The detection of disseminated tumor cells in peripheral blood is limited by the presence of very few tumor cells within a large number of blood cells. Therefore, tumor cell detection calls for enrichment systems with effective depletion of blood cells and high tumor cell recovery.
We compared the new density gradient centrifugation method OncoQuick with the standard method of Ficoll. The enriched cell fractions were quantified. Tumor cell spiking experiments examined the recovery of tumor cells as detected by immunocytochemistry and cytokeratin-20 reverse transcriptase–polymerase chain reaction (RT-PCR). Clinical application of OncoQuick was evaluated in 37 peripheral blood samples of patients with gastrointestinal carcinomas.
The depletion of mononuclear cells (MNCs) in the enriched cell fraction after OncoQuick centrifugation was 632-fold, with an average cell number of 9.5 × 104, compared with Ficoll, with a depletion factor of 3.8 and a mean number of 1.6 × 107 MNCs. The mean tumor cell recovery rates were 87% for OncoQuick and 84% for Ficoll. The increased depletion of MNCs with OncoQuick centrifugation further simplified immunocytochemical evaluation by reducing the number of cytospins and increasing the tumor cell density. Due to the reduced number of co-enriched MNCs by OncoQuick, the blood volume, which could be analyzed in one RT-PCR reaction, was increased up to 30 ml. Examination of peripheral blood samples from 37 patients with gastrointestinal tumors showed a cytokeratin-20 detection rate of 30% and a significant correlation with the presence of distant metastases (P < 0.02).
Despite complete resection of the primary tumor, about 50% of patients with malignant gastrointestinal tumors develop recurrent disease (1). Most likely, this is due to early systemic dissemination of single tumor cells (2). The prognostic value of disseminated tumor cells has repeatedly been described for patients with epithelial tumors (3, 4). However, the detection of minimal residual disease (MRD) does not represent an established prognostic factor and has not reached the level of routine clinical diagnosis in gastrointestinal cancer (5, 6). Therefore, there is great interest in improved methods for the detection of disseminated tumor cells in patients with solid tumors (7).
Methods for detection of disseminated tumor cells are based on monoclonal antibodies directed against epithelial proteins (immunocytochemistry, ICC) (8) or on reverse transcriptase–polymerase chain reaction (RT-PCR) amplifying transcripts expressed by tumor cells (9–11). The specificities of ICC detection and molecular biological analysis of circulating tumor cells, however, are limited by the lack of tumor-specific markers. Therefore, cytokeratins are currently the most valuable markers for the detection of circulating tumor cells (12), although its specificity is influenced by illegitimate transcription, e.g., in granulocytes (13). In addition, the detection of disseminated tumor cells in blood is methodically limited by the small number of circulating tumor cells among a large number of blood cells. Consequently, detection of circulating tumor cells in blood calls for effective enrichment of tumor cells.
Two methods for the enrichment of tumor cells in blood are currently established. The immunomagnetic cell separation separates tumor cells marked with magnetic beads via antibodies from unlabeled cells (14–17). In density centrifugation methods, erythrocytes, platelets, and polymorph nuclear cells are separated in the pellet, and mononuclear cells (MNCs), including tumor cells, gather in the so-called interphase. These interphase cells are used for the further evaluation of tumor cells by ICC or RT-PCR.
The aim of the study was to establish a density gradient centrifugation system that allowed improved enrichment of tumor cells and a high depletion of MNCs in blood for optimized detection of disseminated tumor cells in blood. We therefore evaluated a new density centrifugation system, OncoQuick, and compared it with the current standard method with Ficoll.
MATERIALS AND METHODS
Blood Preparation by Density Gradient Centrifugation
For all experiments, peripheral venous blood was drawn from tumor-free patients by using 10-ml monovettes coated with ethylene-diamine-tetraacetic acid (Sarstedt, Nuernbrecht, Germany) and processed within 2 h after collection. All blood samples were drawn with informed consent of patients according to institutional polices and approval by the institutional review board. All experiments were performed 10 times, and mean values were calculated from all results. MNCs were separated from total blood by Ficoll (Biochrom, Berlin, Germany) and OncoQuick (Greiner BioOne, Frickenhausen, Germany) density gradient centrifugation.
OncoQuick density gradient centrifugation.
Pre-cooled 50-ml centrifugation tubes containing 15 ml of separation medium below a porous barrier separating lower and upper compartments were filled with 10–30 ml of blood and centrifuged at 1600g for 20 min at 4°C in a swing-out rotor (Hettich, Tuttlingen, Germany). The entire volume of the upper compartment including the interphase of MNCs was poured into a fresh centrifugation tube. The porous barrier prevented any contamination of the separated MNCs from the fraction of pelleted blood cells. After the cells were centrifuged with washing buffer at 200g for 12 min at 4°C, the supernatant was aspirated. Pelleted cells were used for further evaluation.
Ficoll density gradient centrifugation.
The 50-ml centrifugation tubes with a porous filter disc at 15 ml (Leucosep, Greiner BioOne) containing 15 ml of Ficoll were filled with 10 ml of blood and centrifuged at 1,000g for 10 min at room temperature. To harvest the MNCs, the entire volume of the upper compartment of the Leucosep tubes was poured into a fresh 50-ml centrifugation tube. The harvested cells were washed in phosphate buffered saline (Böhringer, Mannheim, Germany) at 500g for 15 min at 4°C. The supernatant was aspirated and used for further detection of disseminated tumor cells. The low gravity forces used for the washing steps (500g for Ficoll and 200g for OncoQuick) were chosen to preserve cell integrity for ICC evaluation.
Depletion of Blood Cells After Density Gradient Centrifugation
For quantification of the separated cells by both density gradient centrifugation methods, 30 ml of blood was drawn from 10 tumor-free and healthy probands and divided into three 10-ml aliquots. For each patient, one aliquot was prepared with Ficoll and one with Oncoquick. The third aliquot was not treated. The separated cells by both centrifugation methods and the untreated total blood sample were analyzed with the hematologic analysis device Sysmex SE-9000 (Sysmex, Norderstedt, Germany), which is used clinically for blood cell differentiation. Scatter graphs were created by the resistance measure, with direct current and high frequency alternate current showing the three-dimensional distribution of cell size and internal cell density.
Tumor Cell Recovery After Density Centrifugation
To determine the rates of recovered tumor cells after preparation of whole blood with Ficoll and OncoQuick, we performed spiking experiments with the colorectal carcinoma cell line HT-29. From 10 tumor-free probands, 40 ml of blood was drawn and divided into four 10-ml aliquots. Two aliquots of each proband were spiked with 2.7 × 106 or 2.7 × 105 tumor cells, and the remaining two aliquots were not manipulated. One pair of unspiked and spiked blood sample was prepared with Ficoll, and the other pair with OncoQuick, and both were analyzed with the hematologic analysis device. The difference in total count of MNC between the spiked and the unspiked samples was considered to be due to the spiked tumor cells. Comparison of the number of spiked tumor cells with this calculated number of recovered tumor cells after Ficoll or OncoQuick preparation served as a measure for the rate of recovered tumor cells with either method.
Detection of Disseminated Tumor Cells
Tumor cell detection was performed by ICC and RT-PCR. To evaluate the detection rates of spiked tumor cells in blood prepared with Ficoll or OncoQuick, 10-ml blood samples from tumor-free probands were spiked with increasing numbers of tumor cells (0, 10, 100, and 1000 tumor cells for RT-PCR–based detection and 100 and 1000 tumor cells for ICC detection).
The total number of separated MNCs with and without spiked tumor cells was determined by microscopy in a Neubauer cell chamber. A maximum of 200,000 cells was spun onto each glass slides. The cytocentrifuge preparations were fixed with acetone for 10 min at room temperature, air dried, and preincubated with antibody-free human serum for 25 min to block nonspecific antibody binding. Cytokeratin-positive cells were identified with the mouse anti-cytokeratin monoclonal antibody A45-B/B3 (Micromet, Martinsried, Germany), which detects a common epitope present in a variety of cytokeratin proteins, including cytokeratins 8, 18, and 19, and the monoclonal antibody CK-20 (Dako, Hamburg, Germany), which is directed against cytokeratin 20 (CK-20). After a 45-min incubation with the primary antibody, the reaction was developed with the alkaline phosphatase–anti-alkaline phosphatase technique. The number of cytokeratin-positive cells was microscopically evaluated by two independent observers.
Isolation of total RNA from the cell pellets obtained by Ficoll or OncoQuick preparations was performed with the High Pure RNA Isolation Kit (Roche, Mannheim, Germany). Amplification of CK-20 mRNA (18) was performed in a one-step RT-PCR with the Titan One Tube RT-PCR Kit (Roche, Mannheim, Germany). The PCR was performed in a reaction volume of 50 μl including a maximum of 29.5 μl of template RNA for OncoQuick samples and a maximum of 10 μl of template RNA for Ficoll samples to avoid RNA overloading. By using the set of CK-20–specific primers described by Funaki et al. (19), RT-PCR amplification was performed for all samples in a Varius-V thermocycler (Landgraf, Langenhagen, Germany) as described in a previous study (20). Efficient amplification was consistently obtained from all samples and monitored with a control β2-microglobulin RT-PCR. PCR products were run on a 2% agarose gel. For appropriate negative controls, the template RNA was replaced with nuclease-free water. Extracted total RNA from the colorectal carcinoma cell line HT-29 was analyzed as a positive control.
Specificity of CK-20 RT-PCR in Combination With OncoQuick
To determine the specificity of CK-20 RT-PCR for tumor cell detection, we examined blood samples from 30 tumor-free patients who had no evidence of a cancer disease. Thirty patients with bone fractures, undergoing surgery in our department of trauma surgery, each donated 30 ml of blood. The total blood amount was divided in half for further OncoQuick and Ficoll density gradient centrifugation. CK-20 RT-PCR was performed after RNA extraction from all blood samples.
In addition, we evaluated 15-ml blood samples from five tumor-free patients for CK-20 expression in the cell fractions above and below the “porous barrier” after OncoQuick or Ficoll density gradient centrifugation. The cell fractions above and below the porous barrier of the blood samples were separated after density gradient centrifugation, and total RNA was extracted and examined by CK-20 RT-PCR and real-time CK-20 LightCycler RT-PCR (Roche, Mannheim, Germany) (21) for quantification of CK-20 expression. Lysis of the erythrocytes also was performed in erythrocytes containing cell fraction below the porous barrier.
To validate the clinical use of OncoQuick, we examined 37 consecutive patients who had been staged in our department for gastrointestinal carcinoma. From each patient, two 20-ml samples of peripheral blood were drawn preoperatively. All blood samples were centrifuged with the density gradient OncoQuick. Detection of circulating tumor cells was performed by CK-20 RT-PCR. Detection rates were compared with the presence of distant metastasis and analyzed statistically by chi-square test, with a significance level of P ≤ 0.05.
Quantification and Characterization of the Enriched Cell Fraction
Depletion of blood cells after density centrifugation.
The characterization of the different blood cell fractions from 10 ml of unprepared total blood by the hematologic analysis device showed a mean cell number of 6.0 × 107 white blood cells (WBCs), 4.6 × 1010 red blood cells (RBCs), and 2.3 × 109 platelets. The WBC differentiation is shown in Table 1. The interphase recovered from 10 ml of blood after Ficoll preparation consisted of 1.6 × 107 WBCs, 8.0 × 107 RBCs, and 4.1 × 108 platelets. The harvested interphase after OncoQuick preparation of 10 ml of blood consisted of 9.5 × 104 WBCs, 3.6 × 106 RBCs, and 1.6 × 108 platelets. Using Ficoll centrifugation, the WBC number was reduced by a factor of 4, the RBC number by 575, and the platelet number by 6 compared with the unprepared total blood fractions. In comparison, OncoQuick centrifugation led to a stronger reduction of the harvested cells, with depletion factors of 632 for WBCs, 12,778 for RBCs, and 14 for platelets in comparison with the cell fractions in the unprepared total blood.
Table 1. Total Cell Counts (Mean + Standard Deviation) of 10 ml of Total Blood and Enriched Interphase Cell Fractions After Ficoll and OncoQuick Preparation*
The relation of cell counts before and after density gradient centrifugation is shown as the depletion factor.
n = 10.
Red blood cells
4.6 × 1010 ± 3.5 × 109
8.0 × 107 ± 2.8 × 107
3.6 × 106 ± 1.7 × 106
2.3 × 109 ± 8.6 × 108
4.1 × 108 ± 1.7 × 108
1.6 × 108 ± 5.2 × 107
White blood cells
6.0 × 107 ± 1.6 × 107
1.6 × 107 ± 5.2 × 106
9.5 × 104 ± 1.5 × 104
1.9 × 107 ± 1.1 × 107
1.1 × 107 ± 4.9 × 106
2.7 × 104 ± 1.7 × 104
5.0 × 106 ± 1.5 × 106
3.5 × 106 ± 3.4 × 105
1.0 × 104 ± 3.0 × 103
3.5 × 107 ± 7.0 × 106
8.7 × 105 ± 4.0 × 105
5.4 × 104 ± 2.5 × 104
9.4 × 105 ± 6.2 × 105
4.9 × 104 ± 8.7 × 104
2.0 × 103 ± 2.8 × 103
3.2 × 105 ± 3.3 × 105
2.0 × 105 ± 1.8 × 105
4.0 × 103 ± 3.2 × 103
Tumor cell recovery after density centrifugation.
The recovery rates of 2.7 × 105 and 2.7 × 106 spiked tumor cells were comparable for both centrifugation methods (range = 70–90%), with means of 84% for Ficoll and 87% for OncoQuick, as analyzed by the hematologic analysis device (Fig. 1).
Detection of Disseminated Tumor Cells
Microscopic quantification of the separated MNCs with a Neubauer cell chamber showed means amount of 2 × 105 cells after OncoQuick centrifugation and 1 × 107 cells after Ficoll centrifugation. For further ICC evaluation, a maximum cell number of 2 × 105 cells was spun on each cytospin. Consequently, all MNCs separated by OncoQuick could be spun onto one to two glass slides, whereas about 50 cytospins were necessary for evaluation of all MNCs separated by Ficoll. In spiking experiments with 100 and 1,000 tumor cells in 10 ml of total blood, a mean recovery rate of 42% (range = 25–70%) was determined by ICC regardless of the enriched system used. After Ficoll centrifugation, only 1–2% of the spiked tumor cells per slide were detected, whereas Oncoquick allowed the detection of 25–70% of spiked tumor cells per slide (Fig. 2).
After comparing the two monoclonal antibodies used to identify tumor cells, we found a higher sensitivity for A45-B/B3 than for CK-20 in blood. To determine the detection limit for ICC, we performed spiking experiments with 10 and 100 spiked tumor cells in 30 ml of blood in analogy to the RT-PCR experiments. With the monoclonal antibody A45-B/B3 we found in a series of five twice-replicated experiments one of 10 spiked tumor cells after OncoQuick centrifugation, whereas none of the 10 spiked tumor cells were identified with the monoclonal antibody CK-20 or after Ficoll centrifugation.
The extracted mean total RNA yields were 20 μg after Ficoll and 1 μg after OncoQuick density gradient centrifugation of 10 ml of blood. The CK-20 RT-PCR reaction used in this study was limited to a maximum of 2 μg of total RNA yield, or a maximum of 30 μl of RNA extract. Consequently, only 10% of the RNA yield extracted after Ficoll preparation could be analyzed in a single PCR reaction, equalling the analysis of almost 1 ml of whole blood. For PCR analysis of OncoQuick preparation, about 50% of the obtained RNA yield could be analyzed in a single PCR reaction.
Dilution series with the colorectal cancer cell line HT-29 for sensitivity testing of CK-20 RT-PCR showed a detection limit of 10 spiked tumor cells in 10 ml of total blood for Ficoll density gradient centrifugation (Fig. 3). Due to the superior depletion of blood cells by OncoQuick centrifugation and consecutive smaller amount of “contaminating” blood cell RNA, blood volumes could be increased up to 30 ml for the OncoQuick preparation and RT-PCR analysis, with a detection limit of 10 spiked tumor cells in 30 ml of blood (Fig. 3). In conclusion, target tumor cell RNA was obtained in a higher concentration with less “contaminating” blood cell RNA after blood preparation with OncoQuick than with Ficoll.
Specificity of CK-20 RT-PCR After OncoQuick Density Gradient Centrifugation
Thirty tumor-free blood samples were evaluated for false-positive CK-20 expression due to illegitimate transcription or pseudogenes. All samples depleted by OncoQuick centrifugation were negative by RT-PCR for CK-20 (0%), whereas three of 30 samples (10%) depleted by Ficoll centrifugation were positive, although these patients did not have cancer.
To investigate further, cells below and above the porous layer were evaluated in five tumor-free blood samples after OncoQuick centrifugation. The MNC fraction of the interphase above the porous barrier did not express CK-20 in all five tumor-free blood samples after OncoQuick centrifugation, whereas that below the porous barrier did express CK-20 in all five samples. The examination of 10 ml of blood, which was spiked with 100 HT29 colon cancer cells, did show CK-20 expression in the cell fractions above and below the porous barrier. Additional quantification of CK-20 expression with quantitative LightCycler CK-20 RT-PCR to determine whether there was an increased CK-20 expression in the fraction below the porous barrier after tumor cell spiking showed a mean CK-20 copy number of 3,000–8,000 in the spiked or the unspiked cell fraction. The interphase (cell fraction above the porous barrier) of the tumor cell spiked blood sample contained a CK-20 copy number of 380,000 (data not shown), whereas no CK-20 expression was found in the unspiked cell fraction above the porous barrier. These data demonstrated that no CK-20–expressing tumor cells were below the porous barrier after OncoQuick centrifugation.
In peripheral blood, CK-20 mRNA was detected after OncoQuick density gradient centrifugation in 11 of 37 patients (30%). Twenty-six patients (70%) were negative in both blood samples. Six of 22 patients (27%) with upper gastrointestinal carcinoma expressed CK-20 mRNA in the peripheral blood samples, and five of 15 patients (33%) with lower gastrointestinal carcinoma showed a CK-20 mRNA expression. Correlation with histopathologic data showed that six of 29 patients (21%) without clinical evidence of distant metastasis (pM0) showed CK-20 expression, whereas five of eight patients (63%) with distant metastasis (pM1) expressed CK-20 mRNA (P < 0.02; Table 2).
Table 2. Detection of CK-20 mRNA in Peripheral Blood of 37 Patients with Gastrointestinal Carcinoma with the Use of OncoQuick Density Gradient Centrifugation*
Distant metastases (pM1)
No metastases (pM0)
Tumor patients with distant metastases expressed significantly more CK-20 mRNA (63%) than did tumor patients without evidence of distant metastases (21%, P < 0.02; χ2 test).
The present analysis examined a new density gradient centrifugation based on an altered density gradient separation medium (OncoQuick) and a modified centrifugation protocol. All validation experiments showed that OncoQuick improved tumor cell enrichment, which was achieved by an increased depletion of MNCs and a comparable tumor cell recovery in comparison with Ficoll. The increased tumor cell enrichment provided an opportunity to optimize further tumor cell detection methods. The basis for sensitive and simplified tumor cell detection in blood was the significantly lower amount of MNCs in the enriched cell fraction when using OncoQuick.
These advantages significantly reduced the number of cytospins, which had to be prepared, stained, and evaluated for ICC detection of disseminated tumor cells. In comparison, the large number of remaining MNCs after Ficoll preparation accounted for the vast majority of cells that had to be evaluated on cytospins. As a consequence, the rate of immunocytochemically detected tumor cells per slide was rather low and a large number of cytospins had to be evaluated for tumor cell detection. Therefore, Funke et al. suggested, according to a statistical model, that a screening of 2 × 106 enriched cells is sufficient to provide a representative evaluation of the MRD status of a cancer patient, although this cell amount represented only about 20% of the enriched 107 MNCs by Ficoll (22). Accordingly, with 2 × 105 cells per cytospin, 10 cytospins had to be prepared and evaluated for each patient when using Ficoll, which does not seem practicable in a routine diagnostic setting despite the use of automatic cell imaging systems. In comparison, OncoQuick preparation allowed the screening of 10 ml of blood on only one to two cytospins.
In RT-PCR methods, the larger number of “contaminating” MNCs in the enriched cell fraction by Ficoll could lead to false-positive (non-tumor specific) results due to low-level illegitimate transcription by MNCs (6, 23). The smaller MNC number in the enriched OncoQuick cell fraction significantly reduced the risk of RNA amplification caused by blood cells. Our results showed that OncoQuick completely depletes the CK-20–expressing blood cell fraction (13) and leads to a high specificity for CK-20–expressing tumor cell detection. We demonstrated that, in contrast to Ficoll centrifugation, the CK-20 expressing blood cells were not detected above the porous layer after OncoQuick centrifugation, and CK-20 mRNA detection from MNC fraction of the interphase (above the porous barrier) seemed to be due to tumor cell detection.
The comparison of tumor cell line spiking experiments with clinical blood specimens of unknown tumor cell load, as performed in our study, is difficult but no other model is available to evaluate this problem. Therefore, negative CK-20 RT-PCR in our clinical specimen might well be falsely negative due to, e.g., blood samples containing a very small number of tumor cells (dilution problem). Moreover, CK-20 expression in a single tumor cell remains unknown and may vary immensely. Therefore, further improvements of the sensitivity of tumor cell detection may be achieved by the evaluation of additional molecular markers, e.g., MAGE genes (24).
The evaluation of up to 30 ml of peripheral blood by ICC or RT-PCR with OnoQuick centrifugation provided the means to analyze a more representative sample volume and thus increased tumor cell detection sensitivity. Preliminary data using OncoQuick in combination with a CK-20–specific RT-PCR associated the detection of circulating tumor cells in peripheral blood of patients with gastrointestinal carcinoma with the presence of distant metastasis. The evaluation of the easily accessible peripheral blood in the MRD compartment on the basis of OncoQuick density gradient centrifugation may allow the introduction of MRD detection as a staging parameter, which can be used in an out-patient setting for chemotherapy or postoperative recurrence monitoring, representing the possibilities of MRD analysis (25). In central and mesenteric blood, a prognostic value of tumor cell detection has been described (26, 27). The literature has reported detection rates of 19–65% of circulating tumor cells in the peripheral blood of patients with gastrointestinal carcinoma (9, 19, 28). Mori et al. was one of the first to describe a positive correlation between the detection of cells expressing carcinogenic embryonic antigen (CEA) (19%) in peripheral blood and the occurrence of tumor recurrence (9). Piva et al. examined 126 patients with gastric, pancreatic, and colorectal cancers, detected CEA-expressing cells in the peripheral blood in 34–41% with RT-PCR, and found a positive correlation with increasing tumor stage (28). Although a high specificity was reported for CEA RT-PCR in the study of Mori et al., nonspecificity was found in the study by Piva et al. who reported CEA expression in four of 16 control blood samples (25%). Nonspecificity for the molecular markers CK19 and CK20 was described by other investigators (13, 29, 30).
In summary, OncoQuick provided a rapid, reliable, and effective method for the enrichment of circulating tumor cells in blood. The number of MNCs in the enriched cell fraction was minimized without loss of tumor cells. This decreased the number of slides that had to be evaluated for ICC and increased the total blood amount that could be evaluated in molecular analysis. Processing blood from tumor patients with the use of OncoQuick increased the chance to detect circulating tumor cells and may decrease the risk of false-positive results caused by co-enriched MNCs. We suggest to performing ICC or molecular tumor cell detection with OncoQuick.