Activating ERBB4 mutations in non-small cell lung cancer

Abstract

Recent efforts to comprehensively characterize the mutational landscape of non-small cell lung cancer have identified frequent mutations in the receptor tyrosine kinase ERBB4. However, the significance of mutated ERBB4 in non-small cell lung cancer remains elusive. Here, we have functionally characterized nine ERBB4 mutations previously identified in lung adenocarcinoma. Four out of the nine mutations, Y285C, D595V, D931Y and K935I, were found to be activating, increasing both basal and ligand-induced ErbB4 phosphorylation. According to structural analysis, the four activating mutations were located at critical positions at the dimerization interfaces of the ErbB4 extracellular (Y285C and D595V) and kinase (D931Y and K935I) domains. Consistently, the mutations enhanced ErbB4 dimerization and increased the trans activation in ErbB4 homodimers and ErbB4-ErbB2 heterodimers. The expression of the activating ERBB4 mutants promoted survival of NIH 3T3 cells in the absence of serum. Interestingly, serum starvation of NIH 3T3 cells expressing the ERBB4 mutants only moderately increased the phosphorylation of canonical ErbB signaling pathway effectors Erk1/2 and Akt as compared with wild-type ERBB4. In contrast, the mutations clearly enhanced the proteolytic release of signaling-competent ErbB4 intracellular domain. These results suggest the presence of activating driver mutations of ERBB4 in non-small cell lung cancer.

Introduction

Lung cancer is one of the leading causes of cancer death worldwide.¹ It is often diagnosed at a late stage and has poor prognosis.² Therapeutic targeting of activated oncogenes can significantly increase the survival of lung cancer patients.³ However, aside from activating kinase domain mutations in epidermal growth factor receptor (EGFR) and fusions of the anaplastic lymphoma kinase, the current selection of clinically relevant predictive markers remains limited.³

Recent comprehensive sequencing projects have increased our understanding about the mutational landscape of non-small cell lung cancer and have helped to identify new potential targets that drive tumorigenesis.^{4, 5, 6} One such target is ERBB4, a member of the EGFR subfamily of receptor tyrosine kinases. According to COSMIC (catalog of somatic mutations in cancer),⁷ 5.4% of non-small cell lung cancers harbor ERBB4 missense mutations. More specifically, 5.2% of adenocarcinomas and 6.3% of squamous cell carcinomas have mutations in the ERBB4 gene that encode an amino-acid change in the receptor.

The abundance of ERBB4 mutations suggests that mutated ERBB4 could have a functional role in lung cancer and serve as a potential drug target. In metastatic melanoma, where ERBB4 mutations are particularly frequent,^{8, 9, 10} mutated ERBB4 has been shown to act as a driver oncogene and may serve as a novel drug target.^{8, 11, 12} However, because no mutational hot spots have been identified in ERBB4, and the number of studies comprehensively analyzing lung cancer-associated ERBB4 mutations is limited,¹³ the functional consequences of ERBB4 mutations in lung cancer remain largely unknown.

Here, we have functionally characterized nine somatic ERBB4 mutations that have been discovered from samples of lung adenocarcinoma.⁴ Four mutations were found to be activating, increasing ErbB4 phosphorylation both basally and upon ligand stimulation. The activating ERBB4 mutations targeted both the extracellular domain and the intracellular kinase domain of ErbB4. The increased activity of the ErbB4 mutants was demonstrated to be due to enhanced interaction between receptor monomers in ErbB4 homo- and/or heterodimers. Moreover, the activating ERBB4 mutants promoted cell survival in the absence of serum when overexpressed in NIH 3T3 cells. These results indicate the presence of activating, potentially oncogenic ERBB4 mutations in non-small cell lung cancer.

Results

ERBB4 mutations in lung cancer promote receptor activation

ERBB4 has been identified as a highly mutated gene in lung adenocarcinoma in a study analyzing the exons of 623 potential cancer genes in 188 clinical lung adenocarcinoma samples.⁴ Nine ERBB4 mutations reported in this cohort yield a mutation frequency of 4.8% (Figure 1).⁴ In contrast to the EGFR mutations in lung cancer that typically target the intracellular kinase domain,¹⁴ the reported ERBB4 mutations affect all functional subdomains of ErbB4 protein with no hotspot mutations (Figure 1). Seven out of the nine ErbB4 mutations, N181S, T244R, Y285C, R306S, V348L, D595V and H618P, are located in the extracellular domain and two mutations, D931Y and K935I, in the kinase domain.

Figure 1

Schematic representation of the analyzed ERBB4 mutations. The amino acids targeted by the ERBB4 mutations are indicated by arrows. Different subdomains within the extracellular domain of ErbB4 are indicated by roman numerals. Dark gray indicates the transmembrane domain.

To functionally characterize the ERBB4 mutations, the nine mutations were cloned into a pBABE-puroERBB4-JM-aCYT-2HA retroviral expression vector that encodes HA-tagged ErbB4 JM-a CYT-2 isoform and yields a moderate expression level in stably or transiently transfected cells. To address the effects on ErbB4 activation, wild-type and mutant ERBB4 constructs were transiently transfected into COS-7 cells and analyzed by western blotting. Four out of the nine mutations, Y285C, D595V, D931Y and K935I, were activating, increasing both basal and neuregulin-1 (NRG-1)-induced ErbB4 phosphorylation (Figure 2). The four mutants also promoted increased phosphorylation of endogenous ErbB2 in a NRG-1-dependent manner (Figure 2). The R306S mutation increased basal, but not NRG-1-stimulated, ErbB4 phosphorylation (Figure 2), and did not affect ErbB2 phosphorylation. The mutations N181S, T244R, V348L and H618P did not significantly affect ErbB4 activation (Figure 2), although N181S and V348L enhanced ligand-dependent phosphorylation of ErbB2.

Figure 2

Activity of the analyzed ERBB4 mutants in response to ligand stimulation. COS-7 cells were transiently transfected with constructs encoding wild-type ErbB4 or indicated ErbB4 mutants and treated for 10 min with 50 ng/ml NRG-1. ErbB4 phosphorylation was analyzed by western blotting (W) using phospho-ErbB4 and ErbB4 antibodies. ErbB2 phosphorylation was analyzed by western blotting with phosphotyrosine and ErbB2 antibodies after immunoprecipitation (IP) with ErbB2 antibody. Loading was controlled by western blotting with actin antibody.

Interestingly, all the mutations promoting increased basal and/or ligand-induced ErbB4 activation were structurally located at interaction sites within the ErbB4 monomer and/or between receptor monomers in ErbB4 dimers (see below). The mutations targeted three different functional subdomains of ErbB4; Y285C and R306S the extracellular subdomain II; D595V the extracellular subdomain IV; and D931Y and K935I the intracellular kinase domain (Figure 1). The most active ErbB4 mutants Y285C, D595V, D931Y and K935I were selected for further analyses.

ErbB4 extracellular mutations Y285C and D595V induce receptor dimerization

Extracellular subdomains II and IV, respectively, harboring the Y285C and D595V mutations, are involved in receptor autoinhibition in the absence of ligand and provide the main interactions between dimerized receptors upon ligand binding.^{15, 16} Mutations in these subdomains could potentially affect the integrity of the tethered autoinhibitory conformation that the ErbB4 extracellular domain adopts in the absence of ligand,¹⁵ or increase the affinity of dimerized receptors. In both cases, mutations would increase dimerization and thus activation of the receptor. To investigate whether ErbB4 Y285C and D595V mutations lead to increased receptor dimerization, NIH 3T3 fibroblasts stably expressing wild-type, Y285C or D595V ErbB4 were stimulated with increasing concentrations of NRG-1, crosslinked, and analyzed for the formation of ErbB4 dimers by western blotting. Both mutants formed homodimers more efficiently than wild-type ErbB4 upon activation (Figure 3a), indicating that the mutations indeed increase ErbB4 dimerization.

Figure 3

Analysis of ErbB4 Y285C and D595V dimerization. (a) NIH 3T3 cells stably expressing wild-type or mutant ERBB4 were stimulated for 10 min with 0, 0.5, 5 or 50 ng/ml of NRG-1, crosslinked with membrane impermeable BS₃, and analyzed for ErbB4 dimers by western blotting. (b) NIH 3T3 cells stably expressing wild-type ERBB4 or ERBB4 mutants were treated for 10 min with 50 ng/ml NRG-1 and analyzed for ErbB4-ErbB2 heterodimerization by co-immunoprecipitation and western blotting. (c) Structure of subdomain I/II interaction in the inactive structure of ErbB4 extracellular domain. M60, V28 and V58 of the subdomain I interact with F288 and Y285 of subdomain II. Carbon atoms—gray, oxygen atoms—red, nitrogen atoms—blue, sulfur atoms—yellow. (d) Structure of the dimerization arm docking site in the ErbB4 dimer. The incoming dimerization arm of the opposing receptor monomer is colored cyan. Key interacting residues are indicated. (e) The dimerization interface between subdomains IV around D595. The carbon atoms of the opposing receptor monomer are colored cyan.

ErbB receptors readily form heterodimers with other members of the ErbB family. In particular, the significance of ErbB4 heterodimerization with ErbB2 is well documented.^{13, 17, 18} As the observed Y285C and D595V-mediated increase in ErbB2 phosphorylation (Figure 2) could result from increased heterodimerization, wild-type ErbB4 and the two mutants were co-immunoprecipitated together with ErbB2 using stable NIH 3T3 transfectants stimulated or not with NRG-1. Consistent with the crosslinking experiments, both ErbB4 mutants demonstrated increased heterodimerizaton with ErbB2 as compared with wild-type ErbB4 (Figure 3b).

Structural effects of Y285C mutation on ErbB4 extracellular domain dimer

To understand the observed functional effects of the Y285C mutation on the ErbB4 structure, we examined the likely consequences of the mutation on the known ErbB4 structures. In the monomer structure (PDB code: 2AHX; tethered, autoinhibited conformation) and the dimer structure of ErbB4 (PDB code: 3U7U), tyrosine Y285 is a central, solvent-exposed residue on the surface of subdomain II; Y285 is surrounded by a cluster of mainly aromatic (three phenylalanines) and hydrophobic (proline, two valines) residues, as well as arginine R306, all from subdomain II.

In the monomer structure, both Y285 and F288 of subdomain II interact with subdomain I via aromatic/hydrophobic interactions with methionine M60, valine V28 and valine V58 (Figure 3c). Equivalent valine residues are found in human EGFR and human ErbB2, but otherwise the amino acids at this location are different and none of the residues and specific interactions are conserved in EGFR from Drosophila (PDB code: 1N8Z). In the monomer, the obvious consequences of the mutation Y285C are first to disrupt the subdomain I/II interaction since a key component is missing and this would facilitate the release of subdomain II from subdomain I, which needs to take place to form the active dimer.^{16, 19} Second, a small cysteine side chain at 285 replacing the large tyrosine side chain is ideally placed to form a disulfide bridge with one or more nearby cysteines and, would likely disrupt the normal pattern of disulfide bridges leading to some change in the local structure and residue packing in this region of subdomain II.

In the ErbB4 dimer structure, Y285 remains a central residue participating in critical interactions with three residues (F273, Y268 and P270) from the dimerization arm of the other monomer (Figure 3d). The aromatic ring of Y285 of one monomer directly interacts at right angles with F273 of the arm of the other monomer, and F273 is further stabilized by typical ring stacking interactions with the guanidinium group of R306, which is in turn sandwiched between F273 and V427 of subdomain III. In transitioning from the monomer to the dimer, V427 moves 29Å closer to R306 (based on their Cα atoms). On the opposite side of Y285 is located P294, which packs against the phenyl ring of Y285; and one of the many disulfide bridges present in the extracellular domains of ErbB4 is formed between C293 and C304. When Y285 is mutated to cysteine, the most obvious effects are the loss of the large aromatic side chain and likely interference with the C293–C304 disulfide bridge.

Naturally-occurring clusters of three cysteine residues leading to one disulfide bridge and one unpaired cysteine are very rare but it has been observed in the vascular endothelial growth factors VEGF-C and VEGF-D where the shift in disulfide pairing likely regulates the amount of dimeric growth factor present and hence receptor activity.²⁰ A likely effect of the Y285C change in ErbB4 would be to alter the local conformation, through forming alternative disulfide bridges, for example, with C293 and/or with C304, probably yielding a mixed population of disulfide linkages. The mutation Y285C in ErbB4 does not lead to covalent dimer formation, that is, by interfering with the other nearby disulfide bridge formed by C262 and C289 and causing inter-subunit disulfide pairing. Such a possibility would require multiple larger rearrangements locally in the structure, whereas Y285 when mutated to cysteine appears ideally placed to form a disulfide bridge with C293 and/or C304 within the same monomer. Indeed, the formation of covalent, disulfide-bonded dimers was not detected in western analyses under non-reducing conditions (data not shown).

The subdomain where Y285 is located is structurally relatively static if one compares the native monomer structure versus the native dimer structure of ErbB4, and loss of the key interaction with subdomain I stabilizing the monomer and variations in disulfide linkage could very well shift the equilibrium in favor of a higher population of dimers. Supporting this notion, the mutation results in space for other side chains (and thus the ErbB4 main chain too) to rearrange locally since the very large tyrosine side chain is being replaced by a small cysteine. Indeed, F297 and possibly P294 seem particularly well placed for moving into the space vacated by the loss of the phenyl ring, and thus maintain and even enhance the aromatic/hydrophobic cluster with F273 on the incoming arm of the other monomer (Figure 3d). Thus, several effects are likely to occur: in the monomer, the aromatic/hydrophobic interactions between subdomains I and II would be disrupted leading to exposure of the subdomain II surface that interacts with the dimerization arm in the dimer; interference of the normal disulfide pairing and removal of a large bulky side chain would entail some local main chain and side chain alterations that appear to support a change in equilibrium toward dimer formation.

Structural effects of ErbB4 D595V mutation

The D595V mutation leads to active ErbB4 dimer formation. The region harboring the D595V mutation is not present in the dimeric ErbB4 structure, and thus was modeled based on the structure of EGFR (see Materials and Methods for details). However, it is present in the monomeric ErbB4 structure. Interestingly, the conformation of this hairpin region, residues 592–607, appears to be structurally stabile since the region in the model, based on human EGFR, and the actual structural region in the monomer are very similar; as for many portions of ErbB4, this region is stabilized by a disulfide bridge, here involving C593. In the EGFR dimer structure, and hence in the dimeric model of ErbB4, both monomers directly interact with each other in this region and D595 from both monomers is located at this interface.

In the inactive structure of ErbB4 extracellular domain (PDB code: 2AHX), D585, P594, L597 and K606 of the tether loop of subdomain IV pack against residues Y268, P270 and F273 of the tether loop of subdomain II (Supplementary Figure 1A). D595 is located on the opposite side of subdomain IV, and is not involved in the tether formation. D595 forms a strong electrostatic interaction with K592, which, in the case of D595V mutation, is poised to relocate its long side chain to interact with E591. E591 in turn interacts with Q558, but K592 could still maintain its hydrophobic interactions with F583, which stacks against R612. Thus, the disruption of D595-K592 interaction does not explain the increased dimerization of D595V mutant. Accordingly, mutation of K592 into isoleucine did not increase, but abolished dimer formation (Supplementary Figure 1B).

In forming the dimer, the tether loop of subdomain II from each monomer relocates to the vicinity of Y285 of the other monomer (see above), and the tether loop of subdomain IV of each monomer bind to each other along the opposite face (opposite to where subdomain II binds in the monomeric structure) (Figure 3e). In wild-type dimeric ErbB4, D595 would necessarily form a strong ion pair with K592. This interaction would serve to neutralize the negative charge of D595 at the interface of the two monomers, explaining the effects of the K592I mutation on ErbB4 dimer formation (Supplementary Figure 1B). With the D595V mutation, in the dimer, the side chain of K592 is long enough and flexible enough to move out of this region of the interface. Valine at position 595 is a hydrophobic residue and would promote interactions in the dimer because the surrounding residues are aromatic (F603 and F605; repeated twice) and hydrophobic (P594 and L597, repeated twice) at this symmetrical interface of the two monomers in the dimer.

Taken together, these data indicate that the activating ErbB4 mutation Y285C in the extracellular domain disrupts interactions stabilizing the monomer and likely enhances interactions in the dimer; whereas the D595V mutant clearly would enhance the hydrophobic interactions at a key interface formed in the dimer.

ErbB4 K935I mutation increases transactivation between ErbB4 monomers

Upon receptor dimerization, the kinase domains of two ErbB4 monomers interact (Figure 4a), forming an asymmetrical dimer (PDB code 3BCE; note that to obtain the coordinates for the asymmetric dimer requires application of symmetry operations to the coordinates).²¹ In this dimer, the C-terminal lobe of the activator kinase and the N-terminal lobe of the receiver kinase interact, resulting in allosteric activation of the receiver kinase in trans.^{21, 22} The asymmetric interface between kinase domains extends from one solvent-exposed surface (where a key strong charge-pair interaction takes place between K955, activator kinase domain, and D767, receiver kinase domain), a series of hydrophobic residues from the receiver kinase domain (V763, M766, A769, L770, L788) and from the activator kinase domain (I947, M951, V954, M958, I959), leading to the highly charged interactions at the other solvent-exposed surface. It is at this latter surface where the mutations D931Y and K935I are located within the C-terminal lobe of the activator kinase (Figure 4a).

Figure 4

Transactivation by the ErbB4 D931Y and K935I mutants. (a) Left: schematic illustration of the asymmetric ErbB4 kinase dimer. Right: sequence alignment of the activator kinase dimerization interface. The positions of D931Y, K935I and an artificial V954R mutation disrupting the asymmetric kinase dimer are indicated by arrows. (b) The outline of the transactivation assay. Left: Upon ErbB receptor dimerization, the kinase domains interact in head-to-tail manner, resulting in activation of the receiver kinase by the activator kinase in trans. Middle: ErbB4 mutants harboring either the V954R mutation or the KD K751R mutation are not able to activate the kinase enzyme alone. Right: When the two mutants (V954R and K751R) are co-expressed, a functional kinase dimer interface is formed. The KD ErbB4 K751R serves as the activator kinase, activating the ErbB4 V954R mutant in trans. By introducing D931Y and K935I mutations to the KD ErbB4 background, the effects of the mutations to transactivation can be analyzed. Purple, cyan and red stars indicate V954R, K751R and D931Y/K935I mutations, respectively. w/wo, with or without. (c) Transactivation by the ErbB4 mutants was analyzed in COS-7 cells transiently transfected with indicated ERBB4 constructs using western blotting. Basal ErbB4 phosphorylation was used as a readout of activation. (d) COS-7 cells transiently transfected with constructs encoding ERBB2 and wild-type ERBB4 or ERBB4 mutants were treated for 10 min with 50 ng/ml NRG-1 and analyzed for ErbB4-ErbB2 heterodimerization by co-immunoprecipitation and western blotting. (e) Part of the dimerization interface in the asymmetric kinase dimer harboring D931 and K935. The carbon atoms of the activator and receiver kinase are shown in gray and cyan, respectively.

K935 is fully conserved among human ErbB receptors, whereas D931 is conserved in all ErbB receptors except EGFR where it is serine (Figure 4a). The location of the two activating mutations suggests that on dimer formation the mutations strengthen the interaction between the kinase domains, leading to an increase in transactivation between dimerized ErbB4 receptors. To address this, a point mutation V954R was first introduced into the C-terminal lobe of the full-length ErbB4 kinase domain, disrupting its capability to serve as an activator kinase (Figures 4a and b).^{23, 24} Next, the cancer-associated mutations—D931Y and K935I—were introduced separately into the kinase-dead (KD) ErbB4 K751R background. In principle, neither the V954R mutant, due to the disrupted activator interface, nor the KD ErbB4 is able to activate the kinase enzyme alone. However, when the two mutants (V954R and K751R) are co-expressed, a functional kinase dimer interface is formed. The KD ErbB4 K751R serves as the activator kinase, activating the V954R mutant ErbB4 in trans (Figure 4b). The mutations were introduced into the pcDNA3.1.ERBB4JM-aCYT-2-HA expression vector encoding a HA-tagged ErbB4 JM-a CYT-2 isoform, affording a high expression level favoring ErbB4 homodimerization, and enabling the use of basal ErbB4 phosphorylation as the readout for ErbB4 transactivation.

Using this system, the transactivation potency of the mutants was analyzed in COS-7 cells by transiently transfecting the cells with the V954R mutant alone or together with KD ErbB4 K751R with or without D931Y or K935I mutations. As hypothesized, the K935I mutation increased the transactivation (Figure 4c), suggesting that the mutation enhances the interaction between ErbB4 kinase monomers. Surprisingly, the D931Y mutation, while equally activating as K935I (Figure 2), did not increase the transactivation between dimerized ErbB4 monomers (Figure 4c).

The inability of D931Y mutation to increase the activity of ErbB4 homodimers (Figure 4c) and the apparent ability to promote increased phophorylation of ErbB2 (Figure 2) suggested that there is a difference in the capability of the D931Y mutant to homo- vs heterodimerize. We have previously shown that cancer-associated ErbB4 mutations can promote a qualitative shift in active ErbB4 dimers by favoring the formation of ErbB4-ErbB2 heterodimers over ErbB4 homodimers.¹³ To address whether the D931Y mutation could promote heterodimerization, wild-type ErbB4 or the ErbB4 mutants D931Y and K935I were co-immunoprecipitated together with ErbB2 from transiently transfected COS-7 cells stimulated or not with NRG-1. Both ErbB4 mutants demonstrated increased association with ErbB2 in response to ligand stimulation (Figure 4d). However, K935I, but not D931Y, increased the formation of ErbB4 homodimers in a crosslinking experiment (Supplementary Figure 2). These data indicate that the ErbB4 D931Y mutation induces a qualitative shift in the activity of ErbB4 dimers, increasing the activity of ErbB4-ErbB2 heterodimers, but not ErbB4 homodimers. The ErbB4 K935I mutation, on the other hand, is a potent activator of both ErbB4-ErbB2 heterodimers and ErbB4 homodimers.

Structural effects of D931Y and K935I mutations

The location of D931 and K935 in the activator kinase is on helix 10 at the solvent-exposed interface with the receiver kinase in the ErbB4 asymmetric dimer (PDB code: 3BCE) (Figure 4e). Residues along this helix are identical in ErbB2 (PDB code: 3PP0). From the receiver kinase, two loops approach D931 and K935: K714-E715-T716 (K-E-T in ErbB2) (Figure 4e) and C814-L815-S816-P817 (C-L-T-S in ErbB2). Thus, in the immediate surrounding of D931 and K935 only two residues differ between ErbB4 (serine-proline) and ErbB2 (threonine-serine). If D931Y in ErbB4 would alter the local conformation, then hydrogen bonding to both serine and threonine in ErbB2 could take place and hence stabilize the heterodimer complex in comparison with the ErbB4 dimer where serine and proline are present.

The K935I mutation leaves a main-chain hydrogen bond between K935I and E715 intact, but would break the polar interaction between K935 and D931 (Figure 4e). Nonetheless, the bulky hydrophobic side chain of isoleucine would pack nicely within a hydrophobic ‘cup’ formed by the aliphatic portions of K714-E715-T716 and possibly P790 from the receiver kinase, and D931, L932, E934 and E937 of the activator kinase (Figure 4e). Thus, despite being largely surrounded by polar side chains, the K935I mutation would effectively increase the overall hydrophobic character and affinity of the kinase dimer interface and facilitate dimer formation, which is consistent with the observed increase in transactivation.

Activating ERBB4 mutations promote survival in NIH 3T3 cells

The functional effects of the activating ERBB4 mutations were addressed in NIH 3T3 cells stably transfected with wild-type ERBB4 or ERBB4 mutants Y285C, D595V or K935I. In full (10%) serum, the expression of wild-type or mutant ERBB4 did not have a significant effect on the growth of NIH 3T3 cells (data not shown). However, in the absence of serum, the viability of NIH 3T3 vector control cells and cells expressing wild-type ERBB4 quickly declined, whereas the cells expressing the ERBB4 mutants were significantly more resistant to serum starvation (Figure 5a). Analysis of ErbB4 phosphorylation after 0, 24 and 72 h of serum starvation demonstrated that while wild-type ErbB4 was only moderately phosphorylated in full serum and the phosphorylation level was not detectable after 24 h of starvation, the ErbB4 mutants demonstrated robust ErbB4 phosphorylation in full serum and receptor phosphorylation was sustained for at least 72 h after serum withdrawal (Figure 5b). The sustained ErbB4 phosphorylation of the mutants was reflected to some extent as sustained Erk phosphorylation and PCNA expression (Figure 5b), suggesting prolonged proliferative signaling induced by the ErbB4 mutants in NIH 3T3 cells following serum starvation.

Figure 5

Effect of ERBB4 mutants on survival and intracellular signaling in NIH 3T3 cells. (a) To analyze the effect of ERBB4 mutations on cell survival, NIH 3T3 cells stably expressing wild-type ERBB4 or indicated ERBB4 mutants were plated on 96-well plates in quadruplicates (4000 cells/well). The following day the cells were washed three times and serum-free growth medium was applied. Viability of the cells was analyzed at indicated timepoints using the MTT assay. The mean and s.d. are shown. **P<0.01; ***P<0.001; compared with cells expressing wild-type ERBB4 by one-way ANOVA. (b) To analyze the effects of serum starvation on intracellular signaling, the NIH 3T3 transfectants were grown on 10 cm culture dishes and subjected to serum starvation as in (a). Protein samples were extracted at the indicated timepoints. The phosphorylation of ErbB4, Erk1/2 and Akt was analyzed by western blotting using phospho-specific antibodies, and PCNA expression using PCNA antibody. Loading was controlled using antibodies recognizing total ErbB4, Erk1/2, Akt or actin. (c and d) Analysis of PDGFRA mRNA (c) and protein (d) expression in serum starved NIH 3T3 transfectants. The cells were cultured and treated as in (b), and RNA and protein samples were extracted at the indicated timepoints. PDGFRA mRNA expression was analyzed using real-time RT–PCR (c), and protein level was analyzed by western blotting (d).

In addition to canonical RTK signaling, ErbB4 can be cleaved from the cell membrane through RIP (regulated intra-membrane proteolysis) and the resulting soluble intracellular domain (ICD) can directly translocate into the nucleus and regulate transcription.^{25, 26, 27, 28, 29, 30, 31} Interestingly, all three ErbB4 mutants demonstrated increased ErbB4 cleavage in NIH 3T3 cells both basally and upon serum starvation when compared with wild-type ErbB4 (Figure 4b, long ErbB4 exposure). To address the signaling activity of the soluble ErbB4 ICD, platelet-derived growth factor receptor-alpha (PDGFRA) expression was analyzed in NIH 3T3 cells expressing wild-type ERBB4 and the different ERBB4 mutants. Soluble ErbB4 ICD has previously been reported to associate with the PDGFRA promoter and stimulate PDGFRA transcription in response to serum starvation.³¹ In cells expressing the ERBB4 mutants, PDGFRA mRNA and protein levels were robustly increased already at the 24-h time point after serum withdrawal and were sustained at a higher level than in vector control cells or cells expressing wild-type ERBB4 still at the 72-h time point (Figure 5c and d).

Taken together, these data demonstrate that the ERBB4 mutations activate RIP-mediated ErbB4 signaling and promote the survival of NIH 3T3 cells.

Discussion

ERBB4 mutations are frequent in non-small cell lung cancer,⁷ making ErbB4 a potential novel drug target. However, the functional effects of lung cancer-associated ERBB4 mutations are largely unknown. Moreover, the identified ERBB4 mutations harbor all functional subdomains of ErbB4 protein with no hot spot mutations, which is in contrast to lung cancer-associated EGFR mutations that typically target few sites in the kinase domain.¹⁴ Here, we have functionally analyzed nine ERBB4 mutations previously identified in a study introducing ERBB4 as a highly mutated gene among 623 putative cancer genes in a series of 188 lung adenocarcinoma samples.⁴

Our results indicated that four (Y285C, D595V, D931Y and K935I) out of the nine analyzed ERBB4 mutations were activating, increasing both basal and ligand-induced ErbB4 phosphorylation. The activating mutations were located both in the extracellular ligand binding/dimerization domain (Y285C and D595V) and in the intracellular kinase domain (D931Y and K935I). Interestingly, all activating ErbB4 mutations mapped structurally to interaction sites within the monomer form and between receptor monomers in ErbB4 dimers, suggesting that the enhanced activation was due to effects on receptor dimerization, both uncovering regions in the monomer necessary for dimerization and/or by stabilizing interfaces in the dimer. Indeed, the activating ErbB4 extracellular domain mutations Y285C and D595V increased ErbB4 dimerization in crosslinking experiments, as well as enhanced ErbB4 heterodimerization, as demonstrated by ErbB2 co-immunoprecipitation. Furthermore, the activating kinase domain mutations, D931Y and K935I, enhanced receptor dimerization and transactivation in ErbB4 homo- and/or heterodimers, also indicating increased interaction between receptor monomers in ErbB4 dimers.

The structural locations and functional effects of the activating ErbB4 mutations are similar to those of the recently characterized oncogenic ErbB2 and ErbB3 mutations.^{32, 33} Mutations in the extracellular subdomain II of ErbB2 induce covalent inter-receptor dimers,³³ whereas ErbB3 extracellular subdomain II and kinase domain mutations promote the activation of co-expressed ErbB2.³² Thus, in the case of both ErbB2 and ErbB3, the mutations increase the interaction and activity of ErbB homo- or heterodimers, which is analogous to our results with ErbB4 mutations.

When overexpressed in NIH 3T3 cells, the ErbB4 mutants Y285C, D595V and K935I significantly promoted cell survival upon serum starvation, indicating that activating ERBB4 mutations are potentially oncogenic. Interestingly, in the patient samples harboring the four identified activating mutations Y285C, D595V, D931Y and K935I, the total number of mutated cancer-associated genes was significantly lower than in patients harboring non-activating ERBB4 mutations (4.25 vs 18 mutations in average, P=0.003, t-test).⁴ This finding is in line with our data that the four identified activating ERBB4 mutations were functionally distinct from the other analyzed ERBB4 mutations and consistent with the hypothesis that the activating ERBB4 mutations are functional in these tumors.

The activating ERBB4 mutations were not mutually exclusive with KRAS mutations, as the patient harboring ErbB4 D595V mutation also harbored a KRAS G12V mutation. This is unexpected, because ERBB4 mutations are mutually exclusive with EGFR mutations,⁴ suggesting that mutated ErbB4 acts in the same pathway or has similar functions as mutated EGFR. However, our results indicated that while the activating ERBB4 mutations significantly promoted the survival of NIH 3T3 cells, this effect was only moderately translated to the activation of RAS-MAPK pathway, the canonical RTK pathway used by activating EGFR mutants^{34, 35} as well as by mutant KRAS. In contrast, the activating ErbB4 mutants demonstrated increased cleavage of functionally active ErbB4 ICD from the cell membrane. RIP-mediated release of the ErbB4 ICD is a mode of signaling unique for ErbB4 among the ErbB receptors and ErbB4 ICD has previously been shown to mediate growth-promoting ErbB4 signaling.^{29, 31} Thus, it is possible that activated ErbB4 mutants, potentially through the action of ErbB4 ICD, provide parallel, synergistic signaling in mutant KRAS-driven tumors, rather than alternative means to activate the RAS-MAPK pathway. Interestingly, a recent study demonstrated that in xenograft and transgenic mouse models of non-small cell lung cancer, ligand-mediated ErbB4 signaling has an important role in tumor recurrence after chemotherapy, irrespectively of the KRAS mutation status.³⁶ Furthermore, in metastatic melanoma, the oncogenic ERBB4 mutations typically co-occur with mutations in RAS or RAF family genes,⁸ further suggesting that oncogenic ErbB4 signaling is mediated via alternative pathways other than the RAS-MAPK pathway.

Taken together, this study demonstrates the presence of activating ERBB4 mutations in non-small cell lung cancer and warrants further studies assessing the potential therapeutic value of targeting mutated ErbB4 in this disease.

Materials and methods

Cell culture

COS-7 and NIH 3T3 cells were cultured in DMEM (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FCS (Promocell, Heidelberg, Germany), 50 U/ml penicillin and 50 μg/ml streptomycin.

Expression plasmids

To transiently overexpress ERBB4, pcDNA3.1.ERBB4JM-aCYT-2-HA or pBABE-puroERBB4JM-aCYT-2-HA encoding the HA-tagged JM-a CYT-2 ErbB4 isoform was used to achieve high or moderate expression levels, respectively. The pBABE-puroERBB4JM-aCYT-2-HA plasmid was constructed by cloning a 4-kilobase PmeI fragment containing the whole coding sequence of HA-tagged ErbB4 JM-a CYT-2 from pcDNA3.1.ERBB4JM-aCYT-2-HA into SnaBI-digested pBABE-puro. ERBB4 point mutations were introduced into pcDNA3.1.ERBB4JM-aCYT-2-HA and pBABE-puroERBB4JM-aCYT-2-HA by site-directed mutagenesis as previously described.³⁷ The pcDNA3.1.ERBB2^{(ref. 13)} construct was used to transiently overexpress ERBB2. All constructs were verified by sequencing.

Generation of cell lines stably overexpressing ERBB4

To generate cells stably expressing wild-type ERBB4 or ERBB4 mutants, NIH 3T3 fibroblasts were infected with retroviruses encoding wild-type or mutant pBABE-puroERBB4JM-aCYT-2-HA as previously described.¹³ Infected cells were selected with 2 μg/ml puromycin (Invitrogen) to create virus-positive cell pools. Selected cells were maintained in 1 μg/ml puromycin.

Western blotting

To analyze the activation of ERBB4 mutants, COS-7 cells growing on 6-well plates were transiently transfected with 2 μg of wild-type or mutant pBABE-puroERBB4JM-aCYT-2-HA using Fugene6 transfection reagent (Promega, Madison, WI, USA). The culture medium was replaced by DMEM without serum 3–5 h after transfection. Cells were starved overnight, stimulated with 50 ng/ml NRG-1 for 10 min at 37 °C and lysed in lysis buffer (10 mM Tris-HCl, pH 7.4, 1% Triton-X100, 150 mM NaCl, 1 mM EDTA, 10 mM NaF) supplemented with 2 mM PMSF, 10 mg/ml aprotinin, 10 mg/ml leupeptin, 10 mM Na₄P₂O₇ and 1 mM Na₃VO₄. Cell lysates were analyzed for ErbB4 phosphorylation by western blotting with phospho-specific ErbB4 antibody (#4757, Cell Signaling Technology, Beverly, MA, USA). The filters were reblotted with ErbB4 antibody (E200; Abcam, Cambridge, UK) and loading was controlled by western blotting with actin antibody (sc-1616; Santa Cruz Biotechnology, Santa Cruz, CA, USA).

For ErbB4 transactivation assays, COS-7 cells growing on 12-well plates were transiently transfected with the indicated wild-type or mutant pcDNA3.1.ERBB4JM-aCYT-2-HA constructs using 0.5 μg of DNA for single transfections or 0.25 μg of DNA for each construct in double transfections. ErbB4 activation was analyzed 24 h after transfection by western blotting using phospho-ErbB4 antibody (#4757, Cell Signaling Technology).

For the analysis of ErbB4 signaling in NIH 3T3 transfectants, cells growing on 6 cm dishes were starved for 0, 24 or 72 h in DMEM+0% FCS, lysed and analyzed by western blotting with phospho-ErbB4 (#4757, Cell Signaling Technology), phospho-Akt (#9271, Cell Signaling Technology), phospho-Erk (#9101, Cell Signaling Technology), ErbB4 (E200; Abcam), Akt (sc-1618, Santa Cruz Biotechnology), Erk (#9102, Cell Signaling Technology), PCNA (sc-56, Santa Cruz Biotechnology) or PDGFR-alpha (07-276, Millipore, Billerica, MA, USA) antibodies. Loading was controlled using actin antibody.

Immunoprecipitation and co-immunoprecipitation

For immunoprecipitation (IP) experiments, stably transfected NIH 3T3 cells or transiently transfected COS-7 cells (pcDNA3.1.ERBB2 and/or pBABE-puroERBB4JM-aCYT-2-HA) growing on 10 cm dishes were starved overnight, stimulated with 50 ng/ml NRG-1 for 10 min at 37 °C and lysed in lysis buffer. Lysates were pre-cleared by incubating with 20 μl of G-sepharose beads (Santa Cruz Biotechnology) for 2 h at 4 °C. Pre-cleared lysates containing 800–1000 μg of total protein were subjected to IP with 1 μg of ErbB2 antibody (MA5-14057, Thermo Scientific, Waltham, MA, USA) together with 20 μl of G-sepharose beads. After overnight incubation at 4 °C, the beads were washed four times with lysis buffer and boiled in SDS–PAGE loading buffer. The level of ErbB2 phosphorylation was analyzed from immunoprecipitated samples by western blotting with phosphotyrosine antibody (4G10). The amount of co-immunoprecipitated ErbB4 was analyzed by western blotting.

Analysis of ErbB4 dimers

To analyze the formation of ErbB4 dimers, COS-7 or NIH 3T3 cells transfected with wild-type or mutant pBABE-puroERBB4JM-aCYT-2-HA constructs were starved overnight and stimulated with 0, 0.5, 5 or 50 ng/ml NRG-1 for 10 min at 37 °C. After stimulation, the cells were washed three times with ice-cold PBS (phosphate-buffered saline) and subsequently subjected to chemical crosslinking with 2 mM BS₃ (Pierce, Rockford, IL, USA) in PBS. After 1-h incubation on ice, the reaction was quenched by incubation with ice-cold 50 mm Tris-HCl, pH 7.4, 150 mm NaCl for 15 min on ice. After quenching, the cells were washed three times with ice-cold PBS and lysed in lysis buffer. The formation of ErbB4 dimers was analyzed by western blotting after running the protein samples on 4–20% gradient gels (Bio-Rad, Hercules, CA, USA).

Modeling and structural assessments

Initially, we prepared a composite model based on the 3.30-Å resolution dimer structure of human EGFR (PDB code: 3NJP) covering the extracellular domain, including the regions where Y285C and D595V are located. The sequences of ErbB4 and EGFR are 48% identical, meaning that modeling should give a good representation of the main chain atoms but more variation may be seen in side chain conformations. The model was prepared using Sybyl (Tripos International, St. Louis, MO, USA).

Subsequently, the 3.03-Å resolution X-ray structure of the active dimer of human ErbB4 with bound neuregulin-1 became available (3U7U),¹⁶ and the main chain regions superpose well with a root mean squared deviation of 1.41 Å. Herein we have assessed the effects of mutants based on comparing the ErbB4 structures for the inactive ligand-free monomer (2AHX, 2.4 Å resolution)¹⁵ versus the active dimer with bound ligand (3U7U). In the case of the D595V mutation, the model of that region based on the dimeric EGFR structure (3NJP) was used together with the region from the monomeric structure. For evaluation of the kinase domain mutants, the 2.5-Å resolution structure of human ErbB4 (PDB code: 3BCE)²¹ was used as well as the 2.25-Å resolution structure of human ErbB2 (PDB code: 3PP0).³⁸ In the text, the residue number differs from the numbering used in the X-ray files for ErbB4. Structural analysis, visualization and figures were made using Bodil.³⁹

Cell viability assay

NIH 3T3 fibroblasts stably expressing wild-type or mutant ERBB4 were plated into 96-well plates in quadruplicates at a density of 4000 cells/well in DMEM+10% FCS. The following day, the medium was replaced by DMEM+0% FCS. Cell viability was assessed at indicated timepoints using the MTT assay (CellTiter 96 nonradioactive cell proliferation assay; Promega).

Real-time RT–PCR

RNA samples were extracted⁴⁰ from NIH 3T3 transfectants starved for 0, 24 or 72 h in DMEM without FCS. Real-time RT–PCR analysis for PDGFRA expression was carried out as previously described.^{31, 40, 41}