Clinical characteristics of the patients and target-capture sequencing
The main clinical characteristics of the 27 MBC patients included in the study are shown in Table 1. The mean age at diagnosis was 51.30 years (range, 33–68 years). Most patients had infiltrating ductal carcinoma (clinical stage IV) with lymph node, bone, or hepatic metastasis. In total, 14 (51.8%) patients had a maximal tumor diameter> 3cm, and 11 (40.74%) patients had given birth three or more times. The proportions of ER(+), PR(+), and HER2(+)cases were 62.96%,59.26%, and 62.96%, respectively.
Table 1
Clinical characteristics of MBC patients.
Characteristics | | N (%) |
Diagnostic age (years) | Mean (rang) | 51.30 (33–68) |
Menarche age (years) | Mean (rang) | 14.33 (11–19) |
Stage | III | 1 (3.70%) |
| IV | 26 (96.30%) |
ER status | ER(+) | 17 (62.96%) |
PR status | PR(+) | 16 (59.26%) |
HER2 status | HER2(+) | 17 (62.96%) |
Menopause | YES | 11 (44.44%) |
Size of tumor | ༞3cm | 14 (51.85%) |
| ≤ 3cm | 13 (48.15%) |
Parturitions | ≥ 3 | 11 (40.74%) |
| ༜3 | 16 (59.26%) |
Therapeutic effect* | PR/SD | 16 (59.26%) |
| PD | 11 (40.74%) |
*PR: partial remission; SD: stable disease; PD: progressive disease |
After running an iterative algorithm with multiple databases and optimization by the NimbleGen Design portal, we selected a custom panel covering 119.20kb of the genome. The panel included 961 exons of 835 common driver genes distributed over all chromosomes. Details of the target-capture panel are presented in Table S1. DNA was successfully extracted from all 81 samples and qualified for target-capture sequencing. We obtained an average of 810.45 Mb (range 303.08–1424.64Mb) and 403.28Mb (range 183.51–789.27Mb) of high-quality data for the cfDNA and genomic DNA (gDNA) samples, respectively. The average sequencing depths for cfDNA and gDNA were 6799× (range 2543–11952×) and 3383× (range 1540–6621×), respectively.
Identification of ctDNA mutations and related genes
Some mutations originating from CH-related variants in lymphocytes can also be traced in cfDNA, which may interfere with the analysis of ctDNA. After the comparison and elimination of CH variants, we identified 1182 nonsynonymous mutations from all samples, including frameshift indels, stopgains, and single-nucleotide variants (SNVs). They were distributed in 419 genes on all chromosomes (Fig. 1a). All patients had mutations in FRG1, AQP7, and DNAJC11. Mutations were detected most frequently in FRG1, and the highest mutation burden was seen in MUC16 with 58 mutations. The top 20 genes most frequently mutated in the ctDNA are shown in Fig. 1b. Some typical cancer-related genes such as TP53, PIK3CA, MAPK3K1, KRAS, and PTEN were also included. The number of ctDNA mutations varied markedly between patients (range 38–171, mean 79.96).
To evaluate the influence of biological features by ctDNA mutations, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted for the mutated genes from three subsets (overall, baseline, and after chemotherapy). The top ten subjects were selected for further comparison and analysis. The analysis revealed that two, seven, and six genes in all three subsets were related to biological process (BP), cellular component (CC), and molecular function (MF), respectively, including regulation of cellular component size, cation channel complex, and calmodulin binding. There were more discrepancies among the three subsets according to BP enrichment than CC and MF (Figure S1).
On KEGG analysis, 96, 97, and 108 pathways were enriched in the baseline, after chemotherapy, and over all subsets, respectively. Among the top ten pathways, six were included in all three subsets (i.e., small cell lung cancer, EGFR tyrosine kinase inhibitor resistance, endometrial cancer, PI3K-Akt signaling pathway, endocrine resistance, and cholinergic synapse; Fig. 2). Most pathways were typical and indeed meaningful for cancer development, progression, or therapeutic response. The results revealed no significant discrepancies in overall enriched pathways among the three subsets.
Distribution of gene mutations and clinical characteristics
The patients were divided into different groups according to their clinical characteristics for further analysis. The results show that the ctDNA mutations were significantly associated with tumor size and HER2 status. Patients with tumors> 3cm carried more mutations than those with tumors≤ 3cm (92.29 vs. 66.69, respectively, P = 0.035), while HER2(+) patients carried fewer mutations than HER2(−) patients (68.65 vs. 99.20, respectively, P = 0.029). For single genes, the mutation profiles of ctDNA also exhibited discrepancies. We selected the top 20 genes for further analysis. The results show that mutations in AQP7 and PTEN were significantly increased in patients with a poor therapeutic effect showing progressive disease (PD) (P < 0.05). PIK3CA mutations occurred more frequently in patients with tumors> 3cm (P = 0.039). The same was also observed for DNAJC11, MAP3K1, and PGAP1 in patients with late menarche (≥ 14 years) and KRAS in older patients (age at diagnosis> 51 years, P < 0.05) (Table S2).
A multivariate regression analysis was conducted to examine the relations between mutation burden and clinical parameters (Table 2). The results show that ctDNA mutations were significantly associated with HER2 status. Patients with HER2(+) carried fewer mutations than HER2(−) patients (OR 0.02, 95% CI 0–0.62, P = 0.025). No significant associations were found for other characteristics.
Table 2. Multivariate regression analysis of the relationship between ctDNA mutations
and clinical parameters.
Factors | Coefficient (SE) | Adjusted OR | 95% CI | P value |
Diagnostic age | 0.12 | 0.92 | 0.73–1.16 | 0.475 |
Menarche age | 0.41 | 1.57 | 0.70–3.55 | 0.274 |
ER status, ER(+) vs ER(-) | 2.29 | 2.51 | 0.03-222.34 | 0.687 |
PR status, PR(+) vs PR(-) | 2.07 | 0.07 | 0-4.30 | 0.210 |
HER2 status, HER2(+) vs HER2(-) | 1.70 | 0.02 | 0-0.62 | 0.025 |
Menopause, YES vs NO | 2.02 | 1.48 | 0.03–77.85 | 0.847 |
Size of tumor, ༞3cm vs ≤ 3cm | 1.55 | 10.28 | 0.50-213.38 | 0.132 |
Parturitions NO. ≥3 vs ༜3 | 0.50 | 1.02 | 0.38–2.72 | 0.965 |
Therapeutic effect, PR/SD vs PD | 1.33 | 4.26 | 0.31–58.24 | 0.277 |
Dynamics of ctDNA mutations during chemotherapy
A total of 768 ctDNA mutations were detected in 302 genes at baseline, which decreased to 633 in 291 genes after three courses of treatment. The mean number of mutations for each patient also decreased after three courses of treatment (50.93 vs. 46.59, respectively). To examine the dynamic changes, we compared the distribution of ctDNA mutations in genes before and after chemotherapy (Fig. 3). The mutations reduced sharply after chemotherapy for MUC16, NCOR1, TTN, PIK3R1, and TP53. In contrast, more mutations were detected after chemotherapy for MYO6, FLNC, FMN2, CDH1, and RHO.
An analysis based on clinical information showed different mutation patterns in plasma DNA between baseline and after chemotherapy (Table S2). For baseline ctDNA, mutations in PTEN increased significantly in patients with PD compared to patients with a partial response/stable disease (PR/SD) (P = 0.028). More mutations were detected in gene PIK3CA in patients with tumors> 3cm (P = 0.0039), in MAP3K1 and PGAP1 in patients with late menarche (≥ 14 years, P = 0.006 and 0.008, respectively), and in KRAS in older patients (age > 51 years, P = 0.018). These associations were consistent in all samples, but not in samples obtained after treatment. In addition, fewer ctDNA mutations were found in AQP7 in PR(+) patients (P = 0.023), and in KRAS and PIK3CA in HER2(+) patients (P = 0.041 and 0.030, respectively). In ctDNA after chemotherapy, the number of mutations decreased significantly in TP53 in both HER2 (+) and PR(−) patients (P = 0.011 and 0.024, respectively). Mutations in LUC7L2 were also strongly associated with HER2(−) status (P = 0.020).
ctDNA mutations and clinical outcomes
A Kaplan–Meier analysis was performed to explore risk factors correlated with PFS, defined as the duration from sampling to first disease progression. The mean PFS was 537 days (range 324 to 823 days). All patients were stratified with the mean number of ctDNA mutations as the threshold. The results indicate that HER2 status, therapeutic effect, and number of ctDNA mutations exhibited significant associations with PFS (Fig. 4). HER2(+) patients had significantly longer PFS compared with HER2(−) patients (adjusted HR 0.26, P = 0.038), and PD was related to shorter PFS than PR/SD (adjusted HR 6.54, P = 0.038). Patients carrying more ctDNA mutations overall and in baseline plasma both had poor PFS rates (P < 0.001). No associations were found for other clinical characteristics and mutations in ctDNA after chemotherapy.