Research ArticleCancer

Neuregulin 1–activated ERBB4 interacts with YAP to induce Hippo pathway target genes and promote cell migration

See allHide authors and affiliations

Science Signaling  09 Dec 2014:
Vol. 7, Issue 355, pp. ra116
DOI: 10.1126/scisignal.2005770


The receptor tyrosine kinase ERBB4, a member of the epidermal growth factor receptor (EGFR) family, is unusual in that ERBB4 can undergo intramembrane proteolysis, releasing a soluble intracellular domain (ICD) that modulates transcription in the nucleus. We found that ERBB4 activated the transcriptional coactivator YAP, which promotes organ and tissue growth and is inhibited by the Hippo tumor-suppressor pathway. Overexpressing ERBB4 in cultured mammary epithelial cells or adding the ERBB4 ligand neuregulin 1 (NRG1) to breast cancer cell cultures promoted the expression of genes regulated by YAP, such as CTGF. Knocking down YAP or ERBB4 prevented the induction of CTGF expression by NRG1, as did treating cells with the ERBB inhibitors lapatinib or erlotinib, which reduced ERBB4 cleavage. NRG1 stimulated YAP activity to an extent comparable to that of EGF (epidermal growth factor) or LPA (lysophosphatidic acid), known activators of YAP. NRG1 stimulated YAP-dependent cell migration in breast cancer cell lines. These observations connect the unusual nuclear function of a growth factor receptor with a mechanosensory pathway and suggest that NRG1-ERBB4-YAP signaling contributes to the aggressive behavior of tumor cells.


ERBB4 (also known as HER4) is a member of the epidermal growth factor receptor (EGFR/ERBB) family of receptor tyrosine kinases (RTKs). ERBB4 is essential for normal development and maintenance of the heart, mammary glands, and the nervous system (14). ERBB4 is unusual among RTKs in its ability to undergo regulated juxtamembrane and intramembrane proteolysis to release a soluble intracellular domain (ICD) (5). The ERBB4 ICD relocalizes to the nucleus, where it regulates transcription through its association with transcriptional co-regulators (such as KAP1, TAB2/N-CoR, and AP2) and sequence-specific DNA binding proteins (such as STAT5A and the estrogen receptor) (612). The distinct nuclear functions of the ERBB4 ICD add a dimension to RTK-governed processes and unleash new avenues for signaling.

ERBB4 transcripts undergo tissue-specific alternative splicing (13). ERBB4 CYT-1, but not CYT-2, includes an exon encoding a 16–amino acid peptide distal to the kinase domain with a PPxY motif that is a binding site for p85 PI3K (phosphatidylinositol 3-kinase) and WW domains (14). This small difference endows CYT-1 with substantially different biological properties: in tissue culture and mouse transgenic models, CYT-1 induces differentiation and survival phenotypes, whereas CYT-2 promotes proliferation (15, 16).

The second splice site affects the extracellular domain (ECD). ERBB4 JM-a, but not JM-b, has an extracellular proteolytic cleavage site for TACE [TNF-α (tumor necrosis factor–α)–converting enzyme; also known as ADAM17] (14). Activation of TACE by ERBB4 ligands, phorbol esters, or other agonists releases the ECD of the receptor, leaving a membrane-embedded 80-kD isoform (m80) (17). This enables intramembrane proteolysis at a second, γ-secretase cleavage site, which releases a soluble 80-kD ICD (s80/ICD) (17). Overall, differential regulation of ERBB4 structure by alternative splicing and proteolysis produces receptors with very different signaling qualities. Full-length ERBB4 isoforms signal much like other RTKs at the membrane by binding of downstream proteins to the Tyr-phosphorylated receptor. In contrast, s80 isoforms have entirely novel signaling functions in transcriptional regulation. Epithelial tissues and cell lines appear to express only JM-a, whereas neural and mesenchymal tissues express mostly JM-b or both JM-a and JM-b isoforms (13).

Candidate oncogenic mutations or amplification of ERBB4 occurs with moderate frequency in medulloblastoma, melanoma, and carcinoma. At 2.1% incidence, ERBB4 is the fourth most mutated RTK across 12 major cancer types (18), and overexpression of ERBB4 in mouse mammary epithelium can initiate carcinogenesis (15). However, prognostic associations of ERBB4 expression with breast cancer are variable, with favorable (1923) or unfavorable (2427) associations reported. Part of this inconsistency is likely due to the failure to discriminate among ERBB4 isoforms.

We have recently compared the signaling associated with expression of full-length ERBB4 and the ICD isoforms through transcriptional and chromatin immunoprecipitation–sequencing (ChIP-Seq) analysis (28). The ERBB4 ICD induced numerous YAP-regulated genes. This is consistent with our early transcription profiling studies linking NRG1 (neuregulin 1) and ERBB4 to the transcription of the YAP-regulated gene CTGF, which encodes connective tissue growth factor (29).

The Hippo pathway has emerged as a critical signaling hub that regulates organ growth and size maintenance (30). Dysregulation of this pathway can promote tumorigenesis (31, 32). Hippo signaling inhibits the transcriptional coactivators YAP and TAZ. Hippo pathway kinases MST1/2 and LATS1/2 operate in a kinase cascade that inhibits cell growth and promotes apoptosis under conditions of high cell density (33). LATS1/2 inactivates YAP and TAZ through inhibitory phosphorylation leading to cytoplasmic retention by 14-3-3 binding and proteasome-dependent degradation (34). Under growth-permissive conditions, the Hippo kinases are inhibited such that YAP and TAZ are free to translocate to the nucleus and activate transcription of genes that promote growth and migration (such as CTGF, CYR61, ANKRD1, and AREG) (34). YAP and TAZ do not have DNA binding domains but interact with sequence-specific DNA binding proteins, including TEAD1-4, p73, SMAD, and RUNX (34).

Several other signaling inputs regulate YAP and TAZ activation. The apical-basal polarity proteins NF2, AMOT, and α-catenin regulate YAP at the membrane (34, 35). Disruption of cell junctions releases YAP from these sequestering proteins, enabling nuclear localization. Both the actin cytoskeleton and microtubules also control YAP activation (36) through mechanical stimulation, facilitating proliferation on stiff substrates (37). Agonists for some G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptors (GPCRs) comprise another major class of upstream Hippo regulators that can either activate or inactivate YAP [as is the case for lysophosphatidic acid (LPA) and sphingosine 1-phosphate (S1P), or glucagon and epinephrine, respectively] (38). YAP is also under metabolic control by the SREBP/mevalonate pathway, which activates YAP by inhibiting its phosphorylation through Rho activation (39).

EGFR signaling converges with the Hippo pathway, uniting two major growth regulatory systems (40, 41). The ligand EGF (epidermal growth factor) activates YAP by relocalizing the kinase PDK1 from its scaffolding position in an inhibitory MST1/2-LATS1/2 complex (41). Additionally, YAP can activate EGFR by inducing the expression of AREG, which encodes the ligand amphiregulin (42). YAP binds ERBB4 through an interaction between PPxY motifs in ERBB4 and WW domains in YAP (43). Using artificial GAL-4 luciferase assays, YAP was shown to be required for transcriptional coactivation mediated by the ERBB4 C-terminal fragment (amino acids 676 to 1292) (43, 44). Additionally, YAP and ERBB4 colocalize in the nucleus, where ERBB4 ICD induces genes necessary for lung maturation (45). However, the ability of ERBB4 to activate transcription of YAP target genes was not reported, so the functional impact of the ERBB4-YAP interaction is uncertain.

Tissue-specific ERBB4 cleavage enables the integration of cell proliferation and size control through Hippo- and RTK-regulated pathways and has broad implications in both development and cancer. Here, we investigated the molecular mechanisms and biological consequences of the ERBB4-YAP interaction.


ERBB4 overexpression enriches for a YAP target gene signature

Because ERBB4 binds YAP, we investigated whether overexpression of ERBB4 affects YAP target genes. We analyzed the transcription profile induced by ERBB4 ICD CYT-2 (28) in MCF10A mammary epithelial cells through gene set enrichment analysis (GSEA) using the conserved YAP-dependent gene expression signature identified by Cordenonsi et al. (46). Several YAP target genes were strongly enriched in cells expressing ERBB4 ICD CYT-2 (Fig. 1A), including CTGF, suggesting that ERBB4 activates YAP. We focused on CTGF as a canonical YAP/TEAD-regulated gene in subsequent experiments.

Fig. 1 NRG1 stimulation of ERBB4 activates YAP signaling.

(A) GSEA in MCF10A cells expressing ERBB4 ICD CYT-2. Plot shows enrichment of the conserved signature reported by Cordenonsi et al. (46) in the data from cells expressing ERBB4 ICD CYT-2 (normalized enrichment score = 3.01, P < 0.0001). The table lists genes in the ERBB4 ICD CYT-2 data set with core enrichment for this signature. All enriched YAP target genes are in the top 12.7% of ERBB4 ICD CYT-2–induced genes (1.4 < fold change < 5.7). (B) RT-PCR of CTGF in MCF10A cells expressing pINDUCER20 encoding ERBB4 CYT-1, CYT-2, or a vector control and stimulated with NRG1 (100 ng/ml) for 1 hour with or without an hour of pretreatment with lapatinib (2 μM). MCF10A cells were starved overnight in Opti-MEM with simultaneous DOX (50 ng/ml) treatment. Data are means ± SD from four experiments. (C) Immunoblot of phosphorylated YAP (P-YAP) at Ser127 and CTGF abundance in MCF10A cells expressing pINDUCER20 encoding ERBB4 CYT-1, CYT-2, or a vector control and treated as in (B). (D) Immunoblot showing the time course of NRG1-induced CTGF abundance and YAP dephosphorylation in MCF10A cells expressing pINDUCER20-encoded ERBB4 and treated as in (B). Blots in (C) and (D) are representative of three experiments. (E) RT-PCR showing the time course of NRG1-induced CTGF expression in MCF10A cells expressing vector or pINDUCER20-encoded ERBB4 CYT-2 and stimulated with NRG1 (100 ng/ml) for up to 1 hour. Data were normalized to vector-transfected cells at 0 min. Data are means ± SD from three experiments. n.s., not significant; *P < 0.05; **P < 0.01.

Ligand stimulation of full-length ERBB4 is sufficient for CTGF induction

We next determined if full-length ERBB4 stimulated by its ligand NRG1 is sufficient to induce the expression of CTGF. Doxycycline (DOX)–inducible plasmids encoding JM-a ERBB4 CYT-1 or CYT-2 were introduced into MCF10A cells, a cell line devoid of endogenous ERBB4 (fig. S1). Within 1 hour, NRG1 increased the abundance of CTGF mRNA 5-fold in MCF10A cells expressing ERBB4 CYT-1 and 15-fold in cells expressing ERBB4 CYT-2 (Fig. 1B). CTGF protein abundance increased 1 to 2 hours after stimulation with NRG1 (Fig. 1, C to E). However, the trend for NRG1-induced CTGF mRNA in vector-infected cells was not significant. NRG1 may signal through endogenous ERBB3, which can activate PI3K signaling in collaboration with endogenous EGFR (Fig. 1, D and E). Preincubation with the pan-ERBB inhibitor lapatinib suppressed NRG1 induction of CTGF (Fig. 1, B and C). Because the strongest changes in CTGF abundance were observed in cells expressing ERBB4 CYT-2, we chose to focus subsequent experiments on this isoform.

LATS1/2-dependent phosphorylation of YAP at Ser127 reduces the nuclear localization of YAP by enabling its binding to cytoplasmic 14-3-3. Addition of NRG1 to culture medium reduced the phosphorylation of YAP at this site in MCF10A-pINDUCER20 cells (Fig. 1, C and D). Ser127-phosphorylated YAP decreased over the first hour of NRG1 treatment and remained low during periods of high CTGF expression, consistent with the nuclear function of YAP.

YAP mediates NRG1/ERBB4 up-regulation of CTGF

Because ERBB4 overexpression strongly enriched for a YAP target gene signature, and addition of NRG1 was sufficient to induce CTGF expression, we investigated whether YAP mediated these effects. Dobutamine, a chemical inhibitor of YAP, suppresses YAP-dependent gene transcription (47). Pretreating MCF10A cells expressing pINDUCER20-encoded ERBB4 CYT-2 with dobutamine diminished NRG1-induced CTGF expression at both the protein and mRNA levels (Fig. 2, A and B). We used an inducible YAP shRNA (short hairpin RNA) system to ensure that YAP expression was only suppressed during experiments and avoid potential selection as cells were passaged (Fig. 2C). YAP knockdown in MCF10A cells expressing DOX-inducible pINDUCER20-encoded ERBB4 CYT-2 and pINDUCER10-encoded YAP shRNA greatly reduced the induction of CTGF mRNA by NRG1 (Fig. 2D). Hence, YAP promotes ERBB4-mediated induction of CTGF by NRG1.

Fig. 2 YAP mediates NRG1/ERBB4 induction of CTGF.

(A) Immunoblot of MCF10A pINDUCER20 ERBB4 CYT-2 cells treated with or without DOX (50 ng/ml, overnight in Opti-MEM) and without or with inclusion of dobutamine (30 μM) for the last 4 hours. Cells were then incubated with NRG1 (50 ng/ml) for 1 hour. Blots are representative of two experiments. (B) RT-PCR of CTGF in MCF10A cells expressing pINDUCER20 encoding ERBB4 CYT-2 treated with or without dobutamine (Dob) and NRG1 as in (A). Data are means ± SD from four experiments. (C) RT-PCR (left) and immunoblot (right) of YAP knockdown in MCF10A cells expressing pINDUCER10 encoding YAP shRNA and pINDUCER20 encoding ERBB4 CYT-2. Cells were treated with DOX (1 μg/ml, 72 hours) and then starved (Opti-MEM, +DOX) for 3 hours followed by NRG1 (50 ng/ml) for 1 hour. RT-PCR data are means ± SD from three experiments. Blots are representative of three experiments. (D) RT-PCR of CTGF in MCF10A cells expressing pINDUCER10 encoding YAP shRNA and pINDUCER20 encoding ERBB4 CYT-2. Cells were treated as in (C). Data are means ± SD from three experiments. n.s., not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

We expressed a YAP S127A mutant protein that cannot be phosphorylated at the site responsible for 14-3-3 binding and cytoplasmic retention to test the importance of phosphorylation of YAP at Ser127 in the NRG1-induced response. The induction of ERBB4 CYT-2 expression (+DOX) in MCF10A MSCV (murine stem cell virus) vector control cells resulted in a threefold increase in CTGF expression, with NRG1 present in all conditions (fig. S2). As expected, expression of YAP S127A markedly increased CTGF mRNA production (fig. S2), because it allows constitutive nuclear localization of YAP. Addition of ERBB4 CYT-2 using DOX in MCF10A YAP S127A cells did not significantly increase CTGF. These data demonstrate that ERBB4- and YAP-mediated control of CTGF expression are not additive and suggest that NRG1/ERBB4 regulates YAP through phosphorylation at Ser127 and nuclear localization.

ERBB4 promotes NRG1 induction of YAP target genes

To investigate whether the activation of endogenous ERBB4 induces CTGF production, we used T47D mammary carcinoma cells, which express all four ERBB receptors. In this background, NRG1 induced a 300-fold increase in CTGF mRNA (Fig. 3A). As in MCF10A cells, preincubation with the pan-ERBB inhibitor lapatinib suppressed the induction of CTGF in response to NRG1 (Fig. 3A).

Fig. 3 NRG1 activates endogenous ERBB4 to induce YAP target genes.

(A) RT-PCR of CTGF in T47D cells starved for 3 hours in Opti-MEM and then treated with NRG1 (50 ng/ml) for 1 hour with or without an hour of pretreatment with lapatinib (1 μM). Data are means ± SD from three experiments. (B) RT-PCR (left) of CTGF in T47D cells expressing pLKO encoding ERBB4 shRNA (sh1 or sh3) or scrambled control (scr). Cells were treated as in (A). Immunoblot (right) of ERBB4 knockdown. RT-PCR data are means ± SD from three experiments. (C) RT-PCR of (left to right) YAP, CTGF, CYR61, and ANKRD1 in T47D cells expressing pINDUCER10 encoding YAP shRNA. Cells were treated with DOX (1 μg/ml) for 5 days and then starved overnight (Opti-MEM, +DOX) followed by 1 hour of NRG1 (50 ng/ml) treatment. Data are means ± SD from three technical replicates, representative of three experiments. (D) Immunoblot showing time course of NRG1-induced Tyr1284 phosphorylated ERBB4, ERBB4 ICD, Ser127 phosphorylated YAP, and CTGF abundance in T47D cells. Cells were starved overnight in Opti-MEM and then treated with NRG1 (50 ng/ml) for 0, 15, 30, 60, 120, or 240 min. (E) Quantification of the ratio of phosphorylated YAP at Ser127 to total YAP protein levels from immunoblots in (D). Data are normalized to t = 0 min. Data are means ± SD from three experiments. n.s., not significant; a.u., arbitrary unit; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. All blots are representative of three experiments.

In T47D cells, both ERBB3 and ERBB4 can bind NRG1. We evaluated the importance of ERBB4 in the response to NRG1 with stable shRNA-mediated knockdown of ERBB4 in T47D cells. Partial ERBB4 knockdown greatly diminished both baseline CTGF and NRG1-induced CTGF mRNA expression (Fig. 3B). NRG1 did not significantly induce CTGF expression in vector-infected MCF10A cells, which express endogenous ERBB3 but not ERBB4 (Fig. 1, D and E), providing further support for the involvement of ERBB4 in NRG1 activation of YAP. Quantitative real-time polymerase chain reaction (qRT-PCR) confirmed both NRG1 responsiveness and YAP dependence in the expression of the YAP target genes CTGF, CYR61, and ANKRD1 in T47D cells (Fig. 3C). NRG1 induced CTGF with similar kinetics in both MCF10A and T47D cells, with peak protein induction 2 hours after treatment (Fig. 3D). As in MCF10A cells, the phosphorylation of YAP at Ser127 decreased over the first hour of NRG1 treatment, but rebounded 2 hours later (Fig. 3E). Collectively, these data indicate that ERBB4 mediates NRG1-induced activation of YAP in these mammary cell lines.

ERBB4 binds TEAD1 and governs induction of YAP target genes

The ERBB4 ICD (80 kD), which binds to YAP (43), was most abundant during periods when YAP was dephosphorylated and CTGF expression was increased in T47D cells treated with NRG1 (Fig. 3D). Overexpression of ERBB4 CYT-1 ICD or CYT-2 ICD strongly increased CTGF protein abundance in MCF10A cells cultured at low confluency (Fig. 4A).

Fig. 4 Density dependence of ERBB4 ICD induction of CTGF.

(A) Immunoblot of CTGF protein and YAP phosphorylation at Ser127 in MCF10A cells expressing ERBB4 CYT-1 or CYT-2 ICD seeded to achieve low (30%, lanes 1 to 9) or high (100%, lanes 10 to 18) confluency on the day of cell lysis. Samples are shown in technical triplicate. (B) Immunoblot (IB) showing co-IP of MYC-TEAD1 with ERBB4 CYT-2-V5 in HEK 293T cells. Cells were transfected with pcDNA3.1-ERBB4 CYT-2-V5 and pRK5-MYC-TEAD1 (or TEAD1 Y406A) using Lipofectamine 2000 and lysed 72 hours after transfection. Cells were incubated with NRG1 (50 ng/ml) for 1 hour and then lysed in NP-40 buffer. One microgram of anti-V5 antibody or immunoglobulin G (IgG) control was used to immunoprecipitate ERBB4 from 1 mg of lysate. Five percent input (50 μg of total lysate) was used as MYC control. (C) Immunoblot of CTGF protein induction in T47D cells incubated with NRG1 (50 ng/ml) for 1 hour after lapatinib (1 μM), erlotinib (1 μM), or dimethyl sulfoxide (DMSO) control pretreatment for 30 min. Cells were starved in Opti-MEM for 3 hours before treatment. All blots are representative of three experiments.

Density-dependent growth inhibition functions in part through activation of the Hippo pathway to suppress YAP activity (33). We investigated whether ERBB4 binding to YAP may override YAP inhibition at high density in MCF10A cells expressing ERBB4 ICD. At low confluency (30%, which is permissive for YAP signaling), both ERBB4 ICD isoforms increased the abundance of CTGF (Fig. 4A). In contrast, YAP was much less abundant in confluent cells (likely owing to phosphorylation-dependent proteasomal degradation), and the ERBB4 ICD did not alter CTGF abundance relative to vector-transfected cells. Hence, the ERBB4 ICD does not override density-dependent growth inhibition of YAP.

Suppression of the canonical Hippo signaling pathway activates YAP-dependent gene expression through nuclear localization of YAP, leading to the binding of YAP/TEAD complexes to the promoters of genes, including that of CTGF. Because the ERBB4 ICD, YAP, and TEAD1 all localize to the nucleus, and YAP binds both TEAD1 and ERBB4, we investigated whether ERBB4 forms a complex with TEAD1. In human embryonic kidney (HEK) 293T cells, Myc-tagged TEAD1 coimmunoprecipitated with V5-tagged ERBB4 CYT-2 in total cell lysates (Fig. 4B) and in nuclear fractions (fig. S3). ERBB4 interacts with YAP through the binding of the PPxY domains in ERBB4 to the WW domains in YAP. Distinct sites on YAP mediate its binding to TEAD; thus, it is possible that YAP binds simultaneously to both proteins, bridging ERBB4 and TEAD. However, we found no consistent reduction in the abundance of Myc detected in V5 immunoprecipitates in HEK 293T cells expressing a YAP binding–deficient mutant of TEAD1 (Y406A) (Fig. 4B), suggesting that ERBB4 may bind TEAD independently of YAP. These data support the model that NRG1-ERBB4 signaling regulates YAP target genes through interactions with YAP and TEAD1.

Treating T47D cells with the EGFR inhibitor erlotinib blocked the production of ERBB4 ICD (Fig. 4C) despite failing to effectively inhibit EGF- or NRG1-induced phosphorylation of wild-type EGFR (fig. S4). This coincided with a decrease in the abundance of CTGF (Fig. 4C), suggesting that erlotinib may block the induction of CTGF in response to NRG1 though indirect inhibition of ERBB4 cleavage.

NRG1 activity is comparable to that of other YAP agonists

The ability of ERBB4 JM-a isoforms to couple responses to NRG1 to the expression of YAP-regulated genes means that the expression of ERBB4 JM-a may broadly reprogram the NRG1-induced response to activate YAP (and TAZ) signaling. Hence, we sought to determine how NRG1 compares to canonical agonists for YAP and TAZ. At concentrations that we determined to yield maximal induction of CTGF mRNA, the GPCR-mediated YAP activator LPA induced a sixfold increase in CTGF mRNA abundance, whereas NRG1 induced a maximum 600-fold increase (Fig. 5A). This finding was recapitulated at the protein level in T47D cells, in which saturating doses of LPA increased the abundance of CTGF very weakly compared with NRG1 (Fig. 5, B and C). Although ERBB proteins transactivate when coexpressed, EGF was about threefold less potent in inducing CTGF abundance than NRG1 in T47D cells (Fig. 5B). Differences in CTGF production induced by NRG1, LPA, and EGF were not affected by timing of induction: EGF induced CTGF maximally after 2 hours of treatment, similar to NRG1 (Fig. 5C), but stimulation with LPA did not significantly increase CTGF protein abundance over 4 hours (Fig. 5C).

Fig. 5 Comparison of NRG1 to other YAP agonists.

(A) RT-PCR of CTGF in T47D cells starved overnight in Opti-MEM and then treated with either LPA (left) or DMSO control or NRG1 (right) at the concentrations indicated for 1 hour. Data are means ± SD from three technical replicates, representative of two experiments. (B) Immunoblot of CTGF protein abundance in T47D cells starved overnight in Opti-MEM and then treated with NRG1 (50 ng/ml), EGF (50 ng/ml), or LPA (1 μM) for 2 hours. The control was run in the same experiment but with no treatment. (C) Immunoblot of CTGF protein and YAP phosphorylation at Ser127 in T47D cells over a time course of NRG1, EGF, or LPA treatment. Cells were starved overnight in Opti-MEM and then incubated with saturating doses of NRG1 (50 ng/ml), EGF (50 ng/ml), or LPA (1 μM) for the times indicated. (D) Immunoblot of CTGF abundance in MCF10A cells expressing pINDUCER20 encoding ERBB4 CYT-2. Cells were incubated overnight in Opti-MEM in the absence or presence of DOX (50 ng/ml) followed by 2 hours of EGF (50 ng/ml) or NRG1 (50 ng/ml) treatment. All blots are representative of three experiments.

EGF activates YAP in MCF10A cells, which do not express ERBB4, through EGFR-PI3K-PDK1 signaling (41). To compare EGF with NRG1 in the activation of CTGF in this context, we used MCF10A cells that stably expressed pINDUCER20-encoded ERBB4 CYT-2 and cultured them with or without DOX to induce ERBB4 expression. NRG1 induced CTGF abundance as well or slightly better than did EGF, but only when ERBB4 expression was concomitantly induced with DOX (Fig. 5D).

NRG1 regulates YAP independently of the mevalonate and transforming growth factor–β pathways

ERBB4 induces the expression of several genes in the mevalonate/cholesterol pathway including HMGCR, HMGCS1, and LDLR (28). Because the mevalonate pathway can activate YAP, this raised the possibility that ERBB4 might activate YAP indirectly through induction of the mevalonate pathway. Independent of LATS1/2, the nuclear localization and transcriptional activity of YAP can be blocked by statins (39). Inhibition of HMG-CoA (3-hydroxy-3-methylglutaryl coenzyme A) reductase with simvastatin or lovastatin did not prevent the induction of CTGF expression in response to NRG1 (fig. S5A), so ERBB4 does not appear to regulate YAP signaling via the mevalonate pathway. Knocking down YAP did not block NRG1-induced increases in HMGCR expression (fig. S5B), suggesting that the mevalonate pathway is upstream from or parallel to ERBB4-dependent activation of YAP.

To determine whether transforming growth factor–β (TGF-β) signaling is responsible for the induction of CTGF, we evaluated the phosphorylation of SMAD2 (at Ser465/467) in MCF10A cells expressing pINDUCER20-encoded ERBB4 and treated with phorbol 12-myristate 13-acetate (PMA) or NRG1, respectively. NRG1 or PMA did not affect SMAD activation (fig. S5, C and D), suggesting that TGF-β signaling does not mediate the response to NRG1.

NRG1 activates YAP signaling to induce migration in T47D cells

We next evaluated the biological consequences of NRG1-ERBB4–mediated activation of YAP signaling. YAP drives cell proliferation and migration and can bypass signals from cell-cell contact and mechanical stress that constrain cell division in epithelial sheets when activated or overexpressed. ERBB4 expression and activation promote cell migration (28); therefore, we determined whether migration is mediated by YAP. NRG1 induced a four-fold increase in migration in T47D cells, which was greatly reduced by YAP knockdown (Fig. 6, A and B). Inducible YAP knockdown with DOX was confirmed in the cells used for each migration experiment (Fig. 6C). Inhibition of migration in YAP-deficient cells was not due to reduced cell viability, because cells transfected with either a scrambled control or YAP shRNA were equally viable over the 48-hour experiment (Fig. 6D). Despite high expression of exogenous ERBB4, MCF10A cells transfected with pINDUCER20 did not increase migration when treated with NRG1 (fig. S6). However, YAP knockdown partially reduced migration of these cells. These findings suggest that NRG1 promotes YAP-mediated biological phenotypes, including migration.

Fig. 6 NRG1-induced cell migration is mediated by YAP.

(A) Transwell migration assay of T47D cells expressing pINDUCER10 encoding YAP shRNA or scrambled control. Cells were incubated with DOX (1 μg/ml) for 5 days followed by 3 hours of NRG1 (50 ng/ml) pretreatment, and then allowed to migrate for 48 hours [from 0.1 to 10% fetal bovine serum (FBS)] in the presence of NRG1 and DOX. Percent migration was normalized to scrambled (scr) control without NRG1. Data are means ± SD from three experiments. (B) Migration data from (A) are shown as fold change migration +NRG1/−NRG1. Data are means ± SD from three experiments. (C) Immunoblot of YAP protein knockdown in T47D cells expressing pINDUCER10 encoding YAP shRNA or scrambled control. Cells were treated as in (A). Blots are representative of three experiments. (D) Trypan blue cell viability assay in T47D cells expressing pINDUCER10 encoding YAP shRNA or scrambled control. Cells were treated as in (A). n.s., not significant; *P < 0.05; **P < 0.01; ***P < 0.001.


Although the physical binding between ERBB4 and YAP has been known for several years, suggesting a model in which ERBB4 regulates gene expression through this interaction (4345), the importance of YAP in the Hippo pathway emerged more recently. We found that NRG1 activates YAP target gene expression through ERBB4 to promote YAP-driven biological phenotypes. Hence, ERBB4 connects the ERBB receptor complex with the Hippo/YAP network.

In T47D cells with endogenous ERBB4, NRG1 increased CTGF much more strongly than EGF, and LPA only weakly activated transcription of CTGF. Thus, NRG1 is a bona fide YAP activator and one of only a few known soluble YAP regulators.

ERBB4 may affect YAP in the cytosol through indirect modulation of upstream YAP regulators (such as Hippo kinases and GPCRs). ERBB4 and LATS1, the Hippo pathway kinase that inactivates YAP, both bind to the WW domains of YAP. Binding of YAP to ERBB4 may competitively reduce inhibitory phosphorylation by LATS.

As ERBB4 chaperones STAT5 to the nucleus, another possibility is that ERBB4 diminishes YAP phosphorylation by accelerating translocation of YAP to the nucleus. Mutation of YAP Ser127 to Ala, which would reduce cytoplasmic retention, blocked ERBB4-mediated increases in YAP signaling, consistent with this model, but it has been reported that this mutation also reduces ERBB4/YAP co-IP (43).

Direct interaction between nuclear ERBB4 ICD and YAP may modulate transcriptional activation. ERBB4 coimmunoprecipitates with YAP and TEAD, so it is possible that a ternary complex forms. At least with high-level expression, this does not require the YAP binding site in TEAD, so there may be a direct ERBB4-TEAD interaction. This positions ERBB4 to either aid assembly of the binary YAP-TEAD complex or participate in a ERBB4/YAP/TEAD ternary complex. In the latter scenario, ERBB4 could modulate transcription at TEAD target sites by recruitment or displacement of transcription factors. This is consistent with transactivation activity of ERBB4, which is enhanced in complexes with YAP (43, 44).

Physiological ERBB4 signaling occurs in the presence of other ERBB family members. Because ERBBs efficiently transactivate when coexpressed, the overall scope of ERBB4-YAP signaling must be considered in the context of the other ERBB proteins, particularly EGFR, which activates YAP through an indirect mechanism. It is likely that EGF/EGFR/PDK1 and NRG1/ERBB4 ICD regulation of YAP occur simultaneously (Fig. 7). An additional layer of complexity is that AKT can phosphorylate YAP (48). Therefore, although PI3K activation relocalizes PDK1 and activates YAP, it also activates AKT and may lead to YAP inhibition. ERBB3 contains a PPxY motif and may signal in the nucleus (49), raising the possibility that ERBB3-YAP binding also affects YAP signaling.

Fig. 7 Model of YAP regulation.

The transcriptional activity of YAP may be induced by NRG1- and proteolysis-mediated activation of ERBB4 signaling. Other pathways, such as G protein–mediated signaling and EGF-induced stimulation of another ERBB family member, may indirectly activate YAP by inhibiting the Hippo kinase LATS1.

The Hippo pathway is a major regulator of normal development and cancer. YAP is activated in uveal melanoma through upstream GPCR mutations (50, 51). YAP is dysregulated in breast cancer and other cancers (31, 52). Lapatinib and erlotinib, both U.S. Food and Drug Administration (FDA)–approved drugs, blocked NRG1 activation of CTGF, pointing to potential therapeutic interventions to suppress ERBB4-driven YAP signaling. The YAP/TEAD inhibitor verteporfin is FDA-approved for treatment of macular degeneration (47).

ERBB4 and YAP/TAZ are important for mammary development, tissue remodeling, and lactational differentiation during pregnancy (3, 46, 5355). In mice, the Hippo pathway is required during pregnancy, and YAP hyperactivation leads to defects in terminal differentiation (53). TAZ is required for breast cancer stem cell self-renewal and tumor initiation (46). Furthermore, TAZ can cause lineage switching of mammary epithelial cells from luminal to basal, whereas TAZ depletion in basal cells elicits luminal differentiation (54). In mice lacking mammary ERBB4, lobuloaveoli fail to properly differentiate during pregnancy, and lactation is defective (3). ERBB4 is essential during pregnancy-induced mammary differentiation and lactation at least in part through collaboration with STAT5 (55). NRG1 activation of YAP might coordinate cell growth in the mammary gland by sensing substrate stiffness and directing proliferation. Mechanical sensing may be especially important during reorganization of the mammary epithelium late in lactation and at onset of involution triggered by milk stasis at weaning, where YAP inactivation might initiate differentiation. An siRNA (small interfering RNA) screen for protein tyrosine kinases mediating rigidity-dependent cell polarization suggested that both ERBB3 and ERBB4 are mechanosensitive in fibroblasts (56). Therefore, the ERBBs might regulate YAP in coordination with the noncanonical Hippo pathway involving F-actin and Rho (36, 57, 58).

The unusual mechanism of ERBB4 intramembrane cleavage and nuclear localization provides new inputs into the Hippo signaling pathway that broadens the complexity of ERBB and YAP signaling. NRG1 and other ERBB4 agonists can be added to a short list of known YAP activators, which includes EGF and LPA. Expression of ERBB4 JM-a isoforms will enhance coupling of any ERBB agonist to YAP and TAZ signaling if the cognate ERBBs are coexpressed. As metalloproteinase cleavage of ERBB4 JM-a is rate-limiting for formation of s80 isoforms, a number of metalloproteinase agonists, including ligands for unrelated growth factor receptors, may promote YAP signaling through ERBB4 JM-a. This intersection of ERBB4 and Hippo signaling enables regulation of YAP through alternative ERBB4 mRNA splicing and proteolytic processing of ERBB4, and potentially affects a spectrum of YAP-driven biological processes.



pINDUCER20 and pINDUCER10 plasmids were provided by S. Elledge (Harvard Medical School) (59). ERBB4 JM-a CYT-1 or CYT-2 was cloned into pINDUCER20 using pENTR4. YAP shRNAs were cloned into pINDUCER10 from pGIPZ plasmids [Dharmacon, V2LHS_65509 (sh1) and V3LHS_306099 (sh2)]. ERBB4 JM-a CYT-2 was cloned into pcDNA3.1-V5/His B to add a C-terminal V5 tag. pRK5-MYC-TEAD1 (Addgene, #33109) or pRK5-MYC-TEAD1 Y406A (Addgene, #33047) was used for co-IP experiments. pcDNA3.1-ERBB4 CYT-2/V5 and MYC-TEAD1 were transiently transfected into 293T cells using Lipofectamine 2000 (Life Technologies). ERBB4 pLKO shRNA [Sigma, TRCN0000039688 (sh1), TRCN00000196519 (sh2), and TRCN0000314628 (sh3)] or scrambled shRNA control (Addgene, #1864) was used to generate lentiviruses for stable knockdown in T47D cells. MSCV-YAP S127A-IRES-Hygro and control MSCV-IRES-Hygro were gifts from R. Hynes [Massachusetts Institute of Technology (MIT)] (60) and were used to produce infectious retrovirus.


The following reagents were used: NRG1 (Sigma), EGF (Sigma), lapatinib (Selleck), dobutamine (Sigma), LPA (Santa Cruz), DOX (Sigma), erlotinib (LC Laboratories), simvastatin (Selleck), lovastatin (Selleck), and PMA (Sigma).

Cell culture and gene transfer

MCF10A human breast cancer cells [American Type Culture Collection (ATCC)] were maintained in Dulbecco’s modified Eagle’s medium (DMEM)/F12 (Life Technologies) with 5% horse serum, 1% penicillin/streptomycin (pen/strep), insulin (10 μg/ml; Gibco), EGF (20 ng/ml; Sigma), hydrocortisone (0.5 μg/ml; Sigma), and cholera toxin (100 ng/ml; Sigma). T47D cells were maintained in RPMI (Life Technologies) with 10% FBS, 1% pen/strep, and insulin (5 μg/ml; Gibco). HEK 293T cells were maintained in DMEM (Life Technologies) with 10% FBS and 1% pen/strep.

pINDUCER plasmids were packaged as lentivirus by cotransfecting 293T cells with VSV-G, Tat1b, RaII, and HgPM2 using FuGENE 6 (Promega). Virus was collected in Opti-MEM (Life Technologies) at 48 and 72 hours, pooled, and then concentrated using Centricon Plus-20 filters (Millipore). MCF10A cells or T47D cells were infected overnight in polybrene (4 μg/ml), selected in G418 (500 μg/ml) (pINDUCER20) or puromycin (1 μg/ml) (pINDUCER10) to generate stable polyclonal cell lines. Subculture lines were never exposed to DOX before experimental use to prevent counterselection. ERBB4 knockdown lines were generated in the same way, using lentivirus and appropriate pLKO packaging constructs, and were selected in puromycin (1 μg/ml).

Quantitative RT-PCR

RNA was isolated using the RNeasy Mini Plus Kit with QIAshredder columns (Qiagen). cDNA (complementary DNA) was prepared using the iScript kit (Bio-Rad). RT-PCR was performed with Universal TaqMan Master Mix (Applied Biosystems) coupled with TaqMan FAM-labeled probes and ran on a ViiA 7 RT-PCR machine (Life Technologies). Relative mRNA expression was determined using the 2−ΔΔCt method with GAPDH as the reference gene.

Immunoblotting and IP

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate] supplemented with protease inhibitor cocktail (Roche) and phosphatase inhibitors (Sigma). Protein was quantified by Bradford assay (Bio-Rad) and diluted with 2X Laemmli sample buffer. Samples were loaded onto 4 to 12% bis-tris gradient gels (NuPAGE, Life Technologies) and run in MOPS buffer. For IP, cells were lysed in NP-40 buffer [50 mM tris-HCl (pH 7.4), 150 mM NaCl, 0.5% Igepal]. One milligram of protein was incubated with 1 μg of antibody overnight, followed by 1-hour incubation with protein A/G UltraLink Resin (Thermo Scientific) at 4°C. For isolation of nuclear fractions, cells (10-cm equivalent) were lysed in NP-40 and incubated with agitation for 1 hour at 4°C. Nuclei were isolated by spinning for 3 min at 800g, and then supernatant was removed and nuclei were lysed in two-pellet volumes of nuclear lysis buffer C [20 mM Hepes, 0.42 M NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 1 mM DTT (dithiothreitol)] for 1 hour at 4°C. Nuclear IPs were conducted in the presence of 1 mM DTT using 150 μg of nuclear lysate and 0.5 μg of antibody. IP complexes were washed twice in lysis buffer and twice in buffer ST [50 mM tris-HCl (pH 7.4), 150 mM NaCl].

Protein was transferred to PVDF (polyvinylidene difluoride) membranes at constant amperage (at 500 mA for 1 hour), which worked best for CTGF immunoblotting. Membranes were incubated with appropriate horseradish peroxidase–conjugated secondary antibodies and developed by chemiluminescence (Pierce). Mouse anti-V5 (Invitrogen) was used for IP. Antibodies used for immunoblotting targeted the following: ERBB4 (Santa Cruz, sc-283, rabbit), phosphorylated ERBB4 at Tyr1056 (Santa Cruz, sc-33040, rabbit) or Tyr1284 (Cell Signaling Technology, catalog no. 4757), phosphorylated YAP at Ser127 (Cell Signaling Technology, catalog no. 4911, rabbit), YAP (Cell Signaling Technology, catalog no. 4912, rabbit), CTGF (Santa Cruz, sc-14939, goat), GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (Santa Cruz, sc-25778, rabbit), MYC (Cell Signaling Technology, catalog no. 2272, rabbit), phosphorylated SMAD2 at Ser465/467 (Cell Signaling Technology, catalog no. 3108, rabbit), SMAD2 (Cell Signaling Technology, catalog no. 5339, rabbit), phoshorylated EGFR at Tyr1068 (Cell Signaling Technology, catalog no. 3777, rabbit), and EGFR (Santa Cruz, sc-03, rabbit).

Transwell migration assays

T47D pINDUCER10 YAP KD cells or MCF10A pINDUCER20 ERBB4 CYT-2, pINDUCER10 YAP KD cells were treated with DOX (1 μg/ml) for 3 days (MCF10A) or 5 days (T47D). Cells were pretreated with or without NRG1 (50 ng/ml) for 3 hours and then plated at 1 × 106 cells per well (T47D) or 1 × 105 cells per well (MCF10A) in 24-well plates with 8-μm filter inserts (BD Biosciences) in the presence of DOX (1 μg/ml) and with or without NRG1 (50 ng/ml). Cells were allowed to migrate for 48 hours from 0.1 to 10% FBS for T47D cells, or for 24 hours from 0.05 to 5% horse serum for MCF10A. Membranes were fixed and stained, and cell number per well was averaged from six fields of view (FOV). Technical duplicates for each trial were averaged within three independent biological replicates. Percent migration was calculated on the basis of seeding density and the surface area of each FOV.

Cell viability assay

T47D cells were treated in the same way as for Transwell migration assays, except cells were replated into 12-well dishes after 5 days of DOX treatment. Cells were left to grow for 48 hours to mirror the migration assay, and then were trypsinized and quenched with medium supernatant containing any floating cells. Cells were stained with trypan blue dye and counted for viability using a Countess cell counter (Life Technologies). Assays were performed in technical triplicate and biological duplicate.

Gene set enrichment analysis

GSEA was performed on transcription profiles from MCF10A cells overexpressing ERBB4 ICD CYT-2. Genes significantly altered by ERBB4 (adjusted P value <0.05) compared with empty vector controls were previously reported [(28), GEO GSE57339]. YAP pathway gene sets were manually curated from the MSigDB_v4.0 (Broad Institute) (46, 61). YAP1_UP and YAP1_DN gene sets contain genes up- or down-regulated, respectively, in MCF10A cells overexpressing YAP1. ERBB4 CYT-2 genes (n = 5965) were first rank-ordered by fold change in expression over vector-transfected cells. This list was evaluated with GSEA using the curated YAP target genes under default settings [GSEAPreranked, 10,000 permutations; (62),].

Statistical analysis

Two-tailed Student’s t tests were performed where appropriate. Error bars represent SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.


Fig. S1. DOX-inducible expression of ERBB4.

Fig. S2. Impact of YAP S127A substitution on ERBB4 induction of CTGF.

Fig. S3. ERBB4-TEAD1 interaction in the nucleus.

Fig. S4. Impact of erlotinib on EGFR phosphorylation induced by EGF or NRG1.

Fig. S5. NRG1 regulates YAP independently of mevalonate and TGF-β pathways.

Fig. S6. NRG1 does not induce migration in MCF10A cells.


Acknowledgments: We thank The Hynes Lab (MIT) and Elledge Lab (Harvard) for providing plasmids, and we are grateful to S. Zhang for producing the pcDNA3.1-ERBB4 CYT-2/V5 plasmid. Funding: This work was supported by U.S. Public Health Service RO1 CA80065 and NIH training grant T32GM07223 (J.W.H.). D.X.N. is supported by NIH/National Cancer Institute grant 5R01CA166376. Author contributions: The study was conceived and designed by D.F.S. and J.W.H. Experiments were conducted by J.W.H. The manuscript was written by J.W.H. and D.F.S. Bioinformatics analyses were performed by D.X.N. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Reagents and cell lines produced by this laboratory and described here will be made available for research purposes in accordance with Yale University policies.
View Abstract

Stay Connected to Science Signaling

Navigate This Article