Selection of patients with IBTR for clinical and genomic analysis
A total of 2,770 patients with breast cancer underwent surgery from 1999 to 2018 in our hospital; total mastectomy in 889 patients (32%) and breast conserving surgery (BCS) in 1,881 (68 %) (Fig. 1). After BCS, 98% of patients were treated by irradiation of 50 Gy for whole breast and in cases with residual tumor found at margin of resected specimen had additional 10 Gy for the tumor bed. Furthermore, adjuvant systemic therapies were added according to the guidelines , . Of 1,881 cases with BCS, the ipsilateral breast tumor recurrence (IBTR) had developed in 52 (2.8 %) cases with median follow of 7.7± 4.6 years (ranging 0.9 to 19.7 years).
We reviewed clinical features including age at developing PBC, TNM classifications, stage, histology, immunohistochemical subtypes and interval years from PBC to IBTR (Table 1). Among the 52 cases with IBTR, written informed consent for genomic analysis was obtained in 22 cases. We analyzed genomic profile of paired samples of primary breast cancer (PBC or T1) and recurrence (IBTR or T2) (Fig. 2 a, b).
Clinically defined residual recurrence (RR) and double primaries (DP)
First, we classified those IBTR whether they are clinical residual recurrence (cRR) of the primary tumors or de novo double primaries (cDP) according to the following gross anatomical and microscopic features. Gross anatomical concordance (AC) was judged, whether the IBTR developed in the same quadrant of the index lesion of PBC  .
Histological concordance (HC) was judged by the similarity of pathological findings and immunohistochemical subtypes between the PBC and IBTR . We judged IBTR as clinical residual recurrence (cRR) by belonging to both of AC and HC and the others were judged as clinical double primaries (cDP).
Genomically defined residual recurrence (gRR) and double primaries (gDP)
We classified IBTR into two groups by genome profiling, namely, genomic residual recurrence (gRR) and genomic double primary (gDP) by cancer panel analysis [11-14]. We defined gRR by the existence of identical mutations, especially driver oncogenic mutations, in both tissue of PBC and IBTR, whereas gDP was defined by the absence of shared mutations in both tissues. Correlations of clinically and genomically defined cRR/cDP and gRR/gDP were studied.
Genome analysis of primary breast cancer and IBTR ( 22 cases)
Preparation of samples
The tumor samples from PBC and IBTR were obtained from surgical resections or biopsies. A serial section of 10-μm-thick, formalin-fixed and paraffin-embedded (FFPE) tissue was stained with haematoxylin-eosin and then microdissected using an Arcturus XT laser-capture microdissection system (Thermo Fisher Scientific, Waltham, MA, USA) .
Peripheral blood samples were obtained and buffy coats were isolated following centrifugation of peripheral blood at 820 Í g at 25 °C for 10 min, and subsequently stored at −80 °C until required for DNA extraction. Total DNA was extracted from lymphocytes using the QIAamp DNA blood mini QIAcube kit (Qiagen, Hilden, Germany). The concentration of DNA was determined using a Nano Drop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA).
Targeted deep sequencing
For targeted deep sequencing analysis, as we previously reported , Ion AmpliSeq designer (Thermo Fisher Scientific) was used to design custom primers, which consisted of 2,863 primer pairs in two pools . These primers covered the exons of 53 breast cancer-associated genes (total 287,520 nucleotides) reported by the TCGA project, other studies, and the COSMIC database [18-20] (sTable 1).
Multiplex polymerase chain reaction (PCR) was performed with Ion AmpliSeq Library Kit Plus and two primer pools, as we previously described . PCR amplicons were partially digested with FuPa reagent and subsequently ligated to adapters with Ion Xpress Barcode Adapters. Adaptor-ligated amplicon libraries were purified using Agencourt AMPure XP reagents (Beckman Coulter, Brea, CA, USA). The library concentration was determined by quantitative real-time PCR using an Ion Library Quantitation Kit. Emulsion PCR and chip loading were performed on the Ion Chef with the Ion PI Hi-Q Chef kit. Sequencing was performed on the Ion Proton Sequencer (Thermo Fisher Scientific).
The sequencing data were processed using standard Ion Torrent Suite software running on a Torrent Server as described previously . Raw signal data were analysed using Torrent Suite. The pipeline included signalling processing, base calling, quality score assignment, adapter trimming, PCR duplicate removal, read alignment to human genome 19 reference (hg19), quality control of mapping quality, and coverage analysis. Variants calling and annotations were performed using an Ion Reporter Server System, and peripheral blood DNA was used as a control to detect variants in tumours using filtration of “confident somatic variants” in a Tumour-Normal pair pipeline. The minimum count for mutant allele reads was more than 10 and coverage depth was more than 20. Common single nucleotide polymorphisms were excluded from further analysis.
Annotation of the genome profiling
We compared genome profiling of paired samples of PBC and IBTR.
The significant mutated gene (SMG) which was defined with allelic fraction more than 1 % and the number of driver genes were defined according to the oncoKB (Precision Oncology Knowledge Database, Memorial Sloan Kettering Cancer Center, USA) .
Elucidation of driver gene and pathway to promote recurrence
We tried to elucidate whether any genes and pathways which might be significantly altered might play an important role for the development of recurrences after many years.
Statistical analysis was performed by t test or χ2 test as appropriate using Stat Mate (Atoms, Tokyo, Japan).