BACKGROUND: Circulating tumor cells (CTCs) can be used clinically to treat cancer. As a diagnostic tool, the CTC count can be used to follow disease progression, and as a treatment tool, CTCs can be used to rapidly develop personalized therapeutic strategies. To be effectively used, however, CTCs must be isolated at high purity without inflicting cellular damage.
METHODS: We designed a microscale flow device with a functionalized surface of E-selectin and antibody molecules against epithelial markers. The device was additionally enhanced with a halloysite nanotube coating. We created model samples in which a known number of labeled cancer cells were suspended in healthy whole blood to determine device capture efficiency. We then isolated and cultured primary CTCs from buffy coat samples of patients diagnosed with metastatic cancer.
RESULTS: Approximately 50% of CTCs were captured from model samples. Samples from 12 metastatic cancer patients and 8 healthy participants were processed in nanotube-coated or smooth devices to isolate CTCs. We isolated 20–704 viable CTCs per 3.75-mL sample, achieving purities of 18%–80% CTCs. The nanotube-coated surface significantly improved capture purities (P = 0.0004). Experiments suggested that this increase in purity was due to suppression of leukocyte spreading.
CONCLUSIONS: The device successfully isolates viable CTCs from both blood and buffy coat samples. The approximately 50% capture rate with purities >50% with the nanotube coating demonstrates the functionality of this device in a clinical setting and opens the door for personalized cancer therapies.
Cancer patient morbidity is highly congruent with metastatic dissemination of cancer cells. Many studies have led to the conclusion that the prevalence of circulating tumor cells (CTCs)4 in peripheral blood correlates with disease severity (1–4). It follows, therefore, that the ability to capture CTCs in peripheral blood would potentiate cancer study, diagnosis, and treatment on a patient-to-patient basis (5). The prevalence of CTCs in blood is decidedly low, ranging from a few to several hundred CTCs per 7.5 mL blood (3). Detection of rare CTCs is complicated by the fact that similarly sized leukocytes are much more prevalent, ranging from 3.5 × 106 to 12.5 × 106 per mL (6).
Most current methods for CTC detection begin with Ficoll density centrifugation to separate leukocytes and CTCs from the remainder of whole blood and the use of antibodies or PCR to identify CTCs. CellSearch®, the only FDA-cleared system for CTC enumeration (4), separates CTCs from leukocytes by use of magnetic beads coated with anti–epithelial cellular adhesion molecule (anti-EpCAM) antibodies (3). In a separate approach, CTCs and leukocytes separated by Ficoll centrifugation are lysed, and RT-PCR is performed to detect mutated mRNA or epithelial markers. In both methods, the CTCs are destroyed. The ability to keep CTCs viable would provide not only individualized disease prognosis, but also a means for rapid in vitro analysis of the efficacy of different therapies on a patient-to-patient basis.
Devices to capture viable CTCs have been developed in recent years, generally involving microfluidic chips with CTC-specific antibodies coated onto the inner surfaces. The first generation was the CTC-chip, a chip containing anti-EpCAM–coated microposts (7). An improvement on the CTC-chip device is the GEDI chip (geometrically enhanced differential immunocapture), wherein the microposts are arranged in such a way as to maximize CTC–post collisions and minimize leukocyte–post collisions on the basis of cell size (8). The slow nature of EpCAM binding requires low flow rates for CTC capture, posing a challenge for feasible clinical use.
The microscale flow device previously developed by our group has been used to capture hematopoietic stem and progenitor cells as well as cultured tumor cells from flow without causing any cellular damage to the captured cells (see Supplemental Fig. 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol58/issue5) (9–11). The device consists of a 300-μm (inner diameter) microtube coated with selectins. In vivo, selectin binding is highly specific with rapid dynamics, allowing fast-flowing leukocytes to bind and roll on the endothelium. If appropriately stimulated while rolling, leukocytes will bind firmly to the endothelium via integrins and subsequently extravasate (12). CTCs may use the same process to metastasize to new tissues (13–19). CTCs exhibit stable rolling adhesion to a selectin-coated surface, providing a rapid method for CTC enrichment (20). CTCs can be more specifically targeted with a bimolecular surface consisting of both selectin molecules to induce slow-rolling and CTC-specific antibodies.
Recent work has examined naturally occurring halloysite nanotubes, nanoparticles composed of an aluminosilicate mineral having a typical diameter of approximately 300 nm, variable length that can be >1 μm, and negative surface charge at neutral pH (21). Enhanced capture of cultured cancer cells in buffer is accomplished with the nanotube coating because of the ability of the rough surface to overcome innate physical limitations to the capture of cells in microfluidic devices (22). In this study, we validated the utility of this device by first isolating cancer cells spiked into whole blood. We then demonstrated the potential for use in a clinical setting by isolating CTCs in 14 blood samples from 12 patients diagnosed with metastatic breast, prostate, lung, or ovarian cancer.
Materials and Methods
REAGENTS AND ANTIBODIES
We purchased RPMI 1640 cell culture media, fetal bovine serum albumin, penicillin-streptomycin, 1× PBS, goat anti-mouse IgG Alexafluor 488, and 4′,6-diamidino-2-phenylindole (DAPI) from Invitrogen; recombinant human E (rhE)-selectin and P-selectin fused to IgG from R&D Systems; mouse anti-human prostate-specific membrane antigen (PSMA) monoclonal IgG from Abcam; protein-G from EMD Biosciences; and blotting-grade blocker nonfat dry milk from Bio-Rad Laboratories. Halloysite nanotubes in water (6.6% wt) were provided by NaturalNano.
CELL CAPTURE FROM SPIKED BLOOD
An independent calibration was developed by our laboratory to estimate the total number of cells in the microtube with the following technique. Cells in flow buffer were drawn into a 300-μm Micro-Renathane microtube (Braintree Scientific) blocked with 5% milk protein, and allowed to settle for 10 min. We recorded 20 images at random locations along the length of the microtube and computed the average number of cells per micrograph area. We then eluted all cells within the microtube and measured the total number of cells in triplicate with a hemocytometer. A calibration factor to correlate the number of cells in a micrograph to the total number of cells in the microtube was thus calculated from the experimental data. We used the supplementation of fluorescent KG1a leukemic cells into whole blood and their subsequent capture on the selectin-functionalized microtube, determined previously (20), to estimate cancer-cell capture efficiency.
ACQUISITION AND PREPARATION OF PRIMARY CELLS
All patients and volunteers gave informed consent. Peripheral whole blood (7.5 mL) was collected by BioCytics Inc. from stage IV cancer patients seen at Carolina BioOncology Institute, PLLC. Blood samples were from 6 breast cancer patients (Bre-1 through Bre-6), 3 prostate cancer patients (Pro-1 through Pro-3), 2 lung cancer patients (Lun-1 and Lun-2), and 1 ovarian cancer patient (Ova-1). Patient Pro-2 provided 2 samples, approximately 2 months apart [Pro-2 (1) and (2)]. In some cases, whole blood was collected in BD Vacutainer EDTA tubes (BD BioSciences), and these tubes were immediately shipped overnight at ambient temperature to Cornell for processing. For buffy coat isolation and freezing at BioCytics, whole blood was collected in BD Vacutainer cell preparation tubes (BD BioSciences). The tubes were centrifuged at room temperature for 20 min at 1500g. Plasma was discarded and isolated buffy coat was rinsed once with 1× PBS. After rinsing, buffy coat was frozen in freezing medium (70% RPMI, 20% fetal bovine serum, 10% dimethyl sulfoxide) and transferred to liquid nitrogen until shipment. Frozen buffy coat was shipped to Cornell overnight on dry ice. Frozen samples were rapidly thawed to 37 °C and treated with 20 μg/mL DNase I. Cells were washed twice with RPMI medium, resuspended in medium, and incubated at 37 °C and 5% CO2 under humidified conditions for 3 h.
PRIMARY CTC CAPTURE
Buffy coat samples, washed twice with PBS by centrifugation at 120g for 10 min to remove the culture media, were resuspended in PBS saturated with Ca2+. We coated MicroRenathane microtubes with 1 microtube volume (35.3 μL) of 10 μg/mL Protein G for 1.5 h, followed by 1 microtube volume of a solution containing 5 μg/mL E-selectin-IgG and 50 μg/mL antibodies (anti-PSMA for prostate cancer patient samples, anti-EpCAM for all other cancer types) in PBS. Heparin (1000 U/mL) was then incubated in the microtube for 1 h to block nonspecific adhesion. Introduction of the halloysite nanotube coating has been described (22).
Each primary cell suspension was perfused through the microtube system at a shear stress of 1.5–4 dyn/cm2. We then washed the tubes with 300 μL PBS saturated with Ca2+ at 1 dyn/cm2 to remove unbound and loosely bound cells. Accutase was gently perfused into the microtube and allowed to incubate for 10 min to detach adherent cells, then medium was perfused through the microtube to elute cells. We collected the cells into 1 well of a tissue culture–treated 96-well plate (Becton Dickinson). Medium was changed on days 1 and 3, and the cells were analyzed on day 5.
We fixed cells by incubation with 2% paraformaldehyde in PBS for 20 min. After 2 gently washes in PBS, cells were incubated with 200 μL 10 μg/mL anti-EpCAM (for breast, lung, and ovarian cancer samples) and 200 μL 10 μg/mL anti-PSMA (for prostate cancer samples) for 30 min on ice, then washed with 300 μL PBS. Cells were then incubated with goat antimouse IgG conjugated to an Alexafluor 488 dye for 30 min on ice; DAPI nucleus stain was included with the secondary antibody to stain the nucleus. Cells were then washed twice with 300 μL PBS and analyzed for fluorescence by video microscopy. CTCs in culture were defined as being positive for an epithelial marker (EpCAM or PSMA), positive for DAPI, and between 10 and 25 μm in diameter.
PROCESSING OF HEALTHY BLOOD SAMPLES
Peripheral whole blood (7.5 mL) was drawn from 8 healthy volunteers by venipuncture into BD Vacutainer tubes and processed in a manner similar to that of samples from patients diagnosed with metastatic cancer. Briefly, the buffy coat was isolated by Ficoll density gradient and left at room temperature for 24 h. Each buffy coat sample was then halved and perfused through a smooth tube and a nanotube-coated tube, identically coated with single microtube volumes of 5 μg/mL E-selectin and 50 μg/mL epithelial antibody (anti-EpCAM or anti-PSMA). The buffy coat sample from Norm-6 was processed through a smooth tube coated with anti-PSMA antibody. Samples from participants Norm-1 through Norm-5 were perfused through EpCAM-coated tubes, whereas samples from participants Norm-7 and Norm-8 were processed in both EpCAM- and PSMA-coated tubes. Captured cells were then harvested from each tube and placed in an incubator. The number of viable cells was assessed on day 5.
CTC ENUMERATION WITH CELLSEARCH®
At the same time blood was collected for shipping to Cornell, an additional 10 mL blood was collected in a CellSave tube (Veridex). CellSave tubes contain EDTA as the anticoagulant, as well as a proprietary cell preservative that maintains the morphological and structural features of CTCs. Samples collected in CellSave tubes are stable for up to 96 h. We processed CellSave tubes within 96 h at BioCytics, Inc., with the CellSearch® Autoprep system and the CellSearch® CTC IVD kit according to standard operating procedures developed by the manufacturer (Veridex).
ANALYSIS OF LEUKOCYTE SPREADING
We prepared planar nanotube-coated surfaces in a manner identical to that used for coating microtubes. E-selectin at a concentration of 7.5 μg/mL was incubated on nanotube-coated and smooth surfaces for 2 h followed by blocking with 5% milk protein for 1 h. We isolated the buffy coat from whole blood of 3 healthy volunteers via density gradient centrifugation with 1-Step Polymorphs (Accurate Chemical & Scientific). A total of 105 cells were allowed to settle and contact each surface for 10 min. We washed the surfaces, immediately fixed the remaining cells with 4% paraformaldehyde, and stained the cell membranes with octadecyl rhodamine B (Invitrogen) or CellMask Deep Red (Invitrogen) according to manufacturer's instructions and the cell nuclei with DAPI (Vectashield Mounting Media, Vector Laboratories). We imaged the interface between the surface and an adherent cell with a Zeiss 710 confocal microscope and measured the area of a cell that was in contact with the surface with MetaMorph software (Universal Imaging). Statistical significance was determined with a paired nondirectional t-test.
We recorded 50–150 micrographs of each well at random locations and total cell counts along with CTC counts as described above. We normalized cell counts by the total well area and represented the uncertainty in total cell count estimate by SE. We determined statistical significance by unpaired nondirectional t-test with GraphPad Prism. Purity values were calculated as the fraction of cells that were CTC positive (EpCAM positive or PSMA positive and DAPI positive) compared with the total number of cells that were DAPI positive. P values were compared to a significance level of α = 0.05. The variability within individual samples was quantified via the bootstrap method (23).
ISOLATION OF HEMATOPOIETIC CANCER CELLS FROM MODEL BLOOD SAMPLES
We added known numbers of KG1a cells to 1:1 diluted whole blood and perfused blood samples through microtubes at a flow rate of 4.8 mL/h. The mean number of KG1a cells observed as a function of surface area was converted to the total number of KG1a cells in each microtube based on independent calibration. Tubes coated with P-selectin in addition to anti-CD34 antibody performed significantly better than those coated with anti-CD34 alone. Recovery of approximately 50% was accomplished for the range of spiked samples (Fig. 1).
CAPTURE OF PRIMARY CTCs ON SMOOTH SURFACES
We perfused isolated buffy coat samples from cancer patients through functionalized microtubes. Captured cells were removed from the microtubes and cultured for 5 days. Viable CTC enumeration in culture was accomplished by counting cells that were DAPI positive in addition to EpCAM- or PSMA-positive. We captured 20 to 703 [mean (SE): 172 (47)] viable CTCs, with a mean purity of 0.372 (0.030) (Fig. 2). Reproducibility of the device was evaluated by dividing sample Bre-3 into 2 identical samples and processing in separate devices, and a high degree of congruency was observed (see online Supplemental Fig. 2). Online Supplemental Fig. 3 shows a representative micrograph of captured CTCs after 5 days in culture.
CAPTURE OF PRIMARY CTCs ON HALLOYSITE NANOTUBE–COATED SURFACES
Microtube devices were coated with halloysite nanotubes and then functionalized in a manner identical to that described above. Samples from 6 patients were halved into 2 identical samples for parallel processing through nanotube-coated and smooth surfaces. Significant enhancements in the purity of processed samples were achieved for each trial (Fig. 3). Despite variability between patient samples, the purities of the cell populations captured on the nanotube coating were consistently greater. Specifically, the mean purity of captured primary CTC samples was 0.660 (0.039) on the nanotube surface and 0.372 (0.030) on the smooth surface, demonstrating a significant benefit provided by the nanotube coating (Fig. 4). Online Supplemental Fig. 4 shows a representative micrograph of captured CTCs after 5 days in culture.
CAPTURE OF CELLS FROM HEALTHY BLOOD
Peripheral blood from healthy participants was processed in a manner identical to that of the metastatic cancer samples. Of the samples collected from 8 different healthy participants, 2 contained cells that were identified as CTCs based on positive EpCAM staining on day 5. Norm-4 contained mean 11.6 (median 9.68, range 3.87–23.2) positive cells with both the smooth device and the nanotube-coated device, Norm-6 contained 7.75 (median 5.81, range 0–15.5) positive cells, and Norm-7 contained 3.87 (median 3.87, range 0–11.6) positive cells when processed through a nanotube-coated tube coated with EpCAM (Fig. 5). The number of positive cell samples was significantly smaller than the number of positive cells recovered in any of the metastatic cancer samples, except for Bre-5 processed through a smooth tube. The mean number of positive cells recovered in the healthy blood samples was 1.9 with the smooth tube and 1.7 with the nanotube-coated tube.
CELLSEARCH CTC ENUMERATION
The CellSearch test was used on samples identical to 12 of the 13 samples processed in the selectin-functionalized device. CTC counts ranged from 0 to 218 per 7.5 mL (mean 36.7, 25th–75th percentile 0–44.5). Fewer than 2 CTCs per sample were found in 8 of the 12 samples. For comparison to the selectin-functionalized samples, in which half of the 7.5-mL sample was processed through the smooth device and the other half through the nanotube-coated tube, the CTC counts reported by CellSearch were appropriately normalized (Fig. 5).
ANALYSIS OF LEUKOCYTE SPREADING ON SMOOTH AND NANOTUBE-COATED SURFACES
We measured the area of cell membrane in contact with a smooth or nanotube-coated surface functionalized with E-selectin with confocal microscopy. A significantly larger contact area was formed on the smooth surface than on the nanotube-coated surface [216 (60) μm2 and 110 (40) μm2, respectively, n = 3, pooled mean (SD)] (Fig. 6). The perimeter of the contact area was also compared on the 2 surfaces, and a significant difference was found [77.0 (21.8) μm smooth vs 43.5 (10.1) μm, pooled mean (SD)].
In this study, we demonstrated a straightforward microscale flow device that was able to isolate cancer cells from blood with high specificity and purity at clinically relevant concentrations in blood (Fig. 1) and primary CTCs from patients diagnosed with metastatic breast, prostate, lung, and ovarian cancers (Figs. 2 and 3). CTC capture was facilitated by E-selectin and CTC-specific antibodies, and high purities were accomplished with a halloysite nanotube coating. A notable feature of the device is that it mimics the natural process by which CTCs could adhere to the endothelium; thus the separation process may tend to select for those CTCs that are more able to metastasize (13, 17, 24) and does not inflict cellular damage (25).
We recently reported methods for enhancing cell capture in which the nanoscale topography of a capture surface was altered by use of immobilized nanoparticles (22, 26). The halloysite nanoparticle coating was found to increase the surface area onto which molecules could adsorb, as well as present molecules into the hydrodynamic lubrication layer that slows the sedimentation of cells to the device surface (22). Here, we show that the incorporation of the nanotube coating consistently captures primary CTCs at high purities (Fig. 3).
Cell response to nanostructured surfaces varies widely, depending on cell type, surface composition, surface feature dimensions, orientation, and organization. The modes of action are unclear (27, 28). Studies into the behavior of macrophages on microstructured surfaces have shown both increased spreading (29) and decreased spreading (30). A review of studies into the interaction of immune cells and nanostructured surfaces has concluded that materials with altered nanotopographies can generally be less inflammatory than planar surfaces (31). In the present study, we showed that there was a significant reduction in leukocyte spreading on E-selectin–coated halloysite nanotube surfaces (Fig. 6). This behavior was also observed under flow conditions (data not shown). The presence of spread and activated leukocytes on the smooth surface exhibiting diameters approximately twice those of resting leukocytes on the nanotube-coated surface is indicative of greatly increased adhesion strength, and is a likely explanation for the increased capture purities obtained with nanotube-coated surfaces.
Cancer cells were successfully captured from 2 different types of samples, diluted blood (Fig. 1) and buffy coat (Figs. 2 and 3), to demonstrate the use of the device in clinical settings where the handling of patient samples varies. An advantage of using samples in which CTCs are dispersed in blood is that the samples require only 1 step: a 1:1 dilution with buffered saline. This minimal sample processing reduces potential target cell damage or loss (32). In addition, the nature of blood flow in microvessels promotes margination, which assists CTC capture. A drawback of using whole blood samples is that a sample typically must be analyzed within 24 h of being drawn. Alternatively, buffy coat samples can be frozen, allowing samples to be shipped long distances and stored for extended periods of time. However, cell damage or loss in this process is possible.
Selectin binding is characterized by fast associations, which allows cells to be recruited from fast-moving blood flow to the endothelium (33). An important consequence of this behavior is that selectin-functionalized devices have the ability to process cells at high flow rates, a necessary feature for use in a clinical setting. As such, we report approximately 50% capture efficiency at a flow of 4.8 mL/h. Previously developed systems that rely entirely on antibody binding to capture CTCs achieve similar capture efficiencies and purities, but require a flow rate of 1 mL/h. Subsequent advances to the CTC-chip geometry have resulted in greatly enhanced capture efficiency at low flow rate (approximately 1 mL/h), and approximately 40% efficiency at flow rates near 4 mL/h (34). Several other microfluidic devices of varying geometry and orientation have been developed (8, 35, 36); however, because of their use of antibodies alone, all require reduced flow rates to attain maximal capture purity levels.
The selectin-functionalized device has various parameters that affect capture efficiency and purity, including protein density, flow rate, device length, and surface roughness. Previous work studied the effects of these parameters on cell capture (9, 10, 22, 37). Based on these studies, it is clear that high concentrations of selectin molecules, approximately 20 μg/mL, result in increased leukocyte recruitment. It has also been observed that shear rate can affect cell capture rate; however, shear stress does not have much effect on adherent cells within the physiological range. As such, we chose a flow rate that allows all cells to settle to the surface before exiting the tube.
Of the 8 healthy donor samples processed, 3 had positively stained cells (11.6, 7.75, and 3.87). Two of these 3 samples were processed through tubes coated with anti-EpCAM, whereas only 1 captured a small number of cells with anti-PSMA. These results represent the limit of specificity of the antigens targeted by the device. It is interesting to note that the combination of the nanotube coating with E-selectin and anti-PSMA antibodies resulted in zero positively stained cells detected in control blood (n = 3).
The only FDA-cleared method for quantification of CTCs in patient blood, CellSearch, was employed on samples identical to those processed in the selectin-functionalized device for comparison against this gold standard. There was substantial discord between CellSearch and the selectin-functionalized device: 5 of 12 samples processed by CellSearch were negative for CTCs, whereas these 5 patients were positive for CTCs with the selectin-functionalized device. Altogether, 11 of 12 samples were positive for CTCs with the smooth selectin-functionalized device and 7 of 7 were positive with the nanotube coating, according to the most conservative threshold. Moreover, the CTC yield with the selectin-functionalized device was significantly higher (Fig. 5).
In summary, this study reports on a novel system for isolating viable CTCs from patient samples. This device is a straightforward system consisting of a polyurethane microtube functionalized with a bimolecular surface of selectin and antibody molecules that allow for the reproducible selection and retention of CTCs in the device without inflicting cellular damage. We further report on the application of a halloysite nanotube coating, which leads to enhanced device performance. We believe that this novel bimolecular device is a promising new tool for cancer treatment in the clinical setting.
↵4 Nonstandard abbreviations:
- circulating tumor cell;
- anti–epithelial cellular adhesion molecule;
- geometrically enhanced differential immunocapture;
- recombinant human E;
- prostate-specific membrane antigen.
(see editorial on page 803)
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the Disclosures of Potential Conflict of Interest form. Potential conflicts of interest:
Employment or Leadership: J. Powderly, Carolina BioOncology Institute, PLLC, and BioCytics, Inc.; B. Greene, BioCytics, Inc., and Carolina BioOncology Institute, PLLC.
Consultant or Advisory Role: M. King, CellTraffix, Inc.
Stock Ownership: J. Powderly, Carolina BioOncology Institute, PLLC, and BioCytics, Inc.; B. Greene, BioCytics, Inc.
Honoraria: None declared.
Research Funding: A. Hughes, National Science Foundation Graduate Research Fellowship; M. King, NIH grant CA143876.
Expert Testimony: None declared.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
- Received for publication October 7, 2011.
- Accepted for publication January 9, 2012.
- © 2012 The American Association for Clinical Chemistry