Revisiting Epidermal Growth Factor Receptor (EGFR) Amplification as a Target for Anti-EGFR Therapy: Analysis of Cell-Free Circulating Tumor DNA in Patients With Advanced Malignancies

Purpose To date, evidence for tissue epidermal growth factor receptor (EGFR) overexpression as a biomarker for anti-EGFR therapies has been weak. We investigated the genomic landscape of EGFR amplification in blood-derived cell-free tumor DNA (cfDNA) across diverse cancers and the role of anti-EGFR therapies in achieving response. Methods We assessed EGFR amplification status among 28,584 patients with malignancies evaluated by clinical-grade next-generation sequencing (NGS) of blood-derived cfDNA (54- to 73-gene panel). Furthermore, we curated the clinical characteristics of 1,434 patients at the University of California San Diego who had cfDNA testing by this NGS test. Results Overall, EGFR amplification was detected in cfDNA from 8.5% of patients (2,423 of 28,584), most commonly in colorectal (16.3% [458 of 2,807]), non–small-cell lung (9.0% [1,096 of 12,197]), and genitourinary cancers (8.1% [170 of 2,104]). Most patients had genomic coalterations (96.9% [95 of 98]), frequently involving genes affecting other tyrosine kinases (72.4% [71 of 98]), mitogen-activated protein kinase cascades (56.1% [55 of 98]), cell-cycle–associated signals (52.0% [51 of 98]), and the phosphoinositide 3-kinase pathway (35.7% [35 of 98]). EGFR amplification emerged in serial cfDNA after various anticancer therapies (n = 6), including checkpoint inhibitors (n = 4), suggesting a possible role for these amplifications in acquired resistance. Nine evaluable patients with EGFR amplification were treated with anti-EGFR–based regimens; five (55.6%) achieved partial responses, including three patients whose tissue NGS lacked EGFR amplification. Conclusion EGFR amplification was detected in cfDNA among 8.5% of 28,584 diverse cancers. Most patients had coexisting alterations. Responses were observed in five of nine patients who received EGFR inhibitors. Incorporating EGFR inhibitors into the treatment regimens of patients harboring EGFR amplification in cfDNA merits additional study.


INTRODUCTION
Epidermal growth factor receptor (EGFR), also known as human epidermal growth factor receptor 1 (HER1) or ErbB1, is a receptor tyrosine kinase that belongs to the ErbB family proteins. Along with EGFR, the ErbB family includes HER2 (ErbB2), HER3 (ErbB3), and HER4 (ErbB4). When receptor-specific ligands bind to the extracellular domain of the EGFR, it forms a homodimer (EGFR-EGFR) or heterodimer (eg, EGFR-HER2, EGFR-HER3) that leads to the activation of receptors through ATP-dependent Purpose To date, evidence for tissue epidermal growth factor receptor (EGFR) overexpression as a biomarker for anti-EGFR therapies has been weak. We investigated the genomic landscape of EGFR amplification in blood-derived cell-free tumor DNA (cfDNA) across diverse cancers and the role of anti-EGFR therapies in achieving response. . EGFR amplification emerged in serial cfDNA after various anticancer therapies (n = 6), including checkpoint inhibitors (n = 4), suggesting a possible role for these amplifications in acquired resistance. Nine evaluable patients with EGFR amplification were treated with anti-EGFR-based regimens; five (55.6%) achieved partial responses, including three patients whose tissue NGS lacked EGFR amplification.

Methods We assessed EGFR amplification status among 28,584 patients with malignancies evaluated by clinical-grade next-generation sequencing (NGS) of blood-derived
Conclusion EGFR amplification was detected in cfDNA among 8.5% of 28,584 diverse cancers. Most patients had coexisting alterations. Responses were observed in five of nine patients who received EGFR inhibitors. Incorporating EGFR inhibitors into the treatment regimens of patients harboring EGFR amplification in cfDNA merits additional study.
phosphorylation of tyrosine residues in the EGFR intracellular domain. Activation of EGFR leads to multiple downstream signals, including mitogen-activated protein kinase and phosphoinositide 3-kinase pathways, which enhances cell proliferation and survival. 1,2 Functional activation of EGFR via mutation or amplification/overexpression has been identified in many tumor types, including lung, head and neck, gastroesophageal, and colorectal cancers, and has been associated with proliferation, invasion, and metastasis. 3,4 Alterations in EGFR have also been linked to primary resistance and accelerated tumor growth (designated as hyperprogression) from immune checkpoint inhibitors. [5][6][7] Because of its critical role in tumor aggressiveness, EGFR has been an attractive target for anticancer therapy. 1 To date, there are various anti-EGFR therapies that are US Food and Drug Administration approved, including erlotinib, gefitinib, afatinib, and osimertinib for non-small-cell lung cancer (NSCLC) with specific activating EGFR mutations, 8 cetuximab and panitumumab for colorectal cancer without KRAS or NRAS mutations, 9 cetuximab for head and neck cancer, 10 and necitumumab for squamous cell carcinoma of lung. 11 Biomarkers to predict response to anti-EGFR therapies have been studied extensively. EGFR and KRAS mutation status are widely used in lung and colorectal cancer, respectively. 8,9,12,13 In contrast, EGFR amplification and overexpression in tissue have not been well established as reliable biomarkers for anti-EGFR agents, (selected studies that investigated EGFR amplification/overexpression as a predictive marker for anti-EGFR therapies are summarized in the Data Supplement) 11,[14][15][16][17][18][19] . Overall, a meta-analysis concluded that tissue EGFR amplification status could not be demonstrated to be a consistent biomarker to predict the outcome from anti-EGFR therapies in colorectal cancer. 20 Although it is somewhat surprising that tissue EGFR amplification/expression status has not been established as a reliable biomarker for anti-EGFR therapies, potential reasons include heterogeneity between primary and metastatic lesions, dynamic changes in genomic alterations that may emerge along with therapeutic pressure or progression, presence of genomic coalterations associated with resistance, and potential differences in response to copy number gain due to aneuploidy versus focal EGFR amplification. [21][22][23] Use of plasma-derived cell-free tumor DNA (cfDNA) to assess EGFR status by next-generation sequencing (NGS) could conceivably overcome some of these limitations by detecting tumor-specific alterations that are shed into the bloodstream from multiple metastatic sites as well as the primary cancer. [23][24][25][26][27][28][29] Herein, we examined the genomic landscape of EGFR amplification by interrogating bloodderived cfDNA from 28,584 patients with diverse malignancies using clinical-grade NGS. Furthermore, we investigated the clinical characteristics, concordance between tissue NGS and cfDNA, and therapeutic outcome after anti-EGFR therapies among a subset of 1,434 clinically annotated patients at the University of California, San Diego (UCSD), Moores Cancer Center.

Patients
The genomic landscape of EGFR amplification among 28,584 diverse solid cancers that were referred to Guardant Health from March 2014 to February 2017, were evaluated. Furthermore, we have curated the clinical characteristics of 1,434 evaluable patients with diverse cancers at UCSD who had cfDNA testing at Guardant Health starting in March 2014. All investigations followed the guidelines of the UCSD Institutional Review Board for data collection (Profile Related Evidence Determining Individualized Cancer Therapy; ClinicalTrials.gov identifier: NCT02478931) and for any investigational therapies for which the patients consented (Data Supplement).

NGS for cfDNA and Tissue
All cfDNA analyses were performed at Guardant Health as previously described (Data Supplement). 26 Tissue NGS was performed at Foundation Medicine, as previously described 30 (Data Supplement).

End Points and Statistical Methods
Patient characteristics, prevalence of EGFR amplification, and genomic coalterations were summarized by descriptive statistics. Fisher's exact test was used for categorical variables.
Concordances between cfDNA and tissue DNA were described by percentage of concordance and κ value with standard error. Antitumor response was evaluated using Response Evaluation Criteria in Solid Tumors (RECIST) 1.1. Progression-free survival was defined as the time from treatment initiation to disease progression. Patients who had not experienced disease progression at the time of last follow-up were censored at that time point. Statistical analysis was performed with the assistance of author R.O.

Prevalence of EGFR Amplification in cfDNA Testing in Diverse Cancers
Among 28,584 patients with diverse solid malignancies whose cfDNA was evaluated at a central laboratory, 8 Fig 1A).

Prevalence of EGFR Amplification in cfDNA Testing in Patients With Diverse Cancers From UCSD Cohort
Among the UCSD cohort of 1,434 patients (Data Supplement), overall, 6.8% of patients (98 of 1,434) had EGFR amplification, including 86 patients detected at their first cfDNA evaluation ascopubs.org/journal/po JCO™ Precision Oncology 3       Fig 1B).

Genomic Coalterations Associated With EGFR Amplification (analysis of cfDNA)
Among 98 patients with EGFR amplification at UCSD, the median number of characterized genomic alterations was 5.0 (range, 0 to 17; excluding the EGFR amplification), and the median number of alterations among patients without EGFR amplification (n = 1,336) was significantly less (median, 1.0; range, 0 to 20; P < .001). The most common coalterations associated with EGFR amplification were in the following genes:  Fig A2).

Potential Targeted Therapies for Coalterations Associated With EGFR Amplification
In the UCSD cohort of 98 patients positive for EGFR amplification, 96.9% (95 of 98) of patient tumors had at least one characterized coalteration. All these 95 malignancies harbored at least one characterized coalteration potentially targetable with US Food and Drug Administrationapproved agents as on-or off-label use.

Clinical Characteristics of Patients Who Had Emerging EGFR Amplification With Serial cfDNA Analyses
Six patients who initially tested negative for EGFR amplification in both tissue NGS and cfDNA were found to have emerging EGFR amplification with serial cfDNA analyses. (A total of 324 patients who initially tested negative had serial testing.) Patterns of genomic evolution differed from patient to patient. Emergence of EGFR amplification was seen among patients in a variety of situations, including in 4 ascopubs.org/journal/po JCO™ Precision Oncology  four patients who received checkpoint inhibitors, although two of the four patients also had other intervening therapies (Fig 3; Data Supplement).

Concordance of EGFR Amplification Between cfDNA and Tissue NGS
Among patients from the UCSD cohort whose cfDNA was evaluated, tissue NGS was available in 787 cases. The overall concordance rate for EGFR amplification between tissue and cfDNA NGS was 89.3% (Data Supplement). A shorter interval between the date of tissue biopsy (for tissue NGS) and blood draw (for cfDNA) was associated with statistically higher concordance rate (≤ 6 months: 92.1% v > 6 months: 85.8%; P = .005; Data Supplement).

DISCUSSION
We describe the comprehensive landscape of EGFR amplification in cfDNA among 28,584 patients with varied malignancies whose liquid biopsy was evaluated at a central, clinical-grade laboratory. Overall, 8.5% of patients harbored an EGFR amplification in their blood-derived ascopubs.org/journal/po JCO™ Precision Oncology 5 cfDNA, with EGFR amplifications being most common in colorectal cancer (16.3% of patients), NSCLC (9.0%), genitourinary cancers (8.1%), cutaneous tumors (7.4%), and breast malignancies (7.3%). Having high copy number amplification with greater than 4.00-fold EGFR amplification was seen in 1.0% of patients; most patients had amplification levels between 2.41fold and 4.00-fold (4.6%; Fig 1).

Fig 4.
Efficacy of anti-epidermal growth factor receptor (EGFR)-based therapies among patients with EGFR amplification (amp). Nine patients with EGFR amplification were evaluable for response from anti-EGFR-based therapies. Overall, tumor reduction was seen in six of nine (66.7%), including five of nine (55.6%) patients with partial response per Response Evaluation Criteria in Solid Tumors 1.1. (*) Patient was experiencing mixed response (mixture of disease stability and progressive disease) with pembrolizumab monotherapy after 15 months. Anti-EGFR agents were added for emerging EGFR amplification detected in cfDNA (see also Figure 5). cfDNA, cell-free DNA; CNA, copy number amplification; EGFR, epidermal growth factor receptor; FOLFIRI, fluorouracil, leucovorin, and irinotecan; GE, gastroesophageal; N/A, not applicable; NGS, next-generation sequencing; PFS, progression-free survival; SCC, squamous cell carcinoma; TKI, tyrosine kinase inhibitor. Mutation Burden cfDNA Imaging shown in Figure 5B cfDNA before treatment   A). A 55-year-old woman with metastatic triple-negative breast cancer to bone and lung was treated with pembrolizumab with initial disease stability for more than 1 year, but then progression in the bones and deteriorating performance status requiring a wheelchair. cfDNA obtained before treatment did not detect genomic alterations. cfDNA at the time of bone progression revealed multiple emerging alterations, including EGFR amplification (3+; plasma copy number of 4.88; A). (continued on following page) correlate better with response than protein expression, perhaps because of technical limitations associated with assessment of immunohistochemistry staining. [31][32][33][34] Prior studies looking at the relationship between EGFR amplification and therapeutic response to EGFR inhibitors showed inconsistent results (Data Supplement). However, one potential explanation is tumor heterogeneity, especially between primary and metastatic lesions and even between distinct foci at the same site. Indeed, Pectasides et al 23 demonstrated that, among patients with treatment-naive metastatic gastroesophageal cancers, discordant gene alterations between primary and metastatic tissue were common, being seen in 42% of patients. Interestingly, the discordance rate was higher for gene amplifications. However, among discordant cases, high concordance (87.5%) was seen between metastatic tissue and cfDNA profiling. The concordance rate documented by Pectasides et al 23 is similar to that in the current report that showed an 89.3% concordance rate for EGFR amplification between cfDNA and tissue NGS (Data Supplement). These results suggest that biopsy of a limited tumor focus can misrepresent the overall genomic condition of disease and, thus, may not be a completely accurate guide for targeted treatment. NGS of cfDNA derived from plasma may attenuate this challenge. Consistent with this concept, among our nine evaluable patients who harbored EGFR amplification by cfDNA analysis, anti-EGFR-based therapies led to tumor reduction in 66.7% (six of nine) including 55.6% (five of nine) who achieved a PR (Fig 4). Our data are comparable to those of Maron et al, 35 who showed a 58% (four of seven) objective response rate among patients with EGFR-amplified gastric cancer (all seven patients were positive for EGFR amplification by tissue NGS, and six were positive by cfDNA analysis). Similarly, ERBB2 amplification detected by cfDNA analysis was highly predictive of anti-HER2 targeted therapy response. 36 There were additional noteworthy observations from our patients treated with anti-EGFR-based therapies: three of five patients whose EGFR amplification was only detected in cfDNA (negative on tissue NGS) still demonstrated a PR (including one patient treated with cetuximab monotherapy [patient 57]), PRs were seen across different degrees of EGFR amplification status (from copy number amplification of 2.37 to 143.94), and three of four patients who received dual anti-EGFR inhibitors (coadministration of antibody and tyrosine kinase inhibitor for EGFR) achieved tumor reduction. Importantly, dual inhibition with both an antibody and a small molecule targeting the same receptor has been investigated among patients with HER2-positive breast cancer and reported to have significantly higher response rates when compared with either drug alone. 37 Efficacy of dual-targeted therapy was also seen in patients with HER2-positive colon cancer that showed a 30% response rate with trastuzumab/lapatinib combination. 38 Similarly, early-phase clinical trials with dual-EGFR inhibition (cetuximab/ afatinib-or cetuximab/erlotinib-based therapy) showed favorable clinical outcomes among patients with refractory NSCLC and colorectal cancer. [39][40][41][42][43] The mechanism by which dual inhibition operates is not fully elucidated, but preclinical studies suggest that kinase receptors may function via kinase-dependent and -independent mechanisms. 44,45 Although responses were seen in more than half of the patients with EGFR amplification 8 ascopubs.org/journal/po JCO™ Precision Oncology A 53-year-old woman with metastatic rectal adenocarcinoma to the liver and lungs presented after experiencing disease progression while receiving two lines of therapies (infusional fluorouracil, leucovorin, and oxaliplatin with bevacizumab and fluorouracil, leucovorin, and irinotecan with bevacizumab). Tissue next-generation sequencing was negative for EGFR amplification; however, cfDNA revealed alterations including EGFR amplification (1+; plasma copy number of 2.37). Therapy was started with single-agent cetuximab, and a 44% reduction in tumor burden by RECIST 1.1 was seen (progression-free survival, 6.0 months; C, left to right). (D) Patient 25: Serial CT images while receiving anti-EGFR therapies. A 68-yearold woman was referred with metastatic gastroesophageal junction carcinoma to the liver and lymph nodes. Both cfDNA and primary tumor showed EGFR amplification by next-generation sequencing (EGFR amplification 3+ by cfDNA, plasma copy number of 143.94). Therapy with dual anti-EGFR therapy (cetuximab and erlotinib) was started (patient was also administered one dose of nivolumab on the basis of programmed death ligand 1 positive by immunohistochemistry; however, held because of severe rash). The patient achieved a durable partial response 28 (70% tumor reduction; progression-free survival, 18 months; D, left to right). Serial cfDNA analyses were obtained at 4 months, 12 months, and 17 months after the initiation of therapy and were negative for EGFR amplification. Red arrows and circles indicate presence of tumor. cfDNA, cell-free DNA. treated with EGFR inhibitors, not all patients responded. Our study demonstrates that patients whose tumors harbor EGFR amplifications have considerably more cfDNA genomic alterations than those without EGFR amplification (median, 5.0 v 1.0 genomic coalteration per patient; P < .001). Therefore, primary or secondary resistance could be on the basis of the need to target coexisting activated pathways. Indeed, as seen in Figure 4, of the nine patients with cfDNA EGFR amplification treated with EGFRtargeting agents, the four nonresponders had six to nine genomic coalterations, whereas the five responders had only zero to five coalterations per patient. Furthermore, the patient with the greatest tumor regression and most durable response (Fig 4, 51 have shown that more extensive disease burden corresponds to higher rates of cfDNA detection. The finding of higher numbers of comutations in patients harboring EGFR amplification events could therefore be a possible effect of increased aggressiveness and higher tumor burden (with more extensive disease shedding more cfDNA and thus permitting detection of more alterations), or, alternatively, higher numbers of comutations could be a cause of increased aggressiveness. In this regard, we have recently found that higher percent cfDNA correlates with higher number of alterations. Furthermore, both higher percent cfDNA and higher number of alterations were independently associated with shorter survival after multivariate analysis. This observation suggests that the association between survival and number of alterations is independent of the percent cfDNA (with the latter correlating with disease burden; unpublished data).
In the current report, we also identified patients whose EGFR amplifications emerged in their liquid biopsy with serial testing after a variety of anticancer therapies (n = 6; Fig 3). For instance, one patient who was treated with the EGFR tyrosine kinase inhibitor erlotinib showed emergence of blood-derived EGFR amplification after disease progression (Fig 4), consistent with a previous report demonstrating tumor evolution with EGFR amplification as a potential resistance mechanism to EGFR tyrosine kinase inhibitor administration. 52 Perhaps relevant in this regard, all of our responders had an EGFR antibody included in their regimen. Four patients were found to have emerging EGFR amplification after disease progression while receiving checkpoint inhibitors. Although EGFR alterations are reported to be associated with primary resistance and hyperprogression after immune checkpoint blockade, 6,7 the current observation may suggest that EGFR amplification can also be a possible mechanism for acquired resistance after checkpoint blockades. For those patients with clonal evolution that includes EGFR amplification after anti-PD-1 checkpoint blockade, addition of anti-EGFR therapy may overcome resistance. This is suggested by our representative patient who was treated with pembrolizumab, had a mixed response, and then received erlotinib and cetuximab (in addition to ongoing pembrolizumab) and showed reduction in EGFR cfDNA copy number as well as regression of tumor foci and improvement in pain and performance status (Fig 4, patient 26; Figs 5A and 5B). 39,40 Additional investigation is required to understand the complex interplay of response and resistance associated with EGFR amplifications, EGFR-targeting pharmaceuticals, and checkpoint blockade.
There were several limitations to the current study. First, the investigation of clinical correlates in the UCSD cohort was performed retrospectively. Second, for the large de-identified database of 28,584 patients, sample size bias cannot be excluded, because the number of each cancer type was based on the number of samples sent for cfDNA testing by treating physicians. Moreover, the diagnosis was determined based on the submitting physician's designation. Third, in the de-identified database, we were not able to evaluate the history of systemic therapy that may have affected the dynamics of cfDNA. It is possible that some of the EGFR amplifications emerged because of therapeutic pressure. Last, technological methods and definition of EGFR amplification differed between tissue NGS and cfDNA analyses; thus, direct comparison between those two tests may be challenging. Yet, despite these limitations, the study provides a comprehensive analysis of EGFR amplification detected from plasma-derived cfDNA in a wide range of malignancies.
In conclusion, among patients with diverse cancers (n = 28,584 from a central laboratory), cfDNA interrogated by clinical-grade NGS