| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
1
Semmelweis University, Second Department of Medicine, Szentkirályi u. 46, 1088 Budapest, Hungary
2
Roche Diagnostics GmbH, Nonnenwald 2, 82377 Penzberg, Germany
aauthor for correspondence: fax 36-1-2660816, e-mail ladus{at}bel2.sote.hu
The feasibility of detecting disseminated tumor cells through reverse transcription-PCR (RT-PCR) detection of tissue-specific genetic markers in extra-tumor compartments is currently unclear because of conflicting findings on assay sensitivity and specificity between studies [for reviews, see Refs. (1)(3)]. One potential source of these assay discrepancies includes the different sample preparations and analysis methods used between laboratories (4). Many studies have used nonenriched (NE) whole blood, although the high population of background cells within these preparations is considered to reduce RT-PCR sensitivity and specificity (5). To improve sensitivity and specificity, some investigators have used density-gradient enrichment (DGE) that is directed toward separating tumor cells in a mononuclear cell fraction from peripheral blood according to cell density (5)(6). Others have tested immunomagnetic separation (IMS) to isolate tumor cells with antibody-coated magnetic beads directed against epithelial cells (7)(8). The lack of standardized procedures and the paucity of quantitative analysis and of studies examining these methods in parallel have made the extent of the influence of sample-preparation methods difficult to determine.
To assess the effect of sample preparation on the RT-PCR-based detection and quantification of disseminated tumor cells, we studied healthy donor blood samples with added colorectal cells. We used a newly developed, quantitative RT-PCR assay for the putative gastrointestinal epithelia-specific marker cytokeratin 20 (CK20) (9)(12) to evaluate the sensitivity, specificity, and quantitative PCR effects of DGE, two IMS techniques, and a NE preparation.
For each series, 80 mL of peripheral blood were drawn from the antecubital vein of healthy volunteers with a Vacuette into heparinized tubes (Greiner). To avoid epithelial cells from skin puncture, the first 5 mL were always discarded. We added, per mL, 1000, 100, 10, 1, and 0 HT29 colorectal carcinoma cells to triplicate 5-mL blood aliquots. The 1000 and 100 cells/mL samples were prepared by dilution and the 10 and 1 cells/mL samples by micromanipulation. Each series was processed by either DGE of peripheral blood mononuclear cells with the Ficoll-Histopaque®-1077 (Sigma), IMS with Epithelial Enrich® Ber-EP4-coated microsize beads (Dynal), IMS with a CellSearch Epithelial Cell Enrichment Kit® and anti- Ep-CAM-coated nanosize beads (Immunicon), or without enrichment. The dilution series were repeated three times for each method with blood from different healthy donors. This gave 9 preparations (triplicates, 3 donors) of each of the 5-cell concentrations and 45 determinations for each of the 4-sample preparations, although 1 sample (0 cell/mL) of a DGE series was lost. The replicates served to compensate for the sample-to-sample variation of tumor cell loss during the procedure and also the general sampling problem arising from the problematic detection of low copy numbers. This sampling problem could be related to the probability of obtaining a target rather than being able to detect it (13). At the end of each enrichment protocol, samples were resuspended in 300 µL of PBS, and total RNA was isolated with the High Pure RNA Isolation Kit (Roche Diagnostics). Total RNA for the NE series was isolated directly after cell lysis in Stabilization Buffer for Blood/Bone Marrow (Roche Diagnostics). In all cases, the manufacturer’s instructions were rigorously followed.
We performed RT-PCR analysis of RNA samples with the commercially available CK20 Quantification Kit for the real-time LightCycler PCR Instrument (Roche Diagnostics). From the input of a single RNA aliquot, this method provides a relative quantification of CK20 concentrations in a sample that is comparable between PCR runs. The quantities are expressed as the ratio of CK20 to the housekeeping gene porphorynbilinogen deaminase (PBGD) divided by the CK20-to-PBGD ratio in a calibrator sample. Measurement of PBGD is used as a control for RNA loading and integrity, whereas the calibrator provides an interrun normalizer. RT-PCR analysis was performed according to the manufacturer’s instructions with one exception: the usual reagent calibrator was substituted with 10 ng/µL HT29 colorectal cell-line RNA, thereby setting the ratio of CK20 to PBGD in HT29 cells to 106. Therefore, samples with a relative ratio of 100 have a CK20-to-PBGD ratio 104-fold less than the calibrator. For each enriched sample, 10 µL of RNA, equivalent to 500 µL of whole blood, was analyzed. In the NE series, the 10 µL of analyzed RNA corresponded to only 25 µL of whole blood because of the reduced loading capacity of the procedure.
To examine the impact of sample preparation on CK20 results, we
assessed their effects on CK20 sensitivity, specificity, and PCR
quantification. The results of this study are summarized in Fig. 1
. All methods detected 1000 and 100 cells/mL in 100% of cases, but the
detection of 10 and 1 cell/mL samples was inconsistent for all
preparation methods. Furthermore, for both IMS techniques, this
inconsistency also extended to PBGD detection in the samples with low
cell numbers. These results suggest that, rather than sample
preparation, the statistical chance of capturing low cell and/or copy
numbers may be a more significant determinant of assay sensitivity.
Whether these problems are causes of tumor cell loss during preparation
or inconsistent target-copy capture was not examined in this study, and
further studies using replicates of the same RNA sample will help to
determine this.
|
Although the detection frequencies of 1–10 cells/mL were greater in the enriched than NE series, the inconsistent detection at these low concentrations and the inequality of loading volumes make it difficult to make direct comparisons. Nevertheless, from the results of this study, it appears for the detection of low cell numbers, analysis of triplicate preparations are required to approach a 100% probability of detection.
In samples without added tumor cells, the CK20 signal was seen with the Immunicon IMS (one in six, 17%) and NE (two in nine, 22%) series, whereas it was undetectable from DGE (zero in eight 8, 0%) and Dynal IMS (zero in four, 0%) preparations. The improved CK20 specificity of DGE over NE complies with the finding of Jung et al. (5), who found that DGE was essential for the removal of the CK20-expressing granulocyte fraction in unenriched blood. The result of specific CK20 detection after DGE agrees with many CK20 studies (14)(16), although this has not been a unanimous finding (17). The ability of IMS to improve CK20 specificity has not been reported previously. These results suggest that Dynal IMS but not Immunicon IMS also provides improved specificity, although caution should be observed when making this conclusion because of the single-case nature of the disparity and the inconsistency of detection at this low concentration. Nevertheless, differences in procedure, bead size, and the epitopes targeted between the two methods could potentially influence the number and types of cells entrapped by the two methods, thus influencing specificity. Further studies performing these IMS procedures on healthy-donor blood samples to characterize CK20- expressing cells will help to clarify this disparity.
To assess the effects of sample preparation on PCR quantification, we
restricted our analysis to the 1000 and 100 cells/mL samples to avoid
complications arising from poor quantitative reproducibility at low
copy numbers, (18). Analysis of these samples demonstrated
two distinct quantitative results. For DGE and NE samples, the relative
CK20 ratio decreased with decreasing cell number, suggesting that
results from these preparations may be useful for indicating tumor cell
load. In contrast, the relative CK20 ratio for the IMS preparations
remained constant at 
106 for the same series.
The relative ratio of 106 in the analyzed IMS
preparations suggests that the RNA contained in the samples had the
same CK20-to-PBGD ratio as the calibrator. Because the added cells
targeted by IMS were HT29 cells and the calibrator in this study was
HT29 cell RNA, these results suggest that relative quantification of
IMS preparations may be useful for providing an indication of tumor
cell identity. Closer inspection of the individual CK20 and PBGD
quantities helped to provide an explanation for these two quantitative
patterns. Whereas CK20 quantities decreased in proportion to cell
number for all preparations, PBGD quantities remained constant for DGE
and NE and decreased in proportion with CK20 and cell number for the
IMS preparations (results not shown). This result is consistent with
detection of PBGD from background nonepithelial cells in DGE and NE
preparations and only the isolated epithelial cells in IMS
preparations.
In conclusion, these results suggest that sample preparation may play a role in determining assay specificity and relative PCR quantification concentrations. Assay sensitivity at low cell numbers, however, appears to be primarily determined by the probability of tumor cell and/or target gene capture. The inconsistency of detection at these cell numbers and lack of replicates used by most studies may have made a significant contribution to the controversial results published to date. Careful selection of sample preparation methods providing improved specificity and rigorous replicate and quantitative analyses in future studies may provide a better understanding of the significance of RT-PCR-based micrometastasis detection.
Acknowledgments
This study was financially supported by Roche Diagnostics GmbH, Penzberg, Germany and Semmelweis University, School of Ph.D. Studies, Budapest, Hungary.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  G. Giribaldi, S. Procida, D. Ulliers, F. Mannu, R. Volpatto, G. Mandili, L. Fanchini, O. Bertetto, G. Fronda, L. Simula, et al. Specific Detection of Cytokeratin 20-Positive Cells in Blood of Colorectal and Breast Cancer Patients by a High Sensitivity Real-Time Reverse Transcriptase-Polymerase Chain Reaction Method J. Mol. Diagn., February 1, 2006; 8(1): 105 - 112. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Dandachi, M. Balic, S. Stanzer, M. Halm, M. Resel, T. A. Hinterleitner, H. Samonigg, and T. Bauernhofer Critical Evaluation of Real-Time Reverse Transcriptase-Polymerase Chain Reaction for the Quantitative Detection of Cytokeratin 20 mRNA in Colorectal Cancer Patients J. Mol. Diagn., November 1, 2005; 7(5): 631 - 637. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
