Challenges in enumeration of CTCs in breast cancer using techniques independent of cytokeratin expression

Introduction Given the current postulated plasticity between epithelial and mesenchymal states of migratory cancer cells the detection of non-epithelial CTCs is an important scientific and clinical goal. Methods We used the filtration-based ISET technology to enrich circulating tumour cells (CTCs) in early breast cancer blood samples and identify them using a morphology-based immunocytochemistry (ICC) approach. Results We found greater numbers of putative CTCs by this approach than by the cytokeratin-based CellSearch technology, but a high number of CTC false positives were identified in healthy volunteer samples which were not reduced in successive blood draws. Preliminary work using an oestrogen receptor (ER)-based multiplex ICC method in metastatic breast cancer ISET samples indicated a low number of ER+ CTCs even at this advanced stage. Conclusions This work highlights the challenges in enumerating CTCs without conventional epithelial markers.


Introduction
Every year more than 50,000 women in the UK develop breast cancer, with 11,433 dying from the disease. Globally, it is the most common cause of female cancer death worldwide, with PLOS  more than 500,000 women estimated to die every year from over 1.65 million cases [1,2]. Nearly one quarter of breast cancer patients in the UK die within 10 years, with the majority presenting with distant disease several years after their initial breast cancer diagnosis. This highlights an urgent need for prognostic, predictive and pharmacodynamic biomarkers to identify patients at increased risk of recurrent disease, for early identification of recurrence, to facilitate tailored treatment and to monitor treatment response. Circulating tumour cells (CTCs) isolated from venous blood samples could be a relatively non-invasive real-time liquid biomarker that allows detection, monitoring and phenotyping of breast cancer. CTCs are rare cells (few per 10 ml blood /100 million leukocytes and 50 billion erythrocytes), and therefore highly sensitive assays with techniques to enrich then characterise CTCs are required. Techniques to enrich for CTCs are based either on tumour cell antigen expression or cell size or density. However, all these techniques are limited by their ability to capture all CTCs given the presence of CTC heterogeneity. Once enriched, CTC identification and characterisation can be achieved on a cellular level through microscopy or flow-cytometry, or on a molecular level using RT-PCR, however once again tumour heterogeneity remains the major challenge [3].
Currently the only FDA approved CTC technology is the antibody-based CellSearch system. This identifies epithelial cells through their expression of the epithelial markers Epithelial cell adhesion molecule (EpCAM), cytokeratins 8, 18 and 19, and the lack of CD45 leukocyte marker expression. A similar, widely used technology is AdnaTest BreastCancer (AdnaGen, Langenhagen, Germany), which uses both an anti-EpCAM antibody and an antibody against the epithelial cell surface associated glycoprotein, mucin-1 (MUC-1). However CTCs are heterogeneous, and such positive selection techniques based on expression of epithelial markers are limited by the potential loss of CTC expression of such epithelial phenotype as the CTCs develop a more stem cell-like phenotype and behaviour pattern [4,5] due to the epithelial-tomesenchymal transition (EMT) that occurs to facilitate tumour cell adhesion, motility and subsequent intravasation.
Although CellSearch has widespread popularity in the metastatic breast cancer setting, its use in early breast cancer is more limited. A study of 2026 higher-risk early breast cancer patients (66% node positive) confirmed the presence of CTCs as an independent prognostic marker for disease-free and overall survival. However only 21.5% of patients had CTCs, despite a large blood volume (30ml) analysed [6].
In metastatic breast cancer, a major proportion of CTCs show EMT and tumour stem cell characteristics, with such characteristics being associated with treatment-resistance and worse prognosis [7]. Cytokeratin-negative CTCs may represent a more aggressive CTC subpopulation and a majority of blood-borne tumour cells [8]. Therefore whilst more established CTC enumeration and identification technologies such as CellSearch and AdnaTest employ antiepithelial marker antibodies, more recently there has been a push towards size-based isolation to improve the CTC capture rate and better capture CTC heterogeneity. There is a range of promising non-epithelial based methodologies under evaluation, none of which have yet risen to the prominence that CellSearch has in epithelial-based CTC detection [3].
The ISET device (Rarecells) is a filtration based CTC enrichment technology that collects large blood cells (>8μm) including CTCs and circulating tumour microemboli (CTM) on a filter membrane [9]. Of the 'large cells' captured by this technique, CTCs are routinely identified by morphology. These CTCs and CTMs containing both epithelial and mesenchymal CTC subpopulations can then be subjected to immunomorphological, immunofluorescence, genetic, DNA or RNA analysis. ISET has allowed the identification of CTCs co-expressing cytokeratins and the mesenchymal marker vimentin [10]. Higher rates of CTC positivity by ISET than by Cellsearch using a morphology-based approach have been reported in NSCLC [11,12], pancreatic cancer [13], metastatic prostate cancer [11] and melanoma [14] despite larger cell size criteria. This is consistent with the presence of non-epithelial CTC populations. CTC enumeration by ISET has been correlated with shorter survival in hepatocellular carcinoma [15].
In this exploratory study, we aimed to determine the role of CTC enumeration by a nonepithelial marker dependent technique, using ISET in a high-risk early (non-distant metastatic) breast cancer population, with a view to developing novel biomarkers for response to treatment.
receptor positive MCF-7 breast cancer cells (ATCC HTB-22, obtained directly) spiked into healthy donor blood were used in assay development (S1 Fig). Scoring of CTCs. The mounted ISET spots were scanned using a Bioview automated light microscopy scanning system based on an Olympus BX61 microscope, using Duet software (Bioview, Rehovot, Israel). A circle of radius of 4000μm from the centre of each ISET spot was scanned to capture all cells. Galleries of images were manually reviewed using the Bioview Solo software. In the morphology based approach, CTCs were identified based on a set morphology definition as per previous publications [12][13][14] and under the guidance of a consultant breastspecialist histopathologist (SP). All images were reported by JC, with a subset of the images reviewed by SP for concordance. Cells greater or equal to 16μm diameter, with hyperchromatic nuclei and negative for CD45/CD144 brown chromogen staining were classified as CTCs (Fig  1). For the multiplex immunocytochemistry (ICC) ERα-based approach, ERα positive CTCs were defined as cells !16μm diameter with positive for ERα red chromogen staining and negative for CD45/CD144 brown chromogen staining. Four ISET spots corresponding to 4ml whole blood were ICC stained and scored for each blood sample. The number of CTCs detected was extrapolated to the equivalent of 7.5ml for comparison with concurrent Cell-Search analysis.
CellSearch CTC analysis. CellSave blood samples (7.5ml) were processed by the Cell-Search system as described elsewhere [16]. Briefly, CTCs are immunomagnetically separated from other blood components by EpCAM (epithelial cell adhesion molecule) antibody-conjugated beads and then stained for cytokeratins (CKs 8, 18 and 19) and CD45 in a fluorescentbased approach (S2 Fig). CTCs are defined as CK+ CD45-cells over 4μm in diameter; exact CTC diameters were not measured as the system software does not have this functionality.

Early breast cancer patients
To maximise the chance of finding CTCs, a relatively higher risk group of early breast cancer patients were recruited. At final pathology, 75% were node positive ( Table 1). Of the 16 patients, six had CTCs by CellSearch and all had CTCs by the ISET morphology-based assay at least one time-point (Table 2).
Using ISET, CTCs were detected in all 27 early breast cancer patient blood samples tested, whereas CTCs were only detected in 9 (35%) of the samples analysed by CellSearch. ISET CTC number was greater than CellSearch number in 26/27 samples (Fig 2). The median (range) number of CTCs per 7.5ml whole blood detected by CellSearch and ISET respectively were CellSearch: 0 (0-13) and ISET: 21 (2-62). The CellSearch system does not allow accurate cell size measuring so sizes of CTCs detected by each method were not compared.
There was no correlation between number of CTCs detected by ISET and detected by Cell-Search, either preoperatively, postoperatively or when all sample time-points were analysed together. Number of CTCs (ISET or CellSearch) did not correlate with tumour size, node positivity or timing of blood sampling (preoperation versus postoperation); however the limited sample size is acknowledged. The high number of CTCs found by ISET, independent of clinicopathological risk factors or CellSearch CTC numbers (although possibly reflecting an underpowered sample) raised the concern of false positive identification of CTCs by ISET.
High CTC false positive numbers are detected in healthy volunteers by ISET morphology method ISET filter membranes generated from three sex-matched healthy volunteer blood samples were analysed alongside the early breast cancer samples as presumed negative controls. Surprisingly, high numbers of false positive CTCs were detected, including one sample with 99 'CTCs'/7.5ml blood (Fig 3). These presumed false positives could not be morphologically distinguished from CTCs detected in the early breast cancer samples by the consultant breast histopathologist.

ISET healthy volunteer CTC false positives remain high in successive blood draws
It was hypothesised that epithelial cells may be shed into the bloodstream during the first draw of venous blood collection, which could partly explain the high numbers of CTC-false positives seen in healthy volunteer samples. In three further sex-matched healthy volunteers the number of CTCs detected in first, second and third blood draws were compared (Table 3). A trend towards lower 'CTC' numbers in successive draws was observed, but in all samples 'CTC' number was unacceptably high, highlighting the high false positive rate of the morphology based ISET technique. Low numbers of ER+ CTCs are detected in metastatic breast cancer patients Given the apparent high false positive CTCs identified by ISET-morphology, a similar but non-comparable ERα-based multiplex immunocytochemistry staining assay was employed with the aim of increasing specificity of CTC identification in ER positive breast cancer patients. The methodology was performed on blood taken from three ER-positive metastatic breast cancer patients. The samples from each patient were taken at the same blood draw, and processed for ISET (EDTA) and CellSearch CTC analysis. ISET samples from patients identified as CTC positive on CellSearch were examined. ERα positive CTCs were identified by ISET, however the number of CTCs identified by this methodology was markedly lower than by CellSearch, even in a high-CTC population indicating that this assay would be inappropriate for the low numbers of CTCs found in early breast cancer sample analysis (Table 4).

Discussion
Using the ISET platform and an epithelial-marker independent immunocytochemical assay we detected higher number of putative CTCs in early breast cancer samples than CTCs detected by the EpCAM/cytokeratin dependent CellSearch. However, in an unexpected result, by the same cell morphology based technique we also found high numbers of false positive 'CTCs' in healthy volunteer blood samples. By employing oestrogen receptor alpha as a positive marker we developed an immunocytochemical assay which we applied to a small number of ER+ metastatic breast cancer samples, however this technique identified minimal numbers of CTCs despite evidence of their presence on CellSearch. Much of the previously published literature on the use of the ISET device in detecting supposed CTCs using morphological criteria has found higher numbers of CTCs than are detected by epithelial antigens across a range of cancer types, despite larger size criteria [11][12][13][14][17][18][19]. These studies have identified 'CTCs' by a range of cell and nuclear morphological criteria unlike the standard epithelial-marker based approaches (Table 5). https://doi.org/10.1371/journal.pone.0175647.g003 Table 3. ISET ICC 'CTCs' in healthy normal control successive blood draws. Female healthy volunteer first, second and third draw whole blood samples were analysed by immunocytochemical (ICC) staining of ISET filters. Following venepuncture, three 10ml EDTA bottles were filled in succession from each volunteer, and analysed separately. The number of CTCs detected per 7.5ml blood is shown.  [15] in the first clinical article using the ISET technology described that CTCs/CTMs were detected in 53% of liver cancer samples by cytomorphological analysis but none were detected in 107 non-cancer patients, including many with other types of liver disease. However, the >25μm cell diameter criterion they used is in our view inappropriate for breast cancer. This size would exclude most of the breast cancer CTCs we have encountered by CellSearch and the most commonly used breast cancer cell lines [26]. Furthermore, an independent albeit smaller hepatocellular carcinoma study carried out in our laboratory concurrent to this work found no significant relationships between ISET-derived CTC number and clinical characteristics, raising doubts about the utility of identifying CTCs by morphology alone [19].

Healthy Female
In a subgroup of twenty metastatic breast cancer patients Farace et al. used a nuclear size cut-off of !16μm to define CTCs in ISET, finding 17/20 patients to be CTC-positive [11]. However, no healthy subjects were examined in their study as controls. Hofman et al. in large NSCLC-focused clinical studies that included a breast cancer patient subset have developed their own ISET cytomorphology-based scoring method [23]. Identified 'circulating non-haematological cells' (CNHCs) were further characterised as CNHCs with malignant features (CNHC-MFs, 'CTCs') if they exhibited four of the following criteria: anisonucleosis, nucleus size >24μm, irregular nuclei, presence of tridimensional sheets, and a high nuclear-cytoplasmic ratio. CNHC-MFs/CTCs were identified in 43% malignant disease but only 5% of non-cancer disease patients [24]. Although inter-observer agreement for detection of CNHCs was an impressive 100% (κ = 1) between three assessing cytopathologists, it is noteworthy that inter-observer agreement of CNHCs with benign features was relatively low (κ = 0.35). El-Hilibi et al. replicated their methodology in an alternative filtration based technology (Screen-Cell) and determined that morphological criteria alone were inadequate to distinguish malignant from non-malignant cells [27].
As CellSearch cell size criteria is >4μm, compared to our ISET definition of !16μm, it might be expected that CellSearch would identify higher CTC numbers. However, like several other authors, we found higher numbers with ISET [11][12][13][14][17][18][19]. Although many presume this is because of identification of non-epithelial antigen expressing CTCs, CTC false positivity is an acknowledged problem in filtration based technologies [28]. This is supported by our high CTC identification in normal controls. ISET-based studies have used large cell/nuclear size criteria to avoid misidentification of endogenous nucleated blood cells at the expense of identifying smaller CTCs (Table 5). Small CTCs ( 90 μm 2 ) have been described in prostate cancer [29] with small-nuclear CTCs being associated with visceral metastases [30], and therefore may represent an important subgroup where the ISET technology produces false negatives. There are several possible causes of the false positives seen in this study. It is possible that larger leukocytes may have been misidentified, especially monocytes which can reach up to 20μm in diameter [31], although the CD45 immunocytochemistry method employed has previously been shown to be effective at staining these cells [12]. Another candidate is megakaryocytes, responsible for platelet production with a 37μm mean cell diameter however these are usually discounted due to their round and pale nucleus and high reported CD45 expression [32]. Epithelial cells, which can be collected during peripheral venous blood collection by intradermal needle could also have been present on the filters and misidentified as CTCs [33]. In addition, El-Hilibi et al. suggested circulating endothelial cells as a probable cause of misidentified CTCs in their study [27]. Endothelial cells are released during venepuncture [34], with numbers shown to decline on subsequent draws [35]. This prompted us to examine 'CTC' numbers in successive blood draws as shown in Table 3. Although there was a trend of lower false positives in each subsequent draw, the high numbers seen in the third draw indicates that this is not the sole cause of this phenomenon. Also, by using VE-Cadherin, CD144 as a negative stain, we have gone further than the studies in Table 5 in actively identifying endothelial cells using a specific protein marker. Platelet endothelial cell adhesion molecule-1 (PECAM-1/CD31) as a further endothelial marker and CD61 as a marker for megakaryocytes may provide additional support in correctly identifying CTCs [36,37]. However, approximately 90% of all putative CTCs we observed on stained ISET filters created by standard device operation were sucked into the 8μm pores, which plainly affected their cell and nuclear morphology. This factor undoubtedly contributed to the high rate of false positives seen. Based on these results, morphology based ISET CTC enumeration is not appropriate for early breast cancer.
Oestrogen receptor alpha (ERα) expression in the primary tumour is routinely assessed in breast cancer clinical management [38]. Clinical studies have previously examined ERα expression in cytokeratin/EpCAM positive CTCs [39,40], but to our knowledge we are the first to identify CTCs primarily by using ERα as a positive marker. The lower cell count found Table 5. Published studies using ISET filtration technique to capture CTCs, with a variety of morphological identification techniques. Studies identifying CTCs from patient blood samples using ISET without the use of epithelial markers. All studies utilised nuclear morphology criteria such as nuclear contour irregularity, nuclear-cytoplasmic size ratio and hyperchromasia but with varying cell/nucleus size criteria and use of the leukocyte common antigen CD45 as a negative marker. N/A, Not Applicable; ND, Not Done.

Author (year)
Cancer  [42]. However the lower CTC numbers reported by ISET compared to CellSearch when ERα expression is included may possibly be a result of ISET failing to capture smaller 'true' CTCs. This further highlights the one of the challenges of the ISET technology. In addition, the number of ISET ERα positive identified CTCs in the setting of metastatic disease was surprisingly low compared to ISET 'CTCs' identified by just morphology in the early breast cancer subgroup, as it would be expected that CTC counts would be higher in more advanced disease. Our finding of low CTC numbers in metastatic disease compared to early disease further highlights the likely presence of false positives with the morphology-only technique used in the early breast cancer group. One of the particular challenges we encountered with this methodology is the time required for blood analysis. To ensure the integrity of nucleated cells, whole blood samples in EDTA vacutainers must be diluted in Rarecells buffer within four hours of blood taking, presenting challenges in sample transport and analyst availability. Blood processing and filtration using the ISET technology is non-automated and requires an average of 60 minutes of technician time per sample. However, even more challenging, the images produced from each spot corresponding to 1ml whole blood (up to 10 spots per sample) require 30 minutes to view and score, highlighting the limitations of this technology in its current form.
In summary, we investigated the use of the ISET technology to enumerate CTCs by a noncytokeratin / epithelial marker technique in early breast cancer patients. The high false positive rates based on morphology alone, and the low numbers captured (presumed high false negative) by ERα plus morphology raises doubts on its utility in this low burden disease. This study highlights the need for caution in the use of all non-cytokeratin based CTC enumeration methodologies, with a necessity for extensive examination of normal control samples. However, considering the current postulated plasticity between epithelial and mesenchymal states of migratory cancer cells [43] the detection of non-epithelial CTCs remains an important scientific and clinical goal.