Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Personalized cancer care depends on providing therapy that will be of most benefit and of least toxicity to an individual patient during their disease process. Chemotherapy is given to eradicate systemically shed tumor cells that not only travel to, but also possess the capability of surviving and replicating in distant sites, such as bone, lung, liver, and brain. However, primary breast tumors are heterogeneous, containing tumor cell populations that may differ in metastatic potential, in mechanisms governing metastatic homing to different organs [1–4], and in sensitivity to different chemotherapeutic agents [5]. Among CTCs, which are tumor cells that have migrated into and survived shear forces present within the circulatory system, there are likely subsets of circulating cells responsible for seeding and reseeding metastases. Thus, biological characterization of CTCs should impact both our understanding of metastatic disease progression and our choice of pharmaceutical agents at different times in the course of disease.

With limited number of CTCs present in a tube of blood, one of the challenges in the field is to assess as much information from as little material as possible. Thus, multiplex analyses offer great opportunities for discovering CTC biology. Given the tiny amount of material, such an approach may require some kind of pre-amplification. And for this, special care must be taken to ensure that after such an amplification step, the material still reflects the original composition. Moreover, vast numbers of neighboring leukocytes (106 white blood cells, WBCs, in 1 ml of blood) can potentially contaminate samples and confound molecular assays, and must be taken into account during molecular analyses.

2 CTC Enrichment Using the MagSweeper Technology

The MagSweeper was developed by our multidisciplinary team at Stanford to provide highly purified live CTCs [6] suitable for pooled or single cell analyses and for in vitro or in vivo investigations [7]. In brief, patient blood is drawn into a 10cc EDTA-coated tube to prevent coagulation, then labelled with magnetic beads functionalized with the human BerEP4 monoclonal antibody to epithelial cell adhesion molecule (EpCAM, also designated as tumor-associated calcium signal transducer 1, TACSTD1). Although most of our initial work was done with magnetic beads coated with EpCAM antibodies, labelling CTCs in blood with magnetic beads attached to other CTC cell surface antigens is feasible. Magnetic rods covered by plastic sheaths sweep through the labelled blood at a specified speed, generating a shear force that captures labelled cells while partially separating them from surrounding unlabeled leukocytes. The cells are then transferred to a washing well and finally released into a third well after the plastic sheath is disengaged from the magnetic rod (Fig. 1). A second capture/wash/release cycle produces highly purified cells that can be individually transferred via micromanipulation into a tube for single cell transcriptional analysis, mutation detection, or growth in culture.

Fig. 1
figure 1

Schematic showing the MagSweeper process of CTC capture, wash, and release using internal magnetic rods covered with plastic sheaths. The covered rods first sweep through a sample of immunomagnetically labelled unfractionated whole blood in concentric circular loops, then similarly sweep through a wash solution to remove loosely bound contaminating cells. In the capture solution, the magnetic rods are disengaged from their plastic sheaths and external magnets under the capture well facilitate release of CTCs and excess magnetic particles. An additional round of capture-wash-release eliminates the majority of remaining contaminant cells entrapped by excess magnetic particles. Permission from Proceedings National Academy of Sciences pending [6]

Full size image

3 Sensitive Nucleic Acid and Protein Isolation Techniques

If one aims to characterize multiple markers, it is imperative that all available material from the sample is isolated as purely as possible. This is even more important if it concerns isolation of material from just a few cells, as is often the case in the CTC field. Methods will therefore be preferable that allow separate analysis of genomic DNA, mRNA, microRNAs (miRNAs), and proteins.

New technologies with increased sensitivity are continuously being commercialized, so labs must remain up to date. But once an isolation method for the molecular compartment of interest has been chosen and lysates are used for downstream processing to characterize markers, either at the uniplex or multiplex level, it is sometimes difficult to switch to a different method. This is because any method will have inherent biases, either toward the nucleic acid or the protein side. Which procedure to follow may therefore depend upon the specific research question being studied.

4 Pre-amplification Methods

The next challenge is to find a method that will enable measurement of multiple markers in material isolated from as little as one cell in a linear and homogeneous (unbiased) manner. In view of the ultimate goal to characterize CTCs at the multiple marker level, any marker assay showing as a non-homogeneously amplified outlier in these tests cannot be used for further analysis because the data will not be truly representative of the original sample. Therefore, criteria require high sensitivity combined with a minimum number of non-homogeneously amplified marker assays.

An example of how to address this important issue is given in Fig. 2, which shows that, especially at the level of whole genome amplification techniques such as used by the WT Ovation and Full Spectrum RNA amplification methods, lack of sensitivity (Fig. 2, left panel), as well as homogeneous amplification (Fig. 2, right panel) might present a problem for certain markers [8].

Fig. 2
figure 2

Sensitivity and specificity of pre-amplification methods. Three different linear pre-amplification methods (TaqMan PreAmp from Applied BioSystems/Life technologies, suitable for multiplexing up to 100 gene expression targets; Whole Transcriptome Ovation RNA Amplification from NuGEN; Full spectrum RNA Amplification from System Bioscience) were utilized according to the manufacturer’s instructions on RNA isolated from two epithelial breast cancer cells. The resulting pre-amplified cDNA preparations were analyzed by real-time PCR in a 20 μl reaction volume in an Mx3000P Real-Time PCR System (Agilent), and compared with expression levels measured in unamplified reference cDNA using TaqMan Gene Expression Assays in combination with TaqMan Universal PCR Master Mix No AmpErase UNG according to the manufacturer’s instructions

Full size image

Another challenge in any multiplex approach is to ensure that the multiplexing does not affect the efficiency and specificity of detection. One such example which forced us to change our strategy concerned the measurement of hsa-miR-22 located on chromosome 17 and which had attracted our attention based on its putative regulation of, amongst others, BCAR1 and ERBB3. For reasons still unknown, we were unable to get a quantification cycle (Cq) value for this particular microRNA if measured in a multiplex cDNA reaction that included 30 other microRNAs of our interest, while measurement of this particular microRNA in a uniplex reaction resulted in very nice amplification curves with Cq values in the range of 20–30 and an efficiency close to 100%.

These data demonstrate that—besides reproducibility—reliability must be checked beforehand for each individual marker assay. For now it appears the chance of successful amplification is still higher for target-specific amplification methods such as the PreAmp method from Applied BioSystems/Life Technologies. The disadvantage is that these kinds of methods are restricted to a pre-selected set of markers. To include any new markers will require a new CTC enrichment and downstream processing procedure.

Fortunately, the field of single cell analysis is growing rapidly, and we therefore anticipate that unbiased whole genome amplification will soon—if not already is, such as for example, the PCR-based SMART™ technology (Switching Mechanism At the 5′ end of RNA Transcript) from ClonTech—be possible at the one-cell level.

5 Estimating the Contribution by Remaining Leukocytes

Although most CTC enrichment systems allow capture of CTCs in cancer patient blood by selectively isolating, for example, EpCAM-positive cells followed by quantification of DAPI-and CK-8/18/19-positive cells, there may still be considerable quantities remaining of contaminating leukocytes (DAPI+/CD45+). For the MagSweeper device, contaminating leukocytes range between zero and one per CTC isolated [6]; however, other CTC systems may typically capture both CTCs and about 500–1,000 contaminating leukocytes after enrichment [8]. Thus leukocyte contamination, together with the fact that CTCs appear to occur in small numbers in humans (often fewer than five CTCs per 7.5 ml blood in metastatic breast cancer patients [9]), form pitfalls that cannot be ignored when one is interested in a multiplexed characterization specific for CTCs, no matter what system is used for CTC enrichment.

6 Selection of CTC-Specific Molecular Markers

Different approaches are possible to identify markers that are CTC specific. These include screening markers of interest against publicly available databases such as SAGE (http://cgap.nci.nih.gov/SAGE/AnatomicViewer). An example of this is shown in Table 1, where we illustrate the results of screening 20 putatively interesting gene markers for the expression levels measured in white blood cells, normal breast tissues, and breast cancer tissues. Based on the data presented in this table, one might decide to discard the lower eight markers based on their low expression in breast (cancer) tissue with probably no discrimination possible between leukocyte and breast tissue-derived expression levels. The upper eight markers on the other hand appear promising with levels differentially expressed between leukocytes (low) and breast cancer tissue (higher).

Table 1 Selection of CTC-specific markers
Full size table

Despite these promising screening data for eight and possibly 12 out of the 20 markers, this unfortunately does not guarantee a successful CTC-specific multiplex analysis. For that, the actual levels need to be measured in a clinical cohort that consists of patients that presented themselves with and without CTCs, as well as a decent cohort of healthy blood donors without evidence of disease. Furthermore, it is at this stage important to screen these markers following well-defined standard operating protocols. Not only for the blood sampling, but also for the downstream processing and the performance of these markers in the actual multiplex protocol, which should then include all markers of interest.

Alternatively, one might consider skipping the in silico pre-screen and immediately perform analyses according to well-defined Standard Operating Procedures (SOPs) with all markers of interest included, and discarding markers that in the end do not fulfill the pre-defined criteria. But be aware that this approach still requires a sufficiently large enough control cohort and although more likely to identify additional markers, might in the end be the more expensive option.

7 Data Analysis and Validation

In Fig. 3, we summarize the necessary steps to ensure epithelial tumor cell-specific transcript profiling of CTCs [10].

Fig. 3
figure 3

Step-by-step schematic for reliable measurement of CTC-associated transcripts. After collecting blood samples in EDTA tubes to preserve RNA integrity and enrichment for CTCs (1), a sensitive isolation technique—preferably one that is able to isolate genomic DNA, mRNA, miRNA and protein in separate fractions (2, an example from qiagen.com)—and linear pre-amplification steps (3) are needed to enable detection of molecules in material from as little as one cell. Next, CTC- and leukocyte-specific signals are used to estimate the ratio of the tumor cell-specific signal over leukocyte-derived signal (4). Now, samples can be grouped into those with detectable epithelial-derived signals (5a) and those without detectable epithelial signals (5b), with the latter group comprising both patient samples without detectable epithelial signals and samples from healthy donors. Until consensus has been established on a robust reference miRNA set, normalizing on the mean expression of all expressed miRNAs in both groups (6a and 6b) is probably the optimal method when multiple miRNA transcripts are measured at the same time. Finally, to ensure epithelial tumor cell-specific gene expression profiling of CTCs, levels measured in the samples without detectable epithelial signals (6b) are used as cut-off for the samples with detectable epithelial signals (6a) to calculate the remaining CTC-specific signals (7). Reproduced from Expert Review of Molecular Diagnostics 11(3), 259–275 (2011) [10] with permission of Expert Reviews Ltd

Full size image

The data analysis may prove to be the most difficult part and greatly depends on the question being asked. If the interest is in identifying specific markers for use in monitoring prognosis or therapy response (i.e., if multiplex analysis was set up for screening purposes), a selection can be made from a subset of markers that were identified in the screening phase to be CTC specific and related to response. However, it is necessary that any such marker or set of markers be tested in an independent cohort and according to a protocol that only includes the final selection of markers.

8 Examples of Multiplex and Other Molecular Analyses of CTCs

The amount of RNA in a single CTC is in the picogram range and currently insufficient for affordable and reproducible whole genome microarray analysis. Thus, we have been measuring the expression of up to 96 genes in single cells using a microfluidic qRT-PCR dynamic array from Fluidigm, finding different subpopulations of CTCs within one blood sample. Which genes are selected for expression measurement by the Fluidigm chip will impact the transcriptional profiles and the resulting phenotypic groupings of CTC subpopulations. Cells that express leukocyte-specific genes, such as PTPRC encoding for CD45, are eliminated from analyses.

We have also measured specific mutations—such as exon-specific PIK3CA mutations—in single CTCs isolated from cancer patient blood by the MagSweeper. We have found that within one tube of blood from a single blood draw, there may coexist diverse subpopulations of CTCs, some that are mutant or others that are wild-type. This is consistent with the findings of Dupont Jensen and colleagues [11] who have shown that within neighboring areas of an invasive breast cancer on a single hematoxylin and eosin slide were tumor cells with a PIK3CA exon 9 mutation, other tumor cells with a PIK3CA exon 20 mutation, and a third cluster of tumor cells that were wild-type. They also showed significant mutation discordance between primary tumors and distant metastases.

We have used the MagSweeper to isolate viable CTCs from a mouse xenograft model that contained profound regions of hypoxia and anoxia and produced lung metastases in all mice [7]. Using this model, we isolated CTCs from mouse blood and studied their response to hypoxia by growing the CTCs in culture under normoxic (room air, 21% oxygen) and hypoxic (<0.1% oxygen) conditions. We observed that CTCs had an altered response to hypoxia compared to parental tumor cells as well as distinctive expression and/or induction of anoxia-induced factors and target genes, including those involved in adaptation to nutrition deprivation (as might be present in tumor areas with poor blood supply) and the endoplasmic reticulum stress response. We observed that chronic hypoxia markedly increased colony formation in CTCs compared to parental tumor cells. Moreover, when CTCs were implanted into mouse mammary fat pads, the resulting CTC xenograft tumors showed a more aggressive phenotype, producing larger tumors that developed lung metastases twice as fast as xenografts generated from primary parental tumor cells. Together, the data showing that CTCs have a distinct response to hypoxia in vitro and greater aggression in vivo support the claim that CTCs have a different phenotype than the primary parental tumor cells from which they were derived. Our results suggest that CTCs may be selected for by a combination of tumor hypoxia and nutritional deprivation and/or endoplasmic reticulum stress response [7]. Our study also shows that CTCs captured by the MagSweeper are indeed live cancer cells that produce metastasizing tumors in vivo.

9 Additional Examples of Multiplex Analysis of CTCs

In this last part of our whys, dos, and donts when setting up a multiplex molecular analysis for CTCs, we by no means intend to completely cover the large body of literature currently available on this subject. We have therefore decided to restrict the selection of articles to those published by the co-authors of this book. The examples depicted in Table 2 are merely to give an overview of how researchers started, learned, and with increased knowledge and novel equipment and molecular techniques available, have set out to get as much information as possible from CTCs.

Table 2 Examples of multiplexed analyses from the literature
Full size table

10 Concluding Remarks

Tissues, and therefore CTCs, are rarely homogeneous. Therefore, any expression profile based on pooled CTCs will blend the true expression profiles of its constituent cells to identify the CTCs that are ultimately responsible for the development of a distant metastasis. Single cell analysis rather than analysis of cell populations may be more informative. Indeed, single cell methods have been developed for both microarrays [12–14] and, recently, RNA-Seq [15–17]. Although still too expensive and with questionable reproducibility for cost-effective and accurate clinical use, as whole genome single cell analysis technology develops further, these methods may become preferable for the analysis of small numbers of single cells, and may in particular be useful to study cells that are difficult to obtain in large numbers, such as CTCs.

While the field of multiplexed CTC analyses is now being explored extensively, validation and application to clinical context requires further study in independent clinical cohorts.