Cohort characteristics and detected gene alterations

The cohort included 703 patients diagnosed with HR + HER2− MBC. In detail, the study included 509 patients (85%) with invasive ductal carcinoma (IDC) and 93 patients (15%) with invasive lobular carcinoma (ILC) (Table 1). The most common metastatic site was bone (548 patients, 78%), followed by liver (290 patients, 41%), lymph nodes (262 patients, 37.3%) and lung (221 patients, 31.5%) (Table 1).

Table 1 Clinicopathologic characteristics of the luminal-like MBC cohort.

Endocrine therapy (ET) was the most common previous treatment (452 patients, 74%), CDK4/6 inhibitors were the most frequent targeted therapy (322 patients, 52.8%), followed by mTOR inhibitors (105 patients, 17.2%) and PI3K inhibitors (38 patients, 6.2%) (Table 1).

Across the tested genes, PIK3CA, TP53 and ESR1 were the most altered. As expected, the most likely effect of the detected SNVs was gain of function (GOF) for PIK3CA and ESR1 and loss of function (LOF) for TP53 (Fig. 1A, B).

Fig. 1

Landscape plot of all detectable aberrations in ctDNA samples. A Incidence of the single aberrations [copy number variations (CNV), Fusions (Fus), deletions (Del), insertions (Ins), frameshift (FS), splicing variants (Spl), premature termination codons (PTC) and single nucleotide variation (SNV)] is represented on the left. The mutant allele frequency (MAF) of each mutation is shown in the middle. Effect [gain of function (GOF), loss of function (LOF) and switch of function (SOF)] and pathogenicity [yes, no, unknown (Ukn) and inconclusive (Inc)] of all the detected aberrations are show on the right. Histogram representing different frequency distribution of gene mutations across oncogenic pathways (B). Lollipop plot showing the distribution of ESR1 (C) and PIK3CA (D) mutations by their amino acid coordinates across ER and PI3K domains and across ESR1 and PIK3CA exon sequencies. Oncogenic mutations are highlighted

SNVs alterations were mainly observed in the PI3K (35%), P53 (32%), ER (28%), RAS (8%), RTK (7%) and cell cycle (5%) pathways. Copy number variations (CNVs) were mostly detected in the RTK (20%), cell cycle (15%), MYC (7%) PI3K (8%) and RAF (6%) pathways (Fig. 1B).

ESR1 mutations were detected in 166 patients (24%) and PIK3CA in 214 patients (30.5%) (Additional file 1: Table S1). The most common ESR1 pathogenic mutations among ESR1-mutated patients were found in codons 537 (52 patients, 31%), 538 (34 patients, 20%), 536 (14 patients, 8%) and 380 (12 patients, 7%) (Fig. 1C, Additional file 1: Table S1). The most common PIK3CA codon variants were 1047 (68 patients, 32%), 545 (47 patients, 22%), and 542 (38 patients, 18%) (Fig. 1D, Additional file 1: Table S1).

Other pathogenic PIK3CA SNVs were observed in 28.5% (N = 61) of patients, 16.82% (N = 36) had more than one PIK3CA SNV. Among patients with ESR1 mutations, 33% (N = 54) had polyclonal alterations (Additional file 1: Table S1).

The top 10 genes were CNVs were detected were FGFR1 (12.1%), MYC (9.1%), CCND1 (8.8%), PIK3CA (8.3%), EGFR (6.8%), BRAF (3.8%), CDK6 (3.4%), RAF1 (3.3%), CCNE1 (3.3%) and KRAS (3.1%) (Additional file 1: Table S2).

ESR1 and PIK3CA codon variants are associated with different ctDNA alterations across oncogenic pathways

Specific codon MAF was tested across PIK3CA and ESR1 variants (Fig. 2). No significant differences were observed for ESR1 (Additional file 1: Fig. S1A). PIK3CA codon variants, in contrast, showed a statistically significant difference (P < 0.0001) (Additional file 1: Fig. S1B) with PIK3CA 1047 and 542 showing the highest codon MAF (Additional file 1: Fig. S1B).

Fig. 2

Heat-map showing the association of the main ESR1 and PIK3CA codon variants with concomitant ctDNA alterations in other oncogenic pathways. PIK3CA 1047 was associated with SNVs in the P53 pathway and CNVs in the PI3K, while PIK3CA 542 with CNVs in the PI3K pathway, SNVs in the RTK and SNVs in the RAS and PIK3CA 545 with SNVs in the P53 pathway. ESR1 537 was associated with SNVs in the ER pathway, CNVs in the MYC and SNVs in the RAF, ESR1 538 with SNVs in cell cycle pathway. Color intensity was proportional to Odds Ratio, while P value was described by size (the lower the P value, the bigger the circle). Syn, Synonymous; Ukn, Unknown; CNVs, Copy Number Variations; SNVs, Single Nucleotide Variations

The association between the main ESR1 and PIK3CA codon variants (Fig. 1C, D) and concomitant gene alterations was investigated according to oncogenic pathway classification.

After multivariable logistic regression, PIK3CA 1047 was significantly associated with SNVs in the P53 pathway and CNVs in the PI3K (respectively OR 1.96, 95%CI 1.13–3.39 P = 0.016 and OR 3.25, 95%CI 1.36–7.75 P = 0.008) (Fig. 2, Additional file 1: Table S3).

PIK3CA 542 was significantly associated with CNVs in the PI3K pathway, SNVs in the RTK and the RAS pathways (respectively OR 3.18, 95%CI 1.15–8.76 P = 0.025 and OR 3.87, 95%CI 1.36–10.98 P = 0.011, OR 3.01, 95%CI 1.03–8.77 P = 0.044) (Fig. 2, Additional file 1: Table S3).

PIK3CA 545 was significantly associated with SNVs in the P53 pathway (OR 2.96, 95%CI 1.54–5.66 P = 0.001) (Fig. 2, Additional file 1: Table S3).

ESR1 537 was significantly associated with SNVs in the ER pathway, CNVs in the MYC pathway and SNVs in the RAF pathway (respectively OR 3.00, 95%CI 1.25–7.17 P = 0.014 and OR 2.74, 95%CI 1.20–6.23 P = 0.017, OR 5.08, 95%CI 1.15–22.40 P = 0.032) (Fig. 2, Additional file 1: Table S3).

ESR1 538 was significantly associated with SNVs in the cell cycle pathway (OR 5.27, 95%CI 1.82–15.30 P = 0.002) (Fig. 2, Additional file 1: Table S3).

No concomitant alterations were confirmed after multivariable analysis for ESR1 380, although a numerical difference was highlighted for SNVs in the P53 pathway (Fig. 2, Additional file 1: Table S3).

No associations were observed for ESR1 536.

Alterations in oncogenic pathways and ESR1 / PIK3CA codon variants are differentially associated with sites of metastasis

The association between ESR1/PIK3CA codon variants and alterations in oncogenic pathways were then investigated across the main metastatic sites. Correction for number of lines was applied.

After multivariable logistic regression, ESR1 537 alterations were significantly associated with bone and lung involvement (respectively OR 3.15, 95%CI 1.08–9.23, P = 0.036, OR 1.89, 95%CI 1.01–3.52, P = 0.046), while ESR1 538 alterations were associated with liver metastases only (OR 3.06, 95%CI 1.29–7.29 P = 0.012) (Fig. 3, Additional file 1: Table S4).

Fig. 3

Association of ESR1 and PIK3CA codon variants and alterations in oncogenic pathways with different involvement of the main metastatic sites. (lung, liver, soft tissue, CNS, lymph nodes and bone). The number of concomitant detected aberrations (NDA) of ESR1 and PIK3CA across different metastatic sites was represented next to each metastatic site

PIK3CA 1047 alterations were associated with bone metastases (OR 2.68, 95%CI: 1.02–7.05, P = 0.046) (Fig. 3, Additional file 1: Table S4). PIK3CA 542 alterations were associated with soft tissue and lymph nodes (respectively OR 2.56, 95%CI: 1.14–5.74, P = 0.022 and OR 2.32, 95% CI 1.15–4.68, P = 0.018) (Fig. 3, Additional file 1: Table S4). PIK3CA 545 alterations were associated with soft tissue involvement (OR 3.6, 95%CI 1.76–7.35, P < 0.001) and less associated with liver metastases (OR 0.3, 95%CI 0.13–0.72, P = 0.007) (Fig. 3, Additional file 1: Table S4).

SNVs and CNVs in the RTK pathway were significantly associated with liver metastases (respectively OR 2.47, 95%CI: 1.04–5.85, P = 0.04 and OR 2.12, 95% CI 1.15–3.94, P = 0.017) (Fig. 3, Additional file 1: Table S4). SNVs in the WNT pathway were associated with CNS metastases (OR 4.91, 95% CI 1.17–20.54, P = 0.029) (Fig. 3, Additional file 1: Table S4). SNVs in the ER pathway were less represented in patients with soft tissue involvement (OR 0.2, 95% CI: 0.05–0.86, P = 0.03) (Fig. 3, Additional file 1: Table S4).

The number of concomitant ESR1 and PIK3CA mutations was then analyzed across metastatic sites.

Patients with lung, liver and bone metastases had a significantly higher number of ESR1 mutations (respectively P = 0.0244, P < 0.0001 and P < 0.0001) (Fig. 3), while PIK3CA mutations were significantly higher in patients with lung, CNS, and bone metastases (respectively P = 0.0167, P = 0.0011 and P < 0.0001) and significantly lower in patients with soft tissue involvement (P = 0.0020) (Fig. 3).

ESR1 mutations impact on overall survival together with alterations in PI3K, MYC, RAS and P53 pathways

The prognostic impact of ESR1 and PIK3CA codon variants and number of concomitant mutations was investigated in terms of OS (Fig. 4). Although the detection of ESR1 or PIK3CA mutations had a significantly unfavourable impact on OS (respectively P < 0.0001 and P = 0.0410) (Fig. 4A, B), no differential impact was observed across codon variants (P = 0.3108 and P = 0.3450) (Fig. 4A, B) or the number of concomitant ESR1 or PIK3CA SNVs (P = 0.9414 and P = 0.1301) (Fig. 4C, D). Similar results were observed for patients previously treated with CDK4/6i both for the overall impact of the ESR1 and PIK3CA mutational status (respectively P = 0.0310 and P = 0.0009) and multiple concomitant SNVs (respectively P = 0.8074 and P = 0.3443) (Fig. 4E, F). On the other hand, in patients not previously exposed to CDK4/6i only ESR1 had a significant impact on OS (P = 0.0206) (Fig. 4G, H).

Fig. 4

Kaplan–Meier plots for the impact on overall survival (OS) of ESR1 and PIK3CA mutations. The detection of ESR1 or PIK3CA mutations had a significantly unfavorable impact on OS, however, no differential impact was observed across ESR1 or PIK3CA codon variants (A, B). Similarly, no differential prognostic impact was observed between number of concomitant ESR1 or PIK3CA alterations in the overall population (C, D) and according to previous CDK4/6i exposure (E–H). CNVs, Copy Number Variations; SNVs, Single Nucleotide Variations

Similar results were observed in terms of PFS in patients treated with endocrine therapy (Additional file 1: Fig. S2A–D), apart from a numerical difference according to number of PIK3CA alterations within patients with PIK3CA-mutated MBC (P = 0.0753) (Additional file 1: Fig. S2D).

After multivariable analysis corrected for previous treatment with CDK4/6i, number of lines, lung, liver, bone, and soft tissue involvement, the prognostic impact of ESR1 380 was confirmed (HR 2.48, 95%CI 1.01–6.07, P < 0.048), together with RAS pathway SNVs (HR 1.96, 95%CI 1.16–3.33, P = 0.013), P53 pathway SNVs (HR 1.38, 95%CI 1.17–3.17, P = 0.034), MYC pathway CNVs (HR 1.91, 95%CI 1.08–3.38, P = 0.027), and PI3K CNVs (HR 2.27, 95%CI 1.16–4.47, P = 0.017) (Table 2). Number of lines, liver, bone and soft tissue involvement had an independent impact in terms of OS (Table 2).

Table 2 Association of ESR1 and PIK3CA codon variants and alterations in oncogenic pathways with overall survival (OS).