Detection of circulating tumour cells with a hybrid (epithelial/mesenchymal) phenotype in patients with metastatic non-small cell lung cancer

1Circulating tumour cells (CTCs) have an active role in the formation of metastasis and their detection can provide information on patient prognosis and treatment efficacy. Current clinical results rely exclusively on the detection and enumeration of CTCs expressing epithelial-lineage markers (i.e., the epithelial cell adhesion molecule (EpCAM and cytokeratins) in the peripheral blood. Using this approach, several groups have shown an association between the CTC counts and clinical outcome of patients. In most carcinomas, tumour progression implicates the loss of epithelial features and a shift towards a mesenchymal phenotype. During this process, referred to as the epithelial-mesenchymal transition (EMT) and considered as a pre-requisite to tumour infiltration and metastasis, epithelial carcinoma cells acquire a migratory phenotype and express mesenchymal genes. The phenomenon of ‘partial’ or ‘incomplete’ EMT has also been reported in which invasive and metastasising carcinoma cells may adopt a mesenchymal phenotype and express mesenchymal markers, for example, vimentin and neural cadherin while retaining epithelial charcteristics. However, although EMT must, by definition, have a role in the generation of at least a fraction of CTCs, there has been very few data directly demonstrating the existence of the EMT process in CTCs. Here, we report for the first time the detection of CTCs expressing a dual epithelial and mesenchymal phenotype in the peripheral blood of patients with metastatic carcinoma.

2The present study was approved by our institutional review board and local ethics committee. Informed and written consent was obtained from all patients. Peripheral blood samples were collected from six patients with histologically or cytologically confirmed non-small cell lung cancer (NSCLC). For each patient, 10 ml of blood was collected in EDTA tubes (Terumo, Leuven, Belgium) for CTC enumeration by isolation by size of epithelial tumour cells (ISET). Peripheral blood samples were collected from six healthy volunteers for control experiments.

3Within 4 h after venopunction, blood samples were diluted 1: 10 in an erythrocyte lysis buffer containing paraformaldehyde, filtered on the ISET device using filters with calibrated pores of 8 μM. Filters were air dried and conserved at −20 °C. After thawing, filters were incubated for 20 min at 98 °C in an EDTA buffer (DAKO Target retrieval buffer, pH=9.9; Dako, Glostrup, Denmark), and treated for 5 min by triton X-100 at room temperature, in a humidity chamber for cell permeabilisation. Filters were washed with TBS for 1 min and stained with markers diluted in TBS, including Alexa Fluor 488 conjugated anti-vimentin (Santa Cruz Biotechnology, Heidelberg, Germany) and Alexa Fluor 647 conjugated anti-pan-cytokeratin (Cell Signalling, Danvers, MA, USA) antibodies and SYTOX orange (Molecular Probes, Leiden, The Netherlands) overnight at 4 °C in a humidity chamber and under obscurity. SYTOX orange is a nuclear dye for which the spectra are 545 nm for excitation and 570 nm for emission. After rinsing with TBS, filters were finally mounted using Ultramount (53001; Dako). Imaging is carried out with a confocal Zeiss LSM510 Meta microscope (Carl Zeiss France, Le Pecq, France) with a × 40 or × 63 magnification and an optical slice of 1.9 μm. Using the ImageJ software (NIH ImageJ, Version 1.42-q), a Gaussian blur (radius=0.10 μm) was applied, and contrast was slightly enhanced in every fluorescent channels of acquisitions. CTCs were selected, digitised and validated by a senior and experienced cytopathologist (PV) as cells presenting all the following criterias: (i) nuclear size equal or larger than two pores (16 μM); (ii) irregularity of the nuclear contour; (iii) presence of a visible cytoplasm; (iv) high nuclear-to-cytoplasmic ratio (greater than 0.8). A size threshold of 16 μM was chosen to exclude most normal cells.

4The SKBR3 (breast cancer), MDA-MB-435S (breast cancer) and A549 (lung adenocarcinoma) cultured cell lines were used as controls for markers expression. All cell lines were cultured in 1640 RPMI medium supplemented 10% fetal calf serum and maintained in a humidified incubator in 5% CO2 at 37 °C. For control experiments, 25 cells from each cell line were precisely micro-manipulated under an inverted microscope before being spiked into blood from healthy donors. Blood samples containing spiked cell lines were then diluted using the erythrocyte lysis buffer, filtered and stained as described for patient samples.

5Immunohistochemical stainings were performed on sections of formalin-fixed paraffin-embedded samples, using cytokeratin 7 (Dako) and vimentin (Santa Cruz Biotechnology) according to standard procedures.

6Blood samples from six patients with metastatic NSCLC were processed on the ISET device, which allows CTC enrichment by blood filtration through filters with calibrated pores. Triple fluorescent labelling with anti-vimentin, anti-pan-keratin antibodies and SYTOX orange for nuclear staining was performed to detect CTCs using confocal microscopy. Control experiments were performed in parallel using 25 cells of SKBR3 (pan-keratin positive), MDA-MB-435S (vimentin positive) and A549 (pan-keratin and vimentin positive) cell lines spiked in blood from healthy donors (Table 1). Blood samples from six healthy donors were also processed by ISET and labelled identically. Isolated and/or clusters of immunostained CTCs were morphocytologically identified by an experienced cytopathologist (PV) using bright field images (Figure 1A and F). A strong co-expression of vimentin and pan-keratin antigens (Figure 1B–E and G–J) was observed in almost all CTCs from the six NSCLC patients. Counts of dual CTCs co-expressing pan-keratin and vimentin, of only vimentin positive and only pan-keratin positive CTCs in the six NSCLC patients are presented in Table 1. Dual CTCs expressing both pan-keratin and vimentin were mainly present in clusters and ranged from 5 to 88 cells per 5 ml. CTCs expressing only pan-keratin antigens were not detected in any of the patients, whereas CTCs expressing only vimentin were more rare and detected in three patients (range 3–15/5 ml). The phenotype detected here in CTCs therefore strongly differed from that detected in primary tumours in NSCLC patients, which are positive for pan-keratin antigens and negative for vimentin. As an example, Figure 2A and B shows histological staining from a patient with lung primary adenocarcinoma (patient 6) presenting a keratin-positive and vimentin-negative phenotype. No CTCs were detected in the blood of healthy donors.

7Our exploratory study demonstrates for the first time that the majority of isolated or clusters of CTCs in patients with advanced metastatatic NSCLC harbour a dual epithelial–mesenchymal phenotype, suggesting that EMT is a relevant process for invasion and metastatasis in these patients. CTCs expressing only vimentin (or co-expressing also keratin antigens at a very low yet undetectable level) were also observed. These data underscore the reality of EMT in CTCs and provide experimental evidence to the paradigm of a phenotypical continuum between epithelial and mesenchymal states in CTCs. Low counts of CTCs have been reported in NSCLC patients by methods using EpCAM antigen expression-based enrichment. The partial loss of epithelial markers could therefore lead to an underestimation of CTCs with hybrid phenotype, as previously shown using breast cancer cell lines. EMT phenotype is associated with stem cell properties and resistance to anticancer treatment. A better phenotypical and molecular characterisation of CTCs with hybrid phenotype should provide further insight on the significance of EMT in CTCs and on the molecular mechanism of metastasis in patients with NSCLC.