Clinical & Experimental Metastasis

, Volume 24, Issue 5, pp 317–327

Specific isolation of disseminated cancer cells: a new method permitting sensitive detection of target molecules of diagnostic and therapeutic value


    • Department of Tumor BiologyRikshospitalet-Radiumhospitalet Medical Center
  • Gunhild M. Maelandsmo
    • Department of Tumor BiologyRikshospitalet-Radiumhospitalet Medical Center
  • Hanne K. Hoifodt
    • Department of Tumor BiologyRikshospitalet-Radiumhospitalet Medical Center
  • Heidi Rasmussen
    • Department of Tumor BiologyRikshospitalet-Radiumhospitalet Medical Center
  • Oystein Fodstad
    • Department of Tumor BiologyRikshospitalet-Radiumhospitalet Medical Center
    • Faculty of MedicineUniversity of Oslo
    • University of South Alabama-Mitchell Cancer Institute
Original Paper

DOI: 10.1007/s10585-006-9052-8

Cite this article as:
Tveito, S., Maelandsmo, G.M., Hoifodt, H.K. et al. Clin Exp Metastasis (2007) 24: 317. doi:10.1007/s10585-006-9052-8


Molecular studies of rare cells, such as circulating cancer cells, require efficient pre-enrichment steps to obtain a pure population of target cells for further characterization. We have developed a two-step approach, starting with immunomagnetic enrichment, followed by specific isolation of individual, easily identifiable bead-rosetted target cells using a new semi-automated CellPick system. With this procedure, 1–50 live target cells can now be isolated. As a model system, we spiked a small number of tumor cells into millions of normal mononuclear cells (MNCs). Efficient isolation of pure target cells was obtained by use of the CellPick system, and the nature of isolated, bead-rosetted cells was verified by use of FISH. Single breast cancer cells were picked directly into an RNA preserving lysis buffer, reverse transcribed, and PCR amplified with two cDNA specific primer sets. With the isolated cells we consistently obtained both ubiquitously expressed and tumor cell specific PCR products. We also performed a successful mutation analysis of single cells using PCR and cycling temperature capillary electrophoresis (CTCE). This may have significant clinical implications in cancer and in other diseases, e.g. in characterizing micrometastatic cancer cells in blood and lymph nodes to help identifying patients who most likely will respond to therapies like tyrosine kinase inhibitors and compounds targeting specific mutations. By use of the CellPick system it is possible to specifically isolate bead-rosetted or otherwise labelled target cells from a heterogeneous cell population for further molecular characterization.


MicrometastasisCirculating cellsSingle cellsPure cell populationRT-PCRMutation-detection


The metastatic spread of cancer to organs distant from the primary tumor is the most life-threatening aspect of cancer. To metastasize, tumor cells must detach from the primary tumor, invade either the blood or lymphatic vessels, survive within the circulation, then extravasate and grow in a new organ. It is known that tumors shed a large number of cells into the circulation, but only a fraction of these (∼0.01%) may be able to complete all necessary steps and form new lesions [1, 2]. In solid cancers, circulating tumor cells are thought to represent the origin of metastatic disease, and the presence of disseminated tumor cells has been shown to correlate with poor prognosis [37]. The relative rarity of disseminated cells and the practical problems of isolating them have so far impeded on the molecular characterization of this cell population. Instead, the focus has been on analyzing tumor tissues, both from the primary and metastatic lesions. These studies have identified important molecular alterations associated with carcinogenesis, both in genes that are frequently mutated in many cancers, such as tumor suppressor genes and oncogenes [8], and more specific alterations associated with certain types of cancer, e.g. the HER2/neu amplification in many breast cancers [9]. However, the identification of molecules and genes that are associated with a metastatic end point does not in itself provide information about how they contribute to the metastatic process. While primary tumours are heterogeneous, consisting of many subpopulations of cells with various biological characteristics [10], studies have suggested that metastases can arise from one single surviving cell [11, 12]. In attempts to reveal the alterations that are important in the evolution of a successful metastatic cell it seems logical to examine the characteristics of circulating cells and compare these to those of solid tumor manifestations.

We have previously established an immunomagnetic method for sensitive detection and isolation of disseminated tumor cells in different tissue compartments [13, 14]. Using magnetic beads conjugated with monoclonal antibodies we can easily identify selected cells within larger populations, and isolate the cells with a magnet. The immunomagnetic, positive selection (IMS) method permits rapid screening of as much as 2 × 107 mononuclear cells (MNCs) a much larger number than the 1 × 106 cells usually analysed by immunocytochemical methods. Another advantage is that tumor cells isolated by IMS remain viable all through the process, and can be further characterized in in vitro and in vivo studies. Detection of disseminated tumor cells in bone marrow by IMS has previously been shown to correlate with overall survival in melanoma [6], stages of disease in colorectal cancer [14], and presence of metastatic disease in osteosarcoma [5].

The IMS method provides several logs of enrichment of the target cell population. However, the positively selected fraction still contains a number of normal cells that in sensitive molecular assays may interfere with signals originating from the target cells. As a second step we therefore introduce a new semi-automated CellPick system for specific isolation of individual and easily identifiable bead-rosetted target cells in suspension. Depending on the purpose, 1–50 target cells can be isolated for further molecular characterization. To demonstrate the usefulness of this approach we here present model experiments where limited numbers of cancer cells were spiked into millions of normal hematopoietic MNCs, mimicking clinical samples from patients with micrometastatic disease. The isolated tumor cells were thereafter used for molecular characterization experiments.

Materials and methods

Cells and cell lines

The epithelial cell line SK-BR-3 originates from a human breast adenocarcinoma pleural effusion, and the melanoma cell line COLO 829 was established from a biopsy taken from a metastatic melanoma patient prior to receiving chemotherapy (both from American Type Culture Collection, Manassas, VA). FEMX-I is a melanoma cell line originally established from a lymph node metastasis from a patient with malignant melanoma treated at the Norwegian Radium Hospital [15]. To make FEMX-I RFP the same cell line has been stably transfected with an expression vector encoding the red fluorescent protein [16]. Normal MNCs were obtained from healthy volunteers.

Antibodies and conjugation to beads

The MOC-31 monoclonal antibody (IQ Products, Groningen, the Netherlands) binds to the Ep-CAM antigen, which is consistently expressed in most epithelial cells. The monoclonal antibody 9.2.27 (obtained from Dr. R. Reisfeld, Scripps Institute, La Jolla, CA) recognizes an epitope on the high molecular weight melanoma-associated antigen. Both antibodies have been used in previous studies and show adequate binding specificity [5, 6, 17]. The monoclonal antibody 376.96 reacts with a 100 kD glycoprotein expressed by human melanoma and carcinoma cells [18] and was obtained from S. Ferrone.

Superparamagnetic monodisperse particles with a diameter of 4.5 μm, coated with polyclonal sheep anti-mouse IgG (Dynabeads SAM450, Dynal, Oslo, Norway) were resuspended in 1 ml phosphate buffered saline (PBS) + 0.1% human serum albumin (HSA, Octapharma AG, Ziegelbrucke, Switzerland) and preincubated with one of the antibodies. Typically, 2 μg antibody was added per mg beads. The suspension was incubated at 4°C for 30 min, and the beads were washed in cold PBS + 0.1% HSA to remove excess antibody. The final concentration was 2 × 108 beads/ml. All batches of conjugated beads were tested on cell lines with known binding capacities prior to use.

One μm blue fluorescent microspheres (Molecular Probes, Leiden, the Netherlands) were conjugated with antibody 376.96 and used in a double staining procedure to show that cells magnetically selected with SAM M450 beads and manipulated with the CellPick instrument would still bind latex particles.

Immunomagnetic separation (IMS)

In model experiments freshly isolated MNCs were resuspended to a final concentration of 107 cells/ml in PBS + 0.1% HSA and 1 ml was transferred to 10 ml round bottom tubes. 103 tumor cells were spiked in and immunobeads were added in a concentration of 0.5:1 to the total number of cells. The suspension was incubated at 4°C for 30 min on a rotating mixer. After incubation, the cells were diluted with 4 ml PBS + 0.1% HSA and placed in a strong magnet (Dynal) for approximately 2 min. The supernatant, containing the unbound cells, was removed and the remaining positive fraction was resuspended in PBS + 0.1% HSA and taken to the CellPick instrument for further purification or used directly for FISH slide preparation. Cells are considered positive when they have bound a minimum of five beads and have a diameter that exceeds the diameter of two beads.

The CellPick instrument

The Cellpick instrument was developed in collaboration with Molecular Machines, Inc., Zurich, Switzerland. The system is specially designed for picking individual cells in suspension and consists of a micromanipulator mounted on a microscope (Fig. 1A and B). Both are operated via an attached computer. The immunomagnetic labeled target cells are kept in PBS + 0.1% HSA while viewed under the microscope. Positive cells, surrounded by magnetic beads and referred to as rosetted cells, are selected for picking and marked off on the attached computer screen. Subsequently, an ultra-thin glass capillary is lowered down into the suspension, and the selected cell is carefully sucked into a reservoir of PBS + 0.1% HSA inside the glass-capillary. The selected cells are viable all through the process and can be deposited into any type of liquid for further analysis.
Fig. 1

The CellPick instrument (A) is a semi-automated system for picking cells in suspension. The system consists of a micromanipulator mounted on an inverted microscope and a motorized piston-pump (right side). A ccd camera is mounted on the microscope and allows a real-time view of the sample on the computer monitor. The instrument is operated via an attached computer with appropriate software. (B) is a close-up of the micromanipulator with an exchangeable glass-capillary. A sample of a cell-suspension is placed under the microscope, viewed in the microscope or on the computer screen, and selected cells are aspired by the capillary. Picked cells are kept inside the capillary in a reservoir of PBS + 0.1% HSA until deposited into a desired container

Preparation of slides for fluorescent in situ hybridization (FISH)

For the FISH experiments it was necessary to remove the magnetic beads prior to slide preparation. We therefore used the CELLection™ Epithelial Enrich kit (Dynal), which contains magnetic beads coated with the EpCAM antibody via a DNA-linker to provide a cleavage site for cell release. The positively selected fraction was resuspended in prewarmed releasing buffer (RPMI 1640, 5% FCS, 1× DNAse buffer, (Turbo DNase, Ambion, TX), mixed with 2 U DNase (Ambion) and incubated for 15 min at 37°C. To help release the beads from the cells the mixture was vigorously pipetted and the tube was put on a magnet. The supernatant, containing the cells, was transferred to an eppendorf tube containing fixation solution (3:1 MetOH–HAc + 15 mM EDTA). The EDTA was added to inactivate any remaining DNase. To remove all cells, the beads were washed a total of 3 times in fixation solution, and all supernatants were pooled. After centrifugation the cells were resuspended in a small volume 60% acetic acid + 15 mM EDTA. The cells were applied on to microscopy slides prewarmed to 45–50°C, and left to dry at the same temperature. Slides were stored at −20°C before use.

Fluorescent in situ hybridization (FISH)

The slides were immersed in 75% ethanol at 4°C for 1–2 h before further processing. The chromosome 17p12 (HER2/NEU)/ alphasatellite 17 cocktail probe, direct labeled, from Q-Biogene (Kreatech Biotechnology BV, Amsterdam, The Netherlands) was used for hybridization of the slides. The HER2/NEU specific probe is labeled with rhodamine (Exc. Max. 565 nm, Em. Max. 590 nm), and the chromosome 17 alpha-satellite probe is labeled with fluorescein (Exc. Max. 495 nm, Em. Max. 525 nm). Before hybridization, the slides were pretreated in 2× saline sodium citrate (SSC), pH 7.0/0.5% NP40 and dehydrated in ethanol solutions. Denaturation was performed at 72°C in 70% formamide/2× SSC, pH 5.0 for 2 min, and the slides were dehydrated in series of ice-cold ethanol solutions. 10 μl of the ready probe-mix was denaturated at 96°C, quickly spun down and applied to each slide. Hybridisation was done overnight at 37°C, and never for longer than 16 h. After hybridisation the coverslips were removed and the slides were washed for 5 min at 65°C in 1× PBS solution and air-dried.

To stain cell nuclei DAPI (4′, 6-diamino-2-phenylindole)/antifade (Q-Biogene/MP Biomedicals, Heidelberg, Germany) was applied to all slides before viewed in a Zeiss inverted microscope, Axiovert 200 (Carl Zeiss, Jena, Germany) equipped with single filters for exitation of DAPI, fluorescein and rhodamine. Pictures were composed by the use of Carl Zeiss AxioCam HR, Version 5.05.10 and the AxioVision software. The images taken with different filters were merged with PhotoShop 7.0 (Adobe, San Jose, CA) to produce single multicolored images.

Quantitative real-time RT-PCR

The positively selected cell fraction was diluted and viewed in the CellPick microscope, single bead-rosetted cells were picked and transferred to a PCR tube containing 6 μl lysis buffer (Cells-to-Signal, Ambion). The tubes were briefly vortexed and left at room temperature during the procedure (1–3 h). Pure buffer, or MNCs with no beads, was used as controls.

Reverse transcription was performed in the same tubes in a final volume of 20 μl using the iScript cDNA synthesis kit (BioRad, Hercules, CA) as described in the manual, but with the incubation at 42°C increased to 45 min.

Real time reactions were run on the iCycler (BioRad). All cDNA samples were analysed with two different primer pairs, and all reactions were run in parallel, making it a total of four PCR reactions per cDNA sample. All primers were cDNA specific. The primers for mammaglobin A (hMAM, Forw. 5′-CGGATGAAACTCTGAGCAATGT-3′ and Rev 5′-CTGCAGTTCTGTGAGCCAAAG-3′) amplify a 108 bp fragment, and the primers for YARS, a t-RNA synthetase (Forw. 5′-GCCTACCCAGATCCCTCAAAG-3′, and Rev 5′-ATGACCTCCTC-TGGTTCT-GAATTC-3′) yield a 74 bp product.

For the amplification reaction the iQ™ SYBR®Green Supermix (BioRad) was used, and for both primer pairs the amplification protocol was as follows: 3 minutes initial denaturation at 95°C, 50 cycles of 10 s denaturation at 95°C and 35 s annealing/extension at 60°C. A final melt curve analysis was included to verify that each reaction only gave one specific product.

Detection of BRAF mutations in Colo 829 cells

The melanoma cell line Colo 829 carries the V600 E mutation in exon 15 of the BRAF gene, and the mutation can be detected by cycling temperature capillary electrophoresis (CTCE) as described by Hinselwood et al [19]. We mixed Colo 829 cells with normal MNCs and performed IMS using the 9.2.27 antibody. Single bead-rosetted cells were picked with the CellPick micromanipulator and transferred to PCR tubes containing 2 μl proteinase K digestion buffer composed of 1× One-Phor-All Buffer PLUS (GE Healthcare, Chalfont St.Giles, UK), 0.67% Tween 20 (Merck, Hohenbrunn, Germany), 0.67% Nonidet® P40 (USB Corporation, Cleveland, OH), and 0.67 mg/ml Proteinase K (Sigma-Aldrich, St.Louis, MO) [20]. The cells were incubated over night at 42°C. Normal MNCs with no beads were picked from the same cell suspension and used as controls. The tubes were then heated to 80°C for 10 min to inactivate the proteinase K. BRAF-PCR was performed in the same tubes by the addition of 18 μl PCR-supermix consisting of 2 μl 10× Platinum Taq buffer (Invitrogen, Carlsbad, California, USA), 1.6 μl 50 mM MgCl2, 0.5 μl dNTPs (10 mM), 5 pmol unlabeled primer, 0.25 pmol 6-FAM labeled primer, 1 U Platinum Taq (Invitrogen) and water to a final volume of 18 μl. Primer sequences and cycling conditions were as described by Hinselwood et al. The CTCE was performed on a MegaBACETM 1000 DNA Analysis System (Amersham Biosciences/GE Healthcare, Uppsala, Sweden) as described by Hinselwood et al.

For some samples, we included a whole genome amplification (WGA) step on the DNA obtained from picked cells. We used the GenomiPhi DNA amplification kit from GE Healthcare according to the manufacturer’s instructions. The kit utilizes random hexamers and the Phi 29 DNA polymerase in an isothermal amplification reaction. The amplified DNA was diluted with dH2O before mutation analysis.


To demonstrate the usefulness of the CellPick instrument for obtaining pure cell fractions, we have used a model system where known quantities of tumor cells are mixed with MNCs to mimic a sample from patients with micrometastatic disease. The tumor cells were isolated by immunomagnetic separation (IMS) and the enriched fraction subjected to further purification using the CellPick instrument. Molecular analyses were then performed on the selected cells. An outline of the method is shown in Fig. 2.
Fig. 2

Schematic outline of the immunomagnetic selection procedure with CellPick isolation of target cells. Paramagnetic beads coated with an appropriate antibody are mixed with mononuclear cells isolated from blood or bone marrow. Following incubation at 4°C bead-rosetted cells are magnetically isolated, the positively enriched fraction examined in a microscope, and rosetted cells carefully sucked into a glass capillary by use of the CellPick device. The purified cell suspension can be deposited into any type of liquid or container for further characterization

Isolation of target cells

FEMX-I RFP cells appear bright red when viewed in red fluorescent light. In model experiments, 1 × 107 freshly isolated MNCs were mixed with 103 FEMX-I RFP cells and 5 × 106 magnetic beads coated with the 9.2.27 antibody. After incubation for 30 min the suspension was subjected to immunomagnetic enrichment and the retained cell/bead mixture subsequently resuspended in 200 μl PBS with 0.1% HSA. Twenty μl of the enriched fraction was applied onto microscopy slides for inspection. The IMS analysis yielded an enriched tumor cell population, but, as seen on the light microscopy image in Fig. 3A, several MNCs were still present together with one tumor cell. Under red fluorescence only the bead-covered FEMX-I RFP cell was visible.
Fig. 3

Demonstration of efficiency in obtaining a pure target cell population. (A) shows a magnetically enriched cell fraction containing both mononuclear cells and one bead labeled FEMX-RFP cell. While all cells are visible in normal light microscopy, only the bead-labeled cell expressing RFP is visible in red fluorescence light (lower panel). (B) shows the corresponding CellPick purified cell fraction

We then applied another aliquot of the same cell fraction to new slides for cell isolation with the CellPick instrument. Single bead-rosetted cells were carefully sucked into the capillary and subsequently deposited onto new slides. During the picking procedure the cells were kept in the capillary immersed in PBS with 0.1% HSA. Figure 3B shows a sample after the CellPick procedure, now consisting entirely of bead-rosetted cells (upper panel) which all lighted up under red fluorescence (lower panel). This demonstrates the ability of the CellPick instrument in combination with the IMS procedure to selectively isolate single target cells.

Double staining of CellPick selected cells

FEMX-I cells were mixed with 5 × 106 magnetic beads coated with the 9.2.27 antibody and subjected to immunomagnetic enrichment. The enriched fraction was further purified with the CellPick instrument, and 100 bead-labeled cells were picked with the micromanipulator and deposited in 20 μl PBS + 0.1% HSA. These 100 cells were then mixed with blue fluorescent microspheres coated with the 376.96 antibody and incubated for another 30 min at 4°C. Microscopy examination using a Zeiss Axioscope (Carl Zeiss, Jena, Germany) showed that the picked cells were able to bind the latex particles, (Fig. 4B–C) demonstrating that the targeted cell surface antigens remained intact on cells manipulated with the CellPick instrument. This experiment was also repeated using material from a lymph node of a melanoma patient. The CellPick purified patient cells bound the 376.96 coated latex-beads proving that the picking-procedure does not disrupt or harm the cell-surface (Fig 4E–F).
Fig. 4

Preservation of surface antigens. FEMX-I melanoma cells were immunomagnetically selected with the 9.2.27 antibody and further purified by use of the CellPick micromanipulator. The picked cells were then incubated with blue fluorescent microspheres coated with the 376.96 antibody. (A) A FEMX-I cell rosetted with 9.2.27 magnetic beads only. (B and C) Tumor cells with both magnetic beads and a small number of blue fluorescent microspheres attached to the cell membrane. The same experiment was performed on material from a lymph node of a melanoma patient. (D) Shows a typical positive rosetted cell after the immunomagnetic enrichment. The cells were then picked and further incubated with blue-fluorescent latex beads (E and F). This experiment demonstrates that the CellPick procedure does not harm the expression of cell surface markers

FISH analysis of mixed and pure cell populations

The SK-BR 3 cell line harbours an amplification of the c-erb-B2/HER2/neu gene, which can be visualized by fluorescent in situ hybridization (FISH). It also contains several copies of chromosome 17. In spiking experiments, MNCs were mixed with SK-BR 3 cells and magnetic beads coated with the MOC31 antibody. To demonstrate the malignant nature of the bead-rosetted cells, we performed FISH analysis on both an enriched cell fraction and on a CellPick purified population. A dual probe set with a rhodamine labeled probe for the c-erb-B2/HER2/neu gene and a FITC labeled probe for the chromosome 17 centromere were used. The magnetic beads proved to have very strong binding capacities and stuck to the cells through the entire slide preparation. They also had high auto-fluorescence, making it necessary to modify the protocol slightly by using beads that could be removed from the cell surface before FISH slide preparation. Figure 5 shows FISH slides obtained from an IMS enriched cell population (left panel) and after CellPick target cell isolation. The figure clearly demonstrates that normal cell nuclei (eight smaller ones in left panel) carry two copies of each locus, while a SK BR 3 nucleus displays several signals for both probes. This experiment proves that chromosomal aberrations can be analyzed in individual cells after isolation with the CellPick instrument.
Fig. 5

FISH analysis of enriched and CellPick isolated tumor cells. SK-BR 3 cells, which carry an amplification of the HER2/neu locus, were mixed with normal mononuclear cells and enriched for by immunomagnetic selection. A FISH slide of the enriched population with one SK-BR 3 cell and several mononuclear cells is shown in (A). After CellPick purification and FISH analysis on the isolated cell population, only SK-BR 3 cells can be found on the slides (B). Centromere 17 probes are labeled with SpectrumGreen, c-erb-B2 probes with SpectrumOrange. Normal cells carry two copies of each loci, while the SK-BR 3 cell show several signals for both probes. The normal nuclei are smaller than the nuclei of the tumor cells

RT-PCR from single picked cells

Mammaglobin is strongly expressed in breast tissue, but not or very weakly expressed in other tissues, such as normal blood or bone marrow [21, 22]. To investigate whether we could analyze the expression status of single target cells picked from a heterogeneous cell population, we mixed SK-BR 3 cells with normal MNCs and performed IMS using MOC31 immunobeads. Single, bead-rosetted cells were isolated with CellPick and put directly into an RNA preserving lysis buffer and the entire lysate was used for reverse transcription. Normal MNCs with no beads were used as control. The cDNA from each PCR reaction was split in four individual real-time PCR reactions using two sets of primers. One set amplifies a tyrosyl-tRNA synthetase, YARS, ubiquitously expressed in all cells, while the other amplifies mammablobin. Single, CellPick-isolated SK-BR 3 cells consistently yielded PCR products with both the mammaglobin specific and the YARS primerset. In contrast, only YARS was amplified in the lysate from normal MNCs (typically 3–6 cells picked). Not all samples yielded results in both parallels of each PCR reaction, probably due to the small amount of material. As long as at least one parallel for a primerset produced the correct melt-curve, and a clear positive signal, we scored the sample positive for that marker. Examples of real time PCR results for one SK-BR 3 cell and MNC are shown in Fig. 6. The use of the reference gene amplification was important to verify that the cell lysis and reverse transcription was successful. We did not obtain a signal for hMAM in any sample without also getting a positive signal for the YARS reference gene. Amplification of tissue specific genes, as exemplified here with hMAM, is a valuable analysis to confirm the epithelial origin of the bead-rosetted cells. This set of experiments demonstrates that semi-quantitative expression profiles from a limited number of genes are achievable from isolated single target cells.
Fig. 6

Characterization of single cells by real-time RT-PCR. Single cells were picked with the CellPick micromanipulator and deposited directly into an RNA preserving lysis buffer. After reverse transcription, the cDNA from one cell was used for four individual PCR reactions, two that amplify mammaglobin (hMAM) and two that amplify a t-RNA synthetase, YARS. SYBR-Green was used for real-time detection. (A) The amplification plot and the corresponding melt curve for one single SK-BR 3 cell is shown. Both hMAM and YARS were amplified. The negative signals represent no-template control reactions. (B) A typical signal obtained from 3–6 picked mononuclear cells. Only the reference gene, YARS is amplified, mammaglobin is not expressed

Mutation detection in single picked cells

Cycling temperature capillary electrophoresis (CTCE), is a sensitive, easy and cost effective method for identification of mutations, requiring only PCR and electrophoresis. We wanted to examine whether isolated single cells could be used for mutation detection with this method. The melanoma cell line Colo 829 was used for model experiments, as these cells carry the V600E mutation in exon 15 of the BRAF gene. Tumor cells were mixed with MNCs, and enriched by IMS using the 9.2.27 immunobeads. Single bead-rosetted cells were isolated by use of the CellPick instrument and deposited directly into DNA-lysis buffer.

The BRAF mutation could be detected in 5/5 tubes with only one cell picked directly into proteinase K digestion buffer and PCR amplified for 35 cycles using BRAF specific primers. We also analysed five tubes with two tumor cells deposited, and they all showed the mutation signal. The results were weak for all samples, but the mutation profile was clearly distinguishable. Tubes containing two cells did not yield significantly better results than tubes containing one cell. As a negative control normal MNCs (3–6) were picked and deposited in three tubes. Of these, one gave a mutation signal, and two had the expected homozygous signal (Fig. 7).
Fig. 7

BRAF exon 15 mutations detected by cycling temperature capillary electrophoresis (CTCE). Graph A demonstrates results obtained from Colo 829 genomic DNA. B shows results from one picked Colo 829 cell after linear amplification of the DNA. In graph C are results from 4–6 picked mononuclear cells, also linearly amplified before the BRAF PCR. The presence of a mutation is seen by the appearance of four peaks, where the two first represent homoduplexes and the two last represent heteroduplexes of the PCR products. Graph C have one single peak, consistent with one perfectly matched double strand

To gain more material for PCR-analysis, the DNA obtained from single picked cells was also amplified using a method for WGA. The method consistently produced between 8–10 μg of DNA from a single cell, providing enough material for an indefinite number of PCR analyses. The CTCE analysis of amplified DNA correctly detected homozygous signals from 2/2 negative controls. Mutation signals were correctly identified in 5/5 tubes where two cells had been deposited, and in 3/5 tubes were one cell had been deposited. The fourth tube gave a false homozygous signal, and the fifth was blank, possible due to a failed cell lysis. The possibility to amplify DNA from single cells is beneficial, as it would permit the analyses of many different DNA variations in the same cell. However, when the WGA DNA was re-amplified with the BRAF primerset and analysed by CTCE we could observe a shift in the allele frequency, for which we have no explanation. The mutation was still detected, but the shift could indicate that the WGA procedure does introduce some bias that could influence the mutation detection. Thus, we conclude that this method needs thorough verification before it can be applied to patient material. However, with or without WGA, this experiment demonstrates that single cells may be used for this analysis.


Analysis of micrometastatic or circulating tumor cells is challenging, as a small number of target cells is present among millions of normal hematopoietic cells. The objective of the present work was to develop a new approach for specific isolation of pure populations of such cells, and to perform downstream molecular analyses on isolated single cells. As a model system, tumor cells were added to normal blood to produce a mixture of cells mimicking samples obtained from patients with micrometastatic disease. Effective enrichment of the tumor cells was achieved by the use of immunomagnetic beads coated with specific antibodies directed against surface markers on the target cells. Subsequently, bead rosetted cells were selected by the CellPick instrument, a semi-automated system for specific isolation of single cells in suspension. The procedure results in a pure suspension of cells expressing a given surface antigen, and since the cells remain viable all through the process they are available for any type of molecular analysis. In this study, we have demonstrated the utility and specificity of the procedure by FISH, detection of specific transcripts by RT-PCR, and by analyzing genomic mutations by CTCE of PCR products.

The biology of the metastatic process has been studied for more than a century, but the main question on how metastases arise from primary tumors still remains a subject of discussion. Recent gene expression analysis of primary and metastatic tissue has indicated that primary tumors might be divided into groups of “good prognosis” or “bad prognosis” [2325] based on their expression profile. However, an expression signature of tumor tissue will necessarily reflect signals originating from all cells within the tumor, and thus it would be advantageous to study specifically the molecular characteristics of pure populations of target tumor cells. Moreover, both clinical and experimental evidence show that primary tumours are biologically heterogeneous, with subpopulations of cells that have different characteristics, also with respect to metastatic potential [10, 26], and old experimental data suggested that metastases can originate from single proliferating cells [11, 12]. Thus, to disclose which molecular changes initiate spreading, support survival within the circulation, and finally determine homing and growth in a secondary organ, it may be necessary to study the cells that eventually will be the founders of metastases.

The presence of disseminated tumor cells in bone marrow has been shown to correlate with a poor prognosis in several types of cancer, also in tumor types that rarely metastasize to bone [46, 2729]. Detailed characterization of disseminated tumor cells could facilitate the identification of changes involved during progression of the disease. Such markers would be valuable tools for monitoring disseminated tumor cell load in bone marrow, both to predict metastatic relapse and to observe early response to therapy. Similar studies of individual cells isolated from solid tumor manifestations will allow comparison of alterations found in the different compartments.

Previous attempts to characterize disseminated cells have mainly used cytospins for identification of cytokeratin positive (CK+) cells from bone marrow and then performed conventional comparative genomic hybridization (CGH) to detect gross genomic aberrations [30, 31]. These studies indicated that disseminated cells share few common alterations, both with each other and with the primary tumor. The lack of aberrations shared by the cells could, however, also reflect limitations in the analytical method used. Conventional metaphase CGH requires chromosomal gains or losses of approximately 10 Mb or more for detection. To look for smaller, possibly important, aberrations requires higher resolution, and thus a very accurate amplification step. After picking individual cells with the CellPick instrument, we successfully tested one method for WGA based on the Phi 29 DNA polymerase and observed very good SNP analysis results. However, when the amplified material was applied onto 1 Mb genomic BAC arrays, the background noise was unacceptably high (results not shown). Thus, although our results demonstrated high fidelity with respect to replication, the linearity of the amplification is not yet sufficiently accurate. Preliminary results using diluted DNA (down to 10 pg) were promising, however, suggesting that protocol optimization and improved purification of isolated DNA could enable use of array CGH for single cell analysis.

Molecular analyses of individual or a limited number of cells have so far required known targets, and a future goal will be to develop approaches that allow screening for new markers, including searching for yet unknown properties associated with concepts such as tissue preferred metastasis, cancer cell dormancy, or cancer stem cells. With the analytical tools here demonstrated several questions can be addressed; the cancer cell identity of the specifically isolated bead-rosetted cells can be verified by the use of known RNA or DNA markers that distinguish the target cells from normal haematopoietic cells. Analysis of several disseminated cells isolated from the same patient can investigate the presumptive heterogeneity of the micrometastatic cell population. Comparison of these cells with samples originating from the primary or metastatic lesions may shed new light on the sequence of events occurring during the metastatic process. Finally, with increasingly better protocols for whole genome or RNA amplification, the limitations caused by small amounts of starting material will be overcome, allowing additional molecular analyses to be performed.

The CellPick procedure results in a pure suspension of cells expressing a given surface antigen. The method is equally applicable to samples of disintegrated lymphatic tissue, bone marrow, peripheral blood or other body fluids that might harbor tumor cells. Notably, the approach can also be applied to other heterogeneous cell populations where target cells can be identified by surface antigens, i.e. for selection of stem cells. Moreover, we have previously demonstrated the combined use of immunomagnetic cell isolation and antibody conjugated fluorescent particles to visualize the expression of relevant cell surface markers on micrometastatic cells [5, 6]. Combined with the CellPick approach, this allows the selection and specific isolation of micrometastatic cells with distinct differences in marker expression, and to examine whether these might represent stable cell variants with different metastatic capacities.

The work presented here demonstrates new and practical approaches for real-time PCR and mutation detection in single tumor cells without the interference of signals from non-target cells. The presumed heterogeneity of disseminated cells will require the examination of several cells collected from the same patient. To monitor effect of treatment on circulating cells, real time PCR may be used to follow the expression of tumor markers, such as mammaglobin. Assessment of gene mutations has already proven to be of practical clinical value, and will probably become increasingly important in the future. A recent example is new compounds targeting cells with BRAF mutations in patients with malignant melanoma [32, 33]. Furthermore, a subgroup of patients with non-small-cell lung cancer responds well to gefitinib, an EGFR tyrosine kinase inhibitor [34, 35]. This increased sensitivity is linked to somatic mutations of the EGFR gene in the tumor cells conferring enhanced tyrosine kinase activity. Systemic targeted therapy is likely to become increasingly important also as an adjuvant treatment to eradicate minimal residual disease, and it would undoubtedly be an advantage to examine the cells that the treatment is meant to eradicate. The ability to detect mutations in circulating cells may therefore be used, in parallel with examining the primary tumor, to identify patients that will respond to such new therapies.


We would like to thank Stine Kresse for excellent tutorials in the FISH procedure and Karen Marie Heintz for performing all the CTCE analyses. This work was supported by the Norwegian Cancer Society, the Research Council of Norway and the Anna Lovise Lundeby Memorial Fund.

Copyright information

© Springer Science + Business Media B.V. 2007