Harnessing Institutionally Developed Clinical Targeted Sequencing to Improve Patient Survival in Breast Cancer: A Seven-Year Experience

1. Study population

Patients aged 20 years or older who were diagnosed with advanced or metastatic BC and underwent NGS in SNUH from October 2016 to July 2023 were enrolled. Staging was determined according to the latest TNM classification by the American Joint Committee on Cancer (AJCC) at the time of diagnosis, and clinical staging was used for the patients who had neoadjuvant therapy. All clinicopathological information was retrieved by reviewing electronic medical records, and patients’ survival data were obtained from the Ministry of Public Administration and Security in the Republic of Korea.

While most of the patients were tested by NGS once, a subset of patients in our study underwent clinical NGS tests more than once; this was accomplished as a part of the KRH Project in some, and the others were performed under the reimbursement policy which allowed the second NGS test at the time of disease progression or treatment refractoriness.

2. FiRST Cancer Panel: institutionally developed, internally validated panel

Medical oncologists, pathologists, bioinformatics specialists, and medical geneticists in SNUH collaborated to develop a customized and institutionally validated targeted sequencing platform. As a result, FiRST Cancer Panel (FCP) service finally came into clinical practice in October 2016. In 2017, FCP was readily incorporated into conditional coverage by the National Health Insurance Service (NHIS) of Korea.

The first panel to be in service was FCP ver. 2.0, which covered 148 genes. FCP service was provided upon clinicians’ requests following patients’ informed consent. The results were then discussed at molecular tumor boards (MTB) held twice a month. Our institution continued to collaborate in upgrading and validating the panel, with the latest version, ver. 4.0, being released in January 2022. The updated version now includes 336 genes by DNA panel and 151 genes by RNA panel. The list of genes in each version and the major updated features are summarized in the S1 Table. Details of targeted sequencing and analytic methods are provided in the Supplementary Method.

3. Immunohistochemistry and ERBB2 in situ hybridization

The results of immunohistochemistry (IHC) were retrospectively reviewed from the medical records and pathology reports. The following antibodies were used for IHC: estrogen receptor (ER; clone 6F11, Novocastra Laboratories), progesterone receptor (PR; clone 16, Novocastra Laboratories), human epidermal growth factor receptor 2 (HER2; clone 4B5, Ventana Medical Systems), and Ki-67 (clone MIB-1, Dako/Agilent). Positive stainings for ER or PR in ≥ 1% of tumor nuclei were interpreted as positive [14].

HER2 IHC was interpreted according to the American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) 2018 guidelines [15]. ERBB2 in situ hybridization (ISH) was performed for the samples with equivocal (2+) results of HER2 IHC or upon clinicians’ requests. Fluorescent in situ hybridization was performed using PathVysion HER-2 DNA Probe Kit (Abbott Molecular) from 2016 to March 2022, and silver in situ hybridization was performed using INFORM Her2 Dual ISH probe (Ventana Medical Systems) from April 2022 to July 2023. ISH was interpreted according to the ASCO/CAP 2018 guidelines [15].

4. Statistical methods

Categorical variables were summarized with the frequencies and percentages. Continuous variables were represented with the median values and ranges. Categorical variables were compared by chi-square test or Fisher’s exact test and continuous variables were analyzed by Mann-Whitney or Kruskal-Wallis tests, as appropriate. We used the Benjamini-Hochberg method to correct for multiple comparisons when necessary.

Disease-free survival (DFS) was defined as time from curative surgery to recurrence or metastasis. Progression-free survival (PFS) was defined as the time from the initiation of one line of treatment to the progression of disease after that specific treatment. The overall survival (OS) was evaluated in metastatic BC and was designated as the time from the first diagnosis of metastasis to death from any cause. The survival curves were estimated using the Kaplan-Meier method and compared by log-rank test. For live patients, the survival was censored at the latest date of follow-up. A p-value < 0.05 was considered statistically significant.

1. Clinicopathological characteristics of the study population

Clinical NGS was performed using 548 samples from 522 patients with BC. Characteristics of the 522 patients are summarized in Table 1. Only three patients (3/522; 0.6%) were male, and the median age at diagnosis was 47 years (range, 20 to 76 years). 63.8% of the patients were diagnosed with locally advanced stages II-III, while 23.4% presented with de novo stage IV.

Table 1.

Clinicopathological characteristics of patients and samples in the study

Of 548 tumor samples subjected to FCP, the most common histological subtype was invasive ductal carcinoma (IDC) (91.2%), followed by invasive lobular carcinoma (ILC) (3.5%), and rare subtypes including mucinous, metaplastic, and salivary gland type carcinomas. Sequenced tumors were mostly ER+/HER2– BCs (53.1%), followed by triple-negative BCs (TNBCs) (35.4%), ER–/HER2+ BCs (6.4%), and ER+/HER2+ BCs (5.1%). Tissues for the NGS were obtained from various sites, including 45.1% from primary BC, 9.1% from regional, 2.6% from distant lymph nodes, and 43.2% from distant metastatic organs. Within 6 months before the tissue acquisition, 64.1% of the patients received antitumor treatments, including chemotherapy (31.8%), endocrine therapy (29.0%), targeted therapy (17.0%), and immuno-oncologic drugs (2.4%). The median block age of the samples used for NGS tests was 2.3 months, ranging from 0 to 105 months. 52.4% of the tests were performed with samples collected within 6 months.

2. Mutational profiles of clinical NGS with BC

In 548 samples, a total of 1,701 pathogenic alterations in 144 genes were identified (Fig. 1). Ninety-seven point six percent (535/548) of the tested samples harbored at least one pathogenic alteration, and 70.6% had at least one actionable biomarker annotated by OncoKB [16].

Fig. 1.

The genetic profile of breast cancer patients unveiled by FiRST Cancer Panel. The overall distribution of commonly detected mutations, copy number variations, fusions, as well as microsatellite instability scores and tumor mutational burden are presented alongside clinicopathological features. CTx, chemotherapy; ET, endocrine therapy; ER, estrogen receptor; IDC, invasive ductal carcinoma; HER2, human epidermal growth factor receptor 2; ILC, invasive lobular carcinoma; IO, immuno-oncology; MSI, microsatellite instability; TMB, tumor mutational burden; TNBC, triple-negative BCs; Tx, therapy.

The commonly mutated genes were TP53 (56.2%), PIK3CA (31.2%), GATA3 (13.8%), BRCA2 (10.2%), KMT2C (8.8%), PTEN (8.6%), BRCA1 (7.3%), CDH1 (6.4%), ESR1 (5.5%), and ARID1A (5.3%). Recurrent amplifications were observed in CCND1 (10.8%), FGF19 (10.0%), ERBB2 (9.5%), MYC (9.3%), and FGFR1 (8.0%). None of the tested samples was microsatellite instability–high, and the median tumor mutational burden (TMB) was 5.33 per megabase pair (range, 0.0 to 32.79).

We found distinct genetic alterations in each BC subtype. TP53 mutations were the most frequent in TNBC tumors (83.5%), followed by 76.2% of HER2+ and 33.7% of ER+/HER2– tumors. ESR1 mutations were found exclusively in the ER+/HER2– tumors (30/291; Fisher’s exact test, p < 0.001). Of note, ESR1 mutations were significantly enriched in BCs with prior exposure to aromatase inhibitors (AI) (22.6%, 21/93 in prior AI-treated BCs compared to 4.5%, 9/198 in the patients without prior AI exposure; Fisher exact test, p < 0.001) (S2 Fig). Pathogenic alterations of the homologous recombination repair–associated genes were found in 129 (23.6%) tumors and were significantly enriched in TNBC (56/194, 28.9%) compared to the other subtypes (73/354, 20.6%) (chi-square, p=0.030). We found no significant difference in TMB among BC subtypes (Kruskal-Wallis, p=0.114).

We found 40 tumors with pathogenic BRCA1 mutations of which 32 (80.0%) were confirmed as germline variants by blood tests. BRCA1 mutations were significantly enriched in the TNBC subtype (29/40, 72.5%; chi-square, p < 0.001) compared to ER+/HER2– tumors (11/40, 27.5%). None of the BCs with BRCA1 mutations were HER2+. Among 56 tumors with BRCA2 mutations, 80.4% (45/56) were confirmed as germline. Sixty-seven point nine percentage (38/56) of BRCA2 mutated samples were ER+/HER2–, 23.2% (13/56) were TNBC, and 8.9% (5/56) were HER2+. We also found tumors harboring other germline likely pathogenic/pathogenic mutations involving BC predisposing genes: PALB2 (2/82, 2.4%), TP53 (2/82, 2.4%), and PTEN (1/82, 1.2%). Mutational profiles of the BCs with germline predispositions are summarized in the S3 Fig.

To gain further insights into the distinctive genetic characteristics of Korean BC patients, we conducted a comparative analysis of genetic alterations between our dataset and a Western dataset reported by Razavi et al. [17] from Memorial Sloan Kettering Cancer Center (MSKCC). Key findings are described in the S4 Table. Significant enrichments in the patients from SNUH included TP53 mutations (56.2% in SNUH vs. 35.5% in MSKCC, adjusted p < 0.001), BRCA1 mutations (7.3% in SNUH vs. 1.4% in MSKCC, adjusted p < 0.001), and BRCA2 mutations (10.2% in SNUH vs. 3.3% in MSKCC, adjusted p < 0.001). MAP3K1, CDH1, FOXA1, PIK3CA, ERBB2, ESR1 mutations, and CCND1, FGF19, and FGFR1 amplifications were more frequently found in the MSKCC cohort (adjusted p < 0.05).

3. Diagnostic performance of FCP in determining the HER2 status

ERBB2 amplification is by far the most important copy number variation (CNV) in BC in terms of therapeutic and prognostic implications. To assess the diagnostic performance of the NGS panel in determining HER2 status, we compared the conventionally defined HER2 status by IHC and ISH with ERBB2 amplifications by FCP (Table 2). The concordance rate was 96.7% with only 18 samples (3.3%) having discordant results between the two methods, with the sensitivity, specificity, positive predictive, and negative predictive values of 74.6%, 99.6%, 95.9%, and 96.8%, respectively. The ERBB2 copy number (CN) estimated by NGS were significantly higher in the samples with HER2 IHC 3+ compared to 2+ (Mann-Whitney, p=0.001) (Fig. 2A) and showed significant correlation when compared to the ERBB2 CN assessed by ISH method (p=0.042; Pearson coefficient=0.594) (Fig. 2B).

Table 2.

Comparison between conventional assessment of HER2 status and NGS-based ERBB2 amplification detection

Fig. 2.

ERBB2 copy number assessment by FiRST Cancer Panel compared to human epidermal growth factor receptor 2 (HER2) immunohistochemistry and in situ hybridization. (A) ERBB2 copy numbers were higher in tumors with HER2 immunohistochemistry (IHC) 3+ compared to HER2 IHC 2+. (B) ERBB2 copy number assessed by FiRST Cancer Panel showed a significant correlation with the results of HER2 in situ hybridization. CN, copy number; ISH, in situ hybridization; NGS, next-generation sequencing.

4. Identifying gene fusions with clinical implications

Since the introduction of ver. 4.0 in 2022, the RNA panel gave us a better view of fusion events in patients with BC. In our study population, fusion genes were identified in nine patients (S5 Table). ETV6::NTRK3 fusion was detected in a patient with secretory carcinoma, which has both diagnostic and therapeutic implications. Other potentially targetable fusions included TRIM56::BRAF and RBPMS::NRG1 fusions. Although partner genes have not been reported in the literature, these fusion products were expected to have intact tyrosine kinase and epidermal growth factor–like domains respectively, thus potentially targetable.

We also found a patient with ER+/HER2– ILC harboring ESR1::NBAT1 fusion, which was predicted to have a truncated ligand binding domain (LBD) of ESR1 (Fig. 3A). The patient (case No. 225) experienced disease progression with endocrine therapy, possibly suggesting secondary endocrine resistance. We also report the first detailed human cases of BCL2L14::ETV6 fusion with aggressive TNBC and clinical resistance to conventional chemotherapy (Fig. 3B). Case No. 205 was a 30-year-old female diagnosed with locally advanced TNBC who received window of opportunity trial chemoimmunotherapy with olaparib and durvalumab followed by standard care of neoadjuvant chemotherapy doxorubicin+cyclophosphamide followed by docetaxel, and then curative surgery. Despite the partial clinical response before surgery, this patient could not achieve a pathologic complete response and relapsed with multiple lung metastases with DFS of 1 year and 3 months. Case No. 296 was a 43-year-old premenopausal woman who initially presented with extensively metastatic BC involving the axial skeleton. Despite palliative chemotherapy with docetaxel and cisplatin, the disease rapidly progressed after 3 months with diffuse leptomeningeal seeding and bone metastases. After she received sacituzumab govitecan as salvage treatment for 3 months, the leptomeningeal seeding and extensive bone metastases were improved.

Fig. 3.

Fusion genes identified by FiRST Cancer Panel. (A) The novel ESR1::NBAT fusion gene is expected to cause the distruption of ligand binding domain of ESR1. (B) Fusion gene structures and clinical course of the patients harboring BCL2L14::ETV6 fusions are presented Case No. 205 was a 30-year-old female diagnosed with locally advanced triple-negative breast cancer. The lesions in the right breast, right axillary lymph nodes, and internal mammary lymph nodes (arrows in the first image) decreased after the neoadjuvant treatment with olaparib+durvalumab window of opportunity treatment followed by standard neoadjuvant therapy consisting of doxorubicin+ cyclophosphamide followed by docetaxel (arrow in the second image). However, the disease recurred with lung, pleura, and mediastinal lymph nodes (arrows in the third image) metastases after 1 year and 3 months. Case No. 296 was a 43-year-old premenopausal woman who initially presented with extensively metastatic breast cancer involving the axial skeleton. Despite palliative chemotherapy with docetaxel and cisplatin, the disease rapidly progressed after 3 months in bone and leptomeningeal seeding, however, extensive bone metastases showed metabolic response after 4 cycles of Sacituzumab Govitecan. CTx, chemotherapy; SG, Sacituzumab Govitecan.

5. Bridge to the genetic counseling

With the release of FCP ver. 4.0 in early 2022, we began suggesting additional work-up and genetic counseling through clinical genomic medicine specialists when reporting possible germline pathogenic mutations in cancer susceptibility genes. The suggestions were made in 11.7% (21/180) tests and resulted in the actual clinical actions in 17 patients. Eleven patients were confirmed as the carriers of germline likely pathogenic/pathogenic variants and received genetic counseling. The other six patients were found to have somatic mutations, obviating the need for additional testing of family members.

Importantly, seven out of 11 aforementioned patients who harbored germline likely pathogenic/pathogenic BRCA1/2 mutations were not eligible for insurance reimbursed germline mutations tests under the current Korean health policy. The results of FCP in these seven patients prompted accurate genetic diagnosis and adequate clinical management. One example with germline predisposed BC initially identified by FCP was a 67-year-old woman who presented with metastatic ER+/HER2– BC after 5.5 years of disease-free interval. Clinical NGS revealed two PALB2 mutations along with TP53 and PIK3CA mutations. Two PALB2 mutations were found in the different sequencing reads (Fig. 4A), suggesting bi-allelic inactivation of PALB2. This result prompted confirmatory germline testing, which revealed the germline origin of the PALB2 H276fs mutation. The patient was enrolled in a clinical trial to receive matched therapy with a poly(ADP-ribose) polymerase (PARP) inhibitor olaparib and had a remarkable metabolic response of multiple bone metastasis after 3 months (Fig. 4B) and currently maintaining stable disease by Response Evaluation Criteria in Solid Tumor ver. 1.1 for 1 year and 4 months.

Fig. 4.

Utilization of FiRST Cancer Panel for screening germline predisposition and cancer evolution. (A) Biallelic inactivation of PALB2 revealed by FiRST Cancer Panel (FCP) prompted germline confirmatory testing. The patient was confirmed to harbor germline PALB2 pathogenic mutation and enrolled in a clinical trial using a poly(ADP-ribose) polymerase inhibitor (PARPi), resulting in a remarkable clinical response (B). (C) Sequential FCP tests using samples from two different time points showed a BRCA1 reversion mutation. VAF, variant allele frequency.

6. Clues to the pathogenesis and clonal evolution of BC

Out of 522 patients, 24 patients (4.6%) were tested twice, and one patient underwent sequencing of multiple lesions. Among the 24 patients with serial tests, clinically meaningful differences were observed in five patients (5/24, 20.8%). Two patients acquired ESR1 mutations following endocrine therapy (AI), and the other two patients exhibited intertumoral heterogeneity. The remaining patient was an interesting case that acquired somatic BRCA1 reversion mutations in the absence of prior PARP inhibitor therapy (Fig. 4C).

Multi-regional sequencing was performed on a patient diagnosed with a rare mixed ER+/HER2– IDC and metaplastic carcinoma (spindle cell carcinoma) (Fig. 5). The spindle cell carcinoma component of the resected breast tumor and the IDC component of the metastatic axillary lymph nodes were separately taken for NGS. Both tumors shared PIK3CA, PTEN, and KMT2C mutations, indicating a potential shared ancestral origin despite their morphological differences. Interestingly, distinctive genetic variations were also identified; two different TP53 mutations were detected in the two components, whereas TERT promoter mutation and amplifications of JAK1 and CDK4 were exclusively present within the spindle cell carcinoma component. Additional NGS was done in brain metastasis in which the genetic profiles were identical to the spindle cell carcinoma of the breast, suggesting evidence of clonal evolution in rare cancer.

Fig. 5.

Genetic evolution model of a rare breast cancer by multi-region sequencing. A rare mixed invasive ductal carcinoma and spindle cell carcinoma was subject to multi-region sequencing, where we found evidence of divergent evolution from a common ancestor. AC, adriamycin+cyclophosphamide; DFS, disease-free survival; IDC, invasive ductal carcinoma; LN, lymph node; PD, progressive disease; PFS, progression-free survival.

7. Supplementary test for the accurate diagnosis of rare BC subtype

FCP can aid in diagnosing rare types of BC, like sarcomas, by detecting specific genetic alterations. We have experienced a 53-year-old woman who was initially diagnosed with metaplastic carcinoma with heterologous mesenchymal differentiation by histopathologic analysis (S6 Fig.). FCP was performed and revealed mutations in SUZ12 and NF1. Complete loss of H3K27me3 was confirmed by IHC, leading to a revised diagnosis of malignant peripheral nerve sheath tumor (MPNST).

8. Broadening the therapeutic options including matched therapy and clinical trials

Of 522 patients, 53 (10.2%) received matched therapies or participated in clinical trials based on biomarkers identified from NGS data across various lines of treatment (Table 3). Six patients (1.1%) were enrolled in two trials. Eligible alterations for matched therapy and trials included BRCA1 or BRCA2 mutations (3.6%), ERBB2 amplification (2.7%), PIK3CA mutation (2.1%), high TMB (1.0%), alterations of homologous recombination (HR)–related genes (0.8%), FGFR1 alteration (0.6%), AKT1 mutation (0.4%), and ERBB2 mutation (0.2%). Median PFS and OS starting from day 1 of treatment in this population were 4.7 months and 39.2 months, respectively (Fig. 6A and B).

Table 3.

Matched therapy and clinical trial enrollment guided by clinical sequencing

Fig. 6.

Matched therapy guided by FiRST Cancer Panel. (A, B) Progression-free survival and overall survival (OS) from the day 1 of the matched therapy are presented. (C) Better OS from metastasis diagnosis was observed in patients who received matched therapy compared to those who did not. (D) A remarkable response to poly(ADP-ribose) polymerase inhibitor (PARPi) was observed in a patient with somatic ATM mutation identified by FiRST Cancer Panel. C1D1, cycle 1 day 1; EOT, end of treatment; OS, overall survival; PFS, progression-free survival.

To determine if FCP-guided matched therapy improved survival in BC, we compared OS from the first diagnosis of metastatic disease in patients who underwent matched therapy versus those who did not. Notable differences in OS were observed between the groups (log-rank, p=0.022) (Fig. 6C), indicating that NGS utilization offers tangible benefits to BC patients. One example was a 56-year-old patient (case No. 372) with ER+/HER2– metastatic BC who experienced disease progression over 10 years despite various treatments. After undergoing FCP, a somatic ATM mutation was detected. Consequently, the patient entered a talazoparib trial and showed a partial response within three months (Fig. 6D), remaining progression-free for 16.3 months.

We further investigated the prognostic factors from the FCP results of metastatic BCs (S7 Fig). It was evident that the presence of a TP53 mutation correlated significantly with adverse outcomes of metastatic BCs (log-rank, p < 0.001). Among the TNBC population, mutations in HR-related genes were linked to a more favorable prognosis with marginal significance (log-rank, p=0.080). This trend might stem from the available therapeutic options for DNA damage repair pathway including PARP inhibitors among metastatic TNBCs with HR-related gene mutations compared to the rest (28.1% vs. 5.7%; Fisher’s exact, p=0.001).