K-ras gene mutations are among the most common genetic alterations found in human cancer. Up to 65–100% of pancreatic carcinomas contain K-ras mutations, mainly at codon 12 (Capellá et al., 1991; Villanueva et al., 1996). A similar incidence has been found in cholangiocarcinoma (Levi et al., 1991) while in colorectal cancer this proportion is approximately 40–50% (Bos et al., 1987; Forrester et al., 1987). The high incidence of K-ras mutations and the fact that most of them are restricted to codon 12 have made this genetic alteration an adequate target to evaluate its potential clinical usefulness as a tumour marker.
The development of non-isotopic PCR-based techniques for detection of K-ras mutations has been essential to allow its use in the clinical setting. Several methodologies with relatively high detection limits such as RFLP/PCR (Jiang et al., 1989; Villanueva et al., 1996), single-strand chain polymorphism (SSCP; Orita et al., 1989), denaturing gradient gel electrophoresis (DGGE/PCR; Nedergaard et al., 1997) or others have been used to detect K-ras mutations in tumour samples. Development of PCR-based techniques of lower detection limits has opened the possibility of analysing samples such as pancreatic juices and stool samples which contain a low proportion of tumour cells admixed with a large number of normal cells. These techniques such as intermediate digestion enriched (IDE)-RFLP-PCR (Khan et al., 1991; Nollau et al., 1996; Mora et al., 1998), restriction endonuclease-mediated selective PCR (REMS-PCR; Ward et al., 1998) and continuously enriched by enzymatic digestion-RFLP/PCR (CED-RFLP/PCR; Puig et al., 1999) may detect up to 1 mutant allele among 103–105 normal alleles.
Detection of K-ras mutations in the DNA from colonic and pancreatic cancer cells in the stool or pancreatic secretions has been described (Sidransky et al., 1992; Tada et al., 1993; Smith-Ravin et al., 1995). This may be of help in identifying those subjects at risk for cancer (Kondo et al., 1994; Villa et al., 1996). However, a number of issues have been raised when using these techniques in the clinical setting. Firstly, a high agreement between mutations detected in the exfoliated cells and in tumour cells is required. Secondly, mutations have been detected in the presence of lesions of unknown malignant potential, such as pancreatic duct hyperplasia or colonic aberrant crypt foci, questioning its specificity. Thirdly, the mutation detection system chosen must be reproducible, able to amplify difficult samples (i.e., faecal samples) and should contain dependable internal controls that diminish the possibility of false positives.
Here we analyse the diagnostic utility of the detection of K-ras codon 12 mutation in faecal samples from patients undergoing diagnostic colonoscopy using the CED-RFLP/PCR technique, a robust and sensitive methodology (Puig et al., 1999). Furthermore, we studied its applicability in fine-needle aspirates (FNA) of pancreatic masses from patients with clinical suspicion of pancreatic cancer. We also compared, in both clinical settings, the CED-RFLP/PCR method with 2 other techniques with lower sensitivity: non-enriched (NE) and IDE techniques (Mora et al., 1998).
MATERIAL AND METHODS
Between June 1995 and December 1996, faecal samples collected before colonoscopy were consecutively obtained from patients admitted to the Santa Creu i Sant Pau Hospital to undergo colonoscopy under the clinical suspicion of cancer. Characteristics of the population were 42 male and 24 female; mean age: 67 ± 12. In 1 patient, 2 samples were obtained at different times. Final diagnosis were 12 colorectal adenocarcinomas (3 of them with in situ carcinoma), 25 adenomas, 12 inflammatory bowel disease (4 ulcerative colitis, 3 Crohn's disease and 5 inconclusive), 8 colonic diverticulosis and in 5 cases, other diseases were evidenced (focal cryptic dysplasia, melanosis coli, submucosal lesion, mucosal stenosis and angioplastic lesion). Finally, in 5 patients, no macroscopic or microscopic alteration was evidenced. In 35 of the 67 faecal samples (12 adenocarcinomas, 22 adenomas and 1 focal cryptic dysplasia), tissue samples were also analysed.
In addition, and to further validate our method, we analysed a panel of 61 FNAs from patients with pancreatic masses that have been previously described and analysed for K-ras mutations at codon 12 with NE and IDE techniques (Mora et al., 1998).
Sample processing and nucleic acid preparation
Faecal samples were obtained after laxative treatment of the patients before colonoscopy. Cellular material was collected by serial centrifugation and washed with saline solution. A volume of approximately 300–500 mL of this pellet was digested by proteinase K for more than 12 hr and followed the same procedure as previously described (Mora et al., 1998). We added a final purification step using Wizard DNA Clean-Up System (Promega, Madison, WI) in order to remove PCR inhibitors. DNA from a variable number (3– 6) of 10-μm sections of corresponding paraffin blocks from biopsy tissues was extracted (Villanueva et al., 1996). Tumour cell content was at least 50% in all colorectal tumours analysed. An NP9 human pancreatic cancer cell line derived from a human pancreatic adenocarcinoma that harbours a homozygous aspartic acid mutation was selected as positive control for K-ras codon 12 mutations.
Methodologies for detection of K-ras mutations at codon 12
CED-RFLP/PCR (continuous enrichment for the mutant allele by enzymatic digestion).
This method has been previously described (Puig et al., 1999). Briefly, 0.75–1.5 μL of each sample or control was added to a 50-μL reaction mixture containing a standard PCR mix with 1.5 mmol/L MgCl2, 0.2 mmol/L dNTPs (Promega), 0.1 μmol/L of the external primers K5′ and DD5P, 1 U of Taq DNA polymerase (GIBCO/BRL, Gaithersburg, MD) and 10 U of BstNI (New England Biolabs, Beverly, MA) for 12 cycles as described (Puig et al., 1999) in the presence of PCR buffer supplied by the manufacturer. The upstream primer, K5′ (Jiang et al., 1989), artificially creates a new BstNI restriction site (5′ CCTGG 3′) that is lost when a K-ras codon 12 mutation exists. A volume of 1.5 μL of the first PCR product was reamplified for 35–42 cycles using essentially the same conditions described for the first amplification in the presence of 0.5 μmol/L of the heminested primers K5′ and K3′. K3′ incorporates another BstNI restriction site that serves as an internal digestion control. After enzymatic digestion with BstNI, the products were run on an 8% polyacrylamide gel and stained with ethidium bromide. The relative intensity of the distinct bands observed allows identification of a positive sample with confidence (Fig. 1). The number of PCR cycles in the first-round PCR is critical to achieve an adequate enrichment of the mutant allele without obtaining spurious bands (Puig et al., 1999) that preclude obtaining a conclusive result. All the samples were processed in duplicate and interpretation of the analytical results was performed by naked-eye inspection of 2 independent observers (PP and JM). Only when both observers agreed upon a positive case, a mutation was scored. This method detects 1 mutant allele in up to 105 wild-type alleles.
IDE-RFLP/PCR and NE-RFLP/PCR.
The IDE-RFLP/PCR technique has been previously reported (Mora et al., 1998). It was essentially performed as described above with 2 modifications: the BstNI enzyme was not added to the first PCR reaction that takes place for 12 cycles. Alternatively, 5 μL of the first amplification product was digested in a final volume of 10 μL with 10 U of the restriction enzyme BstNI (New England Biolabs). One to 2 μl of the digestion product was reamplified following the same second-round PCR schedule described above. The small amount of digestion product does not apparently affect PCR efficiency. Finally, enzymatic digestion was performed. With this method, we consistently detected 1 mutant allele among 104 wild-type alleles. The NE-RFLP technique has been also described (Mora et al., 1998) and it was performed as the IDE method without the intermediate digestion step. Using this method, 1 mutant allele is detected when present in up to 102 wild-type alleles. In order to insure the reproducibility and sensitivity of the assay, positive and negative controls as well as serial dilutions of mutant vs. normal alleles (up to 1/100,000) were added in every experiment. Aerosol-resistant tips were used and DNA carryover controls were always added to detect any potential DNA contamination.
Characterisation of mutations
It was performed by means of the SSCP method as described (Mora et al., 1998). Additionally, SSCP analysis was adapted to characterise mutations in IDE-RFLP/PCR amplified samples before the final restriction digestion was performed.
Applicability of CED-RFLP/PCR to the detection of mutations in faecal samples
CED-RFLP/PCR yielded a PCR product in 62 of the 67 prospectively collected faecal samples analysed (93%). In contrast, amplification was possible in 58 cases (87%) with the NE method and in 55 cases with the IDE technique (82%). Using CED-RFLP/PCR, K-ras mutations at codon 12 were detected in 6 of 11 (55%) faecal samples obtained from patients harbouring adenocarcinomas and in 6 of 22 (27%) adenomas. The IDE technique detected mutations in 4 of 11 (36%) adenocarcinomas and in 4 of 16 (25%) adenoma samples. Finally, the NE method detected mutations in a smaller fraction of cases (18% and 22%, respectively). No mutations were detected in inflammatory disease, diverticulosis, other non-neoplastic diseases or in normal colorectal mucosa. (Table I).
Table I. DIAGNOSTIC SENSITIVITY OF THE 3 RFLP/PCR METHODS USED FOR DETECTION OF K-RAS CODON 12 MUTATIONS IN FAECAL SAMPLES (n = 67)
In 35 cases (12 adenocarcinomas, 22 adenomas and 1 focal cryptic dysplasia), corresponding paraffin-embedded samples were available and analysed by the 3 compared methods (Table II). Only in 1 adenoma (case 14) the mutation was exclusively detected by the enriched methods (i.e., CED and IDE), suggesting that a minority of tumour cells were K-ras positive. In this case, mutations were not detected in the faecal sample. Interestingly, the same patient underwent another colonoscopy 3 months later. In this occasion, biopsy evidenced an adenocarcinoma that showed a strong positive signal for the K-ras mutant allele (case 36); at that time, the faecal sample was also positive for the mutation (Table II). When tissues were analysed, mutations were detected in 6 of 12 carcinomas (50%) and in 13 of 22 adenomas (59%). The agreement rate for CED-RFLP/PCR in faecal and tissue samples was 100% in the adenocarcinoma and 65% in the adenoma group (Table II).
Table II. TISSUE AND FAECAL AGREEMENT OF FAECAL ANALYSES IN COLORECTAL SAMPLES ACCORDING TO FINAL DIAGNOSIS1
SSCP analyses of tissue specimens allowed characterisation of 17 of 19 samples. Spectrum was as follows: 10 GAT, 3 GTT, 2 AGT, 1 CGT and 1 TGT (Table II). The 2 cases negative for SSCP were the ones yielding the faintest mutant signal in tissue being the corresponding faecal sample consistently negative. Finally, concordant SSCP results were obtained in faecal and tissue samples.
Applicability of CED-RFLP/PCR to the detection of mutations in pancreatic FNA samples
We analysed a panel of 61 prospectively evaluated FNAs of pancreatic masses previously reported for the IDE and NE techniques (Mora et al., 1998) by the CED-RFLP/PCR method. Both enriched methods (IDE and CED-RFLP/PCR) detected the same number of K-ras mutations (36 positives), although, the PCR yield in the pancreatic cancer group was increased in 2 cases using the CED-RFLP/PCR method (Table III). The use of an increased sensitivity technique did not add false positives and consequently specificity remained 100%. Characterisation was attempted in 29 of the K-ras-positive FNAs by SSCP/PCR and by SSCP of the IDE/PCR samples before final enzymatic digestion. In both cases, concordant results were obtained and the spectrum was as follows: 9 GTT, 8 GAT, 3 CGT and 4 TGT. These results argue against an increased error rate in Taq polymerase when sensitivity was improved.
Table III. SENSITIVITY AND PCR YIELD OF THE DIFFERENT RFLP/PCR METHODS FOR K-ras CODON 12 MUTATION DETECTION IN 61 FNAs OF PANCREATIC MASSES1
In this study, we have analysed the clinical usefulness of a highly sensitive method to detect K-ras codon 12 mutations both in faecal samples obtained from diagnostic colonoscopy and in FNA of pancreatic masses. When compared with 2 previously described methods, the CED-RFLP/PCR method offers a robust alternative for the detection of pre-malignant lesions in faecal samples, which are generally difficult to amplify. In pancreatic cancer diagnosis, its simplified handling may favour its use.
Reliable and sensitive techniques are needed to genetically analyse those non-invasive samples where the low proportion of cancer cells present makes their detection difficult. In patients with clinical suspicion of colorectal cancer, CED-RFLP/PCR gives a high PCR yield, offering the practicability necessary for its use in the clinical setting. Using this technique, a high proportion of K-ras-positive cases, restricted to those patients harbouring colorectal tumours, has been obtained. The percentage of K-ras codon 12 positive adenomas is probably overestimated in our hospital-based study; in this population, a higher incidence of big size adenomas, known to contain more K-ras mutations (Capellá et al., 1990), is observed. Our findings suggest that this technique offers an acceptable alternative for non-invasive diagnosis in a significant proportion of colorectal adenomas or carcinomas. Since colonoscopy is not always complete (Winawer et al., 1997), the molecular diagnosis could be especially useful in those patients in whom colonoscopy cannot be completed or is not accepted.
Specificity of faecal K-ras mutation detection for colorectal tumours has been questioned (Tobi et al., 1994; Ratto et al., 1996; Villa et al., 1996). Previous studies found K-ras alterations in a significant proportion of tissues and/or stools of ulcerative colitis patients (Ratto et al., 1996; Villa et al., 1996). In our study, we did not find any K-ras mutations in the 12 samples obtained from patients with inflammatory bowel disease, ulcerative colitis or Crohn's disease. The low numbers and distinct inclusion criteria used in the different series preclude their comparison. Finally, in our study, no mutation was detected in benign disease or in normal colonoscopy, suggesting that molecular analysis has no false positives when symptomatic patients are evaluated. This is in contrast with a previous study where, in a screening and/or surveillance setting, mutations were reported in the absence of macroscopic lesions (Tobi et al., 1994). Further studies will be needed to clarify the significance of faecal K-ras mutations in asymptomatic patients.
The perfect match evidenced between tissue and faecal samples in the adenocarcinoma group is similar to that previously reported (Sidransky et al., 1992; Smith-Ravin et al., 1995; Ratto et al., 1996). The reasons behind this agreement are still unclear. It has been proposed that colorectal tumour cells could account for up to 1% of the epithelial cells shed from the colon into the stool (Ratto et al., 1996). However, the distinct agreement rates observed in our study between the different techniques utilised (Table II) suggest that this proportion of cancer cells may have been overestimated. Alternatively, it has been suggested that DNA from cancer cells could be more resistant to degradation (Sidransky et al., 1992), offering a better template for the PCR reaction than the rest of epithelial cells. This assumption, however, has not been proved. Furthermore, we have shown that the agreement rate between faecal samples and adenomas remains high (65%). Encapsulation of the pre-malignant tissue and the consequent decrease of the cells shed into the stool could account for the relatively lower agreement observed in adenomas. Altogether, our findings further support the use of faecal K-ras mutation analysis in early non-invasive diagnosis of colorectal tumours. When this methodology was applied to diagnosis of pancreatic masses, the main analytical contribution of the CED-RFLP/PCR approach when compared with the IDE technique, was a better practicability based on a faster procedure and simplified handling. It is of note that the higher sensitivity of the technique offered a better PCR yield without adding false positives, further supporting its reliability in distinct clinical settings that share the clinical suspicion of malignancy.
In conclusion, the CED-RFLP/PCR method for the detection of K-ras codon 12 mutations offers a good alternative for early non-invasive diagnosis of colorectal tumours. In symptomatic patients, the technique does not have false positives and displays a high agreement rate between tissue and faecal samples both in carcinomas and adenomas. This method could offer a good alternative to colonoscopy when this diagnostic procedure cannot be adequately performed. In pancreatic cancer diagnosis, its simplified handling favours its use.
We thank Dr. C. García for supplying DNA from the NP9 pancreatic cancer cell line.