Study design and participants

FUTURE is a phase II, open-label, multicenter, umbrella trial evaluating the efficacy and safety of multiple precision treatments based on molecular subtype and tumor characteristics in patients with heavily pretreated metastatic TNBCs. Eligibility criteria included the following: (1) female patients diagnosed with metastatic breast carcinoma with an ER-negative, PR-negative, and HER2-negative phenotype (the IHC cutoff for ER/PR-negative status was less than 1% staining in nuclei, and HER2-negative status was defined as a score of 0 or 1 by IHC analysis or the absence of ERBB2 amplification by fluorescence in situ hybridization with an IHC score); (2) central pathologic examination of tumor specimens performed by the Department of Pathology at FUSCC; (3) an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 2; (4) at least one measurable lesion according to Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1; and (5) adequate hematologic function, hepatic function, and renal function. Patients with uncontrolled brain metastasis were excluded from enrollment. Full eligibility criteria are provided in the study protocol in Supplementary information, Data S1.

The trial was conducted in accordance with the Declaration of Helsinki and the Good Clinical Practice guidelines of the International Conference on Harmonization. The study protocol was approved by the institutional review board and ethics committee of FUSCC. The ethics committee reference number was 1807188-16. All patients provided written informed consent to participate in this study.

Procedures

Patients who had heavily pretreated metastatic TNBCs and experienced disease progression during or following almost all standard chemotherapies (anthracycline, taxane, platinum, vinorelbine, capecitabine, and gemcitabine) were screened in four centers, including FUSCC, Beijing Cancer Hospital, Liaoning Cancer Hospital and Institute and the First Hospital of Jilin University. Tumor biopsies were obtained to allow IHC staining (AR, CD8 and FOXC1) and NGS (Supplementary information, Table S7)23 in order to classify the tumors into four subtypes (LAR, IM, BLIS, MES),24 and each patient was then enrolled into one of seven arms: (A) pyrotinib (HER1/HER2/HER4 inhibitor) 400 mg orally once daily continuously plus capecitabine 1000 mg/m2 orally twice daily from day 1 to day 14 on a 21-day cycle for LAR subtype with ERBB2 somatic mutation or amplification; (B) AR inhibitor (SHR3680) 240 mg orally once daily backbone therapy for LAR subtype without ERBB2 somatic mutation or amplification; (C) anti-PD-1 antibody (SHR1210) 200 mg intravenously once every 2 weeks with nab-paclitaxel 100 mg/m2 intravenously on days 1, 8, and 15 in a 28-day cycle for IM subtype; (D) PARP inhibitor (SH3162) 150 mg orally twice daily plus VEGFR inhibitor (famitinib) 20 mg orally once daily for BLIS subtype with BRCA1/2 germline mutation; (E) anti-VEGF/VEGFR backbone therapy for BLIS subtype without BRCA1/2 germline mutation; (F) VEGFR inhibitor (famitinib) orally once daily continuously plus etoposide (VP-16) 50 mg orally once daily from day 1 to day 14 in a 21-day cycle for MES subtype without PI3K/AKT mutation; (G) mTOR inhibitor (everolimus) 10 mg orally once daily continuously with nab-paclitaxel by intravenous 100 mg/m2 on days 1, 8, and 15 in a 28-day cycle for MES subtype with PI3K/AKT mutation. Details are provided in Table 5 of Supplementary information, Data S1.

Bayesian predictive probability, sample size, and protocol modifications

Considering different degrees of enrollment efficiency and different population distributions among the treatment arms, we used the Bayesian predictive probability approach to lay out an adaptive design in the new protocol amendment (July 24, 2020). This study was originally designed to have 10–20 patients enrolled in each arm, bringing the estimated sample size to approximately 140 patients. Referring to the historical data of heavily pretreated TNBC patients after multiline chemotherapy,42 if three or more patients in each arm reached CR or PR, then the arm would be considered to have reached the efficacy boundary. Using Bayesian predictive probability, based on the number of patients who achieved objective response (CR + PR) in real time, each arm could be terminated independently according to futility or efficacy boundaries.43 Assuming that the reference objective response rate is p0 = 15%, the prior probability fits the beta distribution (0.05, 0.05). The final threshold value of 0.5 was adopted for the arm to achieve effectiveness, 0.1 was adopted as the threshold value for early termination due to ineffectiveness, and 0.9 was adopted as the threshold value for early termination due to effectiveness. Using Bayesian prediction probability, futility and efficacy boundaries were obtained, and the simulation results under different true values of ORR are shown in the Supplementary information, Data S1.

Due to the difficulty of enrollment and the promising efficacy observed in other arms in the interim analysis, arms A and G could be terminated early at fewer than 10 patients (more details in Supplementary information, Data S1). Arm C was expanded to a maximum of 41 cases based on the efficacy reported in the interim analysis. Considering a dropout rate of 10%, 46 patients needed to be enrolled.

Some modifications have been made in the new version of protocol, and we briefly summarized them as follows: (1) the sample sizes in each arm were set to be more flexible according to Bayesian prediction probability method, and the rationale behind it was explained as above, (2) CTCAE version 4.0 was updated to version 5.0, (3) bevacizumab or low dose VEGFR inhibitors have been explored due to the toxicity issues with the VEGFR inhibitors.

The study design allowed arms to be dynamic so that old arms could be eliminated when finished and new arms could be added. Notably, the investigators used FUTURE as a platform to test the safety and efficacy of potential new drug–biomarker combinations in heavily pretreated TNBC patients.

Outcomes

The primary endpoint was the ORR per investigator according to RECIST v1.1 using imaging at baseline and every two cycles until disease progression.44 A CR or a PR was confirmed with one sequential tumor assessment at least 4 weeks later. Secondary endpoints were PFS (defined as the interval from the start of treatment to disease progression or death from any cause, whichever occurred first, or last PFS assessment for patients alive without progression), OS (first study dose until death from any cause), disease control rate (DCR, proportion of patients who experience a best response of CR or PR or stable disease ≥ 8 weeks according to RECIST version 1.1), and safety and tolerability. For PFS analysis, death before the first progressive disease (PD) assessment was computed as progressed, death between adequate assessment visits was computed as progressed, and death after more than one missed visit was censored on the date of last documented nonprogression. Treatment discontinuation for undocumented progression was censored on the date of last assessment without progression. Safety evaluations included assessments of AEs and serious AEs (SAEs), laboratory safety evaluations, vital signs, and physical examination. AEs were assessed in accordance with the National Cancer Institute Common Terminology Criteria for AEs, version 5.0. For AEs with various grades, the maximum reported grade was used in the summary table.

Biospecimen collection, quality control, and processing

Tumor and matched blood DNA were isolated from tumor samples and peripheral lymphocytes using TGuide M24 (Tiangen, Beijing, China). Absorbance at 260 nm (A260) and 280 nm (A280) was measured to estimate the purity and quantity of the total DNA by a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA). The extracted DNA was considered suitable for subsequent experiments if the A260/A280 ratio was between 1.6 and 1.9.

Sequencing using the FUSCC-BC panel

Details on the sequencing protocol have been described previously.19 The FUSCC breast cancer (FUSCC-BC) panel was used in this study (Supplementary information, Table S7). Both tumor and matched blood samples were sequenced. A KAPA HyperPlus kit (Kapa Biosystems) and Illumina HiSeq X TEN platform (Illumina Inc., San Diego, CA, USA) were used during NGS sequencing. Each alteration identified by the pipeline was manually reviewed to ensure that no false positives were reported.

Genomic biomarker analysis

Somatic mutations were called from the tissue and blood BAM files using GATK4 Mutect2 with the default parameters. The VCF files were annotated using ANNOVAR. The variants and annotation results were transferred into Excel spreadsheets. Oncogenic signaling pathways were defined based on a previous study.45 In post hoc exploratory analyses, PFS for each cohort was analyzed by tissue somatic mutation status.19

TNBC organoid and cell line classification

TNBC organoids were subjected to IHC staining (AR, CD8 and FOXC1) for subtyping.23 We performed hierarchical clustering to determine the TNBC subtype of common TNBC cell lines based on the similarity of expression profiles between patients and cell lines (Supplementary information, Fig. S6a). Patient RNA-seq data along with TNBC subtype annotations were obtained from our previous study.18 Cell line RNA-seq data were derived from the Cancer Cell Line Encyclopedia database and the study of Gong and colleagues.46 Cell lines with specific TNBC subtypes (according to the consistency of the FUSCC TNBC subtype and Lehmann subtype20) were chosen for further experiments.

Cell proliferation assay

We used the human TNBC cell lines BT-549, HCC1187, HCC1599, Hs 578 T, MDA-MB-157, MDA-MB-436, and MFM223 (from ATCC). Cell proliferation assays were performed as previously described.47,48 Briefly, the cells of interest (1 × 103–3 × 103 cells per well) were seeded into 96-well plates overnight in 100 μL of complete growth medium and then treated with the indicated drugs for 5 days in triplicate. Cell viability was tested using the cell counting kit-8 (CCK-8) assay (Dojindo Molecular Technologies, Japan, CK04) according to the manufacturer’s instructions.

Organoid preparation and culture

We developed a biobank for organoid storage as previously described.46 Briefly, fresh breast cancer tissues were minced into small fragments using sterile scalpels. Tissues were digested and resuspended in 10 mL of TAC buffer, incubated for 3 min to remove red blood cells and passed through a 100 mm cell strainer (Corning). For passaging, 5 mL of harvesting solution (Trevigen, 3700-100-01) was used to digest the basement membrane extract, which was incubated on ice for 1 h. Subsequently, the organoids were centrifuged at 350× g for 5 min, washed in digestion buffer and spun down. Next, 3 mL TrypLE Express (Invitrogen) was added, and organoids were incubated at room temperature for 3 min, followed by mechanical dissociation to small cell clusters by pipetting. Organoids were passaged at a 1:2–3 dilution every 2–3 weeks.

Drug response test of TNBC organoids

Drug response testing of TNBC organoids was performed according to a previous paper.49 For organoid drug treatment, organoids in good condition were harvested and digested into single cells. Twenty-five microliters of organoid suspension was added to a cell-repellent black surface in clear bottom 384-well plates (Greiner 781976-SIN) with 1 × 103–3 × 103 cells per well and cultured for another 5–6 days before drug treatments. Organoids with ADCs were cultured for 2 weeks before testing for viability. Organoid cell viability was evaluated by a CellTiter-Glo 3D cell viability assay (Promega, G9683) according to the manufacturer’s instructions.

PD-L1 and Trop-2 IHC

Baseline PD-L1 expression in the FUTURE trial was assessed at a central laboratory and characterized according to the combined positive score (CPS) as reported previously.11 CPS was defined as the ratio of PD-L1-positive cells (tumor cells, lymphocytes, and macrophages) out of the total number of tumor cells multiplied by 100. Available tumor specimens were stained for Trop-2 by IHC as reported previously.37,50 Positivity required at least 10% of the tumor cells to be stained.

Statistical analysis

The primary efficacy analysis population was the intention-to-treat population, including all eligible patients enrolled in the study. Safety was analyzed in all patients who received at least one dose of the study medication. The ORR and DCR with 95% CI were calculated using the Clopper–Pearson method. PFS and OS with 95% CI were assessed using the Kaplan–Meier method. In post hoc exploratory analyses, PFS for each cohort was analyzed by tissue somatic mutation. The association between the HER2 IHC scores and TNBC subtypes was examined using the chi-square test and Fisher’s exact test. Two-tailed Student’s t-test, Wilcoxon’s test and Kruskal‒Wallis test were utilized to compare continuous variables where appropriate. All tests were two-sided, and P < 0.05 was regarded as statistically significant unless otherwise stated.

SPSS (version 20) and R (version 4.1.1) were used for statistical analysis. The full statistical analysis plan is available in the protocol.