Specificity of reverse transcriptase polymerase chain reaction assays designed for the detection of circulating cancer cells is influenced by cytokines in vivo and in vitro.

Several reverse transcriptase polymerase chain reaction (RT-PCR) assays have been described for the detection of circulating tumour cells in blood and bone marrow. Target mRNA sequences for this purpose are the cytokeratins (CK) 19 and 20, the carcinoembryonic antigen (CEA), and the prostate-specific antigen messages. In this study, we investigated biological factors influencing the specificity of the CK19 and CEA RT-PCR assays. Bone marrow, granulocyte colony-stimulating factor (G-CSF)-mobilized blood stem cells and peripheral blood samples obtained from healthy volunteers (n = 15; CEA n = 7), from patients with epithelial (n = 29) and haematological (n = 23) cancer and from patients with chronic inflammatory diseases (n = 16) were examined. Neither CEA nor cytokeratin 19 messages could be amplified from bone marrow samples from healthy subjects and from patients with haematological malignancies. In contrast, specimens from patients with inflammatory diseases scored positive up to 60%. To investigate the influence of inflammation on target mRNA expression, haemopoietic cells were cultured with and without cytokine stimulation in vitro. CK19 messages could be easily detected in cultured marrow cells without further stimulation, CEA messages only after gamma-interferon (gamma-INF) stimulation. In contrast, G-CSF-mobilized peripheral blood stem cells were positive for CK19 messages only after stem cell factor (SCF) or interleukin stimulation. We conclude that transcription of so-called tissue-specific genes is inductible in haemopoietic tissues under certain conditions. These factors have to be considered in future applications of RT-PCR for the detection of minimal residual disease.

PCR) possesses the highest diagnostic sensitivity for the detection of single tumour cells in a -arietv of tissue specimens and body fluids. Another potentially important application is the quality assessment of leukaphereses products. for example purging of grafts before autologous stem cell transplantation of patients treated with high-dose therapy (Kriger et al. 1996a). Eventually. RT-PCR or similar methods v-ill improve staging procedures and. thus. may gain importance for therapeutic management of cancer patients.
Common RNA targets used for these purposes are the carcinoembryonic antigen (CEA) (Gerhard et al. 1994). the cytokeratins 18. 19 and 20 (CK18. CKl9. CK2O) (Kruger et al. 1996: Tschentscher et al. 1997: Soeth et al. 1996b. tyrosinase (Smith et al. 1991) and the prostate-specific antigen (PSA) (Israeli et al. 1995). In contrast to the increasing number of PCR assays published. the gold standard for these applications is currently still the immunocytochemistry (Schlimok et al. 1987: Pantel. 1996. Whereas the reported sensitivity of these RT-PCR assays published is quite similar to one tumour cell in 10mononucleated blood cells (MNCs). specificity of positix-e results is discussed Martmnistrasse 52. 20246 Hamburg. Germany verv controversially ex-en when the same mRNA targets A ere used (Krismann et al. 1995: Zippelius et al. 1997. Furthermore. reports varv x-idelvas to the frequency of positive results. To mak-e things even more complex. preanalytical mechanisms which interfere with RT-PCR assays. the lack of standardization. and different assav design or test performance may make a comparison of the reports impossible. The identical sensitivitof the published assavs could be attributed to the fact that in those studies the analytical performance is adequate. In addition to the analyitical factors that are mentioned aboxe. biological factors influencing RT-PCR results could also be possible. The possibility of induction. alteration. aberrant or low-level expression of targoet rnRNAs under certain conditions has not yvet been investigated. Howexer. differences in test specificity might be a result of modified mRNA expression in haemopoletic tissue.
So far. unspecificity attributed to the specific amplification of the tissue-specific expressed mRNA in haemopoietic tissue was explained by accidental pseudogene amplification of pseudogene sequences for the cytokeratins assay s or by a general unspecificitv of RT-PCR (Neumaier et al. 1995: Zippelius et al. 1997. Most PCR assays haxe been standardized w-ith samples obtained from patients with epithelial malignancies and from healthy 'olunteers. Only small collectixes of healthy people A-ere used as negative controls and to calibrate and standardize assay sensitivits and specificity. How-ever. no sy stematic evaluation w-ith patients sufferinc from non-malinnant diseases has been carried out so far. *Equal contributors. In this study. w-e compared the CEA and CKl9-RT-PCRs by examininc bone marrow. peripheral blood and leukapheresis samples obtained from patients w-ith epithelial malignancies (n = 22). from patients with non-malignant chronic inflammatory diseases (n = 16) and from patients undergoing blood stem cell mobilization and leukapheresis (n = 20).
Furthermore. in vitro experiments with cytokine stimulation of haemopoietic cells w ith and w ithout stroma cells have been carried out to investigate alterations of mRNA expression.

MATERIAL AND METHODS Clinical specimens
WAe investigated 22 bone marrow-samples from patients suffering from different abdominal tumours. and 16 specimens from patients suffering, from chronic inflammatory diseases (CID) such as chronic pancreatitis. Crohnus disease and ulcerative colitis. Bone marrow samples from healthy volunteer donors and from patients with haematological malignancies in remission or chronic phase w-ere used as negative controls. Leukapheresis samples were obtained from healthy donors and from patients suffering from different malignancies. Patient samples were received from the Department of Gynaecology and Obstetrics. the Department of Surgery and the Department of Transfusion Medicine. To exclude bias. no detailed information concerning the diagnosis w-as available to us at the time of analysis.

Cell lines
For reconstitution experiments and sensitivity testing HT29. MCF7. and MDA-MB453 cells were diluted in normal bone marrow or Buffv coat cells of healthy volunteers between 10' and 10-.

RNA purification
Total RNA Awas extracted according to standard protocols (Chomczy nski and Sacchi. 1987). RNA integrity of each preparation was tested by either 32-microglobulin or 3-actin PCR.
Reverse transcription reaction cDNA svnthesis was performed in a 20-pl reaction volume. Ten microlitres of total RNA w as used for first strand cDNA synthesis with Superscript II RT (Gibco BRL Life Technologries). according, to the manufacturer's recommendations.

PCR reactions
Sequences of all primers used are shou-n in Figure 1.

CEA-PCR
A total of 20 g11 of cDNA was used in the first PCR reaction (PCR 1) in a total volume of 50 ji containin, 0.5 mm of primers CEAos and CEAoa. 1.5 mt magnesium chloride. 0.1 mnm Tris-HCl. 0.04 m ammonium sulphate and 2 U of thermnus flavins polymerase (Biozvm Diagnostik. Germnany). Thirty -five cycles were perforned with 1 min at 943C (denaturing temperature). 1 min at 56CC (annealing temperature) and 1 min at 72'C (extension temperature) (extension time prolonged to 10min for the last cycle). Nested-PCR was performed using 3 jl of the first PCR as    (38) template for pnrmers CEAis. CEAia. S-MI and f3-M2 in 100 gl reaction mix using similar conditions as in PCR I. Thirty-fixve cycles wAere performed at 940C denaturingc temperature. 65^C annealing temperature and 72CC extension temperature followed by 10 mn at 72CC.
Cytokeratin 19 PCR PCR was carried out as described earlier (Kruger et al. 1996b). PCR products were *isualized after electrophoresis and ethidium bromide staining on an UV transillumiinator. The sensitivitv of the CKl9 RT-PCR assay w as determrined to be 1: 10-usincg dilutions of breast cancer cell line MCF-7 and MDA-MB453 in mononucleated cells of volunteers. Each sample was in' estigated tw ice with both assay s and judged as positive if there was at least one positiv-e result.
Immunocytochemistry Cells (2 x 106) from mononucleated cell fraction after Ficoll separation were spun onto slides using a Shandon cytospin centrifuge. Cytokeratin-positive cells were detected with antibody KL 1 (Coulter-immunotech). As an additional specificity control and to investigate the influence of cell abdominal surgery. marrow samples from patients suffering and undergoing surgical intervention were subjected to both RT-PCR assays. Tables 1-3 show that CEA as well as CKl9 messages could be detected by PCR in samples obtained from patients suffering from chronic inflammatory diseases of pancreas and bowel. Positivity rate and consistency of results were similar as for cancer patients.
To investigate the influence of cvtokine stimulation on mRNA transcription in vivo and in vitro. G-CSF-mobilized leukaphereses samples were examined. G-CSF-mobilized peripheral blood stem cells from patients without epithelial cancers scored positive by CEA PCR and negative by CKl9 RT-PCR (Tables 1 and 2).
Additionally. 12 samples of G-CSF-mobilized blood stem cells obtained from seven women with stage II and III breast cancer were examined with both assays. From each specimen. 2 x 106 cells were examined for tumour cells by immunocvtochemistrv. CEA message was amplified from all samples. The positivity rate Table 5 PCR amphificaton of cytokeratin 19 and CEA messages from cultured non-stimulated and cytokine-stimulated leukocytes from healthy bone marrow (BM) and G-GSF-mobilized leukaphereses samples (LP), and peripheral blood (PB). PB was examined after 3 days (d3)  for CK19 RT-PCR was 58%: 6 out of 12 (50%) were positive by immunocytochemistry. However, the number of detected tumour cells per sample was very low. with a median of 0.5 (range 0-14) cells per 1.8-2 x 106 MNCs (Table 4). Aliquots of bone marrow harvests and leukaphereses samples from healthy subjects and from patients with non-epithelial malignancies were cultured with and without cytokine stimulation for 7 days. Marrow samples converted from negative to positive in the CKl9 RT-PCR assay after 7 days of culture with and without further cytokine stimulation. In contrast, specific CEA mRNA could only be amplified from bone marrow after 7 days of y-interferon (y-INF) stimulation. Corresponding results for CEA were obtained by stimulation of MNCs derived from peripheral blood with y-INF. Leukaphereses samples scored positive for CKl9 mRNA only after stimulation with SCF, LL-3 and IL-6. According to the results shown in Table 2, examination of CEA expression in stimulated leukaphereses samples was omitted (Table 5).

DISCUSSION
The discussion in the literature regarding the specificity of RT-PCR assays designed for the detection of occult tumour cells in bone marrow and peripheral blood is very controversial.
Suspected reasons for so-called false-positive results of RT-PCR assays are low-level transcription of marker genes in non-epithelial cells, as well as accidental pseudogene amplification. In accordance with our previous reports. no specific amplification of the tissue-specific or epithelial-specific genes in samples of healthy donors was detected (Kruger et al, 1996b;Jung et al, 1997). However, in samples obtained from patients suffering from chronic inflammatory diseases, specific but obviously false-positive amplification of both genes was observed quite frequently. Furthermore, for the CEA RT-PCR assay, a positivity rate of 100% in 14 samples of G-CSF-mobilized peripheral stem cells harvested from women suffering from breast cancer indicates that gene expression of so-called tissue-specific antigens may be altered under certain conditions. In vitro studies showed an up-regulation of the CEA message in bone marrow and peripheral blood cells under stimulation with y-interferon. Stimulation with other cytokines such as IL-3 or IL-6, G-CSF, GM-CSF or SCF as well as cell culture for 7 days did not lead to detectable CEA transcription. Thus, the detected CEA mRNA in materials from CID patients may be induced by cytokines released by inflammation in vivo.
These results are in accordance with data obtained using stimulated HT 29 cells. It is known that y-interferon and tumour necrosis factor a (TNF-a) lead to an up-regulation of the CEA message in HT 29 cells in vitro and that the CEA gene contains a y-interferon responsive element (Takahashi et al. 1993). Both cytokines, y-interferon as well as TNF-a are involved in the cytokine cascade of acute-phase response (Waage and Steinshamn, 1993). These data suggest that the inflammatory process might induce expression of CEA mRNA in haemopoietic cells. Up-regulation of the CEA transcript by y-interferon seems to be very specific because neither other cytokines nor global lymphocyte stimulation with phytohaemagglutinin resulted in an increased CEA mRNA expression.
Cytokeratin 19 messages could also be amplified by specific RT-PCR reaction from 50% of marrow samples obtained from patients suffering from inflammatory diseases. The frequent amplification of CK19 messages from marrow samples of patients with Crohn's disease, ulcerative colitis and chronic pancreatitis suggests induction or stimulation of cytokeratin expression in haemopoietic cells by inflammation. However, a direct liberation of CKl9 mRNA from dying epithelial cells because of inflammation could be another explanation. To investigate the possibility of CKl9 mRNA wanscription in haemopoietic tissues. marrow and stem cell samples were cultured with and without the stimulation of several cytokines. CKl9 mRNA could be amplified from bone marrow samples after a 7-day culture under standard conditions without additional cytokine stimulation. In contrast. in G-CSFmobilized peripheral blood stem cells without typical marrow stromal tissue, CKl9 mRNA could only be detected after additional stimulation with SCF, IL3. IL6 or y-INF.
These results lead us to conclude (I) cytokeratin 19 mRNA transcription is easily induced in bone marrow in the presence of stromal cells; (II) that under specific and very artificial conditions cytokeratin transcription is also possible in haemopoietic precursor cells extracted from peripheral blood; (II) the detected specific cytokeratin mRNA in patients with CID may be induced in stromal cells of the reactive marrow by cytokines involved in the inflammatory process. This is in accordance with reports of Traweek et al (1993). who examined the cytokeratin expression in haemopoietic tissue by RT-PCR. CKl9 mRNA could not be amplified in this study from mononuclear blood cells, from normal bone marrow or from lymph nodes. but could easily be detected in fibroblasts and endothelial cells under cell culture conditions. Additionally, several groups investigated lymph nodes for micrometastases by cytokeratin 19 RT-PCR and confirmed negativity of non-reactive control nodes (Noguchi et al. 1994).
Thus, it seems that in CID different mechanisms lead to specific but misleading positive results for different target genes because specific amplification is usually judged as the presence of circulating epithelial tumour cells in the specimen. For CEA. a member of the immunoglobulin superfamily. it could be speculated that the mRNA is specifically up-regulated by y-interferon. whose function remains unclear so far. CKl9 mRNA seems to be expressed unspecifically by stimulated stromal cells during the inflammatory process. The meaning of these phenomena is unclear and requires further investigation.
The examination of leukaphereses from patients suffering from stage H and HI breast cancer by CKl9 RT-PCR and conventional immunocytochemistry gave discordant results. The median tumour cell load per 1.8-2 x 106 cells was very low with 0.5 per sample. However, only three samples (25%) were negative in both assays. Discordant results were obtained in five (42%) samples. The Poisson distribution of tumour cells in sample aliquots examined by PCR and immunocytochemistry could be responsible for these results. This indicates the necessity to examine clinical samples for contaminating tumour cells by different methods, and. when possible, repeatedly.
We have shown that the pathway to so-called false-positive results obtained by CEA and CK19 RT-PCR assays are completely different. Consequences are (I) both assays are currently not feasible to screen undefined large populations for the presence of tumour cells in bone marrow or peripheral blood: (H) additional markers to discriminate amplification because of inflammatory diseases or cancer should be determined-(Il) RT-PCR assays should be combined with immunocytochemistry in further studies to determine the clinical relevance of circulating tumour cells: and (IV) negativity of clinical specimens in repeated PCR examinations indicates a highly probable absence of tumour cells.
These consequences are quite similar to guidelines established for the diagnostic use of tumour marker detection, such as CEA or Brtfish Journal of Carncer (1998) 78(9), 1194-1198 CYFRA-21.1 on protein level (von Kleist et al, 1980;Wagener and Breuer, 1980). Observations with ntmour markers and immunocytochemistry which have been made during the last two decades cannot be easily applied to mRNA-based RT-PCR assays.