Detection of epithelial cancer cells in peripheral blood by reverse transcriptase-polymerase chain reaction.

Circulating cancer cells in the blood play a central role in the metastatic process. Their number can be very small and techniques for their detection need to be both sensitive and specific. Polymerase chain reaction (PCR) has been successfully used to detect small numbers of tumour cells in haematological cancer in which abnormalities in DNA are sufficiently consistent to make this possible. For most solid tumours this not yet feasible. However, we have found that reverse transcriptase (RT)-PRC for tissue-specific gene expression is a useful technique for identifying small numbers of circulating cells in melanoma and neuroblastoma patients. In this report we describe detection of colon carcinoma cells by RT-PCR using CK 20 mRNA as a marker. Unlike other cytokeratin genes examined (CK 8 and CK 19), CK 20 was not transcribed in normal haematopoietic cells. This suggests a role for RT-PCR in the detection of colon carcinoma metastasis in blood and bone marrow, using CK 20 as the target gene. Future analysis of clinical material will determine the clinical significance of this technique.

Intermediate filaments (IFs) are primary components of mammalian cell cytoskeleton and constitute a multigene family of proteins distinguished by their cell type-specific expression (reviewed by Nagle. 1988). The cytokeratins (CKs), which comprise some 20 different isotypes. are predominantly expressed in epithelial cells, where they show strict lineageand differentiation-associated patterns of expression (Moll et al., 1982;Sun et al., 1984). Malignant cells generally retain the IFs of their progenitor cell type and consequently CKs have been used to characterise neoplastic cells of epithelial origin (Osborn andWeber. 1983: Cooper et al., 1985;Lane and Alexander. 1990).
Some circulating tumour cells result in metastasis and so may have a major influence on patient prognosis. Since the number of circulating cells may be very small, methods for their detection need to be both sensitive and specific. We have used the method of reverse transcriptase-polymerase chain reaction (RT-PCR) to detect both melanoma (Smith et al., 1991) and neuroblastoma (Burchill et al.. 1994) tumour cells in patient blood samples. The success of this technique is dependent on the availability of a specific target which can distinguish tumour cells from haematopoietic cells. As CKs are expressed in a tissue-specific manner by epithelial cells, we have examined the possible use of CK 8. CK 19 and CK 20 genes as targets for RT-PCR detection of disseminating disease in carcinomas. CK 8 and CK 19 were targeted as they are widely expressed by mucosal epithelial tissues (Moll et al., 1982;Sun et al., 1984) and CK 20 was selected as a more discriminatory marker, particularly of cells of gastrointestinal derivation (Moll et al.. 1992(Moll et al.. , 1993.

Cell lines
The three well-characterised human cell lines used in the study were the transitional cell carcinoma-derived RTI 12 cell line, the breast adenocarcinoma    (Wu et al., 1982). CK 20, the most recently identified CK isotype, is expressed by HT29 cells, but not by MCF-7 cells (Moll et al., 1992). All cell lines were maintained in a 1:1 mixture of Dulbecco's modified Eagle medium (DMEM) and RPMI-1640 media supplemented with 5% fetal bovine serum and passaged using 0.25% trypsin in versene (0.02% EDTA).
Blood and bone marrow samples Normal blood or bone marrow samples were obtained from volunteers aged between 18 and 45 years. Samples were taken into EDTA, aliquoted into 2 ml volumes and frozen at -80C until required for RNA extraction.

RNA extraction
Total cellular RNA was extracted from cell lines. normal whole blood, normal bone marrow or spiked normal blood using Ultraspec RNA (Biogenesis, Bournemouth, UK) according to the manufacturer's instructions.

Reverse transcriptase-polvmerase chain reaction (RT-PCR)
The RT-PCR method used was based on that for the detection of neuroblastoma cells (Burchill et al., 1994). Briefly, dilution curves of RNA were DNAse treated and reverse transcribed to produce cDNA using a random hexamer primer. RT products were amplified by PCR for CK 8, CK 19 or CK 20 (primer sequences are given in Table I). RT-PCR products were analysed by agarose gel electrophoresis and ethidium bromide staining. Reverse transcriptase negative controls (RT -ve) in which reverse transcriptase enzyme was omitted were included for all RT-PCR reactions. Water negative controls (W) contained all components for the RT-PCR reaction but no target RNA. Where appropriate, positive controls (+ C) of RNA extracted from HT29, MCF7 or RTI 12 cells were included. Molecular weight markers (qpX 1 74RF DNA/HaeIII, Gibco BRL, Paisley. UK, or 123 bp ladder, Pharmacia, Milton Keynes, UK) were included on all agarose gels.
The quality of RNA was confirmed by amplification of cDNA for glyceraldehyde phosphate dehydrogenase (GAPDH) or 18S probed Northern blot analysis. All primers were purchased from Oswell DNA Services (Edinburgh, UK).  AACAACCTlAGGCGGCAGCTF GCCTGAGGAAGTTGATCTCG GCGGGACAAGATTCTTGGTG CTUCAGGCC-TUCGATCTGCAT CAGACACACGGTGAACTATGG GATCAGCTTCCACTGTTAGACG Primer sequences for PCR were selected using the Dieffenbach Selection Programme. Primers were located within different exons and were either 20-. 21or 22-mers. Specificitv of RT-PCR RT-PCR products were separated on agarose gels and Southern blotted onto nylon membrane (Hybond N+, Amersham, UK). Filters were hybridised with a gamma 32P endlabelled oligonucleotide probe, the sequence of which lay between each primer set. The nucleotide sequence of RT-PCR products was confirmed by dideoxy chain termination sequencing (Sequenase, USS, Canada).

Cell spiking
Cell spiking experiments were used to test the potential sensitivity of this technique for detection of colon carcinoma cells in blood. Known numbers of HT29 cells were added to whole blood samples, mRNA extracted and RT-PCR for CK 20 performed. To 2 ml aliquots of whole blood 10 to I x 1O0 cells were added; an unspiked blood sample was included in each experiment. RNA (100 pg) from HT29 cells was included as a positive control.

Results
RT-PCR detection of bladder, breast and colon carcinoma cells RT-PCR for CK 8 generated a single 244 bp band identified on ethidium bromide-stained agarose gels (Figure la). This fragment was confirmed by Southern blot analysis and sequencing (data not shown) to be the fragment of CK 8 which lies between the two primers used for PCR. The band was detected in 10 pg to 100 ng of total RNA from RTI 12 cells.
RT-PCR for CK 19 generated a single 214 bp band identified on ethidium bromide-stained agarose gels (Figure I b) which was confirmed to be CK 19 by Southern blot analysis and sequencing (data not shown). This band was detected in 1 lOpg to I ng of total RNA from MCF7 cells.
RT-PCR for CK 20 geneated a single band of 370 bp ( Figure lc). This band was confirmed by Southern hybridisation and sequence analysis (data not shown) and detected in 100 pg of total RNA from HT29 cells. In all three cases there was an increase in band intensity with increasing amounts of RNA (Figure 1 Cell spiking In HT29 cell spiking experiments it was possible to detect down to 100 HT29 cells diluted in 2 ml of whole human blood ( Figure 4a). The 370 bp band generated was shown by Southern blotting to hybridise to a 32P end-labelled oligonucleotide probe specific for CK 20 and confirmed by sequence analysis to be that of CK 20 (results not shown). No RT-PCR products were detected in whole blood alone (Figure 4a, 0). RT-PCR products were not identified in reverse transcriptase-negative samples (Figure 4b).
The expression of cytokeratins by epithelial tissues and their expression following neoplastic transformation had made them reliable markers in diagnostic surgical pathology. We have looked at the possible use of CKs as targets for RT-PCR detection of metastasising tumour cells of epithelial origin.
Since CK 8 and CK 19 were expressed in a high proportion of normal peripheral blood samples (88% and 40% respectively), neither would be suitable targets for detection of tumour cells in peripheral blood. These findings for CK 8 confirm the findings of Traweek et al. (1993), though no CK 19 expression was reported by Traweek et al. (1993) in bone marrow or peripheral blood. This is in contrast to the findings reported here. The number of normal peripheral blood samples analysed in this study was greater than in the study %I -C C. 'IO 10-Q b M C o\ 0 to0 i 10 10 Fige 4 Products of RT-PCR for CK 20 mRNA separated by agarose gel electrophoresis and stained with ethidium bromide in blood samples spiked with I -10 HT29 cells. A single 370 bp band was identified when as few as 100 cells per ml of whole blood were analysed (3). No band was identified in unspiked blood (0). RT-negative samples showed no amplified band (b). + C, positive control of RNA extracted from HT29 cells; W. water negative control; M, molecular weight markers.
of Traweek et al.. which may explain the discrepancy in results. The presence of CK 19 pseudogenes (Bader et al., 1986;Savtchenko et al., 1988) further complicates interpretation of data using CK 19. Datta et al. (1994) have recently reported on the use of CK 19 for the detection of breast carcinoma micrometastasis using primer sequences selected to incorporate differences between CK 19 and pseudogene at the 3' end. This would not exclude amplification of other CK 19 pseudogenes that do not differ at the 3' end and would limit the value of CK 19 for this purpose. CK 20 mRNA was not detected in any normal blood or bone marrow samples examined, suggesting it to be the CK of choice for detection of some carcinomas of epithelial origin. CK 20 has been detected in almost all cases of colorectal adenocarcinomas by immunohistochemistry (Moll et al., 1987) and may prove to be a useful target for the detection of disseminating colon carcinoma by RT-PCR. To evaluate further the clinical value of CK 20 as a target gene for the detection of colorectal carcinoma metastatic cells we are examining expression of CK 20 in blood and bone marrow samples from patients.