HER-3 Expression and MEK Activation in Non-small Cell Lung Cancer

analysed tumour samples from 45 patients with HER-3 and HER-2 expression was using well


Abstract
Background HER-3 has a role in pathogenesis and development of resistance to targeted therapies in non-small cell lung cancer (NSCLC).

Methods
We analysed tumour samples from 45 patients with adenocarcinoma of the lung. HER-3 and HER-2 expression was identi ed using immunohistochemistry as well as bioinformatic interrogation of The Cancer Genome Atlas (TGCA).

Results
HER-3 was highly expressed in 42.2% of cases and observed more frequently than HER-2 overexpression. ERBB3 copy number alone did not account for HER-3 overexpression. Bioinformatic analysis of TCGA demonstrated that MEK activity score (a score of downstream pathway activity and surrogate of functional HER-3 signalling) did not correlate with HER-3 ligands. EGFR pathway signalling was predominant with a signi cantly positive correlation of EGFR and AREG expression with MEK activity score. In the ERBB3 expressed group, ERBB3 RNA expression levels were signi cantly correlated with MEK activity score. ERBB3 expression level remained highly positively correlated with MEK activity after adjusting for EGFR expression.

Conclusion
HER-3 expression is common and it is not associated with gene ampli cation nor is it ligand dependent.
HER-3 is an alternative pathway to EGFR activating MEK. HER-3 is a potential therapeutic target in NSCLC by virtue of frequent overexpression and functional downstream signalling.
Background Non-small cell lung cancer Lung cancer is the leading cause of cancer mortality worldwide. There is a poor 5 year survival rate in patients with regional and distant disease in non -small cell lung cancer (NSCLC), of 33% and 6% respectively (1). Systemic treatment with chemotherapy and immunotherapy is standard of care, while a small subset of cancers demonstrate oncogene-addiction with effective targeted treatments. These include EGFR mutated, ALK rearranged and newer targets such as ROS, RET and NTRK. HER-3 (receptor tyrosine-protein kinase erbB-3) is a member of the epidermal growth factor receptor (EGFR) family of receptors. Investigation of HER-3 as a therapeutic target has not been previously prioritised due to its impaired kinase activity (2). HER-3 acts as a heterodimeric partner for other members of the EGFR family, namely EGFR/HER-1 and ERB2/HER-2/Neu (3) (4) (5) (6). It has also been demonstrated that HER-3 can form heterodimers with MET, AXL and IGF1R (7) (8) (9). HER-3 possesses weak tyrosine kinase activity and so preferentially forms heterodimers with kinase-pro cient receptor tyrosine kinases in the presence of its ligands neuregulin (NRG 1 and NRG 2) (2). EGFR has a number of ligands, including amphiregulin, epidermal growth factor and transforming growth factor, while HER-2 has no known ligand and remains in an 'open,' active conformation (10) (11) . However, HER-2/HER-3 dimerisation can occur independently of ligand binding (12). More lately it has been appreciated that HER-3 can also homodimerise on binding to the heregulin ligands, NRG1 and NRG2, so having intrinsic, non-EGFR or HER-2 mediated activity (13) (14). Downstream signalling of HER-3 is mainly via the PI3K/AKT pathway, with MAPK/ERK and JAK/STAT pathways also activated (15)  Evidence for the role of HER-3 in lung cancer pathogenesis is largely drawn from retrospective data sets correlating HER-3 expression with poor prognosis and metastatic development (25) (26) (27) .
Accumulating evidence supports the role of HER-3 in the development of resistance to EGFR targeted therapies (28) (29) (30) (31). From preclinical data, dual targeting of EGFR and HER-3 is capable of overcoming acquired resistance to EGFR inhibition (29) (31). MET ampli cation or overexpression is a mechanism of resistance to EGFR-TKI occurring in 5% of cases (32). HER-3 phosphorylation is observed in EGFR mutant tumours with MET ampli cation, indicating that MET activates PI3K/AKT pathway via HER-3 phosphorylation (33). MEK inhibition causes a transcriptional upregulation of both HER-2 and HER-3 and the formation of heterodimeric complexes in KRAS mutant non-small cell lung cancer (NSCLC) and colon cancer and combined MEK and dual EGFR-HER-2 inhibition has a synergistic effect (30).
Mutations of HER-3 have been described in the extra-cellular domain or tyrosine kinase domain as a rare event in lung cancer and can in uence the activity of HER-3 receptor as well as the dimerization a nity with HER-2 and EGFR. (34) (35) (36).
Here, we explore HER-3 expression levels in adenocarcinoma of the lung (adeno-NSCLC) patients and correlate it with clinical parameters. We also interrogate large genomic datasets to identify the potential mechanism of HER-3 expression and propose that HER-3 could be an alternate pathway for MEK pathway activation.

Patients and tissue samples
The Drug Development Unit at the Royal Marsden (RMH) recruits patients with advanced solid malignancies to early phase clinical trials. Patient with adeno-NSCLC attending for consultation in our phase I clinic provided written informed consent and were enrolled into a tissue for research consent.
Approval was gained from the local research and ethics committee for the use of the clinical data. Archival formalin xed para n embedded tumour blocks were analysed for HER-2 and HER-3 protein expression and genomic analysis for HER-3 copy number was performed. Patient clinical data were retrospectively collected from the RMH electronic patient record system. Immunohistochemistry and scoring system Tissue blocks were freshly sectioned and only considered for immunohistochemistry (IHC) analyses of HER-2 and HER-3 if adequate material was present. Protein expression on 3µm thick FFPE sections was assessed by a Pathologist blinded to the clinical data.
HER-3 and HER-2 antibody validation by siRNA and immunohistochemical assays were performed as previously described.
For HER-2, membranous expression was scored according to College of American Pathologists (CAP) HER-2 scoring guidelines (37). No staining or membrane staining that is incomplete and is faint/barely perceptible and in ≤10% of tumour cells (Score 0); incomplete membrane staining that is faint/ barely perceptible and in >10% of tumour cells (Score +1); weak to moderate complete membrane staining observed in >10% of tumour cells (Score 2+) and circumferential membrane staining that is complete, intense and in >10% of tumour cells (Score +3). For HER-3 staining, the same guidelines were used to generate an ad-hoc HER-2-like score for membranous staining. Additionally, HER-2 and HER-3 cytoplasmic staining was visually assessed, percentage of cells staining weakly, moderately or strongly were noted and H score calculated.

Digital PCR
Digital polymerase chain reaction (PCR) was performed on a QX100 droplet digital PCR system (Bio-Rad). Copy Number Assays were performed with a FAM-ERBB3 validated Bio-Rad assay, reference ID: dHsaCP2500395 (67bp amplicon) and a recommended reference copy number Hex-assay for AP3B1, assay ID: dHsaCP2500348 (60bp amplicon). Digital PCR analysis was performed using QuantaSoft v1.3.2.0 software from Bio-Rad to assess the number of positive droplets. At least two negative control wells with no DNA were included in every run and each assay was run at least in duplicate (see Supplementary Methods).

Statistical methods
Statistical analysis of IHC and clinical data was performed on Graphpad Prism v7.0d. The association between HER-3 expression and other binary data was analysed using two sided Fishers exact test. For survival Kaplan-Meir curves were constructed and Gehan-Breslow-Wilcoxon test compared populations. Survival was taken from diagnosis to death and those alive at follow-up were censored.

Bioinformatic methods
Genomic data and gene expression (RSEM) for 507 adeno-NSCLC was downloaded from cBioportal for analysis (38). MEK activity score is an accumulation measurement of MEK pathway activity based on a previously validated 18-gene transcript signature of MEK signalling (39). The associations between gene expression levels (and with MEK signalling) were assessed by linear regression. All statistical test was performed using R (v1.1.43).

Clinical Characteristics
From December 2010 to December 2018 45 patients with adeno-NSCLC were referred to the Drug Development Unit, consented to tissue analysis and had available archival tumour specimens. The majority of samples were from biopsies (41), the rest being from surgical resection (4). There were relatively fewer males (19) than females (26) in the cohort. The median age at diagnosis was 58 (range 38-72) and the median overall survival was 38 months. The median number of lines of treatment for advanced disease was 3 (range 0-7): 43 patients received chemotherapy (median 2, range 1-4), 22 patients received EGFR tyrosine kinase inhibitor (TKI) (median 1; range 1-3) and 9 patients received immunotherapy. All patients who had an EGFR mutation received an EGFR TKI and 25 patients were enrolled in a phase I trial and received at least 1 phase 1 trial drug.

Correlation of HER-3 protein expression by IHC with clinical characteristics
The median overall survival from diagnosis for HER-3 positive patients was 40.5 months, and for HER-3 negative patients 30.8 months. There was no statistical difference in survival between HER-3 positive and HER-3 negative cases (Gehan-Breslow-Wilcoxon; p = 0.4596) (Fig. 1B). There was no difference in median survival between HER-3 positive and negative cases in those patients with EGFR mutations (median OS 42.6 months HER-3 positive, median OS 43.9 months HER-3 negative; Gehan-Breslow-Wilcoxon; p = 0.4927).

HER-3 gene ampli cation
Digital PCR of ERBB3 was performed in 27 cases. There was low frequency of small copy number gain in HER-3 with 11% (3/27) having copy number variant (CNV) of 3-4 (Fig. 1C). Six out of the seven samples with this highest CNV scored 3 + on HER-3 immunohistochemistry.
EGFR and KRAS mutations and correlation with HER-3 As standard of care EGFR mutational analysis had been performed in all 45 patients and KRAS mutational analysis in 28 patients. EGFR mutation was present in 17.8% (8 cases). HER-3 protein overexpression was present in 8.9% (4 cases) of EGFR mutation and in 14.2% (4 cases) of KRAS mutation. There was no correlation between HER-3 overexpression and EGFR positivity (Fishers exact test p = 0.7043) or HER-3 and KRAS positivity (Fishers exact test; p = 0.6908) (Fig. 2).
Baseline genomic data in TCGA Interrogation of the PanCancer TCGA cancer genome atlas of 507 patients with adeno-NSCLC showed a genetic alteration rate of 31% in the ERBB family with the most common genetic alteration seen in EGFR and ERBB4, which was evident in 16% (79 cases) and 10% (50 cases). In ERBB2 and ERBB3, genetic alterations were less frequent; seen in 4% (19 cases) and 5% (25 cases) respectively (Fig. 3A). Among the different types of genetic alteration, ampli cation was the most common alteration in ERBB3, observed in 48% out of all alterations (12 cases). ERBB3 copy number change demonstrated a positive association with ERBB3 RNA expression (r = 0.26; p = 8 × 10 − 6 ) (Fig. 3B); ampli cation cases had 1.7 times higher expression of ERBB3 RNA than diploid cases (p = 0.004).
ERBB3 ampli cation co-occurred with EGFR ampli cation; the ampli cation of ERBB3 was four times more likely to co-exist with EGFR ampli cation than not, though this did not reach statistical signi cance (p = 0.1). There was no co-ampli cation between ERBB3 and ERBB2, but the expression of ERBB3 was highly correlated with ERBB2 expression levels (r = 0.09 and p = 2.5 × 10 − 23 ) ( Figure S1). ERBB3 expression also positively correlates with ERBB4 expression (r = 0.1 and p = 0.001) but there was no correlation with the expression levels of EGFR or (r = -0.12 and p = 0.5).

MEK activity score correlation with key nodes in ERBB pathway in TCGA data
We performed further analyses of RNA sequencing data from TCGA. We observed ERBB2 and ERBB3 RNA expression levels were higher than EGFR and ERBB4 in adeno-NSCLC (Fig. 3C). We measured absolute gene expression level of ERBB family genes and derived MEK pathway activity using a validated 18-gene MEK signature (39). EGFR RNA expression levels correlated with MEK activity (r = 0.16 and p = 3.0 × 10 − 5 ), ERBB3 RNA expression did not correlate with MEK activity (r = 0.09 and p = 0.78) (Fig. 3D). We further characterised MEK activation score with gene expression levels of ERBB ligand and proteins of ubiquitination (Table 3). AREG (EGFR ligand) and RNF41 (an ubiquitin ligase) had a signi cant positive correlation with MEK activity score. In order to stratify ERBB3 RNA expression effects on MEK activity, we grouped patients into ERBB3 "expression high" (higher 50% of ERBB3 RNA expression cases) and ERBB3 "expression low" (lower 50% of ERBB3 RNA expression cases). In the ERBB3 expression high group, ERBB3 RNA expression levels were signi cant and positively correlated with MEK activity score (r = 0.19, p = 0.006), whereas no correlation was observed in the ERBB3 expression low group (p = 0.12) (Fig. 4). The expression of EGFR did not correlate with MEK activity score in the ERBB3 expression high group (p = 0.11), though did correlate with MEK activity score in the ERBB3 expression low group (r = 0.2, p = 4 × 10 − 5 ). In the ERBB3 expression high group, ERBB3 expression levels still highly positively correlated with MEK activity even after adjusting for EGFR expression effects (p = 0.007). and 86.1% (resected stage 0 and 1A1 adeno-NSCLC) (44). IHC methods and scoring have not been previously standardised. Our data uses a validated and robust scoring system based on optical density and H-score. Although our patient cohort is small, we demonstrate that overexpression of HER-3 protein is common in advanced adeno-NSCLC, occurring in over 40% of patients.

HER-3 protein levels in adeno-NSCLC and mechanism of overexpression
The rate of ERBB3 gene ampli cation in NSCLC is low. The Cancer Genome Atlas Research Group sequenced 230 cases of lung adenocarcinoma and did not identify any ERBB3 ampli cation (45). The cBioportal and GENIE data sets report ampli cation in just over 1% frequency (38). A smaller series of 71 cases reported a 1% ampli cation rate in lung adenocarcinoma (34). In a series of 82 NSCLC patients HER-3 ampli cation was identi ed in 26.8% of patients who were treated with the EGFR TKI ge tinib. ERBB3 ampli cation was more common in those with ERBB2 gain (p = 0.055) and EGFR gain (p = 0.06) but not with EGFR mutations (p = 0.17) (46). It was not possible to perform ERBB3 PCR on all cases in our cohort as tumour tissue was limited, but a sample of 27 cases had su cient tumour sample to pass quality control. There appeared to be clustering of HER-3 overexpressing cases in those with ERBB3 copy number gain but this does not account for the majority of HER-3 overexpression.
Bioinformatic analysis of large, publicly available datasets allows testing of hypotheses. We interrogated the PanCancer TCGA data set of adeno-NSCLC samples. This dataset includes transcriptome (RNA) and genomic (DNA) data. The genomic data supports our nding of low levels of ERBB3 ampli cation, which was the most common cause of ERBB3 genetic alteration. It also provided the con rmation that ERBB3 copy number alteration highly associates with its gene expression.
The MEK activity score is an 18 gene signature derived from an experiment to test a mixed-tumour response to the MEK inhibitor, selumetinib. The MEK activity score is a surrogate marker of MEK pathway activation. It is not speci c for activation through the HER-3 channel as the MEK pathway is also initiated upon HER-2, EGFR and other receptor activation. It also does not take into account HER-3 downstream activation via the PIK3A/AKT/mTOR pathway. However, as HER-3 activity is predominantly via HER-2/HER-3 and EGFR/HER-3 hetero-dimerisation, we inferred that the MEK activity score could be used as a surrogate of, but not speci c to, HER-3 protein expression. If a key node in the HER-3 pathway is found to correlate with MEK activity score it can be concluded that, in that cohort, the node may play a key part in pathway. The lack of correlation with ERBB3 ligands (NRG 1 and 2) expression is an important negative nding. RNF41 also had a signi cant positive correlation with MEK activity score. This is di cult to interpret given the fact that an increase in a degradation protein would be expected to correlate with a lower target protein level.
In conclusion, HER-3 protein is common in adeno-NSCLC, ampli cation levels of ERBB3 are low and on TCGA analysis the mechanism of HER-3 protein overexpression does not appear to be ERBB3 ligand dependent. Further investigation is required to elucidate the mechanism of HER-3 protein overexpression in adeno-NSCLC.

HER-3 expression as a prognostic marker
Yi et al 26 demonstrated that HER-3 overexpression is associated with a poor prognosis in a cohort of 443 patients with advanced NSCLC (27). The median overall survival in our study was not signi cantly different between the HER-3 negative and positive patient group. This phase 1 patient cohort may not be truly re ective of the broader population of NSCLC patients, as these patients tend to be tter with fewer or no co-morbidities.
HER-3 in relation to oncogenic drivers in adeno-NSCLC HER-2 protein overexpression is reported in 6-30% of patients with lung cancer, which in keeping with our results (47) (48) (49). HER-2 expression is less frequently observed in NSCLC than HER-3 expression.
There was some overlap, though no statistically signi cant association between, HER-3 expression with HER-2 expression or KRAS and EGFR mutations. Previous data have demonstrated that HER-3 plays a role in EGFR TKI resistance. Half of the cases in this cohort with EGFR mutation also co-expressed HER-3. No difference in overall survival was seen between HER-3 expressing and non-expressing groups in the EGFR-mutant cases in this cohort though patient numbers were small.
We therefore conclude that HER-3 represents a common target in NSCLC independent of KRAS and EGFR mutations. We have a simple tool in the form of IHC that can be used to "pre-screen" and stratify patients to receive such HER-3 targeting compounds. Furthermore, there are some patients in which HER-3 expression occurs concurrently with driver mutations and the role of HER-3 in these groups, including in drug resistance, requires further investigation. The relationship between HER-3 expression and other drivers in NSCLC (such as ROS, ALK, NTRK and RET) is yet to be determined.

HER-3 and functional activity in adeno-NSCLC
In the TCGA dataset MEK activity score was associated with EGFR and AREG (EGFR ligand) gene expression. This suggests that signalling via EGFR is key to the oncogenic behaviour in lung cancer. EGFR has been previously reported as a key driver of oncogenesis in lung (and other) cancers (50).
Although our initial analysis did not show correlation of ERBB3 with MEK signature, we strati ed ERBB3 expression into two groups; "expression low" and "expression high"; and analysed this along with EGFR gene expression. In the ERBB3 expression high group, ERBB3 RNA expression levels positively correlated with MEK activity score but not in the expression low group. The expression of EGFR correlated with MEK activity score in the ERBB3 expression low group. This supports the hypothesis that HER-3 is an alternative pathway for MEK activation in the HER-3 expression high group. This analysis implies a degree of functional downstream pathway signalling via HER-3. In effect, this suggests that when EGFR drives the cancer, the effect of HER-3 is minimal as EGFR drives the downstream signalling. However, as HER-3 expression increases this becomes the main driver of pathway activation. This nding has relevance in terms of proposing HER-3 as a viable drug target in adeno-NSCLC. As a membrane target HER-3 can be potentially targeted with drugs such as HER-3 target antibody drug conjugates in which expression on the surface is key to activity. However, a degree of functional MEK pathway activation suggests that in some patients drugs that rely on abrogation of signalling (such as HER-3/ HER-2 dimerisation drugs) can also be attempted. The challenge in this patient cohort is to predict patients in which this approach will be effective, our data relies on gene expression but transcriptome analysis is complex in a clinical trial setting.

Conclusion
In conclusion, HER-3 protein is commonly overexpressed in adeno-NSCLC though gene ampli cation levels are low. In the search for non-genomic causes of HER-3 protein overexpression, the MEK activity score was used as a surrogate of HER-3 overexpression; however no correlation with ligands of HER-3 were seen. Further investigation is therefore needed to continue to elucidate the mechanism of HER-3 overexpression. In adeno-NSCLC, signalling via the EGFR pathway is predominant but in the HER-3 expression high group HER-3 is an alternative pathway, with downstream signalling via the MEK pathway.
Since HER-3 overexpression is more common than previously thought and appears to be functional in non-EGFR driven adeno-NSCLC, its utility as a therapeutic target in adeno-NSCLC should be a drug development priority.