EGFR/ERBB2 Amplifications and Alterations Associated With Resistance to Lenvatinib in Hepatocellular Carcinoma

Lenvatinib is a frontline tyrosine kinase inhibitor for the treatment of hepatocellular carcinoma (HCC) capable of inducing tumor responses in one-fourth of patients.1Llovet J.M. et al.Nat Rev Dis Primers. 2021; 7: 6Crossref PubMed Scopus (1792) Google Scholar,2Kudo M. et al.Lancet. 2018; 391: 1163-1173Abstract Full Text Full Text PDF PubMed Scopus (2850) Google Scholar However, many patients demonstrate resistance to lenvatinib, and molecular mechanisms underpinning refractory cancers have been poorly characterized. Recent in vivo studies suggest that hyperactivation of epidermal growth factor receptor (EGFR) signaling may induce lenvatinib resistance, but whether this mechanism is relevant in patients is unknown.3Jin H. et al.Nature. 2021; 595: 730-734Crossref PubMed Scopus (118) Google Scholar,4He X. et al.Sci Rep. 2022; 12: 8007Crossref PubMed Scopus (6) Google Scholar Circulating tumor DNA (ctDNA) profiling of peripheral blood provides a noninvasive method to identify cancer-specific mutations, including heterogenous genomic resistance alterations that arise from clonal selection during treatment. Unlike single-site tissue biopsies, which may only define molecular alterations of cells from a specific geographic region, ctDNA may capture genetic alterations regardless of tumor geography and across all cancer lesions.5Hsiehchen D. et al.Nat Commun. 2022; 13: 7477Crossref PubMed Scopus (1) Google Scholar We analyzed genetic alterations in the EGFR pathway in 46 patients with HCC who had ctDNA profiling using a clinically validated next-generation sequencing-based assay (Guardant360; Guardant Health Inc) before starting therapy and at progression, including 16 patients who received lenvatinib and 30 patients who received immune checkpoint inhibitors (ICIs) including pembrolizumab, nivolumab, or atezolizumab plus bevacizumab (see Supplementary Methods).6Lanman R.B. et al.PLoS One. 2015; 10e0140712Crossref PubMed Scopus (490) Google Scholar At the time of progression to lenvatinib, EGFR amplifications ranging from 2.3 to 4.3 copies were detected in 3 patients who did not have amplifications detected at baseline (Figure 1A). One patient treated with lenvatinib demonstrated an increase in copy number from 4.8 to 7 at progression (Figure 1B). An additional patient who progressed on lenvatinib developed an activating ERBB2 mutation (S310Y) at progression (Figure 1A). Among 30 patients with longitudinal ctDNA assessment treated with ICIs, 2 patients lost EGFR amplifications at progression, 1 patient had a decline in the variant allele frequency of an activating EGFR alteration (A647T), and 1 patient had an ERBB2 amplification that increased in copy number at progression from 2.5 to 3.9 copies (Figure 1A and B). Beyond EGFR family gene alterations, no other ctDNA alterations associated with an increased variant allele frequency or copy number were common across the 5 patients who had EGFR/ERBB2 ctDNA alterations at the time of progression to lenvatinib (Figure 1C). The Cancer Genome Atlas (TCGA) HCC specimens that underwent RNA sequencing and/or analysis by reverse-phase protein arrays confirmed that low- and high-level gains in EGFR/ERBB2 determined by Genomic Identification of Significant Targets in Cancer (GISTIC) had greater RNA and protein levels than diploid HCC (Supplementary Figure 1).7CGAR NetworkCell. 2017; 169: 1327-1341Abstract Full Text Full Text PDF PubMed Scopus (1267) Google Scholar To test whether EGFR/ERBB2 alterations detected before lenvatinib treatment is predictive of early treatment failure, we retrospectively analyzed 227 patients with HCC who underwent baseline ctDNA profiling (inclusive of the original 46 patients) and assembled 3 patient cohorts: patients with EGFR/ERBB2 alterations before to lenvatinib treatment (n = 6), patients without EGFR/ERBB2 alterations before lenvatinib treatment (n = 32), and patients with EGFR/ERBB2 alterations before ICI treatment (n = 17) (Supplementary Table 1). Disease control rates (complete/partial response or stable disease per Response Evaluation Criteria in Solid Tumors [RECIST]) were 20% in cohort 1, 62.5% in cohort 2 (P = .05), and 58% in cohort 3 (P = .09) (Figure 1D). Progression-free survival was longer in cohort 2 (median, 5.9 months; 95% confidence interval [CI], 2.0–9.8) and cohort 3 (median, 7.1 months; 95% CI, 1–13.2) compared with cohort 1 (median, 2.2 months; 95% CI, 1.6–2.8; log-rank P = .002) (Figure 1E). There was also a trend toward longer overall survival in cohort 2 (median, 9.2 months; 95% CI 7.3–11.1) and cohort 3 (median, 16.9 months; 95% CI, 6.2–27.6) compared with cohort 1 (median, 3.9 months; 95% CI, 0.5–7.3; log-rank P = .08) (Figure 1F). In a national cohort of 1616 patients with HCC who underwent ctDNA profiling, EGFR amplifications and activating mutations occurred in 6.5% and 1.5% of cases, respectively, and ERBB2 amplifications and activating mutations occurred in 1.9% and 0.7% of cases, respectively (Figure 1G). Most EGFR/ERBB2 amplifications were low level (<4 copies), with a subset at moderate- or high-level gains (Figure 1H). Among EGFR/ERBB2 mutations, we identified recurrent alterations associated with resistance to earlier-generation EGFR inhibitors (Figure 1I).8Shah R. et al.Clin Lung Cancer. 2020; 21: e216-e228Abstract Full Text Full Text PDF PubMed Scopus (66) Google Scholar These results collectively suggest that EGFR/ERBB2 alterations may be genetic mechanisms that underlie lenvatinib resistance, and their detection in ctDNA may serve as predictive or pharmacodynamic markers. EGFR inhibitors may enhance lenvatinib efficacy in a subset of patients, but the nontrivial number of EGFR gatekeeper alterations in HCC suggests that later-generation EGFR inhibitors may have greater utility in some patients.3Jin H. et al.Nature. 2021; 595: 730-734Crossref PubMed Scopus (118) Google Scholar For instance, ongoing clinical studies including a trial testing gefitinib in lenvatinib-resistant HCC (NCT04642547) and an HCC chemoprevention study testing erlotinib (NCT04172779) rely on EGFR inhibitors, which are ineffective against gatekeeper mutations. If not, it is likely that gatekeeper alterations represent rare subclonal driver mutation events rather than intramolecular mechanisms of lenvatinib resistance because lenvatinib does not directly bind or inhibit EGFR.3Jin H. et al.Nature. 2021; 595: 730-734Crossref PubMed Scopus (118) Google Scholar Combining lenvatinib with ICIs could also be considered in select patients because EGFR/ERBB2 alterations were not associated with ICI resistance. Because patients treated with other tyrosine kinase inhibitors were not examined in this study and only some patients were treated with atezolizumab plus bevacizumab in our ICI cohort, further evaluation is warranted to determine if other anti-vascular endothelial growth factor agents as monotherapy or in combination with ICIs may lead to drug resistance mediated by EGFR/ERBB2 alterations. Further validation of these results is needed given the limited sample size of EGFR/ERBB2-altered cohorts examined. Low-level ctDNA copy numbers of EGFR/ERBB2 associated with lenvatinib resistance differ from larger copy numbers in gene amplifications associated with treatment resistance detected in tissue reported in other contexts (such as MET amplifications in EGFR inhibitor–resistant lung cancers).9Engelman J.A. et al.Science. 2007; 316: 1039-1043Crossref PubMed Scopus (3961) Google Scholar,10Piotrowska Z. et al.Cancer Discov. 2018; 8: 1529-1539Crossref PubMed Scopus (291) Google Scholar However, this could be attributable to differences in the biologic aspects of specimen input because ctDNA captures genetic heterogeneity across cancer lesions, whereas single-site biopsies capture molecular alterations within a limited spatial region. Nonetheless, this study identifies a putative genetic marker of lenvatinib resistance that may guide treatment selection for patients given that multiple frontline regimens currently exist for HCC treatment and studies that have directly compared their efficacy are lacking. Mir Lim, MD (Data curation: Supporting; Investigation: Supporting; Writing – review & editing: Supporting). Joseph W. Franses, MD (Data curation: Supporting; Investigation: Supporting; Writing – review & editing: Supporting). Robin Imperial, MD (Data curation: Supporting; Investigation: Supporting; Writing – review & editing: Supporting). Umair Majeed, MBBS (Data curation: Supporting; Investigation: Supporting; Writing – review & editing: Supporting). Jill Tsai, PhD (Data curation: Supporting; Writing – review & editing: Supporting). David Hsiehchen, MD (Conceptualization: Lead; Data curation: Supporting; Formal analysis: Lead; Funding acquisition: Lead; Supervision: Lead; Writing – original draft: Lead). Patients who had ctDNA testing using a clinical next-generation sequencing assay (Guardant360; Guardant Health) as a component of routine clinical care from August 2019 to July 2022 were retrospectively identified after receiving institutional review board approval at each respective institution. A subset of patients had serial ctDNA analysis done at baseline and after progression to lenvatinib or ICIs, whereas the remainder had ctDNA collected before initiation of lenvatinib or ICIs. Patients who received local regional therapy including transcatheter arterial chemoembolization, ablation, or radiation during systemic therapy were excluded because these treatments could confound ctDNA results. ICI treatment included atezolizumab plus bevacizumab, pembrolizumab, or nivolumab. Clinical data including demographics, treatment history, tumor responses, and survival outcomes were abstracted from the electronic medical record. Cases with corresponding tissue sequencing data were analyzed if tissue specimens were obtained within 6 months of ctDNA collection and without intervening systemic treatments in that time period. Deidentified data from a national registry of HCC that underwent ctDNA analysis using Guardant360 from November 2016 to January 2022 were analyzed to determine the prevalence of EGFR/ERBB2 alterations after receiving an exemption for nonhuman research from the Human Research Protection Program. Clinical history for 2 patients with EGFR inhibitor gatekeeper mutations (EGFR T790M and A547T mutations) was reviewed that demonstrated no concomitant diagnosis of lung cancer or prior history of EGFR inhibitor exposure. Peripheral blood samples were collected in Cell-Free DNA Blood Collection (Streck) tubes and shipped to a Clinical Laboratory Improvement Act–certified, College of American Pathologists–accredited laboratory (Guardant Health, Redwood City, CA). A minimum of 5 ng of cell-free DNA was isolated for library preparation. As previously described, the Guardant360 assay is a targeted high-throughput hybridization-based capture technology for the detection of alterations in 73 or 84 genes.4He X. et al.Sci Rep. 2022; 12: 8007Crossref PubMed Scopus (6) Google Scholar Putative germline mutations including variants identified by allele fractions between 40% and 60%, prior annotation as germline mutations, and manual review were excluded from analysis. Tumor response was categorized according to RECIST 1.1. Progression-free and overall survival were calculated from the time of lenvatinib or ICI treatment initiation. Differences in disease control rates between patient cohorts were assessed using the the Fisher exact probability test. Kaplan-Meier curves were used to estimate progression-free and overall survival and differences were assessed using the log-rank test. Differences in RNA and protein levels between diploid and EGFR/ERBB2-amplified HCC were assessed using the 1-way analysis of variance test. The results are in part based on data generated by the TCGA Research Network (https://www.cancer.gov/tcga).Supplementary Table 1Characteristics of Patients With or Without EGFR/ERBB2 AlterationsCharacteristicsEGFR/ERBB2-altered treated with lenvatinibEGFR/ERBB2 wild-type treated with lenvatinibPEGFR/ERBB2-altered treated with ICIPSex Male5 (83.3)26 (81.3).9014 (82.4).96 Female1 (16.7)6 (18.8)3 (17.6)Age <40 y0 (0)2 (6.3).880 (0).86 41–50 y0 (0)3 (9.4)3 (17.6) 51–60 y3 (50)12 (37.5)8 (47.1) 61–70 y2 (33.3)11 (34.4)4 (23.5) >70 y1 (16.7)4 (12.5)2 (11.8)Race White3 (50)17 (53.1).8312 (70.6).30 Asian0 (0)3 (9.4)1 (5.9) Black2 (33.3)9 (28.1)4 (23.5) Unknown1 (16.7)3 (9.4)0 (0)Line of therapy First3 (50)14 (43.8).8212 (70.6).59 Second2 (33.3)15 (46.9)4 (23.5) Third1 (16.7)3 (9.4)1 (5.9)Cirrhosis Alcohol1 (16.7)5 (15.6).853 (17.6).92 Hepatitis C virus3 (50)13 (40.6)6 (35.3) Hepatitis B virus0 (0)2 (6.3)1 (5.9) Non-alcoholic steatohepatitis2 (33.3)8 (25)6 (35.3) Unknown0 (0)4 (12.5)1 (5.9)Child-Pugh score A4 (66.7)24 (75).6710 (58.8).73 B2 (33.3)8 (25)7 (41.2)Values are n (%). Open table in a new tab Values are n (%).