Colonization of distant organs by tumor cells generating circulating homotypic clusters adaptive to fluid shear stress

Once disseminated tumor cells (DTCs) arrive at a metastatic organ, they remain there, latent, and become seeds of metastasis. However, the clonal composition of DTCs in a latent state remains unclear. Here, we applied high-resolution DNA barcode tracking to a mouse model that recapitulated the metastatic dormancy of head and neck squamous cell carcinoma (HNSCC). We found that clones abundantly circulated peripheral blood dominated DTCs. Through analyses of multiple barcoded clonal lines, we identified specific subclonal population that preferentially generated homotypic circulating tumor cell (CTC) clusters and dominated DTCs. Despite no notable features under static conditions, this population significantly generated stable cell aggregates that were resistant to anoikis under fluid shear stress (FSS) conditions in an E-cadherin-dependent manner. Our data from various cancer cell lines indicated that the ability of aggregate-constituting cells to regulate cortical actin-myosin dynamics governed the aggregates’ stability in FSS. The CTC cluster-originating cells were characterized by the expression of a subset of E-cadherin binding factors enriched with actin cytoskeleton regulators. Furthermore, this expression signature was associated with locoregional and metastatic recurrence in HNSCC patients. These results reveal a biological selection of tumor cells capable of generating FSS-adaptive CTC clusters, which leads to distant colonization.

Lentivirus was generated by transfection of HEK293T cells with pMD2.G and psPAX2.
Transduced cells were selected by using puromycin (5 µg/mL). Knockdown efficiency was examined by using RT-qPCR and Western blot analyses.
Protein Extraction and Immunoblotting. Immunoblotting was performed as described previously 6 . Cells were washed once in ice-cold PBS and were then lysed by adding NP40 cell lysis buffer containing freshly added protease/phosphatase inhibitor cocktail (Sigma Aldrich). After incubation of the cell lysate on ice on a shaker for 15 min, the lysate was removed and centrifuged at 15,000 × g for 15 min to remove insoluble material. Supernatants were stored at -70 °C until use. The protein concentration was determined by using a BCA kit (Pierce Chemical Co., Rockford, IL).
Equal amounts of protein were fractionated via SDS-PAGE and transferred to PVDF membranes (Bio-Rad Laboratories, Hercules, CA). Membranes were blocked with 5% nonfat dry milk and 0.1% Tween 20 (Sigma Aldrich) in TBS (pH 7.4) and were then incubated overnight at 4 °C with antibodies against E-cadherin (#ab1416; Abcam, Cambridge, MA) or β-actin (#A5441; Sigma Aldrich) in 5% BSA (Sigma Aldrich) and 0.1% Tween 20 in TBS (pH 7.4). Several blots were cut prior to antibody hybridization for the simultaneous detection of different proteins in the same sample. After membranes were washed, they were incubated in HRP-conjugated secondary antibodies for 1 h. After this procedure, specific protein bands were detected via an enhanced chemiluminescence system (Amersham Pharmacia Biotech, Buckinghamshire, UK).
Immunoblotting was repeated three times independently with similar results.
Immunofluorescence. Cells were fixed with 3.5% PFA for 10 min at room temperature and then cytospinned onto coated microscope slides (Matsunami Glass Then, genes in identified GO terms ("molecular function" or "biological process") were categorized in terms of the GO biological process by using PANTHER. Lists and information about EBPs and their detailed functional categories were obtained from a previous report 9 .  13 . The network map was manually obtained, with small annotated groups (≤4 and ≤7 gene sets for patient samples and xenograft samples, respectively) and nodes unconnected to any annotations being removed, which resulted in the simplified network map shown in Fig. 8a and Supplementary Fig. S8. Detailed GSEA results (FDR q-value < 0.25) for patient samples are given in the Supplementary Table S6.

Analysis of the Gene Expression
In the current analyses, no statistically significant enriched gene set existed for up-regulated genes in single CTCs, at least in categories such as cellular component (C5cc), molecular function (C5mf), and KEGG pathway (C2kegg).
The gene expression profiles and clinical information for patients with HNSCC were collected from TCGA via the cBioPortal (http://www.cbioportal.org) 14,15 .
To validate the clinical significance of genes that were enriched in clone H (EBP-H; 54 genes), GSEA was performed for a subgroup of patients from the TCGA cohort with clinical information about recurrence (n = 374), pathological stage (n = 434), metastasis in lymph nodes or distant organs (pN or pM; n = 432), or histological grade (n = 483).
Significant gene sets that were enriched were identified by using a nominal p value of <0.05 and FDR q-value of <0.25. We used 20 additional control lists of randomly selected gene sets, of comparable size (100 genes per each set) to address a possible caveat related to gene enrichment in tumors with various clinicopathological features.
None of the random lists was significantly enriched in genes associated with a specific class of tumors (i.e., either LDR or NR) except for some gene sets for "Histological G3/4 vs. G1/2" tests, that still had a high FDR q-value (0.179-0.237).