Genetic landscape and personalized tracking of tumor mutations in Vietnamese women with breast cancer

Abstract

Breast cancer is the leading cause of cancer death in Vietnamese women, but its mutational landscape and actionable alterations for targeted therapies remain unknown. After treatment, a sensitive biomarker to complement conventional imaging to monitor patients is also lacking. In this prospective multi-center study, 134 early-stage breast cancer patients eligible for curative-intent surgery were recruited. Genomic DNA from tumor tissues and paired white blood cells were sequenced to profile all tumor-derived mutations in 95 cancer-associated genes. Our bioinformatic algorithm was then utilized to identify top mutations for individual patients. Serial plasma samples were collected before surgery and at scheduled visits after surgery. Personalized assay tracking the selected mutations were performed to detect circulating tumor DNA (ctDNA) in the plasma. We found that the mutational landscape of the Vietnamese was largely similar to other Asian cohorts, showing higher TP53 mutation frequency than in Caucasians. Alterations in PIK3CA and PI3K signaling were dominant, particularly in our triple-negative subgroup. Using top-ranked mutations, we detected ctDNA in pre-operative plasma in 24.6-43.5% of the hormone-receptor-positive groups and 76.9-80.8% of the hormone-receptor-negative groups. The detection rate was associated with breast cancer subtypes and clinicopathological features that increased the risk of relapse. Interim analysis after a 15-month follow-up revealed post-operative detection of ctDNA in all three patients that had recurrence, with a lead time of 7-13 months ahead of clinical diagnosis. Our personalized assay is streamlined and affordable with promising clinical utility in residual cancer surveillance. We also generated the first somatic variant dataset for Vietnamese breast cancer women that could lay the foundation for precision cancer medicine in Vietnam.

Keywords: circulating tumor DNA; minimal residual disease; mutational landscape; next-generation sequencing; somatic mutation.

Conflict of interest statement

V‐ANH, PLD, NTNT, MLN, NMN, DQN, Y‐TL, M‐DP, HG and LNT are current employees of Gene Solutions, Vietnam. The remaining authors declare no conflict of interest.

© 2022 Gene Solutions JSC and The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies.

Figures

Fig. 1

Schematic of study design and K‐track® assay. (A) 134 patients with primary breast cancer stage I‐III, eligible for curative‐intent surgery were enrolled. Serial plasma samples were collected before surgery and after surgery at scheduled visits. Formalin‐fixed paraffin‐embedded (FFPE) samples of surgically removed tumors were also collected. Patients were followed‐up to record clinical outcomes. (B) Paired FFPE and white blood cells (WBC) DNA of the same individual were sequenced to identify tumor‐specific somatic mutations in 95 cancer‐associated genes. Top 4–5 mutations were selected by our scoring algorithm and tracked in plasma samples using a bespoke multiplex polymerase chain reaction (mPCR) assay and ultra‐deep next‐generation sequencing (NGS).

Fig. 2

Mutational spectrum of 95 genes in Vietnamese breast cancer women. (A) The number of somatic mutations identified per patient was significantly higher in the HR− HER2+ compared to the HR+ groups. (B) The mutation burden was significantly lower in stage I compared to stage II and III. (C) Pie chart showing the distribution of mutation classes identified in 95 genes. (D) The top 25 significantly mutated genes in the cohort. (E) Frequency of top highly mutated genes in breast cancer was compared between our cohort and published datasets of Caucasian and Asian cohorts. (F) Pairwise analysis identified oncogenic alternations significantly associated with different subtypes. Genes in green color were significantly more altered in a particular group compared to the rest while genes in black color were significantly less altered. HR, Hormone receptors; HER2, Human epidermal growth factor receptor 2. *P < 0.05; Kruskal–Wallis and post hoc Dunn's test for (A) and (B); Fisher's exact test for (F). (A), (B): Boxplots with Tukey whiskers.

Fig. 3

Oncogenic signaling pathways and actionable alterations in breast cancer subtypes. (A) The top three signaling pathways with frequent oncogenic alterations in our cohort were genome integrity, PI3K signaling and chromatin SWI/SNF remodeling complex. (B) The top three altered signaling pathways were different among breast cancer subtypes. HR+ HER2− had significantly less alterations in genome integrity pathway compared to other groups. Transcription factor pathway was highly mutated specifically in the HR+ HER2− group. (C) Proportion of patients in each subtype that had actionable alterations predictive of treatment response to a drug at different levels of evidence stratified by OncoKB database. HR, Hormone receptors; HER2, Human epidermal growth factor receptor 2. *P < 0.05; Fisher's exact test for (B).

Fig. 4

Analysis of circulating tumor DNA (ctDNA) in pre‐operative plasma samples. (A) The number of mutations selected to track in each patient was significantly lower in the HR+ HER2− group than in the HR− HER2− group. (B) The percentage of tracked mutations that could be detected in the plasma was significantly lower in the HR+ groups compared to HR‐ groups. (C) Pre‐operative detection rate was associated with breast cancer subtypes and was significantly lower in the HR+ groups compared to HR− groups. (D) The detection rate of PIK3CA H1047R/L in the plasma followed the exact same trend as the overall detection rate even though the mutation was identified at similar variant allele frequency (VAF) in tumors across subtypes. (E) When HR+ groups were stratified by clinicopathological features that increase risk of relapse, the detection rate was significantly lower in the low‐risk than the high‐risk group of HR+ HER2−. (F) Detection was associated with TNM stage as the detection rate in stage I was significantly lower than in stage II and III. Nodal and grade status were not found to affect pre‐operative detection rate. (G) Levels of pre‐operative and post‐operative cell‐free DNA (cfDNA) were not different while VAF of ctDNA significantly reduced after surgery. HR, Hormone receptors; HER2, Human epidermal growth factor receptor 2. *P < 0.05; Kruskal–Wallis and post hoc Dunn's test for (A), (B), (D); Mann–Whitney U test for (G); Chi‐squared test and Fisher's exact test for (C–F). (A), (B), (D), (G): Boxplots with Tukey whiskers.

Fig. 5

Longitudinal monitoring of circulating tumor DNA (ctDNA) and clinical outcomes. (A) Swimmer plot depicting serial ctDNA results over time and incidence of relapse or metastasis of 32 patients that had been followed up for at least 15 months. This was an interim analysis as the clinical study is ongoing. (B, C) longitudinal plot showing the mean VAF of ctDNA, treatment and clinical status over time of patients ZMB022 and ZMB041. CRT, Chemoradiotherapy; Op, Operation. Molecular relapse detection was 13 months earlier than clinically diagnosed relapse in patient ZMB041.