Detection of Ki-ras mutations in tissue and plasma samples of patients with pancreatic cancer using PNA-mediated PCR clamping and hybridisation probes

In the present study, we combined the PCR-clamping approach with melting curve analysis using mutant specific hybridisation probes and wild-type specific peptide nucleic acids (PNAs) to determine the genotypes of the most frequent point mutation in codon 12 of the proto-oncogene Ki-ras in tissue and plasma samples of patients with pancreatic cancer. The sensitivity of our assay was 1–5 × 10−5. The melting curve analysis of tissue samples of four patients revealed two valine mutations, one none-valine mutation and one wild-type sequence. Ki-ras alterations were found in 28% of DNAs (18 out of 64) of nonrelated plasma samples of 10 patients with ductal adenocarcinoma of the pancreas. The valine mutation was the predominantly detected gene alteration (83%). Out of ten patients investigated, four patients (40%) became positive during clinical observation with respect to Ki-ras mutation. All four patients exhibited progressive disease and high levels of tumour marker CA 19-9. In conclusion, the one-step procedure discribed may be a useful clinical tool for analysing Ki-ras point mutations in tissue and plasmas samples. In addition, this method can be adapted for simultanous detection of multiple mutations and quantitation.

Pancreatic cancer remains a major cause of death in western populations. Despite many efforts, little is known about its aetiology (Malats, 2001). Several risk factors have been implicated: male gender, black race, cigarette smoking, diabetes mellitus and meat and fat consumption (Sakorafas et al, 2000;Sternheim et al, 2000;Simon and Printz, 2001).
Ductal adenocarcinoma of the pancreas is a highly aggressive tumour with an early tendency of spreading to other organs (Hermanek, 1998). Pancreatic cancer has a poor prognosis. The median survival is less than 6 months. The overall 5-year survival of 3% associated with pancreatic cancer is largely a result of diagnosis late in the course of the disease. Surgical resection remains the only chance for cure in patients with pancreatic cancer, but only 15 -20% of patients have lesions that are resectable at the time of diagnosis. Current postresection 5-year survival rates are approximately 20% (Wong et al, 2001). Conventional tumour markers, such as CA 19-9, have a high diagnostic sensitivity (80 -90%), but lack a specifity (50 -70%) and are therefore not suited for detecting early tumours (Berndt et al, 1998;Lamerz, 1999;Wenger et al, 1999). More effective screening techniques are urgently needed to improve the poor prognosis of the disease (Mizumoto and Tanaka, 2002).
Pancreatic cancer is among the best-described genetic diseases (Hilgers and Kern, 1999). The development and growth of pancreatic adenocarcinoma involves oncogene activation, loss of tumour suppressor gene function and overexpression of receptor-ligand systems (Goggins et al, 1999;Efthimiou et al, 2001;Shi et al, 2001). The Ki-ras oncogene is one of the three members of the human ras gene family that code for the highly related 21-kDa proteins with guanosine triphosphatase (GTPase) activity. Naturally occurring mutations in the Ki-ras gene have been localised to codon 12 (most frequently), 13 and 61. Owing to an associated inappropriate stimulation signal, the mutant ras transmits a constitutive growth signal to the nucleus (Gansauge et al, 1996;Olson and Marais, 2000). The incidence of ras gene mutations varies among different kinds of tumours, but was found mutated predominantly in pancreatic adenocarcinomas with a frequency of 80 -90% (Sirivatanauksorn et al, 1998;Ghaneh et al, 2002).
Gene alterations of the ras type can be used as molecular markers to screen for the presence of neoplastic cells in different clinical specimens. Enriched PCR methods are able to discriminate mutant from wild-type alleles and also provide a high degree of sensitivity to detect mutant alleles in a large excess of normal DNA. Two PCR-based techniques have been most commonly used: ASA-PCR (allele-specific amplification), which is restricted to the analysis of individual codon-specific mutations and requires several allele-specific oligonucleotides, and the time-consuming RFLP-PCR (restriction fragment length polymorphism), which bears an increasing risk of Taq polymerase-born infidelity (Weber, 1990;Kahn et al, 1991;Takeda et al, 1993;Jacobson and Mills, 1994;Mora et al, 1998).
More recently, the mutation-sensitive hybridisation profile of peptide nucleic acids (PNAs) has been exploited to design novel protocols (PCR-clamping). PNAs are none extendable oligonucleotides where the ribose-phosphate backbone is completely replaced by (2-aminoethyl)-glycine units linked by amide bonds. These synthetic oligomers bind with higher thermal stability to complementary nucleic acids than DNA and RNA. On the other hand, the PNA/DNA duplex is significantly destabilised in the case of single-base mismatch (Ørum et al, 1993;Ratilainen et al, 1998). In PNA-mediated PCR clamping, PNA oligomers suppress amplification of the wild-type sequence confined by a pair DNA oligonucleotide primers (competetive clamping) because PNA oligomers are no substrate for DNA polymerases. In the case of a single-base mismatch, the DNA/PNA duplex is destabilised, which allows strand elongation from bound DNA oligomer to proceed, resulting in the detection of PCR fragments, most of which harbour the variant allele, and the suppression of wild-type genomic sequences during amplification, respectively. This approach has been described for the detection of Ki-ras and p53 point mutations (Thiede et al, 1996;Behn et al, 2000).
Analysis of PCR products via mutation-specific hybridisation probes guarantees the most specific measurement of amplified target sequences. Hybridisation probes consists of two terminally fluorescent-labelled oligomers. Only after binding to the target sequence in close proximity, fluorescence resonance energy transfer (FRET) occur (Lay and Wittwer, 1997) and the real-time fluorescence can be monitored. Hybridisation probes are furthermore suitable for experiments in which the amplification products formed are subsequently quantified using the evaluation software. The combination of detection by internal hybridisation probes and subsequent melting curve analysis expands the spectrum to include mutation analysis by rapid genotyping. One mismatch only, due to a single point mutation, between the internal hybridisation probes and their target sequences, can drastically change the melting temperature of the bound probe (Lyon, 2001). In a recent study, the PNA-mediated PCR clamping method was combined with on-line detection for c-kit proto-oncogene point mutations (Sotlar et al, 2003).
In the present study, we combined the PCR-clamping approach using a wild-type PNA (17-mer) with real-time PCR using mutantspecific hybprobes in single closed LightCycler capillaries. As a preliminary study, plasma and tissue samples of patients with pancreatic cancer were analysed with respect to point mutations in codon 12 of the Ki-ras proto-oncogene by melting point analysis.

MATERIAL AND METHODS Patients
A total of 64 sera from 10 patients (seven male, three female) with pancreatic carcinoma (resectable and nonresectable) were analysed for Ki-ras mutation. In addition, four tissue samples from nonrelated pancreatic tumours were analysed for Ki-ras mutation.

DNA extraction from the plasma
Blood samples (9 ml) were withdrawn from a peripheral vein and placed in tubes containing EDTA. The collected samples were centrifuged at 1500 r.p.m. for 10 min (Minifuge RF, Heraeus, Hanau, Germany). Plasma was stored at À201C until further use. DNA was extracted from plasma with QIAamp spin columns (QIAamp DNA Mini Kit, Qiagen, Hilden, Germany) according to the manufacturer's instruction. Incubation with proteinase K was performed for 10 min at 681C. Extracted DNA from 200 ml of plasma was eluted with 50 ml and stored at À201C.

Isolation of DNA from paraffin-embedded tissue
Tumour tissues were obtained at surgery and fixed with paraffin. DNA was extracted from tissue with QIAamp spin columns (QIAamp DNA Mini Kit) according to the manufacturer's instruction. In brief, not more than 25 mg of paraffin-embedded tissue was placed into a 2 ml microcentrifuge tube. Removal of paraffin was ensured by adding and mixing 1200 ml xylene. After centrifugation at full speed for 5 min at room temperature, the supernatant was carefully removed and 1200 ml of ethanol (96 -100%) was added to the pellet. The mixture was centrifuged at full speed for 5 min at room temperature; afterwards, the ethanol was carefully removed. To remove residual xylene completely, the procedure was repeated. The open microcentrifuge tube was incubated at 371C for 10 -15 min until the ethanol had evaporated and the DNA was extracted was with QIAamp spin columns (QIAamp DNA Mini Kit, Qiagen, Hilden, Germany) according to the manufacturer's instruction. Lysis in the presence of proteinase K was overnight at 561C. AL (200 ml) buffer was added and mixed. DNA was eluted with 100 ml of AE buffer and stored at À201C.

Isolation of DNA from SW480 cells
Cells were resuspended in 200 ml of PBS solution. After adding 20 ml proteinase K and 200 ml AL buffer, the mixture was incubated at 561C for 10 min. DNA extraction was continued as with tumour tissue (see above).
Melting curve analysis was performed at steadily increasing temperature from 40 to 851C with a transition rate of 0.31C s À1 . Fluorescence data obtained were analysed using the LightCycler software (software version 3.5, Roche Diagnostics).

Enriched BstNI RFLP/PCR
Enriched BstNI RFLP/PCR was confirmed as described elsewhere (Mora et al, 1998). Digestion of the amplified products with the restriction enzyme BstNI (New England BioLabs, Beverly, USA) was performed according to the manufacturer's instruction. After two rounds of PCR (heminested PCR) and digestions of the amplified DNA, all PCR products were electrophoresed in a 3% agarose gel (low melting point agarose) and the mutant allele (143-bp band) and the wild-type allele (114-bp band) was visualised by ethidium bromide staining (0.5 mg ml À1 ). All synthetic oligonucleotides used were purchased from TIB Molbiol (Berlin, Germany).

Determination of interassay variability and optimisation of PNA concentration
For detection of the hot spot point mutation of the Ki-ras protooncogene at codon 12, we used PNA-mediated PCR-clamping and real-time PCR with mutant-specific hybridisation probes. The relative positions of primers, hybridisation probes and PNA are shown in Figure 1 (Genbank accession no. K01519; nucleotid positions 1 -164). For PNA and hybprobes, the antisense strand was chosen due to the lower purine content to reduce mismatches. After optimising the PCR conditions, the interassay variability of the melting temperatures was determined. For wild-type and valine-mutated DNA, the variability was 68.01C70.8 (n ¼ 15) and 71.21C70.8 (n ¼ 15), respectively, and therefore allows precise discrimination of wild-type and mutant DNA by melting point analysis.
In appropriate clinical samples the proportion of malignant or premalignant to normal cells is extremely low. As shown in Figure 2, the melting points of 100 pg of mutant DNA extracted from SW 480 colon carcinoma cells in the presence of 1 mg of wildtype DNA (1 : 10 000) were not detectable (curve 2 and 4). However now, in the presence of PNA, the wild-type-specific peak disappears and the mutant allele was detectable in a 1 : 10 000 dilution (mut/wt). For complete suppression of wild-type DNA amplification, raising amounts of PNA were used. In the presence of a PNA concentration of 2.5 mM in the assay (curve 8), only the mutant-specific peak was observed comparable to curve 3 of the mutation-specific peak. Lower concentrations of PNA show a weak shoulder at the wild-type-specific melting temperature (curves 5 -7; arrow), indicating incomplete suppression of the wild-type DNA amplification. Under these conditions, quantification of mutant DNA is impossible. It should be noted that higher PNA concentrations lead to a decrease of the efficiency of DNA amplification and therefore lowers fluorescence intensities (data not shown).
Determination of the sensitivity of the assay and quantification of mutated DNA To determine the sensitivity of our assay, serially an equivalent of 2 -20 000 tumour cells of the colon carcinoma cell line SW 480, which bear the codon 12 mutation glycine (GGT) to valine (GTT) homozygous, were mixed each with 2 000 000 normal peripheral blood cells from a healthy subject. After DNA extraction, one-tenth of the eluted DNA was analysed. As seen in Figure 3A, an equivalent of two mutated cells can be detected in the presence of genomic DNA of an equivalent of 200 000 wild-type cells, which corresponds to a sensitivity of 1 to 5 : 100 000. The melting temperatures for wild-type and mutated DNA are 68.6 and 71.21C, respectively. Additionally, the same dilution experiments were performed using enriched BstNI RFLP/PCR (Mora et al, 1998). After gel electrophoresis of the amplified products, the mutant allele (143-bp band) and wild-type allele (114-bp band) could be visualised after ethidium bromide staining. The detection limit of the mutation was 10 À3 (data not shown). Another favourite advantage of our real-time PCR technology is the quantification of mutated DNA simultaneously, if the amplification of the wild-type DNA was completely suppressed as seen in Figure 3A. After analysing the crossing points by the LightCycler software ( Figure 3B, upper panel), the regression line was linear (slope ¼ À2.878; r ¼ À0.99) in the range of 4 -40 000 copy numbers of mutated DNA in the background of 400 000 copies of wild-type DNA ( Figure 3B, lower panel).  Figure 4 (curve 16) using a ratio of mutant to wild-type DNA of 1 to 10 000. In the presence of the wild-type specific PNA, an additional tissue sample of patient no. 1 (Figure 4, curve 3) was analysed and no fluorescence signal was detected, indicating the wild-type allele. All DNAs were tested by amplification without PNA, first.

Genotyping of plasma samples of patients with pancreatic cancer
Adequate amounts of DNA were extracted from 64 plasma samples of 10 patients (seven male, three female) with pancreatic cancer and three healthy subjects. Blood samples were obtained at least monthly over a period up to 13 months. To determine the amount of DNA per assay, real-time PCR was performed in the absence of PNA (data not shown). Owing to the low amount of extractable plasma DNA, we were able to use approximately 0.5 ng per assay. The PCR method using PNA-mediated PCR-clamping and hybridisation probes showed that the plasma samples of the three healthy control cases were negative for Ki-ras mutations as no fluorescence signals were detectable in the presence of wild-typespecific PNA. Ki-ras alterations were found in DNAs from 18 out of 64 (28%) plasma samples of four out of 10 patients with ductal adenocarcinoma of the pancreas. The mutation from glycine to valine was the predominantly detected gene alteration (15 out of 18; 83%). Other mutations were not further analysed. Data synopsis of patients investigated is summarised in Table 1. The median was 11.5 months (5 -83 months) after diagnosis. Two patients were suitable for surgical therapy. Among all patients metastases were detectable during tumour staging.
Time course analysis of the different clinical samples is shown in Table 2 in context with serum levels of the tumour marker CA 19-9, pharmacological treatments and tumour imaging.
Out of the 10 patients investigated, four patients became positive during clinical observation with respect to Ki-ras mutation (nos. 11, 34, 39, 43). In all these cases, the mutated DNA was steadily detected until death of the patients in association with tumour progression and raising CA 19-9 serum levels (410 4 U ml À1 ). The altered Ki-ras gene was not detected in patient no. 40, while CA 19-9 level was slightly elevated.
Three patients (nos. 13, 16, 38) show raising tumour marker values along with progressive disease, but no mutation was detectable in plasma samples. Patients nos. 29 and 37 showed progressive disease, while CA 19-9 values were low and no Ki-ras gene alterations were found.

DISCUSSION
Our aim was to establish an assay for the detection of the hot spot mutation in codon 12 of the Ki-ras gene in plasma samples using PNA-mediated PCR-clamping and mutant-specific hybridisation probes. In contrast to Thiede et al (1996), where in case of mutant DNA the PCR primer outcompete the wild-type specific PNA, we used wild-type PNA (17-mer) and mutant-specific fluorescentlabelled hybridisation probes. Owing to the higher thermal instability of mutant DNA and wild-type-specific PNA hybrids, the detected fluorescence signal corresponds to the amplified mutant DNA and can be analysed by subsequent melting curve analysis.

A B
neoplasm is uncertain but due to the same gene alterations as the primary tumour it is likely that lysed tumour cells are the origin of naked DNA (Castells et al, 1999). The analysis of tumour cells in peripheral blood is difficult because of the rareness of neoplastic cells. Furthermore, large amounts of nonmutated leucocytes mask the vestiges of mutated cells. We restricted our point mutation analysis to one of the preferred glycine (GGT) to valine (GTT) variant as described in literature (Anker et al, 1997), but our method allows the detection of the other gene alterations at codon 12 (aspartic acid, arginine and cysteine), too. The applicability of this method, first of all the correlation of detected gene mutations with clinicopathological findings, was evaluated by subsequent time course analysis of circulating non-cell-associated DNA in peripheral blood of patients with pancreatic cancer. The occurrence of Ki-ras gene mutations in primary pancreatic tumours is well known from literature (Tada et al, 1993;Sorenson et al, 1994;Mulcahy et al, 1998). We also could detect Ki-ras gene mutations in selected tissue samples of the pancreas using our assay.
The sensitivity of our assay revealed a value of 0.001% being the highest compared with other methods for genotyping like ASA-PCR and RFLP-PCR reaching sensitivities in the range of 1.0 -0.001% (Jacobson and Mills, 1994;Rhodes et al, 1997). Furthermore, the rapid cycle PCR enables quantification of the mutant DNA simultaneously if the fluorescence signal of the wildtype DNA is completely suppressed by binding of the PNA.
Quantification of mutant-type Ki-ras gene in plasma samples is inappropriate due to the low amount of DNA and the absence of cellular equivalents which allows the quantification of single copy genes as a reference. This method allows a very fast detection of Ki-ras point mutations (B1 h) after DNA preparation from clinical samples. As no post-PCR handling is necessary, the possibility of contamination is minimised, which contributes to reduce false positive results.
While all previous studies only analysed one sample per individual after diagnosis of pancreatic cancer, we present a preliminary study of time course analysis of Ki-ras gene alterations. Thus, occurrence of mutations in samples analysed in our study can hardly be compared with incidences found by other investigators (Mulcahy et al, 1998;Castells et al, 1999;Dianxu et al, 2002;Maire et al, 2002). The percentages of plasma samples with mutated Ki-ras gene found to date in the existing studies differ in a wide range of 27 -81%, which might reflect collection of samples at different tumour stages and various sensitivities of the assays for detection of Ki-ras point mutations. Dianxu et al (2002) tested 37 of 41 patients (90.2%) with pancreatic cancer positive when plasma Ki-ras mutation analysis was combined with elevated CA 19-9 serum levels (437 Units ml À1 ). In our study, we detected Ki-ras mutant alleles only in four out of 10 patients with high CA 19-9 levels. These differences might be due to different sensitivities of the detection methods, even though the sensitivity of our method was the highest compared to the others. In general, more clinical samples of patients with pancreatic cancer, chronic pancreatitis and healthy individuals have to be analysed for determination of sensitivity, specificity, negative and positive predictive values of the assay presented in this study. Owing to the limited number of patients analysed, a correlation of the detectable Ki-ras mutations with clinicopathological findings and pharmacological treatments is certainly prematurely, but we can demonstrate the potential of the rapid cycle PCR in the presence of wild-type  Tax  Tax  Tax  Tax  Tax  Tax  PD  PR FS ¼ time of 1st collected sample after diagnosis (in months); Ex ¼ death in months after collection of the 1st sample; NA ¼ no data available; Gly ¼ glycine (wild type); Val ¼ valine (mutation); Mut ¼ mutation different from valine; GFF ¼ gemcitabine+5-fluorouracil+folinic acid; Gem ¼ gemcitabine mono; Tax ¼ taxol; OFF ¼ oxaliplatin+5-fluorouracil+folinic acid; NC ¼ no change; PD ¼ progressive disease; PR ¼ partial regression.
PNA and mutation-specific hybridisation probes for detection of point mutations. We could identify Ki-ras-mutated alleles by this rapid real-time PCR at late stages of carcinogenesis very well and may contribute to therapeutic regimes and clinical practice.