Detection of Circulating Tumor Cell Subpopulations in Patients with Head and Neck Squamous Cell Carcinoma (HNSCC)

Background Since image based diagnostic tools fail to detect early metastasis in head and neck squamous cell carcinoma (HNSCC) it is crucial to develop minimal invasive diagnostic methods. A promising approach is to identify and characterize circulating tumor cells (CTC) in the peripheral blood of HNSCC patients. In this pilot study, we assessed which non-hematopoietic cell types are identifiable and whether their numbers differ in pre- and postoperative blood samples. Methods 20 ml citrated peripheral blood was taken from 10 HNSCC patients before and after curative resection. CTC were enriched using density gradient centrifugation. CTC presence was verified by multi-immunofluorescence staining against cytokeratin (CK; epithelial), N-cadherin (mesenchymal); CD133 (stem-cell), CD45 (hematopoietic) and DAPI (nucleus). Individual cell type profiles were analyzed. Results We were able to detect cells with epithelial properties like CK+/N-cadherin−/CD45− and CK+/CD133−/CD45− as well as cells with mesenchymal features such as N-cadherin+/CK−/CD45− and cells with both characteristics like N-cadherin+/CK+/CD45−. We also observed cells showing stem cell-like features like CD133+/CK−/CD45− and cells with both epithelial and stem cell-like features such as CD133+/CK+/CD45−. The number of CK positive cells (p = 0.002), N-cadherin positive cells (p = 0.002) and CD133 positive cells (p = 0.01) decreased significantly after resection. Kaplan-Meier test showed that the survival was significantly shorter when N-cadherin+ cells were present after resection (p = 0.04; 474 vs. 235 days; [HR] = 3.1). Conclusions This is - to the best of our knowledge- the first pilot study identifying different CTC populations in peripheral blood of HNSCC patients and showing that these individual cell type profiles may have distinct clinical implications.


Introduction
Since image based diagnostic tools fail to detect early metastasis in HNSCC it is crucial to develop minimal invasive diagnostic methods to characterize entities and to find markers that could help choosing the appropriate treatment and monitoring response at an early stage.
Circulating tumor cells (CTC) could serve as a ''liquid biopsy'' for individualizing and monitoring therapy in patients with solid tumors [1,2]. So far, CTC detection methods consisted of enrichment and subsequent identification mostly with anti-cytokeratin (CK) or epithelial cell adhesion molecule (EpCAM) antibodies. CK-positive cells are thought to be absent or to be present in the blood of healthy subjects in very low numbers [3]. CTC have extensively been described in breast and lung cancer and EpCAM-positive CTC quantification has been linked to patient outcome [4][5][6][7]. Up to now only a few reports have been published on the isolation of CTC in HNSCC [8][9][10][11][12][13][14][15].
Standardized approaches with currently available enrichment and detection techniques are based on physical or biological properties of CTC and challenged by their cellular heterogeneity and plasticity. Epithelial-to-mesenchymal transition (EMT) can cause alteration of cellular features and loss of epithelial properties leading to a partial or complete switch to a mesenchymal phenotype. Particularly stem cells have the ability to take on characteristics of other cell types [16].
We recently developed a CTC detection method based on multi-parameter immunofluorescence microscopy (MPIM) that includes but is not solely dependent on epithelial markers such as CK or EpCAM and also detects cells with mesenchymal and stem cell-like characteristics [17]. We were able to show that the individual composition of these CTC profiles correlated to therapeutic success in hepatocellular carcinoma, non-small cell lung cancer and renal cell carcinoma [17][18][19]. In this preliminary study, we used a slightly modified methodology including density gradient centrifugation but no depletion of CD45-positive cells and addressed the question whether different types of CTC are identifiable in the peripheral blood of patients with HNSCC and, if so, whether their distribution may serve as a predictor of treatment response or outcome. With this approach we wanted to scrutinize whether developing a blood-based multi-marker panel for personalized treatment of HNSCC is warranted.

Ethics Statement and Study Population
Written informed consent was obtained from all patients before participating in the study. Blood sample collection and analyses were approved by the Review Board of the Ethics Committee of the regional Medical Association Nordrhein; Germany (2012304) and the Medical Department, University of Essen-Duisburg; Germany (12-5047-BO). We analyzed 10 patients with HNSCC before and after curative surgical resection. The clinico-pathological characteristics of the patients are listed in Table 1. Tumor staging was performed according to the criteria of the TNM Classification by the Union for International Cancer Control (UICC) [20]. Preparation of blood samples and CTC enrichment 20 ml citrated peripheral venous blood were drawn from HNSCC patients directly before and immediately after curative tumor resection and processed within 24 h after collection. Blood sample preparation was done as described previously [17,19]. Briefly, 20 ml of blood were diluted with 10 ml PBS and carefully layered into a Leucosep tube containing 16 ml Ficoll-Paque (GE-Healthcare) below a porous barrier. After buoyant density gradient centrifugation (16006g, 20˚C, 20 min) the interphase consisting of peripheral blood mononuclear cells (PBMNC) and CTC was removed and washed. The cell suspension was spun onto 2-4 glass slides per sample containing a maximum of 6610 6 cells per slide using the Cell Spin II centrifuge (Tharmac, Waldsolms, Germany), air-dried and subsequently fixated with 96% Ethanol. Slides were stored at 4˚C until subjected to immunocytochemical staining (Fig. 1). The presence of CTC was verified by multi-immunofluorescence staining against pan-CK (epithelial), N-cadherin (mesenchymal), CD133 (stem-cell-like), counterstaining with CD45 (hematopoietic) and DAPI (nucleus). CTC and hematopoietic cells were enumerated, normalized and expressed as number of CTC per 1000 PBMNC. Individual cell type profiles were analyzed and correlated to therapeutic outcome in 10 patients.

Identification of CTC subtypes using multi-immunofluorescence staining
Immunofluorescence staining of epithelial, mesenchymal, stem cell-like and hematopoietic cells was carried out as described previously [17]. Briefly, the staining method included fixation of the cells in 4.5% paraformaldehyde for   Table 2) either anti-pan-CK and anti-N-cadherin (or anti-CD133 for CTC and anti-CD45 for hematologic cells overnight at 4˚C, wash in 0,1% Tween, binding of secondary antibodies (Table 2) for 30 min at 37˚C, washing in 0,1% Tween. Subsequently, cells were stained with 496-Diamidino-2-phenylindole dihydrochloride (DAPI; Sigma-Aldrich, St. Louis, MO) for 10 min, mounted with anti-fading medium (Invitrogen) and stored in the dark until evaluation. As described previously, for each test a control slide with a mixture of PBMNC (CD45-positive, pan-CK-negative) from a healthy donor spiked with different cell lines (epithelial, mesenchymal, CD133 positive cells) were used as positive control.
Microscopic evaluation was carried out using the digital Keyence BZ9000 (Biorevo, Osaka, Japan) all-in-one fluorescence microscope with integrated camera and BZ-Analyzer Software. We used pseudo colors to depict cells. Stained slides were manually examined and CTC were detected within the same areas, each consisting of 10 visual fields using a 206 magnification on both slides. Samples from 12 healthy donors were processed and examined under the same conditions in order to define cut-off values for false-positive events.

Statistical Analysis
Statistical tests were performed according to previously published studies by our group [17,21,22]. The associations among CTC subtypes and clinico-pathological parameters were tested with Spearman test for bivariate correlations. Wilcoxon test for paired samples was used to compare differences of various factors at different time points. Mann-Whitney test for independent samples was used to compare differences of various factors in distinct subgroups. Survival curves (500-days cutoff) were plotted according to the Kaplan-Meier method to test correlations of overall survival with CTC subtypes. Significance was tested using

Immunofluorescence based identification of CTC subtypes
For the investigation of cellular subtypes a multi-staining method was required in order to detect various epithelial, mesenchymal, stem cell-like and hematopoietic characteristics. Therefore, we used multi-immunofluorescence staining for CTCsubtype detection. In HNSCC blood samples we observed cells with mesenchymal features such as N-cadherin+/CK2/CD452 and cells with epithelial properties like CK+/N-cadherin2/CD452 and CK+/CD1332/CD452 and cells with both characteristics like CK+/N-cadherin+/CD452. We also detected cells that stained positive for CD133 which is believed to be a cancer stem cell marker and found subtypes such as CD133+/CK2/CD452 and CD133+/CK+/CD452 cells. In addition, we found cells that stained positive for potential markers of CTC and CD45 such as N-cadherin2/CK+/CD45low, N-cadherin+/CK2/CD45low N-cadherin+/CK+/CD45low as well as CD133+/CK+/CD45+, CD1332/CK+/ CD45low and CD133+/CK2/CD45lowcells (Fig. 2). CTC were enumerated and CTC profiles of each patient were examined. We scored the total amount of Ncadherin-positive, CK-positive and CD133-positive cells, and calculated a ratio of mesenchymal to epithelial cells and stem cell-like to epithelial cells, respectively, after density gradient centrifugation. We normalized the enumerated potential CTC against the total PBMNC number detected in the DAPI channel in each visual field and calculated the number of CTC per 1000 PBMNC. Analysis of samples from healthy donors revealed the following cut-off values for false positive events per 1000 PBMNC: 0.07 CD133+/CK+ cells; 0.07 CD133+/CK2 cells; 0.01 N-cadherin+/CK+ cells; 0.03 N-cadherin+/CK2 cells and a total of 0.6 CK+ cells.  Figure 4). To investigate the association between CTC subtypes and tumor staging we used Figure 2. Selection of detected CTC subtypes using multi-immunofluorescence staining. As described above we detected a variety of CTC subtypes. N-cadherin and CD133 were stained on separate slides, respectively, together with CK, CD45 and DAPI. Pseudo colours were used to depict CTC subtypes: Ncadherin (mesenchymal,yellow); CD133 (stem-cell-like, yellow), pan-cytokeratin (CK; epithelial, red), CD45 (hematopoietic, green) and DAPI (nucleus, blue). We were able to detect cells with epithelial and mesenchymal properties like N-cad+/CK+/CD45low, as well as cells with mesenchymal features such as N-cad+/CK2/CD45low. We also observed cells showing mixed characteristics like CD133+/CK+/CD45+ and cells with epithelial characteristics such as CD1332/CK+/CD452 cells. More CTC subtypes were detected, but are not shown in this figure.

CTC-subtypes and clinical outcome
doi:10.1371/journal.pone.0113706.g002  Mann-Whitney test and revealed that the total number of CTC and the number of epithelial CK+ CTC was elevated in stage III compared to stage II patients, but did not differ significantly (both p50.3; Fig. 5 A). Interestingly, the number of mesenchymal N-Cadherin+ (p50.05) and stem cell-like CD133+ (p50.05) cells as well as CK2/CD133+/CD452 cells (p50.05) was significantly increased in Stage III compared to stage II patients (Figure 5 B-D; Table 4).

Discussion
Within this study we examined the individual CTC composition prior to and after curative resection in patients with HNSCC. Multi-immunofluorescence based morphological analysis revealed a variety of CTC subtypes with epithelial, mesenchymal, stem cell-like or mixed characteristics such as N-cadherin+/CK2/ CD452; N-cadherin2/CK+/CD452; CD133+/CK2/CD452 and CD1332/CK+/ CD452 cells. Analyses of individual CTC profiles indicated that the presence as well as the number of mesenchymal N-cadherin-positive and epithelial CKpositive CTC was associated to shortened survival. The presence of stem cell-like CD133-positive cells was correlated to tumor staging. The total number of CTC and detected CTC subtypes were significantly decreased after resection. Kaplan Meier test confirmed that the survival time was significantly shorter when Ncadherin-positive cells were detected after surgery, implying it as marker of poor prognosis. Our data are in line with a study by Nguyen et al. who examined the expression of N-cadherin using immunohistochemistry [8]. They observed high N-cadherin expression in 52 of 80 paraffin-embedded HNSCC cases and reported a significant correlation with malignant behaviors. Subsequently, they suggest Ncadherin to play an important role in malignant behaviors of HNSCC and that cadherin switching might be considered as a discrete critical event in EMT and metastatic potential of HNSCC. Up to now, only few reports have been published on CTC in HNSCC patients. Not long ago, Jatana et al. presented a purely negative depletion process using erythrocyte lysis followed by immunomagnetic labeling with CD45-antibodies and immunocytochemical staining against CD45, CK and DAPI to separate and quantify CTC in HNSCC [9]. Subsequently, Balasubramanian et al. used this negative enrichment method and reported the presence of CTC with mesenchymal and stem cell-like characteristics such as vimentin, EGFR, Ncadherin and CD44 [10]. They emphasized the existence of CTC phenotypes suggesting EMT and described that the presence of CTC in HNSCC correlated with worse disease-free survival. Their initial results add support to our findings that CTC subpopulations with mesenchymal characteristics are not only detectable but also associated to worse clinical outcome in patients with HNSCC when present after surgical resection. A study by Winter et al. showed, for the first time, that almost all patients with advanced HNSCC have circulating cells at the time of surgery [11]. They applied immunomagnetic enrichment and RT-PCR detection of CTC using four epithelial markers (ELF3, CK19, EGFR and EphB4). In a pilot study with samples from 16 patients the CTC detection was not associated with significant differences in overall or disease free survival though. Another study analyzing epithelial EpCAM+/CK+/CD452 CTC using flow cytometry followed by RT-PCR was performed by Hristozova et al. [12]. They reported no significant association of CTC presence to T stage. Here we are able to confirm that in our study the total amount of detected CTC prior to surgery did not show significant association to tumor staging. However, the number of mesenchymal and stem cell-like CTC subtypes was significantly increased in T3 compared to T2 patients prior to resection. Nichols et al. used the CellSearch system and reported that only 6 out of 15 patients (40%) had detectable CTC [13]. Also Buglione et al. applied the CellSerach system and found that a decreased CTC number or their absence throughout the treatment seemed related with nonprogressive disease [14]. The latest study using the CellSearch system was published by Bozec et al. who evaluated the potential detection of CTC in patients with locally advanced HNSCC and intended to identify the clinical factors predictive of CTC presence [15]. They detected CTC in only 8 of 49 patients (16%) before therapy. In contrast to the CellSearch based studies, our findings strengthen the hypothesis that CTC subtype analysis including mesenchymal and stem cell-like properties might reveal more detailed clinical association than epithelial markers alone. EMT-type cells and cancer stem cells (CSC) are believed to play critical roles in cancer metastasis and drug resistance. The formation of CSC and the event of EMT is a dynamic process which is triggered by the interaction of various cellular signaling pathways such as Hedgehog, Notch, PDGF, Wnt, TGF-b, Akt, and NF-k B signaling pathways [23]. Recently the existence of circulating mesenchymal stem cells (MSC) derived from peripheral blood was reported [24][25][26]. It was described that MSC have the ability to migrate from bone marrow to damaged tissue via the circulating peripheral blood to promote regeneration [27]. This process may involve hyper stimulation of bone marrow production using granulocyte colony stimulating factor (G-CSF) resulting in the occurrence of a mixture of MSC, hematopoetic stem cells and other immature progenitor cells [25,28,29]. Furthermore, the literature revealed that MSC migrate to and proliferate within tumor sites [30]. In this study we were able to observe a distinct proportion of cells that stained positive for pan-CK and CD45, a phenomenon already described by Yu and colleagues [31]. The additional CD45+ staining may not be exclusive for hematopoietic cells, but may hypothetically be acquired during the dormant state in the bone marrow or through effects comparable to trogocytosis, i.e. transfer of membrane proteins.
Taken together, our results support the hypothesis that the examination of individual CTC profiles might be used as a follow up measure in addition to clinical data during treatment and to develop novel targeted therapeutic options in HNSCC.