Contralateral breast cancer can represent a metastatic spread of the first primary tumor: determination of clonal relationship between contralateral breast cancers using next-generation whole genome sequencing

Tumor and treatment characteristics are presented in Table 1. Patient 10 was a known BRCA1 mutation carrier. Patient 3 had a local recurrence from BC2 7 months after diagnosis of CBC, treated with surgery and endocrine therapy. Patient 1 had a local recurrence after BC1 12 months before diagnosis of BC2, treated with surgery, radiotherapy, and endocrine therapy. None of the other patients had any local/regional recurrences. Patient 4 developed mucosa-associated lymphoid tissue lymphoma 61 months after BC2. To avoid the risk of misdiagnosing metastases past this date, the diagnosis date of the lymphoma was considered the last follow-up date. For Patients 1 and 7, BC2 was diagnosed during endocrine therapy.

Paraffin material was available for both tumors in Patients 3, 4, 5, 6, 7, and 8 and for BC1 in Patient 2. Here histopathological markers have been reevaluated with immunohistochemistry by a pathologist. The cutoff value for estrogen receptor (ER)/progesterone receptor (PR)-positivity was >10 % positive cells. For the remaining tumors these data are from the patient’s chart. For most of these patients (Patients 1, 9, and 10) ER and PR were measured with immunohistochemistry. For Patient 2 BC1 ER and PR were measured in cytosol (ER 140 fmol/mg protein, PR 170 fmol/mg protein; cutoff value for ER/PR-positivity = 25 fmol/mg protein).

WGS was performed on the 10 CBC pairs employing a strategy that combined low sequence coverage paired-end sequencing with larger fragment sizes for improved physical genome coverage. The median sequence coverage was 6.5 (range 1.8–11.2) and the median physical coverage was 14.7 (range 6.9–28.4) (Additional files 1 and 2). The total number of rearrangements detected in each tumor varied widely, from 15 to 256 with a median of 87 per tumor (Fig. 1a, Table 2; all plots in Additional file 3).

We were first interested in ascertaining how often two unrelated tumors from different individuals display identical chromosomal rearrangements because this would be an indication of overlap occurring by chance or due to incomplete filtering of germline events. For this, all possible pairings of tumors between patients (excluding the true within-patient pairing) were exhaustively compared. This comparison showed that 0–25 rearrangements (median 4) matched between any two random tumors with a combined shared fraction (i.e., the fraction of rearrangements shared between BC1 and BC2 divided by the union of all rearrangements detected in both tumors) ranging from 0 to 18 % (median 2 %; Table 2). For the comparisons of matched patient CBCs, on the other hand, the number of shared rearrangements between BC1 and BC2 ranged between 6 and 68 (median 28) and the combined shared fraction varied between 6 and 76 % (median 13 %).

Patient 8 was a clear outlier and had the most similar CBCs, with 76 % shared rearrangements between the two tumors (68 shared rearrangements), 19 rearrangements unique to BC1, and three rearrangements unique to BC2 (Fig. 1a, Table 2). Patient 8’s extreme degree of similarity of chromosomal rearrangements unequivocally supports BC2 being the result of a metastatic spread from BC1. Furthermore, in a separate study of matched primary and distant metastases using the same experimental approach, we have found that the average primary breast cancer shares well over one-half of its chromosomal rearrangements with the distant metastasis it seeds (results not shown; Tang and Gruvberger-Saal, manuscript under review). Therefore, the 76 % shared rearrangements between CBCs in Patient 8 is highly consistent with known primary-metastasis pairs.

Seven of the remaining CBC patients had much fewer shared rearrangements between BC1 and BC2 with combined shared fractions between 6 and 15 % (median 10 %), most consistent with these CBCs having developed as independent primary tumors (Fig. 1a, Table 2; Additional file 3). However, Patients 2 and 6 had a rearrangement overlap (39 % and 46 %) between that seen for unmatched randomized tumor pairs and true tumor-metastasis pairs (including Patient 8). The cases of Patients 2 and 6 are highly suspicious for contralateral metastasis, but to classify them with certainty normal DNA samples would be needed to rule out any remaining germline rearrangements. Interestingly, for Patient 1, where BC2 was an in-situ lesion and included in the study as a CBC pair presumably representing two independent lesions, 15 % of rearrangements matched between the tumors. This overlap was considerably lower than the overlap for Patient 8 (76 %) but well within the range between CBC pairs when excluding the three most similar patients (Patients 2, 6, and 8) (6–13 %).

The similarities and differences in chromosomal rearrangements between tumors are well illustrated in a barcode plot, where all nonredundant identified rearrangements are arranged and their presence in each tumor is denoted by a line (Fig. 2). For example, the two tumors from Patient 3 were highly eventful yet displayed distinct rearrangement barcodes spread across the whole genome (Figs 1a and 2). Many tumors exhibited localized areas of high density of rearrangements suggestive of one catastrophic shattering event (chromothripsis). Chromothripsis is thought to contribute to oncogene activation and tumor suppressor loss, and as a consequence is an important driving force for early cancer development and would probably be shared between clonal tumors. For Patient 8 this appears to be true; the two tumors share identical chromothripsis-like rearrangement patterns in two localized hotspots on chromosomes 10 and 11 (Figs 1a and 2). On chromosome 11 alone the two tumors share 38 identical intrachromosomal rearrangements, while only five are specific to BC1 and none to BC2. Patient 4, on the other hand, has localized areas on 1q, 6p, and chromosomes 17 and 18 of chromothripsis-like patterns that only appear in BC1, whereas BC2 has specific high-density areas on 6q. Patient 9, with a combined percentage of shared rearrangements of 10 %, has very distinct heavily rearranged areas on 11q and chromosome 12 unique to BC1, while BC2 has areas suggesting chromothripsis on chromosomes 1, 3, and 11p that are not being shared with BC1, further indicative of their nonclonal origin.

A number of cancer-associated genes (e.g., RB1, RARA, CDKN2A, and PTEN) were found to be involved in the enumerated rearrangements [23], probably contributing to tumorigenesis (Additional file 4).

Indicative of a suitable reliability of our analysis pipeline, 68 rearrangements out of 86 (79 %) were confirmed by PCR across the breakpoint junction as either specific to BC1, specific to BC2, or shared. However, nine of these rearrangements (13 %) were also identified in the matched normal DNA, or if unavailable were identified in a pool of 44 normal DNA samples. Therefore, although our pipeline results in a good accuracy of calls based on the sequencing data, owing to the fact that matched normal DNA were unavailable for most patients and thus were not sequenced, our results are underestimating to some extent the contribution of rearrangements that are present in the germline.

CNV profiles were also evaluated using the WGS data (Fig. 1b; Additional file 3). The fraction of shared abnormal copy number events between CBC pairs ranged between 10 and 28 % (median 21 %; Table 2). When exhaustively comparing all pairings of tumors from different patients, the fraction of shared aberrant copy number events ranged between 6 and 22 % (median 13 %), which was only slightly lower than that seen for the matching tumor pairs. Again, the CBCs from Patient 8 had the highest fraction of shared aberrations (28 %) and the CNV profiles appeared similar (Fig. 1b; Additional file 3). Patient 1, with BC2 representing an in-situ lesion, shared 18 % of the aberrant copy number events between the two tumors.

The analysis of genomic clonal similarities between tumors can be visualized in two dimensions (Fig. 3). When looking at chromosomal rearrangements, the majority of the CBC pairs (blue) cluster together just slightly off the cloud of randomly-paired tumors (red) (Patients 2 and 6 showed a somewhat higher similarity as discussed above). Patient 8, on the other hand, is clearly distantly separated from the other CBC pairs (including Patient 1 with an in-situ BC2) and from the randomized pairings.

For the CNV events this separation is not evident and no patient CNV profiles are more similar than some of the tumor pairings from different individuals. Based on the analysis of rearrangements, however, we can conclude that BC2 from Patient 8 in fact is highly clonally related to BC1.