The samples of 464 patients with breast cancer who underwent mastectomy between January 2013 and December 2014 at Itabashi Hospital, Nihon University School of Medicine (Tokyo, Japan) were obtained. Among these, 433 had not undergone preoperative chemotherapy and 14 had been diagnosed with combined lobular and ductal carcinoma (CLDC). All tumor samples underwent routine FFPE at the pathological diagnosis departments of our hospital. The pathological diagnoses were made according to the classification of breast tumors issued by the World Health Organization (WHO) (23,24). Tumors comprising 10–90% of special subtypes were defined as ‘mixed invasive breast carcinoma of no special type (IBC-NST) and special subtypes’ by the WHO in general. However, no distinct subtype or designation has been proposed for mixed IBC-NST and invasive lobular carcinoma. Such tumors have been given several designations, e.g., invasive ductulolobular carcinoma, invasive ductal carcinoma with lobular features, or ductal or lobular carcinoma (25). In the present study, the definition of mixed ductal-lobular carcinoma (MDL) given by McCart Reed et al (25) was used for carcinomas in which the ductal component constitutes at least 10% of the tumor and the lobular component constitutes ≥50%. It is also possible that multiple ductal and lobular lesions are present in the same breast. Therefore, the term, CLDC was used, which was defined by Tazaki et al (26) for breast carcinoma wherein lobular carcinoma coexists with ductal carcinoma on the same side of the breast. Of the 14 CLDC cases, two with tumors comprising lesions that contained sufficient DNA for analytical purposes (1 µg or more) following laser microdissection (LMD), were selected for the present study.

Case 1 comprised a main tumor in the outer cranial region, an MDL lesion consisting of invasive lobular carcinoma (ILC), lobular carcinoma in situ (LCIS), invasive ductal carcinoma (IDC) and ductal carcinoma in situ (DCIS1) (Fig. 1 and Tables I and II). An additional DCIS2 was also present in the outer caudal region. DCIS1 and DCIS2 were discontinuous and separate lesions. Flat epithelial atypia (FEA) was spread throughout the mammary gland. Skin and lymph nodes, which are extramammary non-neoplastic tissues (EMNT), were used as the control.

The samples from case 2 included IDC, ILC, LCIS and FEA. These lesions were located in close proximity to each other (Tables I and II). EMNT was obtained from lymph nodes.

A total of 8 samples were obtained for case 1 and five samples for case 2. Case 1 was a 46-year-old female, and case 2 was a 41-year-old female. Features of the patients, the samples and experiments are listed in Tables I and II.

The present study was performed in accordance with the stipulations of the Declaration of Helsinki and was approved by the institutional review board of the Nihon University School of Medicine Ethics Committee (approval nos. 147, 115 and 250-1). Written informed consent was obtained from all patients. The study was performed according to the Japanese national guidelines ‘Ethical Guidelines for Medical and Health Research Involving Human Subjects’ (the Ministry of Health, Labor and Welfare; Ministry of Education, Culture, Sports, Science and Technology; and Ministry of Economy, Trade and Industry).

FFPE tissue sections (10 µm) were prepared using a DIRECTOR slide (cat. no. 11505158; Leica Microsystems GmbH). Following deparaffinization, the sections were subjected to LMD using an LMD 6000 (Leica Microsystems) according to the protocol provided by the manufacturer. DNA was extracted from FFPE samples using a RecoverAll Total Nucleic Acid Isolation kit (Thermo Fisher Scientific, Inc.) with certain modifications, as described previously (27). DNA quality was determined by performing quantitative PCR (qPCR) for GAPDH using a TaqMan^® Gene Expression Assay kit (cat. no. Hs02758991_g1; Thermo Fisher Scientific, Inc.). As a quality check control, the ratio of sample DNA to frozen tissue DNA was calculated using the delta quantification cycle (ΔCq) method and the results are expressed as 2^−ΔΔCq, as described previously (27–29).

mtDNA polymorphisms are frequently observed in mutational hot spots, designated as hypervariable regions (HV) (30). A total of two HVs were located between bases 16,024-16,383 and 57–372 (30). For PCR, six sets of primers were designed, including forward (F1-F6) and reverse sequences (R1-R6) (Table III). PCR was performed on each sample with eight different primer combinations as follows: F1-R1, F2-R2, F3-R3, F1-R4, F4-R1, F3-R5, F5-R6 and F6-R3. The reaction was performed in 20 µl reaction buffer of Takara Ex Taq Hot Start Version (Takara Bio, Inc.), containing 1 ng of sample DNA, under the following conditions: Preheating at 94°C for 60 sec; 35 cycles of 98°C for 10 sec and 60°C for 30 sec for annealing/elongation, and a final extension at 72°C for 5 min. Subsequently, 10 µl of the reaction products were subjected to 2% agarose gel electrophoresis with GelRed (Biotium, Inc.). Sanger sequencing was performed as described below using the reaction products eluted from electrophoretic bands.

In the absence of a consensus regarding the designation of genetic alterations in mtDNA, definitions in the present study were as follows: A single nucleotide polymorphism (SNP) was defined as a single base difference between individuals, as reported in the National Center for Biotechnology Information (NCBI) reference sequence (https://www.ncbi.nlm.nih.gov/nuccore/NC_012920), where all tissues in each case were identical. A single nucleotide variation (SNV) was defined as a single nucleotide sequence mismatch in a specific harvested tissue, differentiated from SNP. A simple sequence repeat number variation (SSRNV) was defined as a variation of the number of one or several simple base repeat sequences, in a manner that is inconsistent with repetition in EMNT. It was confirmed that the number of SSRNV repeats was not a technical artifact caused by PCR previously (31).

SSRNVs were detected in HV (Mt57-372) and in non-HV (Mt516-525). Regarding SSRNVs of mtDNA HV (Mt303-310), altered repeat numbers were observed, with the coexistence of repeat numbers C8 and C7 in NB and C8 homologically observed in neoplastic lesions. This was designated as loss of variation (LOV). Regarding SSRNVs in non-HV (–525), the homological repeat of CA7 in normal tissue with acquired coexistence of CA6/7 in neoplastic lesions was observed, for which it was designated as a gain of variation (GOV).

DNA from 9 samples (case 1: EMNT, DCIS2, IDC and ILC; case 2: EMNT, NB, FEA, IDC and LCIS) (Tables I and II), were used for amplicon sequencing by an Ion AmpliSeq Comprehensive Cancer Panel (CCP) on an Ion Torrent platform (Thermo Fisher Scientific, Inc.) according to the manufacturers’ protocols with specific modifications as below. Although the manufacturer's protocol recommended to check the amount of DNA using Qbit, the current study checked it according to the method reported by Nakayama et al (27). The panel consists of ~16,000 primer pairs covering all exons of 409 cancer-associated genes (1.6 megabases of target sequences). DNA (120 ng) was used to prepare the libraries with IonXpress barcoded adapters. An Ion Proton Sequencer (Thermo Fisher Scientific, Inc.) was used for sequencing with the ion chip.

Sequencing results were analyzed using the Torrent Suite Software (Thermo Fisher Scientific, Inc.). Candidates for cancer-specific SNVs were selected by Tumor-Normal pair analysis version 4.2 of Ion Reporter (Thermo Fisher Scientific, Inc.) using the following filters: >100 total reads; >10% of mutation rate; and 0 reads of variant in a control non-mammary and non-tumor sample. Interpretation of the clinical significances of detected gene mutations was conducted by referring the oncology database of Precision Oncology Knowledge Base (OncoKB, http://www.oncokb.org/) and Catalog of Somatic Mutations In Cancer (COSMIC, http://cancer.sanger.ac.uk/cosmic).

SNVs detected by amplicon sequencing were validated using SYBR-Green allele-specific qPCR. A number of genetic mutations were detected in case 2; therefore, the highest and second-highest mutation rates were selected for validation analysis, resulting in four genes in case 2 being selected. The target genes of interest, which were validated with SYBR-Green allele-specific qPCR, are presented in Table SI. The small sample size in case 2 was accounted for by pre-amplifying the region targeted for verification using specific primer sets (Table SII). A total of 10 ng of sample DNA were allowed to react with 20 µl of Takara Ex Taq Hot Start Version (Takara Bio, Inc.) in the presence of two sets of case 2 primers (KMT2C and CDH1) as follows: Preheating at 94°C for 60 sec; 45 cycles of 98°C for 10 sec and 60°C for 30 sec as the annealing temperature; and a final extension at 72°C for 5 min. PCR was performed with two sets of case 2 primers (PIK3CA and TIMP3) as follows: Preheating at 94°C for 60 sec; 45 cycles of 98°C for 10 sec and 55°C for 30 sec as the annealing temperature; 72°C for 60 sec; and a final extension at 72°C for 5 min.

Allele-specific qPCR was performed using the Thunderbird SYBR qPCR Mix (Toyobo Life Science) with StepOnePlus (Thermo Fisher Scientific, Inc.) and two primer sets. Allele-specific primers for wild-type and mutant sequences, as well as single reversely-directed primers, were designed by rectifying melting temperatures to a similar temperature, based on the Nearest Neighbor's method for melting temperature calculation (32). PCR was performed in a 10-µl reaction by preheating at 95°C for 60 sec, followed by 40 cycles of 95°C for 15 sec and 60 to 65°C for 60 sec. Annealing temperatures for each primer set are presented in Table SI. The ratio of mutant allele to wild-type allele was calculated as 2^−ΔΔCq, in which ΔCq is calculated by subtracting the Cq value of wild-type PCR from that of mutant PCR. The mutant allele frequency was calculated as 2^−ΔCq/(1+2^−ΔCq).

The primer sets for the wild-type region of genes with mutations were prepared and compared based on a quantitative curve using SYBR green as the fluorescent probe. Mutation analysis was performed based on the differences in amplification efficiency. The primer sequences are presented in Table SI. The results were verified via SYBR Premix Ex Taq (Takara Bio, Inc.) according to the manufacturer's protocol. The mutant allele frequency was calculated as described in the previous paragraph.

PCR products of the amplified mtDNA D-loop region were purified using an Illustra ExoProStar 1-Step kit (Cytiva) and sequenced using the BigDye Terminator version 1.1 cycle sequencing kit (Thermo Fisher Scientific, Inc.). Following purification using the BigDye Xterminator kit (Thermo Fisher Scientific, Inc.), nucleotide sequences were determined using an ABI310 Genetic Analyzer (Thermo Fisher Scientific, Inc.).

A total of two insertion/deletion mutations, namely CDH1 (T>TA at chr16:68842455) of case 1 and KMT2C (AG>A at chr7:151860111) of case 2, detected by amplicon sequencing, were examined via Sanger sequencing. PCR and Sanger sequencing of CDH1 and KMT2C were performed using the same method that was described for mtDNA D-loop using the primer pairs presented in Table SII.

CDH1 and VPS45A were selected for verification of the loss of chromosome 16q and gain of chromosome 1q, respectively. RNase P, a reference assay probe, was used as an endogenous control. The Quant Studio™ 3D Digital PCR System (Thermo Fisher Scientific, Inc.) was used for qPCR. DNA samples (60 ng) were used for the copy number variation (CNV) assay along with the TaqMan^® Copy Number Assay probe reagent for CDH1 (assay ID, Hs05461677_cn; assay location, chr16:68800293; amplicon length, 88 bp) or VPS45A (assay ID, Hs02288810_cn; Thermo Fisher Scientific, Inc.; assay location, chr1:150117372; amplicon length, 86 bp), TaqMan^® Copy Number Reference Assay probe for RNaseP [positon, chr14q11.2(20811565) on NCBI build 37; Thermo Fisher Scientific, Inc.] and QuantStudio™ 3D Digital PCR Master Mix v.2 (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Distilled water was added to adjust the total volume to 15 µl. The negative control reaction, which was performed in a total volume of 16 µl, contained no DNA. This negative control sample was used for each round of thermocycling.

For each sample, 15 µl of each prepared reaction mixture was loaded onto a QuantStudio 3D™ Digital PCR 20K Chip Kit v.2 (Thermo Fisher Scientific, Inc.) using an automatic chip loader and the signal was amplified using the following thermocycling conditions: 95°C for 8 min; 45 cycles at 95°C for 15 sec and 56°C for 1 min; and a final extension step at 60°C for 2 min. Amplification followed by chip imaging and secondary analysis was performed using the QuantStudio 3D Analysis Suit Software (version 3.0.2.2; Thermo Fisher Scientific, Inc.).

The results of the analysis yielded the copy numbers of target genes (CDH1 and VPS45A) and a reference gene (RNAseP) per microliter of each sample, which were used to calculate the ratio of target genes to a reference gene. These calculations were based on the theory that loss of chromosome 16q and gain of 1q resulted in haploid and triploid copy numbers per cell, respectively. The copy number of the target genes (copies/µl) was divided by that of the reference gene (copies/µl) and the obtained ratios of CDH1 or VPS45A to RNAseP in each sample were corrected, so that the ratios of CDH1 or VPS45A to RNAseP of EMNT were 1. Usually, a normal cell has two copies of CDH1 and RNAseP. A single 16q-loss cell has only one copy of CDH1 and two copies of RNAseP. When the total sum of 16q-loss cells and normal cells is 100%, the proportion of 16q-loss cells may be calculated by solving the two sets of simultaneous equations. The same calculations were performed as for 1q gain cells with normal cells having two copies of VPS45A and two copies of RNAseP, while the 1q triploid cells had three copies of VPS45A and two copies of RNAseP.

The results indicated that mtDNA SNPs were present at 13 and 11 sites in the D-loops of cases 1 and 2, respectively (Fig. 2A). SNPs were detected by referring to the NCBI reference database. Only one SNV was detected at the 16,160 locus in ILC samples (case 1; Table IV, Fig. 2B).

SSRNVs were detected in two regions, one of which was in HV (base numbers 303–310), while the other one was in non-HV (base numbers 516–525; Fig. 2B). For SSRNVs in HV, the coexistence of C8 and C7 was detected in NB and DCIS1 (case 1) and NB and FEA (case 2) (Table IV, Fig. 2B). For SSRNVs in non-HV, the coexistence of CA7 and CA6 was detected in ILC (case 1) (Table IV, Fig. 2B).

In case 1, the estimated proportions of cells with a loss of chromosome 16q were as follows: FEA, 37.2%; DCIS1, 0%; DCIS2, 65.0%; IDC, 36.8%; LCIS, 46.2%; and ILC, 47.8% (Tables V and VI). The estimated proportion of cells with a 1q gain was as follows: FEA, 35.0%; DCIS1, 31.0%; DCIS2, 131.4%; IDC, 82.6%; LCIS, 97.8%; and ILC, 75.0% (Tables V and VI).

In case 2, the estimated proportions of cells with a loss of chromosome 16q were as follows: FEA, 0%; LCIS, 31.2%; and IDC, 12.8% (Tables V and VI). The estimated proportions of cells with a gain of 1q were: FEA, 59.4%; LCIS, 39.6%; and IDC, 91.8% (Tables V and VI).

The number of mutations that satisfied adopted filters (described in the Materials and methods) was 5 and 15 in cases 1 and 2, respectively (Table SIII). Data for CCP, gene locus, altered genotype and variant frequency calculated by dividing variant allele coverage by total coverage, are presented in Table SIII. The mutation distribution among different lesions in each case is represented by a Venn diagram (Fig. 3). In case 1, DCIS2, IDC and ILC carried three, one and two mutations, respectively. IDC and ILC shared one mutation, while DCIS2 carried 3 independent mutations. DCIS1 did not share any common mutations with other lesions, including DCIS2. In case 2, FEA, IDC and LCIS carried four, six and 10 mutations, respectively. FEA and IDC shared three, IDC and LCIS shared two, and FEA and LCIS shared one mutation. Furthermore, one common mutation was shared by FEA, IDC and LCIS, and FEA (Fig. 3).

Variant frequencies of gene mutations detected by the CCP are provided in Table SIII and the genes submitted for verification by allele-specific PCR are presented in Table VII. For each sample, the results of allele-specific PCR are presented in Table VIII. In case 1, the calculated mutation rate of CDH1 was 15% in IDC, 30% in LCIS and 34% in ILC. CDH1 mutation (p.Asn174fs; likely oncogenic according to OncoKB, no entry in COSMIC) was detected in IDC and ILC samples. However, it was not observed in any of the other samples, indicating that the same mutation was shared by IDC and ILC. Mutations in RRM1 existed only in ILC, while mutations of AKT1, RALGDS and PIK3CA existed only in DCIS2, confirming the results obtained with the CCP. In addition, CDH1 mutations (T>TA at chr16:68842455) were detected using Sanger sequencing (Fig. 4).

In case 2, a nonsense mutation of CDH1 (p.Gln23Ter, likely oncogenic according to OncoKB, frequent 70 mutations reported in COSMIC) (Fig. 4) existed only in LCIS. Mutation of TIMP3 was shared by FEA and IDC. Mutation of PIK3CA was shared in NB, FEA and IDC; these results were consistent with those from the CCP. Mutations in KMT2C were not detected by SYBR green allele-specific qPCR, supporting the result of the CCP, which indicated a deletion in KMT2C in IDC samples of case 2 (Fig. 4).

A schematic of the progression of mutations in the carcinoma cells of different histological types in each lesion is provided in Fig. 5. In case 1, DCIS1 cancer cells harbored a 1q gain and FEA exhibited additional alterations of 16q loss and mtDNA LOV; this indicated that the two pathways diverged from FEA, wherein an MDL lesion was characterized by an increase in the ratio of 16q loss and acquired CDH1 mutations, as detected in IDC, LCIS and ILC. ILC samples further gained an RRM1 mutation, an SNV at 16,160 and mtDNA GOV. The second pathway was associated with distinct genetic alterations (AKT1, RALGDS and PIK3CA) in DCIS2 that were topologically located in a different area from that of IDC, LCIS and ILC. These genetic alterations were not shared by DCIS1, IDC, LCIS or ILC in an MDL lesion, suggesting that the cancer cells in DCIS2 were of a different clonal origin (Fig. 5). Since FEA had a higher ratio of cells exhibiting 16q loss and 1q gain, as well as a higher homogenization of mtDNA LOV compared to those of DCIS1, FEA diagnosed under the present criteria appeared to be a neoplastic lesion similar to DCIS1.

In case 2, cancer cells in which 1q gain was detected in all lesions diverged into two paths. One branched out to LCIS with mtDNA LOV [homogenization to C8 of SSRNV in HV (303–310)], 16q loss and TIMP3 mutation. The other pathway involved FEA acquiring PIK3CA and TIMP3 mutations. IDC samples acquired additional KMT2C mutations, mtDNA LOV and 16q loss. In this case, LCIS, IDC and FEA were topologically similar and these lesions shared closely related mutations (Fig. 5).