204 and S208 by p38 and ROCK was shown to be important for the full transcription activity of Smad3 (47); in response to TGFβ, CDKs can also phosphorylate S208 as well as T179 and S213, leading to inhibition of Smad3 activity (48, 49). Therefore, regulation of linker phosphorylation and the functional outputs are complex and context dependent and warrant further research.

It is becoming clear that Smad2 and Smad3 have distinct physiologic functions in different circumstances (35). Smad3 is critical for TGFβ–induced EMT and cell migration and promotes metastasis (1, 2, 5, 50). Consistently, we found that TGFβ–induced cell migration is mediated mainly by Smad3, but the antiproliferative effect of TGFβ needs both Smad2 and Smad3. The specific regulation of Smad3 by AKT provides a good example for the contribution of signaling cross-talk to differential function of Smad2 versus Smad3 in mediating TGFβ activity.

In summary, our study unravels a direct link for HER2/EGFR signaling to modulate TGFβ function during breast cancer development. Elucidation of the mechanism underlying the impact of HER2/EGFR on TGFβ–induced cell responses is important to provide guidance to treat breast cancers. Our findings suggest that caution should be exercised over the inhibition of TGFβ signaling to treat breast cancers as intact TGFβ signaling could be beneficial to the therapy by inducing cytostasis of cancer cells in the presence of HER2/EGFR inhibitors. Moreover, Smad3 S208 phosphorylation induced by EGF was also observed in non–small cell lung cancer, gastric cancer, and esophageal cell lines, indicating that the impact of HER2/EGFR on the TGFβ function switch may apply to other types of cancers.

No potential conflicts of interest were disclosed.

Conception and design: F. Huang, Q. Shi, Y. Li, Y.-G. Chen

Development of methodology: F. Huang, Q. Shi, C. Xu, X.H.-F. Zhang, X.-H. Feng, J.-D.J. Han, S. Luo, Y.-G. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Huang, Q. Shi, Y. Li, L. Xu, C. Xu, F. Chen, H. Wang, H. Liao, Z. Chang, S. Luo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Huang, Q. Shi, Y. Li, L. Xu, C. Xu, X.H.-F. Zhang, X.-H. Feng, J.-D.J. Han, S. Luo, Y.-G. Chen

Writing, review, and/or revision of the manuscript: F. Huang, Q. Shi, Y.-G. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Liao

Study supervision: Y.-G. Chen

This work was supported by grants from the National Natural Science Foundation of China (31730056 and 31330049) and the National Key Research and Development Program of China (2017YFA0103601) to Y.G. Chen.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.