Bone marrow provides a reservoir for DTCs that may evade therapy, remain dormant, and can cause overt metastases over time [9]. Hence, developing our understanding of the origin, nature, and biology of this cancer cell “type” is important. In contrast to previous studies that primarily applied low-resolution metaphase-chromosome or microarray comparative genomic hybridization, we employed single-cell genome sequencing to shed new light on the cells isolated as “DTCs” from bone marrow aspirates according to established immunocytochemical and morphologic criteria. Specifically, we demonstrate the existence of three major classes among the isolated cells: (i) true DTCs, which share both CNAs and single nucleotide substitutions with the tumor; (ii) aberrant cells of unknown origin which have CNA profiles that do not match those of the observed tumor or metastasis and that lack the somatic substitutions of the tumor; and (iii) normal cells having copy neutral genomes without tumor-specific mutations.
Of the 19 single cells immunocytochemically and morphologically classified as TCs, ten could be conclusively categorized as true DTCs (14 out of 24 including doublets). In contrast, only one of the 21 single cells classified as uncertain, PHC, or HC turned out to be a true DTC (three out of 26 including doublets). While the adaptation of the staining—as required for downstream genomic analysis [26]—may have affected the precision of the morphologic classification of the immuno-detected single cells, these findings still underscore its value (true positive rate 52.6?±?11.5%, false negative rate 4.8?±?4.7%). Notably, we found true DTCs for only three patients (MicMa003, 083 and 107), all of which went on to develop distant metastases (Table 1). In contrast, only AU and N cells were found in MicMa017, 019, and 044, of which only MicMa019 showed systemic recurrence.
Our results suggest that previous studies based on morphologic criteria may have been underrepresented for “true DTCs”, as defined here by genetic profiling. Cells identified in those studies, usually from patients without distant metastases, carrying a smaller number of chromosomal aberrations [23, 25, 33] distinct from one another and from those in the tumor under consideration [23, 33], and usually consisting of whole-chromosome gains or losses [23], likely correspond to AUs. These cells may have been interpreted as genuine DTCs, thus supporting a parallel progression model of the disease. The idea has been put forward that AU-like cells, in their ectopic site, obtain a genomic landscape similar to the primary tumor in a macroevolutionary process resulting from evolutionary shifts as could be induced by telomere crisis or the inactivation of a tumor suppressor such as TP53 [34]. To the best of our knowledge, we provide the first comprehensive investigation with modern sequencing technologies of these cells. Our data show a complete absence of tumor-specific truncal mutations in the AUs, supporting the notion that these cells do not derive from the MRCA nor any other observable progenitor of the sampled breast tumor. Indeed, direct evidence implicating AU-like cells as the precursors of overt metastases is scant [35].
The AUs (and perhaps also the normal cells) in our study may either represent a likely epithelial cell type of breast or non-breast origin homing to the bone marrow or derive from the hematopoietic cell lineage. Alternatively, these cells may originate from another neoplasm in the patient, an undetected synchronous primary breast tumor, or an undetectable or unsampled tumor cell clone residing in the primary tumor. Our results suggest that at least part of the AUs derive from hematopoietic cell populations, most likely plasma cells. Interestingly, studies employing single nucleus sequencing have previously reported cells with random gains or losses of single chromosomes or chromosome arms in diploid genomes, similar to many of the AUs [6, 36, 37]. These “pseudodiploid cells” were observed at rates of 1–6% in different normal tissues (brain, liver, skin, and breast) and 6–8% in breast tumor stroma, suggesting that even normal tissues display low-level aneuploidy, likely due to mitotic segregation defects. While pseudodiploid cells may account for about half of the AUs, those harboring recurrent CNAs (e.g., gain of chromosome 1q or 5 and loss of 9) are more likely to represent clonal expansion of a (pre)malignant subpopulation. At least two studies using SNP-array analyses of bulk DNA have now documented the emergence of subclonal hematopoietic cell lineages containing large CNAs within the blood of cancer cases as well as cancer-free controls [38, 39]. Like the fraction of AUs, the frequency of this subclonal genetic mosaicism increased with age in cancer-free individuals. By using single-cell sequencing, we may be witnessing the diversity of genomically aberrant hematopoietic cells that exist in low numbers within the bone marrow.
Using the CNAs as a guide, true DTCs were mapped onto the phylogenetic trees of the breast cancers, inferred by Battenberg from the bulk tumor DNA, exposing their evolutionary origins. The single-cell genome sequences enabled us to add considerable detail to the phylogeny of the solid tumors of patients MicMa083 and MicMa107. For patient MicMa083, the 50, 30, and 21% primary tumor subclones could only be nested confidently within one another because the single-cell DTC sequences reported coexistence of all three CNA events. Without knowledge of co-occurrence of these rare CNAs in the same cell, it would have been impossible to order these events in molecular pseudo-time. Similarly, the phylogenetic tree of the lymph node metastasis in patient MicMa107 could be reconstructed in more detail owing to the single-cell DTC sequences.
Our results provide clear insights into the origin of DTCs in breast cancer. All true DTCs in this study derive from either the MRCA (MicMa003), subclones of the MRCA (MicMa083), or subclones in an axillary lymph node metastasis that was seeded by a subclone of the MRCA of the primary tumor (MicMa107). We did not observe any DTC that carried only a subset of somatic changes present in the MRCA of the primary tumor, which would point towards early dissemination. Taken together, these results support a model whereby the ability of breast cancer cells to disseminate to the bone marrow arises relatively late in tumor evolution. However, we cannot rule out the possibility that continuous seeding and replacement is occurring: early disseminating tumor cells are replaced by new arrivals from the tumor as they compete for a bone marrow niche.
Interestingly, in patient MicMa083, all DTCs originated from the same 21% subclone. In MicMa107, however, multiple subclones hold the ability to disseminate to the bone marrow. Notably, only subclonal populations in the lymph node metastasis that have undergone considerable evolution since the MRCA are observed in the bone marrow. Hypothetically, metastatic potential (or the ability to disseminate to the bone marrow as a proxy to that) emerged with the MRCA in MicMa003 and significantly later in MicMa083 and MicMa107. While our sample size is small and further study is needed, these results raise the hope that early detection strategies can lead to diagnosis and treatment before the emergence of such metastatic clones.