Stability and variability of molecular subtypes: comparative analysis of primary and metastatic triple-negative breast cancer

Real-world cohorts of patients with early-stage TNBC (eTNBC)

A total of 465 consecutive Chinese patients who were diagnosed with malignant TNBC and treated in the Department of Breast Surgery at Fudan University Shanghai Cancer Center (FUSCC) from 1 January 2007 to 31 December 2014 were enrolled according to the following defined criteria: 1) female patients diagnosed with unilateral invasive ductal carcinoma with an ER, PR, or HER2 phenotype using < 1% positively stained cells as the cut-off value for ER/PR negativity based on immunohistochemistry (IHC)²²; 2) no evidence of distant metastasis at the time of diagnosis; and 3) availability of a complete medical history for further analysis. Patients with breast carcinoma in situ or inflammatory breast cancer were excluded. Follow-up evaluations within this early-stage cohort of patients were completed on 6 December 2023 and the median follow-up was 68.1 months [interquartile range (IQR), 46.8–100.9 months].

Real-world cohorts of patients with mTNBC

We retrospectively recruited patients with mTNBC treated at FUSCC from October 2018 to November 2020. The eligibility criteria for patients were as follows: (1) female patients diagnosed with metastatic breast carcinoma with an ER, PR, or HER2 phenotype by the Department of Pathology at FUSCC; and (2) a sufficient number of tumor biopsies obtained for IHC staining (AR, CD8, and FOXC1)¹⁹ to classify the tumors into four subtypes (LAR, IM, BLIS, and MES)¹⁸ based on a central pathologic examination of tumor specimens performed by the Department of Pathology at FUSCC; and (3) complete medical history obtained for further analysis. Patients with multiple primary tumors or solitary lesions who had undergone radiotherapy treatment were excluded from this study. The clinicopathologic characteristics included age, number of metastases, treatment history, disease-free interval (DFI), duration of first-line therapy, and ER, PR, HER2, and Ki-67 status. Follow-up evaluations within this cohort of patients were completed on 6 December 2023 and the median duration of follow-up was 38.9 months (IQR, 27.1–61.1 months).

All tissue samples included in the present study were approved by the FUSCC Ethics Committee [Approval No. 2011226-17 (2010-ZZK-31)] and each patient provided written informed consent.

Immunohistochemical staining

We performed immunohistochemical staining on 4 paraffin-embedded sections (4 μm thick) of tumor specimens to evaluate the expression of AR, CD8, and FOXC1. We used the following primary antibodies: anti-AR (ab133273, 1:200 dilution; Abcam, Cambridge, UK); anti-CD8 (SP57, undiluted; Ventana, Tucson, Arizona, USA); and anti-FOXC1 (ab227977, 1:500 dilution; Abcam). All four antibodies used were rabbit monoclonal antibodies and were recommended by the manufacturers for immunohistochemical experiments involving paraffin-embedded sections. The tumors were classified into IHC-based subtypes according to the staining results. Detailed descriptions of the IHC-based classification methods were published previously¹⁹.

Statistical analysis

The data distribution was characterized by frequency tabulation and summary statistics. The relationships between the TNBC subtypes and clinicopathologic parameters were determined with Pearson’s χ² test or Fisher’s exact test, as indicated. Relapse-free survival (RFS) was defined as the time from diagnosis to first recurrence or a diagnosis of contralateral breast cancer. OS was defined as the time from diagnosis of mTNBC to death from any cause. The objective response rate (ORR) was defined as the percentage of participants with an objective complete response (CR) or partial response (PR) according to RECIST (v1.1) among all participants. Patients without events were censored from the date of the last follow-up evaluation. OS and RFS were compared among the four TNBC subtypes using the Kaplan‒Meier method and a Cox regression model. The Wilcoxon test was used to compare continuous variables. All P values were two-tailed, and a P < 0.05 was considered to indicate statistical significance. R software (version 4.3.2; The R Project for Statistical Computing, https://www.r-project.org/) was used for calculations and analyses.

Research cohorts and study design

To unravel the heterogeneity of TNBC and explore precision treatment strategies, we established molecular subtypes based on a multiomics cohort of primary TNBC tumors. The TNBC populations were divided into the following four subtypes: LAR; IM; BLIS; and MES¹⁸. To deepen our understanding of the clinical characteristics of this novel TNBC classification system and provide valuable guidance for clinical practice, we assembled a real-world clinical cohort of TNBC patients and conducted a comprehensive analysis of TNBC subtypes from a temporal perspective (Figure 1A). These cohorts included an eTNBC cohort comprised of 465 individuals and a metastatic cohort comprised of 415 patients with TNBC treated at FUSCC. Subtypes were determined based on primary tumors in the eTNBC cohort, while subtypes were determined based on metastatic lesions in the mTNBC cohort. Moreover, we retrospectively collected 40 paired samples of primary tumors and paired metastatic lesions within the mTNBC cohort.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Schematic of the study and overview of the cohorts. (A) Schematic of the study; (B) Clinical features of our early-stage cohort; T stage, patient age, Ki-67 status, histology, and LN status at first diagnosis are depicted; (C) Distribution of relapse-free intervals in the early-stage cohort (n = 68); (D) Metastatic sites of the patients in the early-stage cohort (n = 126); (E) Metastatic sites in the metastatic cohort (n = 1267); (F) Distribution of disease-free survival in the metastatic cohort (n = 341); (G) Kaplan‒Meier curve for overall survival in the metastatic cohort (n = 413); two patients were excluded due to loss of follow-up after the initial diagnosis; n, number of patients; TNBC, triple-negative breast cancer; FUSCC, Fudan University Shanghai Cancer Center; LN, lymph node; CI, confidence interval; see Figure S1.

We conducted a comprehensive analysis of eTNBC and mTNBC subtypes, which revealed important insights into the stability and variability of molecular subtypes. The clinical dataset, which consisted of 880 TNBC patients, included subtype information, clinical data (age and metastatic sites), pathologic information (Ki-67 and lymph node status), treatment history (chemotherapy and radiotherapy), and prognostic information (RFS and OS). In the current study the previously published RFS data for eTNBC subtypes were updated, and for the first time, OS data are presented for each mTNBC subtype in a real-world cohort.

Poor outcomes of primary and metastatic TNBC patients in a real-world cohort

The median duration of follow-up evaluations in the eTNBC cohort was 68.1 months (IQR, 46.8–100.9 months). Most tumors [91.6% (426/465)] were invasive ductal carcinomas (IDCs) and 40.3% (186/462) of the TNBCs were lymph node-positive (Figure 1B). As of 6 December 2023, a total of 68 patients with TNBC had recurrences (Figure 1C), 69.1% (47/68) of whom relapsed within 3 years. The most common sites of metastasis were bone [47.1% (32/68)], lung [44.1% (30/68)], and lymph nodes [42.6% (29/68); Figure 1D].

The median duration of follow-up evaluations in the mTNBC cohort was 38.9 months (IQR, 27.1–61.1 months) and the median DFI was 17.0 months [95% confidence interval (CI), 15.2–18.0; Table S1]. Of the mTNBC patients, 86.5% (359/415) were < 60 years of age and 85.8% (356/415) of the patients received adjuvant chemotherapy. Additionally, 56.4% (234/415) of the TNBC patients had a heavy tumor burden at the time of enrollment with ≥ 2 metastatic sites (Table 1). Lymph nodes were the predominant biopsy site in this cohort (Figure S1A), possibly because of the frequent involvement as the primary site for initial metastasis (Figure S1B). The most common sites of metastasis were lymph nodes [75.2% (312/415)], bones [49.6% (206/415)], and chest wall [49.6% (206/415); Figure 1E]. Detailed clinical information for the mTNBC cohort is shown in Table 1. The median DFS of patients with mTNBC was 17.0 months (95% CI, 15.2–18.1 months; Figure 1F). Importantly, the median OS in these patients was only 27.2 months (95% CI, 24.4–30.2 months; Figure 1G). The above data demonstrated that patients with TNBC have a poor prognosis, which highlights the urgent need for further research to improve patient outcomes.

Stability of molecular subtypes (excluding the MES subtype) between primary and metastatic TNBC

The molecular classification of primary TNBC tumors involves full exploration of the heterogeneity of the tumors and serves to guide clinical treatment for improving patient prognosis^18,20. We conducted a comprehensive longitudinal study to determine the clinical characteristics and clinical value of TNBC molecular subtypes. According to a previously established IHC-based classification method, we performed IHC staining and classification in the TNBC primary and metastatic cohorts, followed by a comparative analysis (Figure 2A). The molecular subtypes in the eTNBC cohort were distributed as follows: LAR (25.1%); IM (19.6%); BLIS (38.2%); and MES (17.1%). Proprotions of patients in the mTNBC cohort with the LAR, IM, BLIS, and MES subtypes were 34.9%, 13.5%, 42.4%, and 9.2%, respectively. Notably, compared with patients in the primary cohort, the proportion of patients in the metastatic cohort was significantly greater for the LAR and BLIS subtypes, while the proportion of patients in the MES and IM subtypes was lower (P < 0.001; Figure 2B). These findings indicated that the original TNBC molecular subtypes were present in both primary lesions and metastases, albeit in varying proportions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Stability of molecular subtypes (excluding the MES subtype) between primary and metastatic TNBC. (A) Schematic of the TNBC classification system; samples were taken from primary breast cancer lesions in the early-stage and metastatic cohorts; (B) Comparison of TNBC subtype distributions in the early-stage and metastatic cohorts; P values were calculated using the chi-squared test; (C) Schematic of the TNBC classification system; samples were taken from primary tumors and paired metastases in the metastatic cohort; (D) Agreement between primary and metastatic TNBC subtypes; (E) Sankey plot showing the molecular subtype switch between the primary and paired metastatic sites; (F-H) Protein expression of AR (F), CD8 (G), and FOXC1 (H) in primary and paired metastatic TNBC tissues determined by immunohistochemistry; representative images of AR, CD8, and FOXC1 staining are shown; P values were calculated using a paired Wilcoxon test; TNBC, triple-negative breast cancer; LAR, luminal androgen receptor; IM, immunomodulatory; BLIS, basal-like immunosuppressed; MES, mesenchymal-like; see Figure S2.

To further investigate the clinical characteristics of the different molecular subtypes involved in disease progression, we obtained paired samples of primary and metastatic lesions from 40 TNBC patients and classified the lesions (Figure 2C). The distribution of the different molecular subtypes at each biopsy site is shown in Figure S2. Overall, 77.5% (31/40) of the patients maintained consistent molecular subtypes during disease progression, with LAR, IM, and BLIS concordance rates of 80.0%, 100.0%, and 78.6%, respectively, in the paired samples (Figure 2D and 2E). Furthermore, we performed an analysis of key molecules associated with each subtype. Our findings revealed that the levels of AR, CD8, and FOXC1 (key molecules in the subtyping system) protein expression remained stable throughout disease progression and showed no significant changes (Figure 2F–H). Surprisingly, the consistency of the MES subtype was poor because most patients transitioned to other subtypes upon recurrence or metastasis. Furthermore, the transition process was unidirectional with minimal evidence of other subtypes converting to the MES subtype during disease progression (Figure 2D and 2E).

In general, molecular subtype stability was observed between primary and metastatic TNBC except for the MES subtype. The subtype of TNBC metastatic lesions can be predicted by the molecular subtype of the primary TNBC among TNBC patients with LAR, IM, or BLIS. Furthermore, additional research is warranted to determine the predisposition of patients with the MES subtype during subtype transition.

A precision treatment strategy guided by the MES subtype of target lesions improves treatment efficacy

We conducted the FUTURE trial (ClinicalTrials.gov identifier: NCT03805399), a phase II umbrella trial, to evaluate the potential efficacy of subtype-based precision strategies in patients with mTNBC. As of 31 March 2022, a total of 15 mTNBC patients with the MES subtype were enrolled in the FUTURE trial. Among the 15 patients, 12 received precision treatment specific to the MES subtype, for which efficacy data are available. Among the 12 patients, 8 were diagnosed with the MES subtype based on IHC assessment of the target lesion (matched group), while 4 patients were diagnosed with the MES subtype based on the primary tumor or non-target lesions (unmatched group; Figure 3A). Interestingly, compared to the patients in the unmatched group, the patients in the matched group had longer progression-free survival (Figure 3B) and achieved an ORR of 50% (Figure 3C). None of the patients in the unmatched group achieved an ORR (Figure 3C). Furthermore, patients in the unmatched group experienced faster disease progression and a shorter duration of response (Figure 3D and 3E). Typical imaging examples of the matched and unmatched groups are shown in Figure 3F and 3G, respectively. In general, a precision treatment strategy guided by the MES subtype of target lesions improved treatment efficacy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

A precision treatment strategy guided by the MES subtype of target lesions improves treatment efficacy. (A) A sketch map of the FUSCC subtype-based precision strategy and grouping method; (B) K‒M curves for progression-free survival in patients with the MES subtype with matched or unmatched target lesions.; (C) Proportion of confirmed responses in the matched and unmatched MES subtype groups; (D) Longitudinal change in the sum of the longest target lesion diameters from baseline among patients with matched and unmatched MES subtypes; one patient was excluded due to the emergence of new metastases before the initial evaluation of target lesions. (E) Durability of treatment among patients with matched and unmatched MES subtypes; (F) Example of patients in the matched group. Yellow arrows indicate metastatic lesions. A 52-year-old woman with mTNBC was identified with the MES subtype based on a liver lesion biopsy. Baseline imaging was performed on the liver lesion, which achieved a PR after the initiation of therapy. Baseline images of intrahepatic diffuse metastases and typical images of tumor regression during treatment are shown; (G) Example of patients in the unmatched group. Yellow arrows indicate metastatic lesions. A 23-year-old woman with mTNBC was identified with the MES subtype based on primary tumor biopsy. Baseline imaging was performed on the lung lesion, which continued to progress after the initiation of therapy. Baseline images of diffuse lung metastases and typical images of tumor progression during treatment are shown. RECIST, response evaluation criteria in solid tumors; CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease; AE, adverse event.

The original TNBC subtypes exhibit prognostic value in eTNBC and mTNBC patients

The differences in subtype proportions between primary and metastatic TNBC cohorts imply diverse clinical outcomes linked to specific subtypes. The previously published RFS data for eTNBC subtypes were updated and the long-term OS data from a large-scale cohort of patients with mTNBC subtypes were presented for the first time. Overall, patients in the IM subgroup of the eTNBC cohort exhibited longer RFS than patients in the MES or LAR subgroup [MES vs. IM, hazard ratio (HR): 3.2, 95% CI: 1.2–8.2; P = 0.02 and LAR vs. IM, HR: 2.7, 95% CI: 1.1–7.0; P = 0.04; Figure 4A and 4B]. A significant difference in OS was detected among the four subtypes in the mTNBC cohort (P = 0.001; Figure 4C). The molecular subtype was shown to be an independent prognostic factor for mTNBC based on multivariate Cox analysis and the IM subtype exhibited a longer OS than the other subtypes (35.4 months, 95% CI: 20.9–49.9 months; Table S2). The median OS times of patients with the LAR and MES subtypes were 29.6 (95% CI: 22.6–36.5 months) and 25.4 months (95% CI: 20.3–30.5 months; Table 2), respectively. BLIS tumors exhibited the worst OS (19.9 months, 95% CI: 14.6–25.2 months; Table 2) and were associated with a significantly poorer prognosis than IM tumors (BLIS vs. IM, HR: 2.0, 95% CI: 1.3–3.1; P = 0.001; Figure 4D), which corroborated previous literature on the poor prognosis of BLIS patients¹⁸. Additionally, the discovery of the inferior prognosis associated with BLIS tumors shed light on the observation that patients with the BLIS subtype developed metastasis at a younger age (< 60 years) and had a greater tumor burden (multiple metastases) compared to patients with other subtypes (Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

The original TNBC subtypes exhibit prognostic value in both eTNBC and mTNBC patients. (A) Kaplan‒Meier curves for relapse-free survival grouped by eTNBC subtypes; (B) Association of eTNBC subtypes with relapse-free survival assessed by Cox regression analyses; the hazard ratios are shown with 95% confidence intervals; the data involved in the Cox analysis were statistically tested to meet the proportional hazards assumption; (C) Kaplan‒Meier curves for overall survival grouped by mTNBC subtypes; (D) Association of mTNBC subtypes with overall survival assessed by Cox regression analyses; hazard ratios are shown with 95% confidence intervals; the data involved in the Cox analysis were statistically tested to meet the proportional hazards assumption; (E) Distribution of metastatic sites in mTNBC subtypes; (F) Proportion of brain metastases among all metastatic sites in each mTNBC subtype; P values were calculated using Fisher’s exact test; (G) Examples of MES-subtype patients with brain metastasis; yellow arrows indicate metastatic lesions; LAR, luminal androgen receptor; IM, immunomodulatory; BLIS, basal-like immunosuppressed; MES, mesenchymal-like; see Figures S3 and S4.

In addition, we examined the trends in the expression of three molecules (AR, CD8, and FOXC1) among the four subtypes in both the primary and metastatic TNBC cohorts. FOXC1 was highly expressed in the metastatic cohorts of all subtypes except the MES cohort, suggesting that high expression of FOXC1 indicates poor prognosis in patients with the LAR, IM, and BLIS subtypes (Figure S3A–D). Organotropic metastasis of patients with TNBC subtypes was then examined and an intriguing phenomenon was observed. Despite representing a relatively small proportion of patients in the overall cohort, patients in the MES subgroup exhibited a significantly greater proportion of brain metastasis, accounting for 25.0% of the population (Figure 4E and Figure S4). Furthermore, among the four TNBC subtypes, the MES subgroup had the highest incidence of brain metastasis (Figure 4F and 4G). These findings strongly suggest the propensity of the MES subtype for brain metastasis and highlight the importance of enhancing brain surveillance.