Communication to the Editor
Customizing CellSearch platform
Article first published online: 2 AUG 2013
Copyright © 2013 International Society for Advancement of Cytometry
Cytometry Part A
Volume 83A, Issue 7, pages 595–598, July 2013
How to Cite
Rossi, E., Facchinetti, A. and Zamarchi, R. (2013), Customizing CellSearch platform. Cytometry, 83A: 595–598. doi: 10.1002/cyto.a.22292
- Issue published online: 2 AUG 2013
- Article first published online: 2 AUG 2013
- Manuscript Accepted: 9 MAR 2013
- Manuscript Revised: 1 MAR 2013
- Manuscript Received: 11 DEC 2012
To the editor:
We read with interest the recent manuscript from Lowes et al. , published in Cytometry A, reporting on the developing and optimizing protocols for the characterization of circulating tumor cells (CTCs) on the CellSearch platform for two proteins of interest, CD44 and M-30.
We totally agree with the authors that appropriate steps must be taken for proper optimization of user-defined protein marker assays on this system and this paper undertakes the laudable effort of defining a working method.
However, we have significant issues with the overall study design and the conclusions derived from this study.
A major concern regards the fact that the optimization experiments have been conducted on spiked samples only, without any further validation in ex vivo samples.
To detect tumor cells in peripheral blood, a general consensus exists that an optimal immunocytochemical method has to fulfill specific criteria, including, among others, proven clinical significance. To prove the clinical significance of a new method for tumor cell detection and characterization, this should be optimized in a pilot set of patients .
The CellSearch platform is a closed system that uses cytokeratin (CK) 8, 18, and 19 to identify tumor cells of epithelial origin. The widespread distribution of these keratins makes them useful markers for the great majority of carcinomas arising from simple epithelia.
At single cell level, the cytoplasmic CK staining can be strong and evenly distributed; more often, CK expression patterns are heterogeneous, with strongly- and weakly-stained parts [3, 4] asymmetrically distributed. Similarly, the expression of M30, a neo-epitope in CK18 that becomes available at a caspase cleavage event during apoptosis , is heterogeneous as well.
We were therefore disappointed by the study design pursued by Lowes et al. who, by assuming that M30 expression in an in vitro drug-conditioned cell line (usually showing high intensity CK-staining) would equal that of CTCs coming from a patient sample, limited the integrated assay optimization steps to cell lines (added to normal donor blood) only, and evaluated the percentage of M30-positive CTCs as the only mean to determine M30 monoclonal Antibody (mAb) suitability with the CellSearch system. To support such broad conclusions about the suitability of M30 in conjunction with the CellSearch system in cancer patients, samples representing more than one kind of cancer should have been tested.
To meet this criterion, when we customized M30 CTC assay, we started from cell lines spiked into whole blood samples, but afterwards we studied M30 expression on CTCs in a group of 122 patients affected by different kind of tumors, namely breast, colorectal, and renal cancer .
Overall, the results of this study by the CellSearch platform showed that 64 (52%) cancer patients had at least one CTC, and 56 (87.5%) CTCs-positive patients had one or more M30-positive CTCs (4). These data were consistent with the results obtained by different methods, by which apoptotic CTCs were detected in breast , prostate [6, 7], and lung  cancer patients; moreover an association between overall survival and presence of apoptotic CTCs has been also highlighted in CRPC .
Furthermore, by applying the M30 mAb in conjunction with the CellSearch assay, we have recently shown that the changes in live/apoptotic CTCs predict treatment failure in a group of 53 metastatic renal cell carcinoma (mRCC) .
A second key point concerns the choice of flow cytometry (FCM) as a reference method.
Despite the author's declaration that FCM cannot be used to compare CellSearch results, they de facto use FCM throughout the study for testing the suitability of user-defined markers.
User-defined markers in conjunction with the CellSearch system have been used either to detect treatment targets in CTCs [4, 10, 11] or to characterize CTCs [6, 7] in cancer patients; in none of these papers has FCM been used to assess the quality of the integrated assay, but only to evaluate the expression of the user-defined marker in that specific cell line, used as positive control. Moreover in these papers [4, 6, 7, 10, 11], the feasibility of the test was further assessed both in cell line spiked-samples and in ex vivo samples, by evaluating the quality of images on signal-to-background basis.
Due to the fact that the CellSearch system works at significant lower frequency than FCM, a close consistency between the percentages measured by the two cytometric assays is not expected. For CTC enumeration in cancer patients, CellSearch and FCM were directly compared , showing that, when CTCs are defined as EpCAM+ DNA+ CD45−, then FCM finds more CTCs (3.3-fold increase of sensitivity).
However, the addition of the positivity to cytokeratin staining, required for the definition of a CTC, as well as the required morphologic criteria, are expected to reduce this difference, because of both disruption/loss of fragile cells during complex immunostaining and more stringent selection criteria by CellSearch method .
Furthermore, taking into account that the CellSearch procedure has been clinically validated and FDA approved, a consensus has recently been reached that, for CTC-based biomarker development purposes, CellSearch should be used  as reference assay.
Therefore, it does not seem advisable to use the percentage of positive events measured by FCM with the purpose of assessing a high-standard quality CellSearch-based new test. As in the case of the CD44 in the Lowes et al. article, this may lead to inconsistent conclusions with the experimental results.
Due to the fact that the percentage of CD44-positive cells as determined by the CellSearch with anti-CD44 fluorescein isothiocyanate (FITC) significantly differed from the corresponding FCM “control”, Lowes et al. recommend to use the CXC kit. It is not clear however why the CXC kit designed for lower density antigens (about 50,000 antigens/cell) should work better than CTC standard kit (100,000 antigens/cell) in the case of CD44 expression on MDA-MB-468 cell line, that showed comparable high mean fluorescence intensity (MFI) values with both FITC and phycoerythrin (PE) mAbs by FCM.
In the CellSearch procedure, the anti-EpCam enrichment step precedes the immunostaining; hence, it does not seem conceivable that the contaminating quote of leukocytes in the enriched CTC fraction (about 1,000 cells per sample  as absolute number) can affect adequate CD44 visualization in the staining condition of large antibody excess (4 μg/ml) used in the Lowes article.
Finally, Lowes et al. raise criticism on the choice of M30 epitope as apoptosis marker and on its potential use as indicator of tumor response.
In principle, we believe that the choice of any user-defined marker for CellSearch assay should be guided by both compatibility with the standard test and by its clinical relevance.
As for apoptosis, several assays devised for detecting different components of the apoptosis signaling cascade or specific apoptotic features (including DNA fragmentation, caspase activity, membrane alterations, and mitochondrial changes) are currently available. Whenever possible, using more than one method is strongly recommended because of the limited sensitivity and specificity of these tests. Conversely, choosing the most appropriate apoptotic assay should be based on sample type (tissues or cellular effusions) .
It is precisely on the basis of this last item that we considered M30 as a good biomarker of treatment efficacy to be used in conjunction with the CellSearch standard assay. In fact, anti-M30 mAb defines an epitope of CK18 disclosing early phases of apoptosis , when, despite caspase cleavage, CK18 is still retained in a filamentous network and tumor cells still satisfy morphological features of CTCs according to CellSearch requirements. Moreover, a clinical correlation was reported between serum levels of these CK fragment and tumor load and prognosis both in breast and colorectal cancer (4).
The documentation of the clinical relevance of M30 analysis in cancer patients [4, 9] has been needed to demonstrate whether M30 had been a good choice for quantifying apoptosis in the CTCs.
Which would be the best strategy to assess the efficiency of a new CTC test?
Recently we have re-edited the γH2AX-CTC assay (firstly published by Wang et al.  as a pharmacodynamic biomarker) as an on-line procedure to be used in multicenter clinical trials. Data obtained during the procedure development confirmed and extend previously published data on M30-CTC assay (Fig. 1), indicating that M30-CTC assay recognizes early apoptosis in CTCs.
We measured the apoptotic fraction in drug-conditioned MCF7 cells by different FCM assays (Annexin V/PI, M30, and γH2AX) and mAb-integrated CellSearch assays. As shown in Table 1, as absolute events' number we obtained significant different results for early and late apoptosis (P < 0.001 in both cases) by the three methods.
These data are not surprising because it was reported that M30 and γH2AX recognize different phase in the apoptotic cascade [14, 15]; moreover, in the FCM analysis different gates have been used in the Forward Scatter/Side Scatter (FS/SSC) plot for Annexin V/PI staining (all events) and M30 and γH2AX staining (whole cells, nude nuclei excluded).
However, when we considered the fraction of both early and late apoptosis events obtained with each method (Ann/PI FCM, M30/γH2AX FCM, and M30/γH2AX CellSearch) consistent frequencies were obtained with Ann/PI FCM (P = 0.875) and M30/γH2AX CellSearch (P = 0.897) in comparison with M30/γH2AX FCM frequencies, which were used as expected value.
On this basis, for the initial evaluation of the performances of a new integrated CellSearch assay, we propose to use an internal positive control rather than an external one, in order to permit a comparison of FCM and CellSearch results.
In conclusion, we cannot help but stress the absolute need for shared criteria for the evaluation of any innovation in the CTC field, namely the transferability in a clinical investigation plan and the assessing of its clinical relevance.
1Oncology Section Department of Surgery Oncology and Gastroenterology University of Padova, Padova, Italy
2IOV-IRCCS, Padova, Italy
- 9Dynamic changes of live/apoptotic circulating tumour cells as predictive marker of response to Sunitinib in metastatic renal cancer. Br J Cancer (in press)., , , , , , , , , , et al.