Research ArticleNon–small cell lung cancer

Blocking NRG1 and Other Ligand-Mediated Her4 Signaling Enhances the Magnitude and Duration of the Chemotherapeutic Response of Non–Small Cell Lung Cancer

See allHide authors and affiliations

Science Translational Medicine  06 Feb 2013:
Vol. 5, Issue 171, pp. 171ra18
DOI: 10.1126/scitranslmed.3004438


Although standard chemotherapies are commonly used to treat most types of solid tumors, such treatment often results in inadequate response to, or relapse after, therapy. This is particularly relevant for lung cancer because most patients are diagnosed with advanced-stage disease and are treated with frontline chemotherapy. By studying the residual tumor cells that remain after chemotherapy in several in vivo non–small cell lung cancer models, we found that these cells have increased levels of human epidermal growth factor receptor (HER) signaling due, in part, to the enrichment of a preexisting NRG1HI subpopulation. Neuregulin 1 (NRG1) signaling in these models can be mediated by either the HER3 or HER4 receptor, resulting in the differential activation of downstream effectors. Inhibition of NRG1 signaling inhibits primary tumor growth and enhances the magnitude and duration of the response to chemotherapy. Moreover, we show that inhibition of ligand-mediated Her4 signaling impedes disease relapse in cases where NRG1 inhibition is insufficient. These findings demonstrate that ligand-dependent Her4 signaling plays an important role in disease relapse.


For most cancers, poor response to or relapse after chemotherapy is a major cause of mortality. This is particularly relevant for non–small cell lung cancer (NSCLC) because chemotherapy is used to treat patients with all stages of disease. More than 50% of patients present with advanced disease and are treated with frontline chemotherapy (1). However, despite the aggressive use of chemotherapy in the treatment of NSCLC, the 5-year survival rate remains at 3.8% for patients with distant disease (2).

Partial responses to chemotherapy and relapse after chemotherapy suggest that tumor cells are heterogeneous in their drug sensitivities; some tumor cells are effectively killed by a given agent, whereas others are spared. Cancer stem cells (CSCs) have been an area of intense study in recent years, and several groups have reported that CSCs show enhanced resistance to conventional chemotherapeutic agents and radiation treatment (38). However, we previously showed that the tumor cells that drive disease relapse in several NSCLC models are not enriched for stem cell properties (9). Therefore, to avoid making any assumptions regarding the nature of these cells, we refer to them herein as “tumor-reinitiating cells” (TRICs).

It is believed that the use of targeted therapies may eradicate tumor cell populations that are resistant to conventional chemotherapeutics. Epidermal growth factor receptor (EGFR) inhibitors are frequently used to treat NSCLC patients and are effective in treating the majority of patients whose tumors harbor an EGFR-activating mutation. Activation of EGFR signaling is a relatively frequent event in lung adenocarcinoma (10). EGFR is the prototypical member of the ErbB or human epidermal growth factor receptor (HER) family of tyrosine kinases, which includes EGFR (HER1), HER2, HER3, and HER4. Recent evidence shows that other HER family members may also play a role in NSCLC. However, their contributions to the disease are less well characterized and most studies have focused on their ability to activate EGFR signaling (1114). Neuregulin 1 (NRG1) is a ligand for the HER3 and HER4 receptors. NRG1 autocrine signaling has been shown to regulate lung epithelial cell proliferation and play a role in human lung development (15) and has been implicated in insensitivity of NSCLC to EGFR inhibitors (12).

Here, we sought to elucidate the mechanisms underlying resistance to and relapse after conventional chemotherapies. We used models of NSCLC that undergo stasis or tumor regression followed by disease relapse after chemotherapy treatment to characterize the residual tumor cells (TRICs). Here, we identify NRG1 as being highly enriched in residual tumor cells and show that naïve tumors contain a small preexisting population of NRG1-expressing cells. We show that inhibition of NRG1- and other ligand-mediated Her4 signaling has variable effects on primary tumor growth, but consistently and significantly enhances the response to chemotherapy and delays tumor regrowth after cessation of treatment.


Activation of HER signaling in TRICs

To characterize TRIC populations, we used the LSL-K-rasG12D genetically engineered mouse model (GEMM) of NSCLC (16) crossed to the Z/EG Cre-reporter strain (17) and three human xenograft models. Cisplatin treatment of LSL-K-rasG12D mice results in a reduction in tumor burden but does not prolong survival, indicating that tumors resume growing after therapy (18). Lungs from LSL-K-rasG12D were collected 1 week after the final dose of cisplatin, and green fluorescent protein (GFP)–positive tumor cells were isolated by fluorescence-activated cell sorting (FACS). To characterize differences between vehicle-treated and residual tumor cells, we isolated RNA from the tumor cells (PI and GFP+) and performed expression profiling. Of all the genes on the array, NRG1 showed the most significant enrichment in TRICs. Using independently generated samples, we carried out quantitative real-time polymerase chain reaction (qRT-PCR) to confirm the microarray results and extended our analysis to additional HER ligands and receptors (Fig. 1A). In addition to NRG1, expression of Her2, Her3, Her4, betacellulin (BTC), and heparin-binding EGF (HB-EGF) was also highly enriched in the TRICs.

Fig. 1

Enrichment of HER receptors and ligands and HER pathway activation in TRICs. (A) NRG1 mRNA is enriched in TRICs from the K-rasLSL-G12D mouse NSCLC model. Analysis is shown for one microarray probe (n = 6 per group). Expression of HER ligands and receptors was assessed by qRT-PCR on independent samples (mean ± SEM). (B to D) Expression of HER ligands and receptors in vehicle-treated and residual chemotherapy-treated tumor cells was assessed by qRT-PCR in the H1299 (B), Calu3 (C), and H441 (D) models (mean ± SEM). (E) qRT-PCR analysis of NRG1 expression in vehicle-treated and residual gemcitabine- and vinorelbine-treated Calu3 and H441 tumor cells. (F) Western blots of vehicle-treated and residual chemotherapy-treated Calu3 and H441 tumor cells. (G) Immunohistochemistry (IHC) for p-HER3 on residual chemotherapy-treated and vehicle-treated Calu3 tumors. Red arrows indicate small nests of p-Her3–positive tumor cells in the vehicle-treated tumor. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

This finding led us to assess the expression of HER pathway components in TRICs isolated from xenograft models in which tumors regress or undergo stasis in response to cisplatin + paclitaxel doublet chemotherapy and relapse after the cessation of treatment. We generated GFP-expressing sublines of the Calu3, H441, and H1299 human NSCLC xenograft models. GFP+ tumor cells were isolated from the Calu3, H441, and H1299 xenograft models by FACS after treatment with carboplatin and paclitaxel, a standard chemotherapy doublet used in frontline treatment of NSCLC. We observed strong enrichment of distinct but overlapping HER pathway components in all of the models (Fig. 1, B to D). In the case of the H1299 model, only HER4 itself was highly enriched in TRICs (Fig. 1B), whereas the Calu3 model showed enrichment of multiple HER4 ligands (Fig. 1C) and the H441 model showed enrichment of multiple HER ligands and receptors (Fig. 1D). Among the ligands, NRG1 showed the highest level of enrichment, and in all cases, the combination of ligand and receptor enrichment was indicative of increased HER4 activity.

In addition, we assessed NRG1 expression levels in TRICs isolated from the Calu3 and H441 xenograft models after treatment with other chemotherapeutic agents commonly used to treat NSCLC. Treatment with gemcitabine resulted in tumor regression in each of these models, and the residual tumor cells were highly enriched for NRG1 (Fig. 1E). However, Calu3 and H441 tumors continued to grow upon treatment with vinorelbine, and treatment did not result in enrichment of NRG1.

We also assessed NRG1 protein levels and activation of the HER3 receptor in the small number of residual tumor cells (<20,000) that could be isolated from the Calu3 and H441 models after treatment with carboplatin and paclitaxel. NRG1 and phospho-Her3 (p-Her3) levels were consistently higher in TRICs than in vehicle-treated tumor cells (Fig. 1F). Finally, we examined the activation of HER3 by immunostaining tumors for p-HER3. The majority of tumor cells in the residual tumors were p-HER3–positive, whereas the vehicle-treated tumors showed only scattered clusters of cells (Fig. 1G). Thus, residual tumor cells express NRG1 and show enhanced receptor activation, demonstrating increased NRG1 signaling.

Generation of NRG1-blocking antibodies

We next set out to evaluate the contributions of NRG1 and its receptors, HER3 and HER4, to primary tumor growth and relapse after chemotherapy. Anti-HER3 and anti-HER4 antibodies that block ligand-induced receptor activation were previously developed and characterized (19, 20). Because function-blocking antibodies to NRG1 were not available, we generated two high-affinity, phage-derived monoclonal antibodies (mAbs) to NRG1: YW538.24.71 and YW526.90.28. The NRG-1 gene gives rise to numerous protein isoforms. All isoforms contain an EGF-like domain, composed of a common exon that undergoes alternative splicing to generate a functional EGF-like α or β domain (NRG1α and NRG1β), which is necessary and sufficient for receptor tyrosine kinase (RTK) activation (21, 22). The antibodies were targeted to the common portion of the EGF-like domain. Both YW538.24.71 and YW526.90.28 had similar affinity for NRG1α isoform [dissociation constant (Kd) of 1.32 and 0.587 nM, respectively], but YW538.24.71 had a significantly higher affinity for NRG1β than YW526.90.28 (Kd of 0.00192 and 1.09 nM, respectively) (Fig. 2, A and B).

Fig. 2

Generation of high-affinity NRG1-blocking antibodies. (A and B) Affinity measurements of YW538.24.71 and YW526.90.28 for NRG1α and NRG1β isoforms using BIAcore. (C and D) Cellular IC50 for YW538.24.71 and YW526.90.28 against NRG1α (C) and NRG1β (D) as evaluated by kinase receptor activation assay (KIRA). (E) Enzyme-linked immunosorbent assay (ELISA) analysis of binding of YW538.24.71 and YW526.90.28 to EGF-family ligands. (F) Western blot showing inhibition of ligand-induced Her4 phosphorylation in H522 cells by YW538.24.71.

We then tested the ability of the mAbs to inhibit Her3 phosphorylation in response to exogenous NRG1 ligand in a cell-based assay. Both YW538.24.71 and YW526.90.28 inhibited Her3 activation in response to NRG1α stimulation with a similar median inhibitory concentration (IC50) (14 ± 2.8 nM and 13.9 ± 4.8 nM, respectively) (Fig. 2C). However, YW538.24.71 was a more potent blocker of NRG1β-induced Her3 activation than YW526.90.28 (IC50, 0.12 ± 0.017 nM and 1.44 ± 0.5 nM, respectively) (Fig. 2D).

To confirm antibody specificity, we assessed binding of the related ligands: EGF, HB-EGF, and BTC. There was no binding of these ligands to either antibody (Fig. 2E). Like NRG1, HB-EGF and BTC are also ligands for the HER4 receptor. Therefore, we tested the ability of YW538.24.71 to inhibit BTC- and HB-EGF–induced Her4 phosphorylation in H522 cells. Although YW538.24.71 potently inhibited NRG1β-induced Her4 phosphorylation, it had no effect on BTC- or HB-EGF–induced phosphorylation (Fig. 2F).

A preexisting subpopulation of NRG1-expressing cells

The increased expression of NRG1 observed in residual tumor cells could be due to the enrichment of a preexisting NRG1-expressing subpopulation. Alternatively, chemotherapy might induce NRG1 expression in cells, making them resistant to its cytotoxic effects. To distinguish between these possibilities, we used YW538.24.71 to determine whether NRG1-expressing cells were present in untreated tumors by FACS. NRG1 is a transmembrane protein that undergoes proteolysis resulting in release of the soluble ligand (23) and, therefore, may be detectable on the cell surface by FACS. Both Calu3 xenograft tumors and tumors from the LSL-K-rasG12D GEMM contain a small NRG1Hi subpopulation (0.54 ± 0.04% and 0.083 ± 0.05%, respectively) (Fig. 3, A and B). These findings were confirmed by qRT-PCR, which revealed increased NRG1 transcript levels in the NRG1Hi cells (Fig. 3B). Moreover, evaluation of freshly resected, treatment-naïve NSCLC patient samples revealed the presence of a small NRG1Hi tumor cell population in three of the five tumors analyzed (Fig. 3, C and D).

Fig. 3

Preexisting NRG1Hi subpopulation in treatment-naïve tumors. (A and B) Flow cytometric determination of the NRG1Hi subpopulation in naïve tumors from the Calu3 model (A) and LSL-K-rasG12D GEMM (B). One of five independent samples is shown. qRT-PCR of sorted cells shows relative expression of NRG1 in positive and negative populations. (C) Flow cytometric determination of the NRG1Hi subpopulation in a representative treatment-naïve patient tumor. (D) Percentages of EpCam- and NRG1-expressing cells in patient samples. (E) qRT-PCR analysis of tumor cell NRG1 mRNA levels in Calu3 tumors of various sizes and times after chemotherapy.

To determine whether NRG1 expression is dynamically regulated by tumor size or exposure to chemotherapy, we analyzed NRG1 expression levels in tumors of different volumes and at various times after chemotherapy by qRT-PCR (Fig. 3E). We did not see any increase in NRG1 mRNA levels after a single dose of chemotherapy (cisplatin + paclitaxel). With the exception of the residual tumors, NRG1 levels were equivalent at all of the times and volumes tested. These results demonstrate that NRG1 expression is not induced by chemotherapy or influenced by tumor size, consistent with the enrichment of a preexisting subpopulation of NRG1-expressing cells.

NRG1-HER signaling in NSCLC

We next interrogated the mediators of NRG1 signaling in NSCLC. We examined the expression of NRG1 and its receptors in several human NSCLC cell lines to identify models that coexpress both the ligand and its receptors. Although the expression levels of NRG1α and NRG1β transcripts were heterogeneous among cell lines, they were much higher in the majority of cell lines when compared to normal lung (Fig. 4A). Surprisingly, although NRG1 was enriched in H441 TRICs in vivo, cultured H441 cells did not express detectable NRG1 transcripts, highlighting the differences in the properties of cells grown in vitro and in vivo. Western blot analysis of the four HER receptors revealed heterogeneous expression among the six human NSCLC lines (Fig. 4B).

Fig. 4

Autocrine NRG1-HER signaling in NSCLC cell lines. (A) qRT-PCR analysis of NRG1α and NRG1β in human NSCLC cell lines and normal human lung. (B) Western blot analysis of HER receptors in human NSCLC cell lines. (C) Cell viability assay of cultured Calu3 cells upon treatment with YW538.24.71. (D and E) Western blots for p-HER proteins and downstream effectors in Calu3 cells (D) and H522 cells (E) treated with YW538.24.71. (F) Western blot for p-HER proteins and downstream effectors upon stimulation of H441 cells with exogenous NRG1. (G) Analysis of HER receptors in individual murine lung tumors by Western blot. WT, wild type. (H) Western blot for p-HER proteins and downstream effectors in LKPH1 cells treated with YW538.24.71.

On the basis of receptor expression patterns, we selected two cell lines with which to evaluate the possible downstream mediators of NRG1 autocrine signaling. Calu3 is HER2-amplified and had the highest in vitro expression level of all four HER receptors relative to other cell lines. Treatment of Calu3 cells with YW538.24.71 resulted in a dose-dependent reduction in cell viability and dose-dependent decreases in p-HER3, p-HER4, and p-AKT levels. There were no detectable changes in the levels of p-HER2 or p-MEK1/2 (phospho-mitogen-activated or extracellular signal–regulated protein kinase kinase 1/2) in these lysates (Fig. 4, C and D, and fig. S1A). H522 cells lacked detectable HER3 but showed abundant HER2 and HER4 expression. Treatment of H522 cells with YW538.24.71 resulted in decreased levels of p-HER4, p-HER2, p-MEK1/2, and p-S6 (Fig. 4E and fig. S1B). Thus, the downstream effectors of NRG1 signaling appear to vary between tumors, likely governed by differences in HER receptor expression.

Because H441 cells did not express any NRG1 transcript in vitro, we assessed the mediators of NRG1 signaling in these cells by stimulation with exogenous NRG1. Upon NRG1 stimulation of serum-starved cells, there was an increase in p-HER3 and p-AKT (Fig. 4F) but no detectable changes in the levels of p-STAT3 (signal transducer and activator of transcription 3) or p-MEK1/2 (fig. S1C).

NRG1 receptor expression and effector pathway usage were also evaluated in primary murine lung tumors and a murine lung cancer cell line. Analysis of individual microdissected tumors revealed expression of Her2 and overexpression of Her3 relative to total normal lung. Expression of EGFR was not detected (Fig. 4G). Effects of NRG1 inhibition were assessed in the murine LKPH1 cell line. Treatment of LKPH1 cells with anti-NRG1 resulted in decreased levels of p-HER3 and p-AKT (Fig. 4H). Changes in the levels of p-MEK1/2 and p-STAT3 were not detectable by Western blot. Together, these data suggest that Her3 and Her4 differentially activate downstream effectors in response to NRG1-mediated activation.

Effect of NRG1 inhibition on chemotherapy response in NSCLC xenograft models

Next, we sought to determine the effects of NRG1 inhibition on primary tumor growth and relapse after chemotherapy. In the Calu3 model, both YW538.24.71 and YW526.90.28 significantly inhibited tumor growth as single agents (Fig. 5A), with YW538.24.71 providing a more potent tumor growth inhibition (TGI) relative to vehicle-treated tumors (69% and 42% TGI, respectively). Furthermore, both antibodies enhanced the response to chemotherapy, resulting in 72% and 101% TGI over chemotherapy alone for YW526.90.28 and YW538.24.71, respectively. However, only YW538.24.71 provided a significant increase in time to progression over chemotherapy alone (Fig. 5A). The differences in efficacy seen with YW526.90.28 and YW538.24.71 demonstrate that it is necessary to inhibit both NRG1α and NRG1β isoforms.

Fig. 5

Effect of anti-NRG1 on chemotherapy response in vivo. (A) Tumor growth curves for mice with established Calu3 tumors that were administered YW526.90.28 or YW538.24.71 alone or in combination with chemotherapy. (B and C) Tumor growth curves for mice with established LKPH2 tumors treated with YW538.24.71 alone or in combination with chemotherapy. Chemotherapy consisted of carboplatin + paclitaxel (B) or gemcitabine (C). (D) Tumor growth curves for mice with established H596 tumors treated with YW538.24.71 alone or in combination with chemotherapy. Tumor growth curves are presented as linear mixed effects (LME) fit analysis of tumor volume graphed as cubic splines with autodetermined knots. Kaplan-Meier curve showing progression-free survival (progression defined as a tumor reaching twice its starting volume). P values were calculated by Gehan-Breslow-Wilcoxon test.

Chemotherapy treatment of Calu3 tumors resulted in substantial tumor regression, which is not typical of patient response. Therefore, we further evaluated the effects of single agent and combination anti-NRG1 treatment in two additional NSCLC models that undergo stasis in response to chemotherapy. As seen with the Calu3 model, YW538.24.71 treatment of LKPH2 tumors resulted in significant inhibition of tumor growth as a single agent, and the combination was superior to chemotherapy alone (Fig. 5B). Treatment with YW538.24.71 substantially reduced p-Her3 levels in tumors, demonstrating effective inhibition of receptor activation (fig. S2). Moreover, YW538.24.71 also enhanced the response of LKPH2 tumors to gemcitabine treatment, another chemotherapeutic agent commonly used to treat NSCLC (Fig. 5C). Although there was no single-agent activity of YW538.24.71 on H596 xenografts, the antibody significantly enhanced the magnitude and duration of the response of these tumors to chemotherapy (Fig. 5D).

HER receptor usage in NSCLC

To understand which HER receptors are required for NRG1 signaling in NSCLC, we evaluated the effects of HER3 and HER4 knockdown on tumor growth. We generated doxycycline-inducible shHER3 (Calu3-shHER3) and shHER4 (Calu3-shHER4) sublines. In the presence of doxycycline, HER3 and HER4 levels were decreased in Calu3-shHER3 and Calu3-shHER4, respectively (Fig. 6, A to C). The extent of p-AKT down-regulation observed in Calu3-shHER3 was much greater than that seen in Calu3-shHER4.

Fig. 6

Role of HER3 and HER4 in tumor growth in vivo. (A) qRT-PCR analysis of HER3 expression in cultured Calu3-shHER3 and Calu3-shLuc cells. (B) qRT-PCR analysis of HER4 expression in cultured Calu3-shHER4 and Calu3-shLuc cells. (C) Western blot analysis of HER3, HER4, and p-AKT proteins in cultured Calu3-shHER and Calu3-shHER4 cells. (D and E) Tumor growth curves for mice with established Calu3-shHER3 (n = 14 per group) (D) and Calu3-shHER4 tumors (n = 12 per group) (E) treated with sucrose or doxycycline (Dox). (F) Tumor growth curves for mice with established Calu3 tumors treated with anti-HER3 (n = 9 per group). (G) Viability of H522 cells upon siRNA-mediated knockdown of individual HER family receptors. NTC, nontargeting control siRNA; RLU, relative luciferase units. **P < 0.01 using unpaired t test. (H) Tumor growth curves for mice with established H522 tumors treated with YW526.90.28, YW538.24.71, or anti-HER4. Tumor growth curves are presented as LME fit analysis of tumor volume graphed as cubic splines with autodetermined knots.

To assess their roles in vivo, we performed studies using Calu3-shHER3 and Calu3-shHER4 xenograft models treated with either sucrose or doxycycline. There was substantial inhibition of Calu3-shHER3 tumor growth (Fig. 6D and fig. S2) in doxycycline-treated mice, similar to that seen with single-agent anti-NRG1 treatment (Fig. 5A). We did not see a notable inhibition of tumor growth upon doxycycline treatment in the Calu3-shHER4 in vivo study (Fig. 6E and fig. S3). The Calu3-shHER3 findings were confirmed in an independent study using anti-HER3 in this model (Fig. 6F). Thus, the in vitro receptor analysis (Fig. 4) and in vivo studies indicate that despite high HER4 levels in this model, it is dispensable for primary tumor growth.

In contrast, HER4 signaling appears to be a key driver of tumorigenesis and mediator of NRG1 signaling in the H522 model. We tested the requirement for each HER family member in H522 cell proliferation in vitro by small interfering RNA (siRNA)–mediated knockdown. Only knockdown of HER4 resulted in decreased cell proliferation (Fig. 6G), and treatment of H522 cells with YW538.24.71 reduced the levels of p-HER4 (Fig. 4D). Treatment of mice bearing H522 tumors with YW538.24.71, YW526.90.28, or anti-HER4 all resulted in tumor stasis, demonstrating that HER4-mediated NRG1 signaling is a key driver of tumor growth in this model (Fig. 6H). Thus, both HER3 and HER4 can act as drivers of primary tumor growth in NSCLC.

Effect of Her4 inhibition on chemotherapy response

Her4 and multiple Her4 ligands were enriched in TRICs in the K-ras–driven GEMM of NSCLC (Fig. 1A). Therefore, we used a ligand trap approach to sequester all Her4 ligands, thus preventing their binding to receptors in vivo. HER4-ECD (extracellular domain) will bind all Her3 and Her4 ligands with high affinity (24). We generated a fusion of the human HER4-ECD fused to murine immunoglobulin G2A (IgG2A) Fc. Her4 protein could not be detected by Western blot in murine NSCLC cell lines, but the addition of HER4-ECD to serum-starved murine NSCLC cells in vitro inhibited ligand-induced HER3 signaling as demonstrated by diminished p-Her3 levels (Fig. 7A).

Fig. 7

Effect of HER3/4 inhibition on chemotherapy response in vivo. (A) Western blot analysis of p-HER3 levels in HER4-ECD–treated GEMM tumor cell lines LKP2, LKPH1, and LKPH2. (B) Pre- and posttreatment μCT scans showing tumor burden in representative mice from each treatment group. The posttreatment scan was taken at the 4-week time point. (C) Tumor growth curves for LSL-K-rasG12D;p53Fl/Fl mice treated with vehicle + control IgG, vehicle + HER4-ECD, cisplatin + control IgG, or cisplatin + HER4-ECD. Graph represents the average percent change from starting tumor burden ± SEM. (D) Tumor growth curves for mice with established H1299 tumors treated with YW538.24.71, anti-HER3, anti-HER4, or anti-HER3 + anti-HER4 alone or in combination with gemcitabine. Tumor growth curves are presented as LME fit analysis of tumor volume graphed as cubic splines with autodetermined knots.

In the LSL-K-rasG12D;p53Fl/Fl mouse model, cisplatin treatment provides a survival benefit but does not result in tumor regression (18). We used the HER4-ECD in this model to determine the role of Her3/4 in primary tumor growth and relapse. Lung tumor–bearing mice were imaged by x-ray micro–computed tomography (μCT) at the start of the study (day 0) and segregated into four treatment groups of equal average tumor burdens. Only the cisplatin + HER4-ECD combination resulted in a significant inhibition of tumor growth relative to the vehicle control at all time points (Fig. 7, B and C). Furthermore, there was a significant reduction in tumor burden in the cisplatin + HER4-ECD–treated mice relative to cisplatin + control IgG2A–treated mice at the 4-week time point. In contrast, there was no effect of HER4-ECD treatment alone on tumor growth, further supporting a unique role for ligand-mediated Her3/4 signaling in chemoresistance and/or tumor regrowth.

Finally, we used blocking antibodies to evaluate the contributions of NRG1, HER3, and HER4 to primary tumor growth and relapse in the H1299 xenograft model, which showed robust enrichment of HER4 in TRICs (Fig. 1B). Treatment with anti-NRG1, anti-HER3, anti-HER4, or anti-HER3 + anti-HER4 showed modest to no inhibition of primary tumor growth. In contrast, when given in combination with gemcitabine, anti-HER4 was unique in its ability to completely prevent tumor relapse after the cessation of chemotherapy, resulting in a 94% TGI relative to gemcitabine alone (Fig. 7D). Thus, this study demonstrates that HER4 signaling alone is sufficient to confer resistance to conventional chemotherapy.


Here, we sought to address mechanisms of innate chemoresistance and drivers of cancer relapse. This topic is of particular importance in NSCLC because the majority of patients will receive chemotherapy as their first line of treatment (25). However, not all patients respond to chemotherapy, and most that do quickly relapse. We reasoned that by studying the small number of residual tumor cells that are left behind after chemotherapy, we could gain new insights into disease mechanisms that could ultimately have a major impact on patient survival. Here, we have shown that expression of the HER4 receptor and several of its ligands, most notably NRG1, is highly enriched in TRICs. Moreover, inhibition of ligand-mediated HER3/4 signaling with either NRG1-blocking antibodies, HER4 ligand trap, or HER4-blocking antibodies enhances the response to standard-of-care chemotherapies in several NSCLC models, delaying the time to relapse. In addition, we find that NRG1-mediated HER4 signaling can be a potent driver of primary tumor growth in certain HER4-overexpressing lung cancers.

We have shown that inhibition of ligand-mediated HER3/4 signaling delays relapse in several NSCLC models with varying responses to chemotherapy, including both GEMM and human xenograft models. Although we observed robust effects when inhibiting HER3/4 signaling in conjunction with chemotherapy in all the models studied, the effects on primary tumor growth were variable; in some cases, HER3/4 inhibition had no effect on primary tumor growth. These findings suggest that there are differences in the key drivers of tumor cell proliferation and/or survival in the naïve state versus when tumors are challenged with chemotherapy. Because neoadjuvant is not an established treatment paradigm for NSCLC, we were unable to obtain a sufficient number of patient samples to determine whether enrichment of HER3/4 ligands and receptors in residual tumor cells also occurs in patients undergoing chemotherapy treatment. This remains an important topic for future assessment.

Members of the HER family of RTKs are believed to play an important role in lung cancer. Amplification, overexpression, and mutations in EGFR and HER2 occur frequently in NSCLC (11, 2631). EGFR is a proven therapeutic target in NSCLC, and EGFR kinase–inhibiting drugs were first introduced for the treatment of NSCLC in 2003 (32). Furthermore, a recent study analyzing somatic mutations in a panel of human lung adenocarcinomas identified numerous HER4 mutations, suggesting a role for HER4 in lung cancer (14), but the biological consequences of these mutations were not determined. Here, we show that HER4-mediated NRG1 signaling drives primary tumor growth in the H522 lung adenocarcinoma model. Moreover, we show that ligand-dependent HER4 signaling is required for disease relapse after chemotherapy in the H1299 xenograft model. Thus, we demonstrate an oncogenic role of HER4 in NSCLC, validating it as a therapeutic target.

NRG1 autocrine signaling has been reported in cell lines from lung (33, 34), breast (3537), ovarian (38), and head and neck squamous carcinoma (39). However, the role of NRG1 autocrine signaling in NSCLC has not been well characterized. It has largely been investigated in the context of insensitivity to EGFR inhibitors, where recent studies show that it may contribute to therapeutic resistance (12). We show that NRG1- and other ligand-mediated HER3/4 signaling contributes to resistance to conventional chemotherapies. This may be mediated in part through activation of the phosphatidylinositol 3-kinase/AKT pathway, which has been shown to underlie resistance to HER pathway inhibitors (40) and to promote cisplatin resistance in NSCLC and other cancer cell lines (41, 42). In the GEM model, single-agent HER4-ECD treatment has no effect on tumor growth but enhances the durability of the response to cisplatin. Although the ligand trap approach can also sequester some EGFR ligands, EGFR is not expressed in these tumors, and it was previously shown that EGFR tyrosine kinase inhibitors do not enhance cisplatin response in the model (43). Therefore, this effect can only be due to inhibition of Her3/4 signaling. The HER3 antibody used in our studies is a fully human antibody, and the HER4 antibody has not been tested for cross-reactivity to mouse Her4, precluding their use in the GEM model. Therefore, we were unable to determine the precise contributions of Her3 and Her4 to chemoresistance in this model. Likewise, we show that NRG1-mediated HER3/4 signaling contributes to chemoresistance in the Calu3, H596, and LKPH2 models, but the eight-arm studies required to determine the precise roles of HER3 and HER4 in these models were beyond the scope of the current study.

Our findings suggest that antibodies blocking HER4 signaling may be effective in enhancing response to chemotherapy and delaying disease relapse after chemotherapy. Although none of the models used in our studies carried EGFR mutations, we did observe HER3/4-mediated resistance in both K-ras mutant (H441, LKPH2, and LSL-K-rasG12D;p53Fl/Fl) and K-ras wild-type tumors (Calu3 and H596), suggesting that inhibition of this signaling axis in combination with conventional chemotherapy could be a therapeutic strategy for a broad NSCLC patient population for whom no targeted therapy currently exists. In addition, a recent study showed that NRG1 is sufficient to confer resistance to the anaplastic lymphoma kinase (ALK) inhibitor TAE684 in NCI-H3122 cells, harboring an ALK translocation, and to the MET inhibitor crizotinib in MET-amplified gastric cancer cell lines (44). Hence, NRG1- and/or HER3/4-blocking antibodies might also be an effective combination therapy in ALK- and MET-driven NSCLCs.

Materials and Methods

Cell lines

NSCLC cell lines were obtained from the American Type Culture Collection. Calu3 and H441 cell lines were transduced with TZV–β-actin–eGFP lentivirus. After multiple passages, GFPHi cells were sorted and expanded, and these sublines were described as Calu3-GFP and H441-GFP. Mouse NSCLC cell lines LKPH1 and LKPH2 were derived from two independent tumors from a KrasLSL-G12D/+;p53FL/+;Z/EG lung tumor–bearing mouse. The LKP2 cell line was derived from a tumor from a KrasLSL-G12D/+;p53FL/FL;Z/EG lung tumor–bearing mouse.

Patient specimens

Freshly resected NSCLC tumor samples were obtained from Bio-options Inc. Specimens were shipped overnight on wet ice. Specimens were collected in compliance with all applicable federal and state laws and regulations and ethical guidelines.

Generation of anti-NRG1 antibodies

See Supplementary Materials.

Kinase receptor activation assay

Serum-starved MCF-7 cells were treated with fixed amounts of NRG1α or NRG1β and increasing amounts of test materials (YW538.24.71 and YW526.90.28) for 15 min in a CO2 incubator. After decanting the medium, cells were lysed and analyzed in an ELISA previously described (45).

Inducible short hairpin RNA lentivirus

Hairpin oligonucleotides used in this study are as follows: shHER4, 5′-GATCCCCGATCACAACTGCTGCTTAATTCAAGAGATTAAGCAGCAGTTGTGATCTTTTTTGGAAA-3′ (sense) and 5′-AGCTTTTCCAAAAAAGATCACAACTGCTGCTTAATCTCTTGAATTAAGCAGCAGTTGTGATCGGG-3′ (antisense); shHER3, 5′-GATCCCCAAGAGGATGTCAACGGTTATTCAAGAGATAACCGTTGACATCCTCTTTTTTTTGGAAA-3′ (sense) and 5′-AGCTTTTCCAAAAAAAAGAGGATGTCAACGGTTATCTCTTGAATAACCGTTGACATCCTCTTGGG-3′ (antisense). The complementary double-stranded short hairpin RNA (shRNA) oligonucleotides were inserted into our Tet-inducible viral gene transfer vector as described (46). The luciferase shRNA construct was previously described (46).

In vitro studies

Anti-NRG1 studies were performed as follows: LKPH1 cells were incubated in 0.1% fetal bovine serum (FBS) ± antibodies for 48 hours. Overnight serum-starved H522 or Calu3 cells were incubated with conditioned medium containing 1% FBS ± antibodies for 20 min. LKP2, LKPH1, and LKPH2 cells in 0.1% serum were treated with HER4-ECD (1 mg/ml) for 48 hours. Addition of NRG1 to H441 cells was performed as follows: Cells were serum-starved for 18 hours. Recombinant human NRG1β1 (1 μM) (R&D Systems) or anti-ragweed IgG2A (1 μM) was added for 10 min. After the indicated times, cells were washed and lysed with radioimmunoprecipitation assay buffer containing Halt protease and phosphatase inhibitor cocktails (Thermo Scientific).

In vivo xenograft tumor studies

Tumor cells (10 to 20 million) were transplanted into the right flank of athymic nude mice. When tumor size reached ~200 mm3, the mice were divided into treatment groups. The chemotherapy dosing regimen was paclitaxel [20 mg/kg, intravenously (iv), every 2 days for five doses] and cisplatin [5 mg/kg, intraperitoneally (ip), every week for two doses for the Calu3 model and every 2 weeks for two doses for the H441 model]. Regressed tumors and time-matched vehicle controls were collected ≥1 week after the last dose of chemotherapy. Tumors were dissociated and GFP+ tumor cells were collected by FACS. For anti-NRG1 efficacy studies, the antibodies were dosed at 20 to 25 mg/kg ip, every week, for the duration of the study. For the H596 and LKPH2 studies, the chemotherapy regimen was carboplatin (60 mg/kg ip) and paclitaxel (20 mg/kg iv, every 4 days for five doses). LKPH2 gemcitabine study was dosed 100 mg/kg ip, every 4 days for four doses. For the H1299 study, the treatment regimen was as follows: YW538.24.71 was administered at 20 mg/kg, anti-HER4 at 25 mg/kg, and anti-HER3 at 50 mg/kg ip, once a week, for the duration of the study; gemcitabine was administered 100 mg/kg ip, every 4 days for five doses. All animal experiments were approved by the Institutional Animal Care and Use Committee at Genentech Inc.

Xenograft tumor growth analysis

A mixed-modeling approach was used to analyze tumor volumes (47). Cubic regression splines were used to fit a nonlinear profile to the time courses of log2 tumor volume for each treatment group. Kaplan-Meier analysis was used to determine progression-free survival. Progression was defined as a doubling in the initial tumor volume. Log-rank analysis was used for statistical analysis to compare treatment groups.

In vivo LSL-K-rasG12D;p53Fl/Fl Her4-ECD study

LSL-K-rasG12D;p53Fl/Fl mice were infected with Adeno-Cre. Baseline CT scans were performed at 16 weeks after tumor induction (day 0 of study), and mice were grouped such that the average starting tumor volume per group was equal. Mice were dosed with phosphate-buffered saline or cisplatin (7 mg/kg ip, every week for three doses) and biweekly with control IgG2A or HER4-ECD-Fc (25 mg/kg ip) for the duration of the study. Longitudinal CT scans were performed at days 14, 28, and 42.

Supplementary Materials

Materials and Methods

Fig. S1. Effector pathways mediating NRG1 autocrine signaling.

Fig. S2. Phospho-Her3 levels in YW538.24.71-treated LKPH2 tumors in vivo.

Fig. S3. Effect of doxycycline treatment on Calu3-shLuc tumor growth.

References and Notes

  1. Acknowledgments: We thank L. Johnson, A. Lima, P. Hamilton, R. Molina, and J. Long for providing tumor-bearing K-rasLSL-G12D;p53Fl/Fl mice; Z. Jiang for bioinformatics support; C. K. Poon, J. Borneo, L. Gilmour, and J. Cupp for FACS consultation and expert technical assistance; M. Huey for IHC support; and I. Mellman for helpful discussions. Funding: Genentech Inc. Author contributions: E.L.J., G.V.H., and C.C.d.l.C. designed the project, conducted in vivo and in vitro experiments, and wrote the manuscript. Y.Z. and E.A.S.-C. generated TRICs from the GEMM. Y.W. oversaw and N.A., C.C., S.S., and Z.J. generated anti-NRG1 antibodies, and C.C. conducted in vitro binding assays. L.R. and S.-P.T. conducted and analyzed KIRA assays. C.C.d.l.C. and N.A. conducted NRG1 FACS experiments. L.S. conducted H522 in vitro experiments and G.S. designed and analyzed H522 experiments. L.C., S.R., M.M., J.T., N.A., and D.G. conducted, designed, and/or analyzed in vivo studies. S.K.W. and R.A.D.C. analyzed μCT data. Competing interests: All authors (except Y.Z. and E.A.S.-C.) are paid employees of Genentech Inc. Genentech has filed a patent application related to this work on using anti-NRG1 antibodies to treat cancer. Data and materials availability: Data and materials are available from Genentech Inc. under a materials transfer agreement.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article