High Sensitivity Immunomagnetic CTC Isolation as Compared to Alternative Isolation Methods 1. Introduction: An overview of CTC isolation methods 2. Challenges for direct comparisons of CTC recovery 3. Immunomagnetic isolation depends on antigen expression levels and system sensitivity 4. Patient data for high sensitivity immunomagnetic CTC isolation 5. Conclusions The IsoFlux System is a high sensitivity immunomagnetic enrichment system that utilizes a microfluidic technology to isolate circulating tumor cells (CTCs) in a format optimized for downstream molecular analysis. 1. Introduction: An overview of CTC isolation methods The adoption of circulating tumor cell (CTC) analysis in clinical research is increasing due to the ability to monitor the status of important molecular cancer biology biomarkers, longitudinally, without requiring a tissue biopsy for each test. CTCs are tumor cells that have entered the vasculature and traveled away from the primary tumor site. They are very rare (1 s cells per 1cc blood tube) and separating them from surrounding white blood cells (WBCs) is challenging. CTCs differ from surrounding WBCs in terms of pathway activity and expression profiles. While DNA abnormalities also exist, expression differences result in directly observable size differences and in varying expression of protein markers. There are two main principles that have been proposed for CTC isolation from blood and other fluids. First, immunological selection, which is dependent on antibody binding to a known biomarker (protein) expressed and presented on the CTC surface. An antibody may be used to either bind magnetic beads (, IsoFlux, immunomagnetic separation) or solid substrates ( herringbone chip, On-Q-Ity, BioCept) that have been functionalized to bind antibodies to a specific surface marker. In FACS, fluorescently labeled antibodies have been used to identify and separate CTCs. Immunological selection has most often employed binding to the epithelial marker EpCAM, but has also been demonstrated using other markers such as N-cadherin, vimentin, or a combination of markers. Second, selection based on cellular morphology and physical properties that includes size-based filtration and dielectrophoretic separation. Proponents of separation based on physical characteristics often tout the technique as a marker-independent approach (Farace et al., Krebs et al.). However, like surface protein expression, cell size is also regulated by cellular pathways and expression profiles. As an example, the PI3K/mTOR pathway has been shown to regulate cell size as well as apoptosis (Kozma et al, Fingar et al). Current research shows mammalian cell size to be dependent on pathway activation and gene expression, so this physical property depends on its own set of markers that are not explicitly known for the tumor cells in question. Like cell size, immunological selection is based on upregulation of certain pathways and changes in protein expression profiles. The changes include presentation of a larger amount of a set of protein markers at the cell surface. Antibody binding to these surface proteins is then used to separate CTCs from leukocytes. Different antibodies may be used in order to select distinct CTC populations or optimize the capture process for different indications. Similarly, dielectrophoretic force separation relies on physical differences between CTCs and WBCs; the cellular process changes that lead to different responses to electric field forces are not well understood.
This report examines the difference between CTC recovery rates for different platform technologies and explores the likely causes for the observed differences. 2. Challenges for direct comparisons of CTC recovery There are a number of factors that make consistent comparisons between different CTC isolation methods difficult. First, it is difficult for a number of different methods to be tested using matched patient samples, since each platform typically requires a 1cc blood tube as the starting volume. This means that most publications only compare 2 different technology platforms using matched samples, which is the best way to control for biological variability. Punnoose et al. were able to test 3 different platforms side by side. One notable exception is the work that Bayer AG (Berlin, Germany) has undertaken (Stresemann, et al.), where at least 6 different technologies were tested using matched samples, including the IsoFlux System. This work was enabled by the development of leukapheresis techniques that extract a large amount of buffy coat from patients, permitting about 3 different experiments to be performed at quantities on the order of a 1cc blood draw tube per experiment. Secondly, much of the available counting data on matched samples do not employ exactly matched CTC identification criteria. Because the system was the first entrant to this market, their identification criteria (immunofluorescence methods identifying CK+, CD45- and DAPI+ cells) have become somewhat of a standard for CTC identification. While the debate over the exact definition of the CTC continues, it is our belief that in order to compare tumor cell recovery across platforms, this definition should still be used. The IsoFlux and data presented in this white paper use the above definition of a CTC, while other platforms use slightly different methods. An overview of the CTC counting criteria used is presented in Table 1. Finally, because a number of platforms are a pre-commercial stage or only offer in-house CTC isolation services, multisite evaluation of the technology is not possible. Testing at different sites by a number of different operators is the best way to accurately determine the performance of any isolation method. Because both the Veridex platform and a number of other emerging technologies use immunomagnetic separation, we will focus on platform-to-platform differences using magnetic bead based separation followed by a comparison to filterbased approaches. 3. Immunomagnetic dependence on antigen expression can be minimized An important aspect of immunomagnetic separation that is often overlooked is the dependence of recovery efficiency on antigen expression, in most cases EpCAM expression. Patient tumor samples and tumor cell lines both show significant variability of greater than 1x in antigen expression that is the case in CTC populations as well (Punnoose et al.). Figs. 1 and 2 summarize the reported variability in EpCAM expression for both cell lines and tumor tissue from a number of groups. Platform Study Method CTC Definition (immunomagnetic) All studies using patient samples CellTracks CK+, CD45-, DAPI+ (nucleated) Prometheus collaboration Semiautomated flourescence microscopy (using Profile Kit) CellTracker (prelabeled spike-in cells) IsoFlux (immunomagnetic) All studies using patient samples Semiautomated flourescence microscopy CK+, CD45-, DAPI+ (nucleated) Fluxion analytical samples Semiautomated flourescence microscopy CellTracker (prelabeled spike-in cells) ISet (filter based) Farace et al. Microscopy, brightfield and flourescent IHC: large nucleus > 16µm, nucleus to cytoplasm to cytoplasm ration >.8, irregular shaped nuclei
Figure 1. A. EpCAM expression in tumors can span 1-2 orders of magnitude. B. Expression of EpCAM in relationship to other epithelial and mesenchymal markers in breast cancer cell lines (adapted from Punnoose et al. 21). In order to characterize recovery efficiency analytically, several groups have performed experiments looking at model cancer cell lines of varying EpCAM expression levels. When characterized using fluorescent EpCAM antibody binding (either by microscopy or flow cytometry), tumor cell lines present a wide range of expression, from about 2x to 5x the signal from non-epcam expressing controls (Fig. 1). Therefore, cell lines can be characterized as having low, medium, or high EpCAM expression (i.e. MDA-MB-231, PC3 and SKBR3 respectively - Fig. 2). For cells present in solid tumors, the expression levels are significantly more heterogeneous and span a large range (Fig. 1). Figure 2. EpCAM expression in cell lines. Cytometry and immunofluorescence have been used to characterize EpCAM expression in commonly used tumor model cell lines as compared to controls by Sieuwerts et al. (A) and Ozkumur et al. (B). Some of the cell lines where cross-platform recovery data is available are shown in (C) and fall into low (MDA-MB-231), mid (CAL-12, PC3) and high (SKBR3, MCF1) categories. While cell lines don t reproduce in vivo heterogeneity, spikein experiments have the advantage of yielding an absolute % recovery metric for the systems employed. Another advantage is unambiguous counting results if cells are fluorescently labeled before the spike-in step. A great majority of the analytical data characterizing EpCAM based CTC isolation has been obtained on the system until recently, this was the only commercially available immunomagnetic separation instrument. That system has been shown to perform very well (recovery > ) for cell lines that are high EpCAM expressers like SKBR3 and MCF7. Recovery is dependent on the amount of
EpCAM expressed however. When comparing recovery across cell lines with varying levels of EpCAM expression, the literature consistently reports lower recovery for cell lines that are in the mid to low expresser category (Sieuwertz et al., Punnoose et al.). Immunological assessment of antigens in breast cancer cell lines with different intrinsic subtype characteristics and circulating tumor cell recovery Intrinsic No. of Flow cytometry, MFI, % subtype cell CD45 CD24 CD44 EpCAM cells recovered lines (95% CI) Normal-like 6 <5 <5 > <5 2 ( to 6) Basal-like 5 <5 5-2 2-2-2 48 (36 to 61) Luminal 5 <5 5-2 5-2 2-2 75 (62 to 89) HER2-positive 3 <5 5-2 <5 2-2 86 (61 to 18) Table 2. Measurements of EpCAM based recovery. model cell recovery is dependent on EpCAM expression (adapted from Sieuwertz et al.). MDA-MB-231, a low expresser model cell line, is part of the Normal-like subtype in the table above, whereas SKBR3 is part of the HER2-positive category and a high EpCAM expressers. A number of publications and recent data report recovery in the range of 2- for mid level expressers (CAL-12, PC3) (Punnoose et al, BioCytics/Fluxion internal) and 12% and below for low expressers like MDA-MB-231 (Sieuwerts et al.). Low analytical recovery from a number of cell lines using the system has led a number of groups to conclude that EpCAM-based CTC isolation is a low sensitivity technique that misses a significant percentage of CTCs present in patient samples (Farace et al., Sieuwertz et al., Punnoose et al.). In contrast to that conclusion, recent studies are finding that for lower EpCAM expression levels, isolation efficiency is gated by the sensitivity of the immunomagnetic separation system used (Ozkumur et al., Stressman et al.). Immunomagnetic recovery depends on a number of systemspecific parameters such as the magnetic field gradient in the separation region, fluid flow, magnetic bead reagents, and antibody binding efficiency. It is not surprising therefore that improved microfluidic technologies are demonstrating significant recovery improvements for low to mid EpCAM expressing cell lines, as well as CTCs from patient samples. Recent data produced by IsoFlux users and others have demonstrated that it is possible to efficiently recover cells of much lower EpCAM expression by designing immunomagnetic separation systems with much higher sensitivity (Table 3 and Fig 3). Cell line EpCAM Expression % Microfluidic immunomagnetic References SKBR3 Hi 8 85 Punnoose et al. / Fluxion data CAL-12 Mid 25 NA Punnoose et al. / NA PC3 Mid 42 9 BioCytics / Fluxion data MDA-MB-231 Low 12 74 Sieuwerts et al. / Fluxion data Table 3. Recovery of cell lines depends on the sensitivity of immunomagnetic separation systems. IsoFlux recovery is significantly higher for both mid and low EpCAM expressers, and analytical recovery data is independent of CTC definitions or counting bias. Percent 9 8 7 6 5 4 3 2 1 Percent Recovery ( cell spike-in) IsoFlux Low (MDA-MB-231) Mid (PC3) Hi (SKBR3) Figure 3. Recovery comparisons for 3 types of model cell lines. IsoFlux cell line recovery was measured by fluorescently labeling target cells before spiking into blood, making the results independent of counting bias. PC3 recovery for both platforms was measured using exactly matched protocols. MDA- MB-231 and SKBR3 data is reported in literature using similar spiking experiments (Punnoose et al., Sieuwerts et al.) This data provides an analytical explanation for the improved CTC recovery observed in patient samples using the IsoFlux System. In addition, measuring analytical recovery measurements has the significant advantage of being free of counting bias, because fluorescent labeling is well understood for cell lines, and cells can be labeled with CellTracker dyes before spike-in. For MDA-MB-231, the lowest expresser tested, recovery increases from 12% for to 74% for the IsoFlux instrument. IsoFlux data on MDA-MB-231 is shown in Fig. 4B. The strongest data available concerns the mid level EpCAM expressing PC3 cell line. For this set of experiments the same spiking and imaging/counting protocols were used for both platforms by Prometheus, a user of the IsoFlux platform. For, cells were recovered using the profile kit (after
separation) and counted using the same hardware, staining, and image analysis protocols as IsoFlux samples. The results are presented in Fig. 4A. Percent 12 8 6 4 2 25 2 Percent Recovery ( cell spike-in) PC3 Count - IsoFlux PC3 Count - S1 S4 S3 S4 Mean y =.64x y = 1x 4. Patient data for high sensitivity immunomagnetic CTC isolation Based on improved recovery for low to mid EpCAM cell lines for microfluidic platforms like the IsoFlux, we expected improvements in CTC recovery from patient samples. For the samples enumerated here, CTC identification followed the same definition employed by the System: tumor cells are defined as being CK+, CD45-, and DAPI+ (nucleated) by microscopic evaluation. As expected, CTC counts from patient samples also improve dramatically with higher sensitivity immunomagnetic separation (Fig 5). This is likely due to the heterogeneous EpCAM expression levels in CTCs, combined with the higher sensitivity of the microfluidic system to any EpCAM being present. Recent publications indicate a lowering of EpCAM expression levels when tumor cells enter the circulation, making the isolation of the low expressing population very important. 18 8 5 15 2 25 SPIKE IN MDA-MB-231 Cells Low EpAM Figure 4. IsoFlux analytical recovery data. A comparison of % recovery for PC3, a mid EpCAM cell line, in matched samples is presented alongside recovery and linearity data for MDA-MB-231, a low EpCAM expressing line. Comparatively, literature reports only 12% recovery for MDA-MB-231 using (Sieuwerts et al.) Literature data supports the possibility for much higher efficiency CTC capture using immunomagnetic methods paired up with microfluidic technology for controlling flow and force distribution. For example, a recent publication from the Toner and Haber labs also reports significantly higher recovery of mid expressing PC3-9 cells using a microfluidic immunomagnetic separation approach as compared to (Ozkumur, et al.). A B 9 7 5 3 1 9 7 5 3 1 Bladder Allard et al % 5 CTCs % 1 CTCs Bladder Breast Colorectal Lung Pancreatic* Prostate Total Breast Punnoose et al. Breast Farace et al. Colorectal Allard Lung Tanaka et al. Lung Farace et al. Pancreatic Kurihara et al. % 5 CTCs % 1 CTCs Prostate Danila et al. Prostate Farace et al. Figure 5. Patient CTC recovery data. In (A), a comparison of % patients that are CTC positive (5 cell and 1 cell cutoffs) using IsoFlux. In (B), literature values are shown using the system. IsoFlux has increased the % patient samples with >5 CTCs recovered, from approximately to above on average. For IsoFlux healthy controls (n = 8, data not shown) none of the patients display counts above 5 CTCs. *Pancreatic samples were obtained using leukapherisis collection. Note that literature data for recovery is surprisingly
consistent across publications from different sites using the system. The gains in recovery (% patients with at least 5 CTCs per tube) for high sensitivity immunomagnetic separation are most dramatic for lung, colorectal and pancreatic indications. Data for healthy controls analyzed using the IsoFlux system showed a median of 1 cell (for nonzero CTC count samples) and no healthy controls contained more than 5 counted CTCs. To tie it all together, our data indicates that for the IsoFlux system, better sensitivity to cell lines of low antigen expression (i.e. low EpCAM, Fig. 6B) translates into higher CTC recovery for patient samples (Fig. 6A). Overall, more than of patients tested were CTC positive (defined as >=5 CTCs per 7.5ml blood draw) across all indications. A % Patients >5 CTCs Healthy Pancreatic Lung Bladder Colorectal Breast Prostate B Percent Recovery ( cell spike-in) 9 8 7 6 5 4 3 2 1 Low (MDA-MB-231) Mid (PC3) Hi (SKBR3) Figure 6. Patient sample CTC recovery in relation to cell line recovery data. The higher CTC recovery shown in (A) for patient samples mirrors increased sensitivity to low to mid EpCAM expressing tumor cell lines (B). IsoFlux For other recovery methods, like filter-based isolation, studies to date have also shown significantly higher CTC counts in comparison to. A majority of studies have drawn the conclusion that lower reported numbers are due to the absence of EpCAM expression and thus that size-based separation is more efficient at CTC isolation than immunological separation. By contrast, our data indicates that when sensitivity to EpCAM expression is increased, recovery is significantly improved with respect to. In some cases, the CTC numbers as recovered using filter technology are comparable to IsoFlux recovery, while for others, filter-based recovery did not perform as well. A summary of the data for breast, prostate, and lung cancer is shown in Fig. 7. While size-based CTC isolation works well for some indications, for others the smaller CTC size affects recovery (Fig. 7B, C). The numbers reported are adjusted for a 7.5 ml blood sample, but ISet instruments can only process 1ml blood per filter and the results are pooled together. For molecular analysis, filter-based approaches have the additional challenge of removing the cells from the filter substrates and retaining more white blood cells onto the filter. The other challenge in drawing a comparison stems from the fact that cells are counted via microscopy while still on % Patients over threshold 9 7 5 3 1 IsoFlux CTC Recovery Breast Fluxion Prostate Fluxion Lung Fluxion collaborator collaborator collaborator % Patients over threshold 9 7 5 3 1 Filter-based (ISet) CTC Recovery Breast Farace et al. Prostate Farace et at. Lung Farace et al. % Patients over threshold 9 7 5 3 1 CTC Recovery Breast Farace et al. Prostate Farace et at. Lung Farace et al. % 1 CTCs % 5 CTCs % 1 CTCs % 1 CTCs % 5 CTCs % 1 CTCs % 1 CTCs % 5 CTCs % 1 CTCs 1 IsoFlux % 5 CTC ISet % 5 CTC % 5 CTC Breast Prostate Lung Figure 7. Size-based CTC recovery data as compared to IsoFlux and. (A) A comparison of % patients that are CTC positive (1, 5, and 1 cell cutoffs) for ISet (filter-based), IsoFlux and technologies across three different pathologies as reported by published reports. (B) Percent patients about our mutation detection LOD cutoff of 5 cells for the three different separation technologies across three indications. Note that both ISet and IsoFlux have improved performance for prostate and lung patients, but filter-based recovery is lower for breast cancer. A possible explanation is the lower CTC size observed for some breast cancer patients.
the filter and the CTC definition advanced by ISet doesn t match the definition of CK+, CD45-, and DAPI+ (see Table 1). There are no literature reports that include CTC enumeration after removal from the filter substrates, so it s hard to directly compare this approach to others in terms of cells recovered outside the isolation device. Both and IsoFlux count cells recovered outside the principle immunomagnetic device by either spotting onto a standard microscopy slide (IsoFlux) or the counting cartridge (Magnest). 5. Conclusions While a number of techniques have been used for CTC isolation in a research setting, commercial instruments exist for immunomagnetic (), and immunomagnetic microfluidic (IsoFlux), and size-based separation. (ISet, other filters). In analytical validation work, IsoFlux instruments demonstrate improved EpCAM based CTC recovery of cell lines that are low to mid EpCAM expressers with respect to. The improved sensitivity translates into higher CTCs recovered from patient samples as compared to the system. For patients across all indications tested, IsoFlux EpCAM-based recovery results in significant increases in the % patients presenting 5 CTCs (median 84% of patients across all pathologies compared to median of 3- for ). For both systems, the traditional CTC definition of CK+, CD45-, DAPI+ was used with samples evaluated via fluorescence microscopy. Size-based separation also demonstrates higher CTCs recovered as compared to across most indications, with the exception of breast cancer where a higher overlap in CTC and WBC size was observed. ISet to count comparisons are somewhat confounded by a divergent CTC definition employed by ISet, including nuclear size and shape and excluding CK+ expression. Previous literature studies attributed low CTC recovery rates to a lack of EpCAM expression, but all data was generated with the low sensitivity system. For both the IsoFlux system and other experimental high sensitivity microfluidic systems, EpCAM based isolation yields much higher CTC recovery rates. The high sensitivity IsoFlux System enables a wide range of downsream molecular studies using CTCs. The increased sensitivity leads to a greater percentage of patients across all carcinoma indications (>), that can have enough CTCs collected for analysis. Additonaly, Fluxion has validated several downstream assays that can be used with IsoFlux CTC samples. One example is a mutational profiling assay for common oncogene targets (KRAS, BRAF, EGFR, etc.) that uses a standard qpcr instrument. Please visit www.fluxionbio.com/isoflux for more information. 6. References Allard et al., Clin Cancer Res (24) 1(2):6897-94 Danila et al., Clin Cancer Res (211) 17: 393-3912 Farace et al., British Journal of Cancer (211) 15: 847 853 Fingar et al., Genes Dev. (22) 16(12): 1472 1487 Kozma et al., Bioessays (22) 24(1): 65-71 Krebs et al., J Thoracic Oncology (212) 7(2): 36-315 Kurihara et al., J Hepatobiliary Pancreat Surg (28) 15:189 195 Ozkumur et al., Science Transl Med (213) 5: 1789ra47 Punnoose et al., PLOS One, (21) 5(9): e12517 Sieuwerts et al., J Natl Cancer Inst (29) 11: 61 66 Stresemann, et al. A comprehensive comparison study: Capturing of CTCs by different technologies followed by molecular analysis. Poster presentation, ATCC Conference (212) Tanaka et al., Clin Cancer Res (29) 15(22): 698-6986 385 Oyster Point Blvd., #3 South San Francisco, CA 948 www.fluxionbio.com About Fluxion Biosciences Fluxion Biosciences provides cellular analysis tools for use in critical life science, drug discovery, and diagnostic applications. Fluxion s proprietary microuidic platform enables precise functional analysis of individual cells in a multiplexed format. Products include the BioFlux System for studying cellular interactions, the IonFlux System for high throughput patch clamp measurements, and the IsoFlux System for rare cell access. Fluxion s systems are designed to replace laborious and difficult assays by providing intuitive, easy-to-use instruments for cell-based analysis. 213 Fluxion Biosciences, Inc. All rights reserved. IsoFlux and CellSpot are trademarks of Fluxion Biosciences, Inc. Rev 1.