Exploring potential molecular resistance and clonal evolution in advanced HER2-positive gastric cancer under trastuzumab therapy

Abstract

HER2-positive gastric cancer (GC) makes up 15-20% of all GC incidences, and targeted therapy with trastuzumab is the standard of treatment. However, the mechanisms of resistance to trastuzumab are still not fully understood and presents a significant challenge in clinical practice. In this study, whole exome sequencing (WES) was performed on paired tumor tissues before trastuzumab treatment (at baseline) and at progressive disease (PD) in 23 GC patients. Clinicopathological and molecular features that may be associated with primary and/or acquired resistance to trastuzumab were identified. Lauren classification of intestinal type was associated with a more prolonged progression-free survival (PFS) than diffuse type (HR = 0.29, P = 0.019). Patients with low tumor mutation burden (TMB) showed significantly worse PFS, while high chromosome instability (CIN) was correlated with prolonged OS (HR = 0.27; P = 0.044). Patients who responded to treatment had a higher CIN than nonresponders, and a positive trend towards increasing CIN was observed as response improved (P = 0.019). In our cohort, the most common genes to acquire mutations are AURKA, MYC, STK11, and LRP6 with four patients each. We also discovered an association between clonal branching pattern and survival, with an extensive clonal branching pattern being more closely related to a shorter PFS than other branching patterns (HR = 4.71; P = 0.008). We identified potential molecular and clinical factors that provide insight regarding potential association to trastuzumab resistance in advanced HER2-positive GC patients.

Fig. 1. Mutational landscape for 23 GC patients at baseline and at PD.

Only top sixteen gene with somatic mutations and top eight genes with CNVs were shown. Genes were ordered according to the frequency of appearance. CR complete response, PR partial response, SD stable disease, PD progressive disease, CNV copy number variation, IHC immunohistochemistry, FISH fluorescence in situ hybridization.

Fig. 2. Cox proportional hazard model for clinical factors.

A Forest plot for clinical factors. The characteristic used for comparison was labeled “Reference”. A black box denotes the mean hazard ratio for each comparison to the reference. Error bars on forest plots represent the 95% confidence interval for each hazard ratio. Significant P values were shown in bold. Kaplan–Meier curves were shown for B Lauren classification, C differentiation, and D peritoneal metastasis. Log-rank test was used to determine significance. ECOG PS Eastern Cooperative Oncology Group performance status.

Fig. 3. Characteristic and significant alterations for 23 patients at baseline.

Survival analysis of TMB for patients with A PFS or B OS as endpoint. C Survival analysis for patients with high and low CIN with PFS or D OS as endpoint. E Boxplot of CIN distribution for response and no response patients. Wilcoxon ranked sum test was used to determine significance. F Trend test for significance of CIN across different response groups. Two-tailed Jonckheere’s trend test was used to determine the significance. G Kaplan–Meier curve of a patient with NOTCH1 deletion using PFS and H OS as endpoint. CR complete response, PR partial response, SD stable disease, PD progressive disease, WT wild type, TMB tumor mutation burden, CIN chromosomal instability.

Fig. 4. Acquired mutations and CNVs for 19 patients with acquired resistance to trastuzumab.

A Oncoprint for acquired mutations and CNVs across 19 patients with acquired resistance. Only alterations with ≥3 supporting samples were selected to be shown. B Forest plot for the hazard ratio of patients who acquired somatic mutations (Top) or CNVs (Bottom) vs. WT based on their PFS. A black box denotes the mean hazard ratio for each comparison to the reference. Error bars on forest plots represent the 95% confidence interval for each hazard ratio. Significant P values were shown in bold. Log-rank test was used to test for significance. C Survival curve of ERBB4 mutations based on PFS or D OS of patients. Kaplan–Meier curves for acquired FAT4 mutations were shown according to E PFS or F OS. WT wild type, CNV copy number variation.

Fig. 5. Clonal evolution pattern predicts response to progression-free survival in patients.

A Example fishplot for patients with extensive branching, single branching, and linear evolution patterns. Radiology imaging of the tumor site in each corresponding patient was displayed in relation to the time point when the image was captured. Sex, age, best response to treatment, PFS, and OS for each patient was shown above the corresponding fishplot. B Hazard of extensive branching pattern over other evolutionary patterns with PFS or C OS as the endpoint. SD stable disease, PR partial response, PD progressive disease.