Patients recruitment and sample collection
The study had been prospectively approved by the Research Ethics Committee from IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) "Dino Amadori" (Italy) (reference, IRSTB017) according to the recommendations of the declaration of Helsinki, patient consented to the project were enrolled at IRCCS-IRST between 2013 and 2017.
Eligible patients were newly diagnosed with metastatic or locally advanced EC with treatment protocols of first line chemotherapy and neoadjuvant chemo-radiotherapy followed by surgery respectively. Patients were staged and managed standard treatment protocols were in accordance with international guidelines 39. Clinical data including age, tumor staging, treatments protocols and clinical outcomes were collected. Disease-free patients were defined as those who resulted negative for tumor detection at CT scan after surgery for at least five years.
Metastatic patients’ blood was collected at study entry and at CT scan evaluations. Blood drawn of non-metastatic patients was collected before the administration of neoadjuvant therapy, after neoadjuvant therapy at CT-scan evaluation and a month after surgery concurrently with CT-scan. All blood drawn volumes intended for CTC analysis ranged between 18–20 ml; first 3 ml of blood were not included in CTC enrichment to avoid epithelial cells contamination.
Cell Culture
The MDA MB 436 breast cancer cell line and MCF10A epithelial breast cell line were purchased from ATCC. EC cell line OE33 was kindly donated by Prof. Nickolas Stoecklein (University Hospital Düsseldorf). MDA-MB 436 cell line was cultured in Leibovitz's L-15 medium (GIBCO, USA, USA) with 10 ug/ml insulin (Sigma-Aldrich, USA Aldrich), 100U/ml penicillin-streptomycin (GIBCO, USA, USA), 2mM glutamine (GIBCO, USA, USA) complemented with 10% FBS (GIBCO, USA), in a non-free gas exchange condition.
MCF10A cells were cultured in DMEM/F-12, GlutaMAX™ medium (GIBCO, USA) complemented with 5% Horse Serum (New Zealand origin) (GIBCO, USA), 100 U/ml penicillin-streptomycin (GIBCO, USA), 2 mM L glutamine (Invitrogen, USA), 100 mg/ml human epidermal growth factor (EGF) (Merck Millipore, UK), 1 mg/ml hydrocortisone (Sigma-Aldrich, USA), 1mg/ml cholera toxin (Sigma-Aldrich, USA Aldrich) and 10 mg/ml insulin (Sigma-Aldrich, USA Aldrich), 100U/ml penicillin-streptomycin (GIBCO, USA) as mentioned 40.
OE33 cell line was cultured in RPMI 1640 medium (GIBCO, USA) complemented with 10% FBS (GIBCO, USA), 100U/ml penicillin-streptomycin (GIBCO, USA).
All cell lines were maintained as a monolayer in a 37°C incubator with 5% CO2 and sub-cultured at 70–75% confluence (approximately twice weekly).
We induced epithelial-to-mesenchymal transition (EMT) on MCF10A cell line by TGF beta treatment as reported 41. Briefly, cells were treated with 10ng/ml of TGF beta1 (Proprotech) every 48 hours for ten days. Morphological changes observation and images were performed by light microscopy Axiovert 200 (Zeiss).
Grab all assay and DEPArray analyses were performed on cells collected at four time points: before treatment, after 4, 6 and 10 days.
CTC enrichment, Grab-all assay and single cell isolation
Blood samples (18-20ml per sample) were collected in EDTA vacutainers (BD) and processed within 3 hours. Label-independent enrichment of CTCs was carried out by density gradient centrifugation with OncoQuick® as we previously used 17,42.
Next, Grab-all assay was performed on CTC samples. In details, samples were fixed in 2% paraformaldehyde (Miltenyi Biotech, USA) and incubated 30 minutes at 4 C° with cell membrane directed antibodies: EpCAM-PE (clone HEA-125) (Miltenyi Biotech, USA) 1:10; E-Cadherin-PE (clone 67A4) (Miltenyi Biotech, USA) 1:10; N-Cadherin AlexaFluor 647 (clone 8C11) (BD Pharmigen, USA) 1:20; CD44v6-APC (clone 2F10) (R&D Systems, USA) 1:20; ABCG2 (clone 5D3) (R&D Systems, USA) 1:10); CD45-AlexaFluor 488 (clone GA90) (Life Technologies, USA) 1:10. We performed intracytoplasmic staining with pan-CKs-PE (clone C11) (Aczon, Italy) 1:10, Hoechst 33342 (1 µg/mL, Life Technologies, USA) diluted in a solution of 1X PBS/1% BSA (Sigma-Aldrich, USA-Aldrich) 0.3% Triton™ X-100 (Bio-rad, The Netherlands).
Stained samples were resuspended in 14 ul of SB115 Buffer (Menarini-Silicon Biosystems SpA, Italy), loaded into DEPArray™ chip (Menarini-Silicon Biosystems SpA, Italy) and analyzed by DEPArray platform.
In order to avoid any technical bias related to acquisition of images, first we defined time exposure of DEPArray acquisition. We stained leukocytes only with CD45 antibody acquiring images of all fluorescent channels to define the maximum exposure time of FITC channel for saturation and to verify the absence of any autofluorescence signals on PE and APC channels. We followed the same approach to define epithelial and stem-mesenchymal tags (PE and APC channels respectively) by staining a mix of MDA MB 436 and leukocytes cells. MDA MB 436 cells were stained with epithelial plus CD45 antibodies and subsequently with stem/mesenchymal plus CD45 antibodies. We configured time exposure of channels to prevent autofluorescence signals especially on FITC channel as well as saturation signals.
All acquired patient’s cell images were manually reviewed, cells classified as CTCs had nucleus rounded- or spindle- shaped morphology and were negative for CD45. Epithelial CTCs were positive only for epithelial tag (EpCAM/CKs/E-Cadherin), mesenchymal/stem CTCs were positive only for stem/mesenchymal tag (N-Cadherin/CD44v6/ABCG2) and bi-phenotipic CTCs when cells were positive for both tags. All routable CTCs were sorted as single cells and single leukocytes were sorted as reference diploid chromosome set for further molecular analysis. Cells were sorted into 0,2ml tubes, and stocked at -80°C after manual volume reduction and 1X PBS (GIBCO, USA) wash until whole genome amplification (WGA).
Single CTC WGA and quality control (QC):
We performed WGA using Ampli1 WGA Kit (Menarini-Silicon Biosystems) according to manufacturer’s instructions except for the cell lysis step, which was conducted overnight into thermal cycler. DNA samples were stocked at -20°C until use.
All the WGA products were quality checked adopting Ampli1™ QC Kit (Silicon Biosystems) as per manufacturer’s instructions. Briefly, we performed a multiplex PCR to amplify four amplicons of different lengths. The amplification of the regions was assessed on a 2% agarose gel using Chemidoc instrument (Bio-rad, The Netherlands).
Library preparation and next generation sequencing (NGS)
DNA libraries for low-pass whole genome sequencing were generated using a streamline kit of WGA products named Ampli1™ LowPass Kit for Ion Torrent (Menarini-Silicon Biosystems) as reported 18. WGA products of some samples were re-amplified by Ampli1™ ReAmp/ds following manifacturer's protocol for the generation of appropriate libraries. Concisely, 10 µl of WGA products were linked to barcoded adaptors for Ion Torrent System by a PCR amplification. Next, libraries were quantified by Qubit™ 3.0 Fluorometer (Invitrogen, USA) and fragment length was assessed using Bioanalyzer High Sensitivity DNA Kit (Agilent). We combined libraries in an equimolar pool, and selected fragments of the pooled libraries ranging between 300 bp to 450 bp using E-Gel™ SizeSelect™ II Agarose Gels, 2% (Invitrogen, USA) on E-Gel™ iBase™ and E-Gel™ Safe Imager™ (Invitrogen, USA).
Whole genome sequencing of single CTCs and matched leukocytes was carried out on Ion PGM™ and Ion GeneStudio™ S5 System (Thermo Fisher), loading 8 pooled samples on 318 and 520 chip, respectively. Sequencing was conducted following the guidelines reported in the protocol (emulsion PCR for 400-base-length libraries and 525-flows sequencing). For sequencing experiments conducted on Ion S5, chip preparation and emulsion PCR were performed on the Ion Chef instrument, using Ion 510™ & Ion 520™ & Ion 530™ Kit reagents.
CNA calling from low-resolution whole genome sequencing:
The output BAM files were given as input at Control-FREEC for CNA calling. We used two different non-parametric tests (Mann-Whitney and Kolmogorov-Smirnov) to assess the statistical significance of each putative CNA. The whole genome has been subdivided in windows (bin), whose size depends on the quality of the sequencing. Then, we generated plots showing copy numbers mapped for each bin, highlighting the points which are included in CNAs with a significance lower than 0.05 in both the non-parametric tests we applied.
Generation of Jaccard Indexes and statistical analysis
We computed a Jaccard Index (JI) for the detection of similarity, for what concerns CNA content, among each CTC against five leucocytes which have been analyzed with the same workflow. JI has been computed applying the following formula:
Starting from a list of cytogenetic bands which were of specific interest for the disease
5, we extracted the number of CTCs with at least a CNA in the specific region, with a further discrimination for access time and patient response. A similar procedure has been applied to a list of genes included in the same cytogenetic bands.
For each sample, we extracted the list of genes located inside a significant CNA detected by the previously described pipeline. From this list, using R package EnrichR
43 we performed an enrichment analysis exploring GO Biological Process dataset and we filtered out terms with adjusted p-value higher than 0.05. Finally, for each dataset, we generated networks using R package igraph
44. Inside these networks, nodes are the enriched terms while edges give the information about the presence of more than 10 shared genes among the two linked nodes. Moreover, the size of each node is related to the number of genes, while the sections of the pie chart inside nodes are proportional to the percentage of samples of access time.