Phase Ib Study of Lumretuzumab Plus Cetuximab or Erlotinib in Solid Tumor Patients and Evaluation of HER3 and Heregulin as Potential Biomarkers of Clinical Activity

HER3 is a key dimerization partner of HER family members that activates several oncogenic signal transduction pathways, particularly the PI3K/Akt pathway (1). Recent studies indicate that HER3 pathway activation is important in the development of resistance to EGFR- and HER2-targeting treatments (2–6). The role of HER3 as a prognostic marker remains controversial. Increased HER3 protein expression as measured by immunohistochemistry (IHC) has been described as an adverse prognostic factor in many solid tumor types, including breast, gastric, lung, ovarian, and colon cancers (7–11), and targeting HER3 can sensitize refractory tumor models to EGFR inhibitors (12). In contrast, higher expression levels of HER3 mRNA were associated with increased progression-free survival (PFS) in HER2-positive metastatic breast cancer treated with pertuzumab plus trastuzumab plus docetaxel (13) and in patients with platinum-resistant ovarian cancer treated with gemcitabine plus pertuzumab (14). However, adding pertuzumab to chemotherapy in platinum-resistant ovarian cancer patients selected on low HER3 mRNA expression did not increase PFS significantly (15). Autocrine loops involving the HER3 ligand HRG and leading to HER3 activation have been described in squamous cell carcinoma of the head and neck (SCCHN), non–small cell lung cancer (NSCLC), and ovarian cancer (8, 16–20).

Lumretuzumab is a humanized, glycoengineered immunoglobulin G1 antibody that selectively binds with high affinity to the extracellular domain of HER3. Prevention of HRG binding to HER3 by lumretuzumab resulted in almost complete inhibition of HER3 heterodimerization and phosphorylation as well as inhibition of tumor growth in cell line–based mouse xenograft models (21). In a phase I study, the safety of lumretuzumab was evaluated in patients with advanced solid tumors. No DLTs were observed at doses of 100 to 2,000 mg every 2 weeks (as a flat dose) and an MTD was not reached. Downregulation of HER3 membranous protein was observed in on-treatment tumor biopsies from 200 mg and higher doses, and a target-independent pharmacokinetic profile was observed at ≥400 mg doses (22).

Multiple preclinical models have shown that combinations of anti-HER3 antibodies with anti-EGFR therapies lead to enhanced antitumor activity as compared with the single agents, and even complete tumor regression has been described in different studies (12, 21, 23, 24). Preclinical studies have also shown that HER3 pathway activation and overexpression of HRG predicted response to HER3-targeting therapy (25, 26), and preliminary clinical data from early studies with anti-HER3 antibodies are consistent with these reports (22, 27, 28). We therefore hypothesized that (i) the combination of lumretuzumab with EGFR inhibitors could improve clinical activity versus EGFR inhibitors alone, and (ii) higher tumor expression levels of HRG mRNA could indicate improved clinical outcome and could ultimately serve as a predictive biomarker for HER3-targeting therapy.

The current study evaluated the safety and clinical activity, and potential biomarkers of lumretuzumab treatment when combined with the EGFR-targeting agents cetuximab or erlotinib. To evaluate the clinical activity of lumretuzumab plus erlotinib in a biomarker-enriched population, we recruited an extension cohort of sqNSCLC patients, as sqNSCLC has been shown to express higher HRG mRNA levels compared with non-sqNSCLC (data on file; refs. 29, 30).

This was a multicenter, phase Ib, open-label, nonrandomized, dose escalation and extension study (ClinicalTrials.gov Identifier: NCT01482377) investigating the safety (including the MTD and/or optimal biological dose), pharmacokinetics, pharmacodynamics, and clinical activity of lumretuzumab in combination with cetuximab or erlotinib in patients with metastatic or advanced HER3-positive carcinomas. The study consisted of a dose escalation phase following a standard “3 + 3” study design and an extension phase conducted for the cetuximab plus lumretuzumab as well as the erlotinib plus lumretuzumab combination.

For lumretuzumab monotherapy, pharmacokinetics was linear at dosages ≥400 mg, indicative of target-mediated drug disposition saturation, and clinical pharmacodynamic activity was demonstrated (22). Therefore, the dose of 400 mg was used as the starting dose for the combination treatments with cetuximab and erlotinib. Two administration schedules were applied: every 2 weeks and every 3 weeks; the latter only applied for a subset of patients in the lumretuzumab plus erlotinib part.

Standard doses of cetuximab (400 mg/m² for the first infusion followed by 250 mg/m² weekly infusions) and erlotinib (daily dose of 150 mg orally) were used.

Patients continued treatment until disease progression, unacceptable toxicity, or withdrawal of consent.

Local ethics committee approval was obtained, and all patients provided written informed consent. The study was conducted in accordance with Good Clinical Practice guidelines and the Declaration of Helsinki in nine centers in Spain, the Netherlands, Denmark, and South Korea.

Patients had to have a histologically confirmed diagnosis of an advanced or metastatic HER3-expressing carcinoma that was refractory to standard treatment or for which no standard therapy existed. Eligible patients were ≥18 years of age, had an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 2 and had adequate hematology, blood chemistry, and renal and liver function. Patients eligible for enrollment underwent a fresh (pretreatment) tumor biopsy that was used to assess the level of HER3 protein expression by IHC using central pathology review. Discernible HER3 membrane staining in any neoplastic cell, provided that a minimum of 100 tumor cells were present in the biopsy specimen, was considered diagnostically positive for HER3 protein expression.

For the extension phase of the cetuximab combination part, only patients with advanced HER3-expressing SCCHN, NSCLC, and colorectal cancer (proven to be KRAS wild-type and expressing EGFR as per local assessment by the investigator) were recruited. For the extension phase of the erlotinib combination part, only patients with advanced HER3-expressing sqNSCLC and patients with nonsquamous lung carcinomas with a documented NRG1 somatic gene fusion (determined by local PCR and sequencing assays) were recruited to enrich for patients more likely to benefit from lumretuzumab. Higher expression of HRG mRNA in sqNSCLC as compared with non-sqNSCLC had been demonstrated through mining of data extracted from TCGA and analysis of HRG by qRT-PCR in formalin-fixed, paraffin-embedded (FFPE) tumor specimens from the Roche clinical tumor bank and from preclinical HER3 xenograft response modeling studies. In addition, higher HRG levels positively correlated with response to lumretuzumab treatment in preclinical models (data not shown). Recently, gene fusions were identified involving the gene encoding for HRG (NRG1; ref. 31) activating HER3 in an autocrine and paracrine manner as a novel mechanism for HER pathway dysregulation in lung adenocarcinomas (32). Tumors harboring these NRG1 fusions may be particularly susceptible to HER3-targeting therapy. The protocol was therefore amended to allow inclusion of these rare patients to explore signals of clinical activity. Because of the differences in the underlying biology, it was decided to analyze the patients with NRG1-rearranged tumors separately and not include them into the analysis of the sqNSCLC cohort.

Responding patients were tested for EGFR mutations with Qiagen therascreen EGFR Pyro assay (Source BioScience).

Patients received premedication 30 minutes prior to the start of the first lumretuzumab infusion consisting of paracetamol (500–1,000 mg orally) and diphenhydramine (25–50 mg orally or inravenously, or an alternative antihistamine). Corticosteroids were allowed in case of ≥ grade 2 infusion-related reactions (IRR). Lumretuzumab was administered as an intravenous infusion. Cetuximab and erlotinib were administered according to standard posology, and dose reductions were allowed according to the label. Study drugs could be delayed to assess or treat adverse events (AE) for up to 14 days. One reduction to the previously assessed lower dose level was allowed for lumretuzumab.

Tumor response assessment using RECIST version 1.1 (33) was conducted at screening and every 8 or 9 weeks thereafter for the every-2-week or every-3-week schedule, respectively.

Safety assessments included physical (ECOG performance status, vital signs) and laboratory examinations and electrocardiogram. AEs were defined according to the Common Terminology Criteria for Adverse Events, version 4.0 (CTC-AE v4.0).

A dose-limiting toxicity (DLT) was defined as an AE occurring during the first two cycles of treatment (i.e., 28 days) with lumretuzumab that was considered to be study drug related and was either: grade 4 neutropenia (i.e., absolute neutrophil count <0.5 × 10⁹ cells/L for a minimal duration of 7 days); grade 3/4 febrile neutropenia; grade 4 thrombocytopenia; grade 3 thrombocytopenia associated with bleeding episodes; or grade ≥ 3 nonhematologic toxicity. IRRs and grade 3 skin and/or epithelial toxicities associated with EGFR inhibition were not considered DLTs for the cetuximab combination. For the erlotinib combination, IRRs, grade 3 skin, and/or epithelial toxicities associated with EGFR inhibition and grade 3 diarrhea, which reduced to less than grade 2 within 14 days by supportive treatment or dose reduction of erlotinib or interruption of erlotinib, were not considered DLTs.

Fresh tumor biopsies were collected during screening and on day 14 of cycle 1. HER3 protein expression was assessed using an IHC assay and scored semiquantitatively and reported as an immunoreactive score (IRS, range 0–300) as described previously (22).

HRG mRNA expression was measured by a qRT-PCR assay in FFPE sections obtained from fresh tumor biopsies collected from all patients at screening prior to initiation of treatment. In addition, HER3 mRNA expression was determined using a separate qRT-PCR assay. Both assays were performed at Roche Molecular Systems.

Total RNA was isolated from FFPE tumor tissue sections using the Cobas RNA Isolation Kit. TaqMan probes were designed to detect HRG or HER3 mRNA and respective reference genes simultaneously. All reagents were prepared at Roche Molecular Systems, and qRT-PCR was performed using the Cobas 4800 system. Calculation of the cycle-to-threshold (C_t) for each fluorescent channel was done using LC480 software, and the relative HRG or HER3 mRNA log expression was reported as ΔC_t, where ΔC_t = C_t(reference) − C_t(target). Where feasible, biopsies with less than 50% tumor content underwent macrodissection guided by pathologist annotation of adjacent H&E-stained sections. No reference ranges were defined a priori for HRG or HER3 mRNA expression using the initial research-grade assay.

During the course of the clinical study, the HRG mRNA assay was transitioned to a prototype diagnostic assay, which differed in that reagents were prepared in a GMP facility, the assay run using a z480 PCR system, and calculation of C_t values was performed using validated diagnostic software (Heregulin Expression Test Analysis Package for use with Cobas 4800 System Release 2.1). Predose HRG mRNA expression data derived using both the research and prototype diagnostic assays were available for 37 of 120 patients. Overall, the research assay reported higher expression than the diagnostic prototype due to differences in the data analysis parameter settings, albeit there was strong correlation between the two assays (mean difference 1.53 ΔC_tR² = 0.96).

The prototype diagnostic assay was used to define a reference range for HRG expression using 150 primary FFPE tumor biopsy samples from sqNSCLC patients (obtained from a Roche tumor bank and an external vendor). Median HRG log expression corresponded to a ΔC_t of −1.92 with a 75th percentile of −0.92. The median was used as a threshold above which tumor biopsies derived from patients were defined as HRG-high.

All patients who received at least one dose of study medication were included in the safety population. Descriptive statistics were used for demographics and safety.

To explore biomarker levels and correlations, a pharmacodynamic population was also defined, comprising of all patients receiving at least one dose of study medication and providing a pharmacodynamic sample at baseline. Because of the distinct histology, invasive mucinous NSCLC, and potentially distinct biology, lung cancer patients with NRG1 gene alterations were included in the sqNSCLC cohort but were analyzed separately. For some patients, biomarker data were not available, either due to lack of evaluable tumor tissue or for technical reasons. Therefore, for each analysis, the number of evaluable patients (n) is specified.

Exploratory statistical analyses were used to evaluate differences in biomarker levels for patients with different tumor histologies (nonsquamous vs. squamous carcinoma) or lung cancer diagnosis and changes from baseline levels using the Wilcoxon rank sum test or Wilcoxon signed rank test as appropriate.

The Spearman rank correlation coefficient (ρ) was used to determine the association between biomarker levels.

The information on patients' histology (squamous and adenocarcinoma) was available for the vast majority of the patients or could be inferred through the tumor type (e.g., patients with colorectal cancer have adenocarcinoma): Only 4 patients had to be excluded in the analyses of biomarker differences between adeno and squamous carcinoma patients as the histology was either unknown, both histologies were present, or the histology was of transitional cell type, although all available data were utilized to assess biomarker correlations.

For efficacy, RECIST response was evaluated and summarized by study part (cetuximab/erlotinib combinations), and the objective response from the sqNSCLC patients of the extension cohort was related to biomarker levels via a multinomial logit model. All available patients were included in the analyses, as all dose levels were considered pharmacodynamically active.

Patient demographics and baseline characteristics are presented in Table 1. In the dose escalation phase of the cetuximab part, 27 patients were enrolled into five dose cohorts, that is, 400 (n = 5), 800 (n = 6), 1,200 (n = 5), 1,600 (n = 5), and 2,000 mg (n = 6) of lumretuzumab every 2 weeks. In the extension phase of the cetuximab part, 22 patients received 2,000 mg. In the dose escalation phase of the erlotinib part, 37 patients were enrolled into four every-2-week dose cohorts, that is, 400 (n = 6), 800 (n = 6), 1,600 (n = 7), and 2,000 mg (n = 8) of lumretuzumab, and 10 patients were enrolled into two every-3-week dose cohorts, that is, 800 (n = 7) and 1,600 mg (n = 3). In the extension phase of the erlotinib part, 34 patients received 800 mg of lumretuzumab every 2 weeks (Fig. 1). Two of those patients had NSCLC adenocarcinoma with an NRG1 fusion, and 32 were sqNSCLC patients.

Overall, 102 patients (85.0%) discontinued the study due to progressive disease or death, 6 patients (5.0%) were withdrawn due to an AE, 6 patients (5.0%) refused further treatment, 1 patient (0.8%) withdrew consent, and 5 patients (4.2%) discontinued for other reasons.

Three patients with DLTs were reported, one in the cetuximab part (dehydration grade 3 in the 800-mg cohort) and two in the erlotinib part (diarrhea grade 3 and hypokalemia grade 3 in one patient of the 1,600 mg cohort and blood bilirubin increased grade 3 in one patient of the 2,000 mg cohort). The MTD was not reached up to the highest dose tested (i.e., 2,000 mg) in both treatment parts. A total of 1,452 AEs of any grade were reported in 120 patients (100%; Table 2). Most AEs (87.3%) were grade 1 or 2 in intensity. A total of 184 grade ≥3 AEs were reported in 88 patients (73.3%). No patients in the cetuximab part and 3 patients (4.2%) in the erlotinib part died from an AE (pneumonia, cerebrovascular accident, and hemoptysis with one patient each, all considered unrelated to lumretuzumab treatment). The most frequent AEs included diarrhea [38 patients (77.6%) in the cetuximab part and 54 patients (76.1%) in the erlotinib part] and rash [26 patients (53.1%) in the cetuximab part and 32 patients (45.1%) in the erlotinib part]. The most frequent ≥grade 3 AEs included diarrhea [6 patients (12.2%) in the cetuximab part and 11 patients (15.5%) in the erlotinib part], hypomagnesemia [5 patients (10.2%) in the cetuximab part and no patients in the erlotinib part], and dermatitis acneiform [5 patients (10.2%) in the cetuximab part and 4 patients (5.6%) in the erlotinib part]. Diarrhea was the only AE for which a potential dose dependency was observed during dose escalation. In the cetuximab part, 1 of 6 patients had grade 3 diarrhea only at the top dose of 2,000 mg lumretuzumab. In the erlotinib part, 1 of 6 patients at 800 mg, 3 of 7 patients at 1,600 mg, and 2 of 8 patients at 2,000 mg lumretuzumab had grade 3 diarrhea.

The severity of diarrhea was increased at 1,600 and 2,000 mg, particularly in combination with erlotinib. Although these diarrhea events did not qualify for DLTs as per protocol, a lower lumretuzumab dose of 800 mg every 2 weeks was selected as the recommended phase II dose for the combination with erlotinib, whereas a dose of 2,000 mg every 2 weeks was selected for the combination with cetuximab based on the observed safety profile.

Similarly as described for monotherapy with lumretuzumab (22), a linear pharmacokinetic profile was observed with doses of lumretuzumab of ≥400 mg every 2 weeks, indicating target-mediated drug disposition saturation (data not shown).

Tumors of all enrolled patients were HER3 positive based on IHC analysis. In line with previously published data (22), membranous HER3 was downregulated in tumor and skin biopsy samples collected on day 14 of cycle 1 compared with baseline (Supplementary Tables S1 and S2).

In baseline tumor samples, HER3 protein and HER3 mRNA were generally more highly expressed in adenocarcinoma compared with those of squamous cell histology [HER3 IHC: median IRS 2.00 (n = 61) and 0.07 (n = 50), respectively, P < 0.0001; HER3 mRNA: median ΔC_t 1.88 (n = 61) and −0.27 (n = 49), respectively, P < 0.0001, Supplementary Fig. S1]. Furthermore, HER3 protein expression was significantly correlated with HER3 mRNA expression (ρ = 0.61, P < 0.001, n = 112). Similarly, for the subset of patients with NSCLC, HER3 protein was more highly expressed in the nonsquamous compared with the squamous subtype [HER3 IHC: median IRS 2.25 (n = 14) and 0.06 (n = 36), respectively, P < 0.0001, Supplementary Fig. S2].

Conversely, analysis of baseline tumor biopsy samples using the research HRG qRT-PCR assay showed an overall higher expression of HRG mRNA in tumors of squamous cell histology compared with those of adenocarcinoma histology (P < 0.0001; Supplementary Fig. S3). Similarly, within NSCLC samples, median HRG mRNA log expression (ΔC_t) was higher in sqNSCLC as compared with non-sqNSCLC (P < 0.0001, Supplementary Fig. S4).

In addition, HRG mRNA expression across tumor histologies was inversely correlated to HER3 protein (ρ = −0.24, P = 0.024, n = 89) and HER3 mRNA (ρ = −0.24, P = 0.022, n = 91).

In the cetuximab part, the ORR was 6.1% and the disease control rate [DCR, i.e., the percentage of patients with a best response of stable disease (SD), partial response (PR), or complete response (CR)] was 40.8% (Table 3).

Two patients with colorectal cancer showed a PR and one patient with SCCHN had a CR (Supplementary Fig. S5). The SCCHN patient had not been treated with an EGFR inhibitor previously. The duration of response in this patient was 16.9 months. The patient's tumor expressed relatively high levels of HRG mRNA (ΔC_t: 0.09, median ΔC_t in the cetuximab part: −2.44, n = 44) and EGFR protein (IRS: 3.0) and relatively low levels of HER3 mRNA (ΔC_t: −1.06, median ΔC_t in the cetuximab part: 1.43, n = 46) and HER3 protein (IRS: 0.0004) as determined on fresh baseline FFPE tumor biopsies. Overall, there was no detectable relationship between any of the exploratory biomarkers and clinical response defined by RECIST (data not shown).

In the erlotinib part, the ORR was 4.2% and the DCR was 43.7% (Table 3).

Three patients (4.2%) had a PR, one with ovarian cancer (Supplementary Fig. S6) and two with sqNSCLC (Fig. 2). The ovarian cancer patient had a duration of response of 19.8 months. This tumor showed high levels of HRG mRNA (ΔC_t: −0.36, median ΔC_t in the erlotinib part: −1.4, n = 47) and HER3 mRNA (ΔC_t: 1.49, median ΔC_t in the erlotinib part: 0.28, n = 68) and HER3 protein expression (IRS: 2.00) as determined on baseline biopsies. All partial responders in this cohort were EGFR inhibitor naïve, and neither of the two sqNSCLC patients nor the ovarian cancer patient harbored an EGFR-activating mutation.

Overall, there was no detectable relationship between any of the analyzed exploratory biomarkers and clinical response according to RECIST (data not shown).

In both the cetuximab and the erlotinib part, phosphorylated HER3 and Akt were expressed at low levels in predose FFPE biopsy samples; expressions of both were reduced in on-treatment tumor biopsy samples in the cetuximab and erlotinib arm, reflecting the mechanism of action of lumretuzumab (Supplementary Table S2). There was no significant downregulation of total membranous EGFR nor a decrease in proliferative activity based on Ki67 staining, which is in line with findings from others (34). Other exploratory biomarkers were unchanged following treatment with lumretuzumab.

The 2 patients with a SLC3A2-NRG1 fusion gene were heavily pretreated with five therapy lines for metastatic disease, including prior erlotinib treatment. Both patients had SD as their best response.

Altogether, 32 patients with sqNSCLC were recruited into the extension cohort of the erlotinib part. HRG mRNA expression data were available for all patients, and RECIST response data were available for 29 patients (90.6%). Overall, the ORR was 6.3%, and the DCR was 56.3% (Fig. 2).

Twelve patients were considered HRG-high using the median ΔC_t as the cutoff, determined from the analysis of an external cohort of 150 FFPE tumor samples from sqNSCLC patients. Among the 12 HRG-high patients, one had a PR as best response with a duration of 3.8 months. Hence, the ORR in the HRG-high population was 8.3%, while the DCR was 75.0% (9/12 patients) and the median PFS (range) was 3.7 (1.0–8.0) months, compared with an ORR of 5.0% (1/20 patients, with a response duration of 4.9 months), a DCR of 55.0% (11/20 patients), and a median PFS (range) of 1.8 (0.2–6.5) months for the HRG-low patients (Fig. 3). In the same cohort, low HER3 mRNA expression (using the median value of the cohort as a threshold to define high and low expression, corresponding to a ΔC_t of −0.3) was associated with an ORR of 12.5% (2/16 patients) and a DCR of 43.8% (7/16 patients) and median PFS (range) of 1.8 (0.2–6.5) months as compared with an ORR of 0%, a DCR of 68.8% (11/16 patients), and a median PFS (range) of 2.9 (0.3–8) months for patients with high HER3 mRNA.

HRG expression was also confirmed on FFPE tumor biopsies using in situ hybridization and IHC assays. Relative HRG expression was consistent across assays, and no differences were observed in the clinical response relationship (data not shown). In this cohort, there was no evidence that higher HRG or lower HER3 mRNA levels were correlated with improved response to lumretuzumab and erlotinib treatment.

This study describes the safety, pharmacokinetics, pharmacodynamics, and clinical activity of the combination of lumretuzumab, an anti-HER3 mAb, with cetuximab and erlotinib, respectively. The study aimed to detect signals of clinical activity (i) of anti-EGFR/HER3 combination therapy per se, and (ii) activity of this combination in a histologic entity enriched for higher HRG expression levels, that is, sqNSCLC.

The most common AEs in this study were gastrointestinal and skin toxicities, similar to the well-known side effects of cetuximab and erlotinib. The addition of lumretuzumab may have caused an increase in the incidence of diarrhea compared with the incidence for cetuximab monotherapy [77.6% (any grade) and 12.2% (grade ≥3) compared with 12.7% (any grade) and 1.2% (grade ≥3) for cetuximab monotherapy (35)] and erlotinib monotherapy [76.1% (any grade) and 15.5% (grade ≥3) compared with 55% (any grade) and 6% (grade ≥3) for erlotinib monotherapy (36)]. The severity of diarrhea seemed increased at the highest lumretuzumab doses tested, that is, at 1,600 and 2,000 mg. A similar diarrhea side effect profile as in the current study was seen for combination treatments of other HER3-targeting antibodies with cetuximab (37, 38) and erlotinib (38–40) in early clinical studies. Effects might be caused by a disinhibition of HER-modulated chloride channels in colonic epithelial cells leading to secretory diarrhea (41, 42).

Although not reaching a defined MTD for both combinations, diarrhea was the only AE for which a potential dose dependency was observed. On the basis of the pharmacokinetic profile and safety assessments, a dose of 2,000 mg in combination with cetuximab and a dose of 800 mg in combination with erlotinib were determined as the recommended phase II doses.

Overall, the efficacy seen in the cetuximab (ORR = 6.1%, DCR = 42.8%) and erlotinib (ORR = 4.2%, DCR = 43.6%) combination parts does not exceed what has previously been reported for cetuximab monotherapy (43, 44) or erlotinib monotherapy (45). Previous clinical studies on combination treatments of HER3-targeting antibodies in combination with erlotinib in NSCLC have demonstrated comparable response rates (40, 46, 47).

Subsequently we attempted to assess the clinical activity of lumretuzumab plus erlotinib in a patient population that was more likely to respond to HER3-targeting therapy. Preclinical studies (25, 26) as well as early clinical studies (22, 27, 28, 48, 49) reported an association between higher HRG mRNA expression levels and susceptibility to HER3-targeting therapies. Modest but statistically significant prolonged median PFS was reported for HRG-high–expressing patients, that is, a 0.2-month increase in a study reported by Sequist and colleagues (40) and a 1.6-month increase in a study reported by von Pawel and colleagues (46).

In the current study, we observed higher HRG mRNA expression levels in squamous carcinomas as compared with adenocarcinomas, which is in line with the described analysis from internal tumor bank samples and TCGA data. We therefore chose to enrich for higher HRG mRNA expression in the lumretuzumab plus erlotinib extension cohort by selectively enrolling sqNSCLC patients. However, lumretuzumab plus erlotinib did not demonstrate higher response rates, longer duration of responses, or more pronounced disease stabilization than what would have been expected with erlotinib alone, even when taking HRG and HER3 mRNA expression levels into account. Slightly higher DCR and PFS were observed in the HRG- and HER3-high–expressing sqNSCLC patients, suggesting some biological evidence for the hypothesis of HRG and HER3 expression levels and HER3 blockade. However, the short-lived responses and disease stabilization suggest that these tumors are not dependent on HER3 signaling and that alternative signaling pathways may be involved as escape mechanisms. In accordance to our findings, a phase II/III study in NSCLC patients treated with the HER3-targeting mAb patritumab and erlotinib was stopped recently due to a lack of efficacy (ORR of 2.2% for patritumab plus erlotinib compared with 6.3% for erlotinib alone and median PFS of 1.9 months for patritumab plus erlotinib compared with 2.7 months for erlotinib alone in patients considered to have high HRG levels; ref. 47). Similarly, no correlation was found between HER3 expression and response or clinical benefit in several phase I and II studies testing HER3-tageting therapies (27, 46, 50, 51). Recently, a dual-action antibody targeting EGFR and HER3 also failed to improve PFS and overall survival compared with cetuximab therapy in SCCHN patients in a randomized phase II trial, even in the HRG-high–expressing subpopulation (52).

In conclusion, combination treatment of lumretuzumab with cetuximab or erlotinib was manageable but demonstrated modest clinical activity. Apart from two exceptional responders, a patient with SCCHN achieving a prolonged CR and a patient with ovarian cancer achieving a prolonged PR, both showing higher expression levels of HRG mRNA, the current study was not able to identify a robust biomarker signal that could serve as response prediction biomarker for lumretuzumab in the future. Furthermore, in the presented sqNSCLC cohort, even higher HRG or HER3 mRNA expression levels were not associated with increased ORR or clinically meaningful durations of response, refuting our initial hypothesis. Thus, the response and resistance mechanisms to lumretuzumab-based therapy may be multifaceted or simply undetectable in our cohort of sqNSCLC. Together with the results from randomized trials in SCCHN (52) and NSCLC (47), however, our data provide further evidence that adding HER3 to EGFR-targeting therapy is not sufficient to derive clinically meaningful benefit.

W. Jacob, S. Vega-Harring, C. Adessi, and M. Weisser ownership interest (including patents) in Roche. A. Cervantes reports receiving speakers bureau honoraria from Roche. M. Lolkema reports receiving commercial research grants from Astellas, Johnson & Johnson, and Sanofi. E. Felip reports receiving speakers bureau honoraria from AstraZeneca, Novartis, and Pfizer and is a consultant/advisory board member for Boehringer Ingelheim, Eli Lilly, MSD, Pfizer, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Meulendijks, W. Jacob, E.E. Voest, J.H.M. Schellens, M. Thomas, G. Meneses-Lorente, I. James, M. Nguyen, F. Michielin, B. Bossenmaier, M. Weisser, U.N. Lassen

Development of methodology: W. Jacob, M.P. Lolkema, J.H.M. Schellens, M. Thomas, M. Ceppi, G. Meneses-Lorente, M. Weisser, U.N. Lassen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Meulendijks, E.E. Voest, M. Mau-Sorensen, M. Martinez-Garcia, A. Taus, T. Fleitas, A. Cervantes, M.P. Lolkema, M.H.G. Langenberg, M.J. De Jonge, S. Sleijfer, J.-Y. Han, A. Calles, E. Felip, S.-W. Kim, J.H.M. Schellens, I. James, R. Dua, L. Steiner, U.N. Lassen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Jacob, E.E. Voest, M.P. Lolkema, M.J. De Jonge, A. Calles, J.H.M. Schellens, S. Wilson, M. Thomas, M. Ceppi, G. Meneses-Lorente, I. James, S. Vega-Harring, M. Nguyen, C. Adessi, F. Michielin, M. Weisser, U.N. Lassen

Writing, review, and/or revision of the manuscript: D. Meulendijks, W. Jacob, E.E. Voest, M. Mau-Sorensen, M. Martinez-Garcia, A. Taus, T. Fleitas, A. Cervantes, M.H.G. Langenberg, M.J. De Jonge, S. Sleijfer, J.-Y. Han, A. Calles, E. Felip, S.-W. Kim, J.H.M. Schellens, M. Thomas, M. Ceppi, G. Meneses-Lorente, I. James, S. Vega-Harring, C. Adessi, F. Michielin, B. Bossenmaier, M. Weisser, U.N. Lassen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Meulendijks, W. Jacob, F. Michielin, U.N. Lassen

Study supervision: D. Meulendijks, W. Jacob, M.J. De Jonge, M. Weisser, U.N. Lassen

Other (responsible for clinical development): M. Weisser

The authors thank the patients and their families for their participation in this study, and the staff at the study sites.

This study was funded by F. Hoffmann–La Roche Ltd.

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