Tobacco Documents Online

hit. J. Cancer (Pred. OncoL): 74, 162-170 (1997) © 1997 Wiley-Liss, Inc. DETECTION OF LOW-FRACTION USING A SENSITIVE METHOD B088 HX434 15~ Z~T d ~ANCER 97 (C)NILEY-LZSS KEOH nai union Against Cancer aationale Contre le Cancer K-ras MUTATIONS IN PRIMARY LUNG TUMORS Phouthone KEOHAVONGhS'*, Dan ZHU~, Alea C. MELACRINOSI. Mary Ann A. DEMICHELE2, Robert J. WEYANT3, James D. LUKETICH4"5, Joseph R. TESTA6. Madelyn FEDDER6 and Jill M. SIEGFRIED2"5 ~ Departtnent of Em'iromnental and Occupational Health, University of Pittsburgh, Pittsburgh, PA 2Departtnent of Pharmacolog3; Universi~." of Pittsburgh, Pittsburgh, PA 3Department of Dental Public Health, University of Pittsburgh. Pittsburgh, PA 4Department of Cardiothoracic Surgeo', Universit3' of Pittsburgh, Pittsburgh, PA SUniversity of Pittsburgh Cancer bzstitute, University of Pittsburgh, Pittsburgh, PA 6Department of Medical Oncolog3; Fox Chase Cancer Center, Philadelphia, PA i i | Mutations in the K-ras gene are often identified in lung tumors and are implicated in the development of lung cancer. We used a sensitive method to analyze low-fraction muta- tions occurring in codon 12 of the K-ras gene in 114 primary lung tumors, including 77 adenocarcinomas, 31 squamous cell carcinomas and 6 adenosquamous carcinomas, which had previously been shown to be negative for codon 12 K-ras mutation in a first screening using less sensitive methods. Sixteen of these tumors were found to contain a low-fraction mutation, including 9 mutations among the adenocarcino- mas, six mutations among the squamous cell carcinomas and one mutation among the adenosquamous carcinomas. Our study also showed that the occurrence of low-fraction muta- tion was associated with a positive smoking history, as was previously found for the occurrence of high-fraction muta- tion. Patients with low-fraction mutations were younger (mean age 58.8 years) than those with either high-fraction mutations (63.2 years) or no mutation (66 years). Patients with low-fraction mutations were more often stage I (8 of I 0) than patients with either high fraction mutations (22 of 44) or no mutation (33 of 71). Moreover, the overall survival was better for the group with a low-fraction mutation than both the high-fraction mutation group and the group with no K-ras mutation, but due to small sample size, the difference was not statistically significant. Our results suggest that using highly sensitive methods of K-ras mutant detection in tumor DNA could obscure differences between patients in whom the mutation is found throughout the tumor, those in whom the mutation is only present in a small subpopulation and those who have no muta~;ion. Int. J. Cancer 74:162-170, 1797. © 1997 Wiley-Liss, Inc. Point mutations of the K-ras gene are common genetic alter- ations in human lung tumors (Barbacid, 1987; Bos, 1989) and are potentially useful diagnostic and prognostic biomarkers for lung cancer. The frequency of K-ras gene mutations in lung adenocarci- nomas has been found to vary among studies from 15-20% (Kobayashi et al., 1990; Suzuki et al., 1990; Sugio et aL, 1992) to about 30% (Rodenhuis et al., 1988; Rodenhuis and Slebos, 1992; Keohavong et al., 1996). On the other hand, these mutations have been reported at a lower frequency (Mitsudomi et al., 1991; Rodenhuis and Slebos, 1992; Wachtenheim et al., 1995; Keo- havong et al., 1996; Rosell et al., 1993, 1996) in squamous cell carcinomas. Studies of K-ras mutations have used different assays capable of detecting cells with a mutated gene present at a traction of at least 10% among cells containing genetically normal alleles of the gene in tumors. Since K-ras mutation is usually found in only one allele, the expected highest mutant fraction is 50%. Previously, we had used polymerase chain reaction (PCR) (Mullis and Faloona, 1987; Saiki et al., 1988) in combination with a second technique, denaturing gradient gel electrophoresis (DGGE) (Fischer and Lerman, 1983), to analyze mutations in the K-ras gene in human primary lung tumors. We reported that this method allowed us to detect K-ras mutations in tumor DNA containing at least 2.5% of mutated cells among cells with a genetically normal allele of the K-ras gene. We found that 31.5% of the lung adenocarcinomas analyzed contained a K-ras gene mutation, almost all of which THIS ARTICLE IS FOR INDIVIDUAL USE ONLY AHD I~AY HOT BE FURTHER REPRODUCED OR STORED ELECTRONICALLY 14ITHOUT HRITTEN,o PERMISSION FROI'I THE COPYRIGHT HOLDER./~L. (fHAf~Tf~R~D REf~RODUGT.r-Of~ I~A~[ RESULT,[ff J ZN FINANOZAL AND OTHER PEHALTZES. V~ (93%) occurred at codon 12 (Keohavong et al.. 1996). Further- more, our results also indicated that the frequency of K-ras mutations in lung adenocarcinomas might be actually higher than 31.5% among the tumors analyzed because some of these tumors showed inconclusive data, indicating that they might also contain a K-ras mutation, but at a mutant traction just below the limit of the detection of the PCR + DGGE method. To test the possibility that K-ras mutations may be present at a relatively low mutant traction in some lung tumors, a method of detection with a level of sensitivity significantly higher than that of the PCR +DGGE method was needed. It is known that DNA polymerases used to catalyze PCR also produce errors, mostly point mutations, in the amplified DNA fragment, giving rise to mutant alleles which can obscure the mutations occurring in the original tumor DNA (Loeb and Kunkel. 1982; Keohavong and Thilly, 1989). Therefore, the sensitivity of the PCR +DGGE method to analyze a heteroge- neous cell population where only a small fraction of the cells are expected to contain a mutation is limited by the polymerase- induced mutant background. Knowledge of the frequency as well as the mutant fraction of K-ras mutations in lung tumors would help in understanding the prevalence, the involvement and the prognostic and diagnostic values of these mutations in the development of lung cancer. In the present study, we improved the sensitivity of the PCR + DGGE method further by including a step of restriction enzyme digestion to enrich codon-specific mutant alleles (MAE) in the PCR- amplified fragments before DGGE analysis. We applied this approach to search for K-ras mutations that may be present at a low percentage in primary lung tumors. The results of this study are discussed in relation to the mutant frequencies of K-ras mutations reported in earlier studies using different methods of detection, and also in relation to clinical behavior. MATERIAL AND METHODS Tissue samples The tissues were fresh-frozen lung tumors obtained following lung resection at the University of Pittsburgh Medical Center and the Fox Chase Cancer Center, Philadelphia. PA. Cases were collected and stored sequentially between 1989-1993. Tissues were collected and processed by a certified pathologist, and histopathology was reviewed to confirm presence of malignant cells prior to collection. Diagnosis of carcinoma primary to the Contract grant sponsor: NIH. contract grant number I-RO3-CA71609- 01; Contract grant sponsor: American Cancer Society. contract grant number IRG-58-32: Contract grant sponsor: NCI. contract grant numbers P20 CA-58235. NO I-CN- 15393-02. *Correspondence to: Department of Environmental and Occupational Health. University of Pittsburgh. 260 Kappa Drive. Pittsburgh. PA 15238. Fax: (412) 624 1020. E-mail: phol .vms.cis.pin.edu Received 17 June 1996: revised 24 December 1996 I | i I ! !

K-ras MUTATIONS IN lung was confirmed by consulting surreal pathology reports for each patient to rule out any tumors metastatic to the lung from other organs. The cases examined for K-ras mutation included 77 adenocarcinomas, 6 adenosquamous carcinomas and 31 squamous carcinomas. These samples had been screened previously' for K-ras mutations occurring at codons 12 and 13 using the PCR + DGGE method, which can detect mutations present at a mutant traction of at least 2.5%, and were classified as being negative for these mutations (Keohavong et aL. 1996). Of the cases reported previ- ously as being negative, all but 14 were reexamined. In these 14 cases, no more DNA was available for analysis. We also included in our clinical analysis an additional 6 cases of adenocarcinoma with a high-traction K-ras mutation for which clinical data were not previously available, making a total of 45 cases of adenocarcinoma with a high-fraction K-ras mutation. Clhzical correlations Clinical information was obtained from each patient's medical record. Patients with any type of adenocarcinoma were tbllowed for outcome. This was not done for squamous carcinomas because of the small sample size (31) and the number of patients lost to follow-up (5,). During follow-up of the adenocarcinoma group, 62 patients were alive (censored) and 57 deaths occurred. Nine adenocarcinoma patients with wild-type K-ras, as well as 3 patients with high-traction K-ras mutations, were lost to tbllow-up. Mean follow-up period was 38 months for survivors (Table II). Recur- rence was documented by a physical examination, radiological tests and standard diagnostic procedures. Parameters examined in relation to the presence of K-ras mutations included age. gender, TMN stage, degree of differentiation and smoking history. Signifi- cance of clinical parameters was assessed by chi-square test or Fisher's exact test. Kaplan-Meier survival curves (Kaplan and Meier, 1958) were genei'ated for overall survival, and prognostic significance was assessed by the logrank test and Wilcoxon test. DNA isolation and amplification in vitro For DNA isolation, lung tissues were digested with ribonuclease A1 and proteinase K followed by phenol-chloroform extraction. This method consistently yields high-purity DNA suitable tbr PCR. Normal human genomic DNA (Promega, Madison. WI) was used as negative control. The first DNA amplification was performed in a 100-lal reaction mixture contaiping 1-2 lag genomic DNA in 10 mM Tris-HCl, pH 8.5, 1.5 mM MgCI,_, 50 mM KCI. 100 laM each dNTP. 0.5 laM each primer and 2.5 units of Taq DNA polymerase (Perkin Elmer. Norwalk, CT). The first round of PCR was carried out for 30 cy- cles (94°C for 1 min, 53°C for 2 min and 72°C for 2 min) using a DNA Thermal Cycler (Perkin Elmer) and the primers PKII-I (sense): 5'-TATTATAAGGCCTGCTGAAA-3' and PKI2-1 (anti- sense): 5'-ATCAAAGAATGGTCCTGCA-3'. The PCR products were purified from the reaction mixture and resuspended in Tris-EDTA buffer. Restriction enzyme digestion and DGGE analysis The codon 12 sequence of the human K-ras gene, 5'-GGT-3', and its flanking codons do not correspond naturally to the site tbr any restriction enzyme. However, the sequence formed between codons 12-13, 5'-GGT GGC-3', closely resembles the restriction enzyme site for Ban I. 5%-GGT GCC-3'. We used a mismatch-- primer, PKB: 5'-AGGCACTCTTG-~CTAC GGCA-3', to substi- tute the second G of codon 13 with a C to cr~~te the site for the restriction enzyme Ban I in the above amplified fragment IFig, 1) using the approach previously described IJiang et al., 1989; Kumar and Dunn, 1989; Mitsudomi etaL, 1991 ). DGGE separates mutant alleles based on the melting difference of the duplex form compared to wild-type as a consequence of mutation occurrence. To be suitable for analysis by DGGE. a DNA fragment must contain two contiguous regions, a high and a low temperature melting domain (Fischer and Lerman. 1983). Such a structure can occur naturally. Otherwise, such as with exon I of the LUNG CANCER 163 K-ras gene. the artificially high temperature melting domain must be added using PCR and a primer containing a short GC-rich sequence. PKGC: 5'GCCGCCTGCAGCCCGCGCCCCCCGTGC- CCCCGCCCCGCCGCCGGCCCGGCCGCCTATAAGGCC- TGCTGAAAATG-3' (Cariello et al., 1990; Sheffield et al.. 1991 : Keohavong and Thilly. 1992). Mutant alleles differing by only a point mutation in their low temperature melting domain will separate from each other and from wild-type mutant alleles when they migrate through a polyacrylamide gel containing an increasing gradient of denaturants, which imitates an increasing temperature from the top to the bottom of the gel. To enhance their separation by DGGE, mutant alleles are boiled and reannealed with the wild-type allele, which gives rise to 2 additional mutant/wild-type heterodu- plexes. These latter molecules are less stable than the original mutant and wild-type homoduplexes and will separate further from them in lower denaturant concentrations of the gel. Even without the additional denaturation-reannealing step, a fraction of the PCR-amplified fragments was usually found as mutant/wild-type heteroduplexes which occurred during the last denaturation- reannealing step of PCR when DNA amplification was no longer at an exponential increase (Keohavong and Thilly, 1989). For these reasons, when genomic DNA contained a mixture of wild-type and a sufficiently high traction of mutant alleles. 4 major bands representing a mutant and a wild-type homoduplex and the 2 respective mutant/wild-type heteroduplexes were usually detected by direct DGGE analysis. Therefore. to create a suitable template both for digestion with Ban I and for DGGE analysis, 10"~ copies of the first round- amplified fragment was used as template for a second round of PCR for 16 cycles using primers PKB and PKGC in a 25-121 mixture containing 10 mM Tris-HC1. pH 8.5. 1.5 mM MgCI2, 50 mM KCI, 35 12M each dNTP. 0.5 laM each primer, 0.25 121 of 3-'P[c~-dCTP] (3,000 Ci/mmole, NEN, Boston, MA) and 1 unit of Taq DNA polymerase. A 10 121-aliquot from each mixture was diluted to a final 100-121 1 × Ban I buffer, and digested with 20 units Ban I (New England Biolabs, Beverly, MA) at 37°C for 2 hr. The DNA was recovered by ethanol precipitation and migrated through an 8% polyacrylamide gel Ibis/acryl = 1/19). The gel was exposed against X-ray film. The position of the DNA in the gel was located by superimposing the autoradiogram on the gel. For DGGE analysis, the portion of the gel containing the DNA fragment resistant to Ban I digestion was excised from the gel and directly transferred into the wells of a denaturing gradient gel which consisted of a 12.5% polyacrylamide gel (bis/acryl = 1/37.5) containing a 20-40% gradient of denaturants (a 100% denaturant contains 40% urea (w/v) and 40% (v/v) formamide). The gel was subjected to electrophoresis for 12 hr under the conditions de- scribed elsewhere (Cariello et al.. 1990; Keohavong and Thilly, 1992) and, afterward, dried and subjected to autoradiography. Mutant alleles present in the DGG were compared and isolated from the gel and characterized by sequencing (Keohavong and Thilly, 1992). RESULTS Sensitivi~." of mtttation (tete('ti(m We first carried out reconstruction experiments to test the level of sensitivity of the PCR + MAE + DGGE approach to detect K-ras mutant alleles in a background of wild-type alleles, in comparison with the sensitivity of the original PCR + DGGE. Genomic DNA containing 10~' molecules of wild-type K-ras codon 12 (GGT) was mixed with different amounts of a known mutant genomic DNA containing a G to T transversion at the second G of the K-ras codon 12 (GTT) to generate DNA samples containing a mutant fraction of 10-L 10-2. 10-3, 10-4. 10-5 and 0 ~,negative control). Exon 1 of the K-ras gene was anaplified from each DNA sample using the appropriate primers and steps as described for Figure t in Material and Methods section. The amplified DNA was subsequently analyzed by DGGE either directly, or after enrich- ment for codon 12 mutant alleles by digestion with Ban I. The

164 KEOHAVONGETAL A PKII-1 eodonl2/13 3' (exonl = 111 bp) PKI2-1 , (181 bp) 5,' , :3' (126 bp) B K-ra# exon 1 with normal codon 12113. K-ras exon 1 with mutated codon 12 3'--L----CCA CCG- ..... 5' 3'- ....-CAA CCG ..... 5' Step 2 Step. 3 5'M----GGT GCC ....... 3' 3'- ...... -CCA CGG- .... 5' Ban I site ,l, PCR using a mismatch primer 5'------GTT GCC .......... 3' 3'- ..... CAA CGG ...... 5' without Ban I site ,~+ Ban I 5'-------G 3'------CCACG GTGCC ..... 3' G ........ .-5' cut 5'------GTI' GCC. 3' 3'- ....CAA CGG ......... 5' uncut Step 4 Separation of uncut fragments from cut fragments by gel electrophoresis Analysis of the uncut fragment by DGGE FIGURE 1 -- PCR + MAE + DGGE protocol. (a) Schematic presentation of steps of PCR preparation of the DNA template. Top: The I I 1 -bp exon 1 of the K-ras gene and surrounding intron sequences (dotted lines). The first PCR reaction using primers PKII-I and PKI2-1. which hybridize to the intron sequences, produces a 181-bp fragment. This latter fragment is amplified by a second PCR reaction using primers PKGC and PKB, which gives rise to the 126-bp fragment. (b) Introduction of Ban I restriction site at the level of codon 12/13 of the K-r~ts gene in the DNA template, and treatment with Ban I. Step I, PCR-amplification of K-ras exon I from genomic DNA extracted from primary lung tumors to produce the 18 l-bp fragment shown in (a). Step 2, the PCR-product is further amplified using the mismatch-primer PKB to introduce the site for the restriction enzyme Ban I and a GC-clamp primer PKGC to create the 126-bp fragment suitable for analysis by DGGE. and .~ZP[c~-dCTP] to label the PCR-product. Step 3, digestion of the amplified 126-bp fragment produced in step 2 with Ban I. followed by gel electrophoresis of the DNA. Step 4, autoradiography of the wet gel. Portion of the gel containing the remaining uncut fragment is located, excised from the gel and directly loaded and electrophoresed on a denaturing gradient gel. The gel is dried and autoradiographed. results of DGGE analysis are shown in Figure 2. When the amplified DNA was directly analyzed by DGGE (Fig. 2a). a major band corresponding to the homoduplex wild-type fragment (wt) and a pattern of background mutant/wild-t.vpe heteroduplexes. formed between wild-type fragment and polymerase-induced mu- tant fragments, appeared in all samples. In addition, the GTT mutation was able to be clearly detected as 3 bands including a mutant homoduplex ~mut. horn.} and the 2 respective mutant/wild-

K-ras MUTATIONS IN LUNG CANCER 165 het.2 het.l holil o 10"I 10-2 10"310"410-5 0 I 10"1 10-2 10"310-4 10-5 0 • het.l FIGURE 2 -- Reconstruction assays to detect K-ras codon 12 mutations by DGGE. Wild-type genomic DNA ( 10~ copies) was mixed with mutant genomic DNA containing a GGT (glycine) to GTT (valine) substitution at codon 12 of the K-ras gene to generate samples containing a mutant 1 "~ ~ 4 5 t~ fraction of 10% (10-), 1.% (10--), 0.1% (10-.), 0.01% (10-). 0.001% (10-), and 0.0% (0). Exon I of the K-izts =ene was amplified from the mixed DNA samples and processed following the steps outlined in Figure 1. (at ,An aliquot from each of the amplified fragments obtained in step 2 was analyzed directly by DGGE. (b) The fragment amplified in step 2 was further digested with Ban I. followed by analysis of the remaining undigested fragment by DGGE, as outlined in steps 3 and 4 in Figure 1. For each sample, the amount of DNA used for restriction enzyme digestion and analyzed in (b) was 5-fold higher than that used for direct analysis by DGGE shown in (at. because of the expected low fraction of the GTT mutation. The mutant fractions from 10-~ to 0 (negative control) in the samples analyzed in both (at and ~b) are as indicated. The position of the wild-type homoduplex (wt), the mutant homoduplex (horn.) and the 2 respective mutant/wild-type heteroduplexes ~het. I and het.2) corresponding to the valine mutation are indicated in lane 10- ~ of (at. and in lane 0 of (b). In (b), the 4 mutant/wild-type heteroduplex bands, each indicated by a line in lane 0, correspond to 2 background mutations, a GAT (aspartate) and an AGT (serine), produced by Taq in the K-~z~s codon 12. type heteroduplexes (het. 1 and het. 2) in the sample containing a mutant fraction of 10-L This same mutant was also detected as faint bands in the sample containing a mutant fraction of 10-~-, but were completely obscured by the background of Taq-induced mutant sequences in samples containing a mutant fra6tion of 10-3 and smaller. When the amplified fragment was first digested with Ban I and only the remaining uncut fragments were analyzed by DGGE (Fig. 2b), the bands corresponding to the wild-type homoduplex (wt in Fig. 2at and most of the mutant/wild-type heteroduplexes produced by Taq seen in Figure 2a disappeared from each lane. On the other hand, the mutant band corresponding to the GTT mutant allele was able to be detected in the samples containing as few as 0.1% mutant cells (Fig. 2b, lane 10-3). The bands corresponding to the mutant fraction of 10-4 were detected as 2 faint mutant/wild-type heterodu- plex bands (lane 10-4), while the mutant fraction of 10--s was totally obscured by the DNA polymerase-induced mutant se- quences (lane 10--st. In addition, a similar pattern of mutant/wild- type heteroduplex bands was reproducibly detected in all samples (indicated by arrows in lane 0), although they appeared less intense in the sample containing a mutant fraction of 10-~, than in all other samples including the negative control (lane 0). These bands corresponded to mutations induced by Taq polymeruse. They were isolated from the gel and further characterized by sequencing analysis and were found to originate from 2 mutant alleles each containing a mutation, including a GGT to GAT and a GGT to AGT, in codon 12 of the K-ras gene. Comparison of the intensity of the mutant bands appearing in Figure 2b between the GTT mutation and the Taq-induced AGT and GAT mutations showed that these latter 2 mutations each occurred at a mutant traction between 10-3-10-4 under the PCR conditions used. including the composition of the reaction mixture and the total number of PCR cycles which can affect error rate and frequency of polymerase- induced mutations (Loeb and Kunkel, 1982: Keohavong and Thilly, 1989). Therefore, the use of the restriction enzyme digestion step to eliminate the wild-type allele and to decrease the polymerase- induced mutant alleles containing normal K-ras codon 12 from the amplified fragments should easily allow detection of a K-ras codon 12 mutation present at a mutant fraction of 10-3. The detection of some mutations present at a mutant fraction lower than 10-3 by this method would be hampered by the Taq-induced noise. Analysis of tumor~ We then applied this PCR + MAE + DGGE method to screen for K-ras codon 12 mutations in 114 primary lung tumors, all of which had been screened previously using the PCR + DGGE approach and found to be negative for K-ras mutations (Keo- havong et al., 1996). Figure 3 illustrates an example of DGGE analysis of 6 of these tumor samples without (Fig. 3at or with (Fig. 3b) mutant allele enrichment. Results in Figure 3a showed that these samples revealed no obvious mutant sequences among the DNA polymerase background noise. When these same DNA samples were further digested with Ban I before analysis by DGGE (Fig. 3b). three of them danes 4-61 each clearly revealed a pattern of additional mutant sequences which were found to correspond to a GAT, TGT and GTT mutation, respectively. Altogether, our study revealed that 16 of 114 tumors analyzed contained a K-ras codon 12 mutation ~vhich failed to be detected

166 KEOtlAVONG ET AL. wt 6 1 2 3 4 5 6 ~..-'~"~"~ ....77-. , . holD.. Fi~URI~ 3 - A.n example of DGGE screening for mutations in codon 12 of the K-ras gene. Shown is a DI3GE analysis of the fragment amplified from 6 tumor samples either without (a) or with (b) enrichment of codon 12 mutant alleles by Ban I digestion, as shown in Figure sample shows a major band corresponding to the wild-type homoduplex Iwt~ and a pattern of mutant/wild-type heteroduplexes produced by Taq. (b) After mutant allele enrichment by restriction enzyme digestion, the wild-t3 pe homoduplex and most of the mutant/wild-type heterodul~lexes disappeared. Samples 4-6 each revealed a new mutation in addition to those induced by Taq. They corresponded to a GAT (aspartate) in lane 4, a TGT (cysteine) in lane 5 and a GTT (valine) in lane 6. For each sample, the amount of DNA used for restriction enzyme digestion and analyzed in (b) was 5-fold higher than that used for direct analysis by DGGE shown in TABLE i - LOW-FRACTION K-ras MUTATIONS IN THE K-r~ts GENE CODON 12~ IN PRIMARY LUNG TUMORS Histology TGT GAT GTT GCT q~al Number of (Cys) (Asp) {Val) (Ala) mutations~ ca.,~es analyzed Adenocarcinomas 5 I 2 1 9 77 Adenosquamous 1 0 0 0 1 6 Squamous 5 1 0 0 6 31 rThe normal sequence tbr codon I2 is GGT (gly). previously by the less sensitive PCR + DGGE method. Table I summarizes the kinds and positions of the mutations and the resulting amino acid changes for the 16 mutations. They consisted of 13 G to T transversions (81%), including I 0 TGT (cysteine) and 3 GTT (valine). 2 G to A transitions (10%), all of which were a GAT (aspartate) and 1 G to C transversion (alanine, GCT). It was important to ensure that these low-fraction mutations actually corresponded to a low percentage of mutated cells within these tumors, and did not derive from a cross-contamination from other samples that contained a K-ras mutation during the process of DNA preparation and analysis, although all necessary precautions were taken to avoid this potential problem. For this purpose, we compared both the type of mutations and the date of DNA preparation and analysis tbr these 16 tumor samples that contained a low-percentage mutation with those of the tumors which had been previously found to contain a higher percentage of K-ras mutations (Keohavong et aL. 1996). When there was a possibility that a low-fraction mutation might originate from a cross-contamination from a neighboring sample, that mutation was further reexamined at the level of the original stock of tumor. Using this approach, we were able to confirm that these 16 low-fraction mutations were indeed detected in the original tumor samples Idata not shown). Tumor samples were also examined histologically at the time of freezing to verity' that malignant cells were present. Normal tissue components were found to make up no more than 10% of these tumor specimens. Cli2ffcal data analysis. The clinical significance of the low-fraction mutations occurring in adenocarcinomas was examined by comparing clinical character- istics ~Table II) and outcome those with either high-traction mutations or a wild-type K-ras. The group with high-fraction mutations ~vas largely described previ- ously IKeohavong et aL. 1996). It also includes 6 additional patients whose clinical intbrmation was not known previously. Clinical correlations were not examined in squamous cell tumors because of the small sample size. The mean age of patients with a low-fraction K-ras mutation 158.8 years) was significantly lower than the age of either the high-fraction K-ras group (63.2 years) or the wild-type group (66 yearsl, p = 0.03. There was also a trend toward a less advanced stage in the low-fi'action group (8 of 10 cases were sta,,e I compared with ~'~ of 44 in the hieh-fraction group and 33 of 71 in the wild-type group, p = 0.09~. As was observed previously for the high mutant fl'action group, there was an association with.positive smoking history and a low-fraction K-ras mutation compared with the wild-type group (p = 0.03). i/nplying that these low-fraction mutations also arise alter exposure to the carcinogens in tobacco sntoke. The degree of differentiation of the tumor and the gender of the patient ~vere not significantly different among groups. Because of the possibility that neoadjuvant therapy might ulter the components of the tumor, the therapy status of the patients prior to surgery was examined. Only 2 stage IV patients received neoadjuvant therapy: neither of these patieuts had a detectable K-ras mutation, while still having macroscopic tumor at the time of surgery, Some differences were noted in survival, but did not reach the 95% confidence level of significance. There were fewer recur- rences in the low-fraction group than in the other 2 groups and

K-tax MUTATIONS IN LUNG CANCER TABLE II - CLINICAL INFORMATION ON 128 LUNG ADENOCARCINOMA AND ADENOSQUAMOUS CARCINOMA~ PATIENTS USED FOR K.rets MUTATIONAL ANALYSIS Wild type Mutated high fractmn Mutaled Iov, fraction Number of patients n = 73 n = 45 n = 10 Mean age +-- SEM (years) 66 --+ 1.2 63.2 ~ 1.5 58.8 ~ 3.24 Gender Male 41 (56.2%) 2~ (60.0%) 3 (30.0%) Female 32 (43.8%) 18 (40.0%) 7 (70.0%) Stage-' I 33 (45.2%) 22 (48.9%) 8 (80.0%) II 11 (15.1%) 9(20.0%) 1 (10.0%) Ilia 15 (20.5%) 8 (17.8%) I (10.0%) llI~ 4 (5.5%) I (2.2%) 0 IV 8 (I 1.0%) 4 (8.9%) 0 Unknown 2 (2.7%) 1 (2.2%) 0 Degree of differentiation Poor 28 (38.4%) 14 (31. I% ) 5 (50.0%) Moderate 25 (34.2%) 18 (40.0%) 3 (30.0%) Well 15 (20.6%) 12 (26.7%) 2 (20.0%) Unknown 5 (6.8%) 1 (2.2%) 0 Smoking history (packs-year)5 Nonsmoker 13 (17.8%) 1 (2.2%) 2 (20.0%) <25 18 (24.7%) 9 (20.0/%) 1 (10.0%) >25 39 (53.4%) 33 (73.3%) 7 (70.0%) Unknown 3 (4. 1%) 2 (4.5%) 0 Number of recurrences during follow-up3 41 of 64 (64.1%) 25 of 42 (59.5%) 4 of 10 (40.0%) Number of cancer deaths during tbllow-up3 35 of 64 (54.7%) 20 of 42 (47.6%) 2 of 10 (20.0%) Mean follow-up period for deceased patients 18.9 - 24.2 14,1 --+ 12.2 23.5 --- 17.7 (months) Mean follow-up period for survivors (months) 38.7 +-- 20.2 38.2 -4- 17.0 38.3 +" 16.3 tAdenocarcinoma included the lbllowing confirmed diagnosis: adenocarcinoma, adenosquamous carcinoma and bronchioloalveolar carcinoma. Sequential cases were collected between 1989-1993. This analysis includes 45 patients with high-traction mutations.~TNM pathological staging determined at time of biopsy.~Follow-up is from time of biopsy. Follow-up data exclude 9 patients with wild-type K-ras and 3 patients with high-fraction mutated K-ras who were lost to follow-up.-4p = 0.03 compared with other 2 groups.--~p = 0.03 for positive smoking history for both high- and low-fraction K-ras mutation groups compared with wild-type. 167 fewer cancer deaths (2 of 10 vs. 20 of 42 in the high-fraction group and 35 of 64 in the wild-type group (p = 0.09). This probably reflects the fact that 8 of the 10 patients with a Iow-lt'action mutation were stage I. Kaplan-Meier survival analysis showed a better survival for the low-traction mutation group compared with the other 2 groups (Fig. 4a, p = 0.06). but when the stage I patients were analyzed,separately, this difference was not observed (Fig. 4b, p = 0.50). The better overall survival of the low-fraction group compared to the other 2 groups is best explained by the large percentage of early-stage cancers in this group. DISCUSSION Sensitive methods of mutant detection are required to examine nontumor tissue for possible precancerous lesions and to screen for mutant subpopulation in tumors. We improved the sensitivity of the PCR + DGGE method to analyze mutations occurring at a low mutant fraction in codon 12 of the K-ras gene in primary lung tumors. This improvement was achieved by digesting the PCR- amplified fragment with a restriction enzyme to selectively enrich K-ras codon 12 mutant alleles, prior to DGGE analysis. This approach allowed detection of mutations occurring in codon 12 of. the K-ras gene at a mutant traction as low as 0.1% among cells containing genetically normal K-ras codon 12. The detection of mutations present at a mutant fraction lower than 0.1% can be limited by the background of Taq-induced mutations. Our results showed that Taq induced a G to A mutation at both Gs of codon 12 (GGT) of the K-ras gene, giving rise to a mixture of AGT and GAT which appeared among other K-ras gene codon 12 mutation,; selectively enriched by Ban I digestion. Therefore. among the 6 possible single base-pair substitutions that may involve either the first G (AGT. CGT or TGT) or the second G (GAT. GCT or GTT) of the wild-type K-ras codon 12. the presence of an AGT or a GAT mutation at a mutant fraction of at least 10-3 in the original genomic DNA would be detected by an increased intensity of the mutant bands corresponding to either an AGT or a GAT induced by Taq relative to those in the negative controls (genomic DNA without such mutations). An example of the detection of a low-frequency GAT mutation in one lung tumor sample using this MAE + DGGE method is shown in Figure 3b, lane 4. On the other hand, other types of mutations involving one or both Gs of K-ms codon 12, including the remaining 4 single base-pair substitutions, double base-pair substitutions and deletions of one or both Gs. present at a mutant traction of at least 10-3. are expected to be easily, detected by our method since the patterns of these mutations in DGGE are different from each other and from those of the AGT and the GAT mutations (Keohavong et aL, 1996). Using this sensitive method. 16 of the 114 "'wild-type'" lung tumors analyzed (14%) were tbund to contain a low-fraction K-ras mutation, including 9 among 77 adenocarcinomas (11.6%), 6 among 31 squamous cell carcinomas (19.4%), and I among 6 adenosquamous cell carcinomas 116.7%). These low-fraction muta- tions consisted mostly of GGT to TGT (cysteine) (62.5%) and of G to ":F transversions (81%). As a comparison, our earlier study of this same pool of lung tumors using the less sensitive PCR + DGGE method had identified 40 mutations occurring in codon 12 of the K-ras gene among 173 tumors. These mutations included 12 GTT (valine), t2 TGT (cysteine). 11 GAT (aspartate), 2 of each serine (AGT) and arginine ~CGT) and I phenylalanine ~TTT) (Keo- havong et al., 1996). There was thus a predominance of G to T mutations (25/40 or 62.5%) which changed the wild-type glycine mostly m a valine, a cysteine and an aspartate. It is notable that 37.5% (6/16~ of these low-fraction mutations were identified in squamous cell carcinomas. Therefore. our study showed that about 20% (6/31) of squamous cell carcinoma's

1.0 0.9 0.8 0.7 0.6 03 0.4 O.3 0.2 0.1 0.0 A 0 10 20 30 40 50 60 70 80 90 Wild Type High Fraction Mutant Low Fraction Mutant Survival Time (Months) 0.7 "1 I 0.6 0.5 0.4 0.3 0.2 0.1 0.0 B 0 10 20 30 40 50 60 70 80 Wild Type High Fraction Mutant Low Fraction Mutant Survival Time (Months) FIGURE 4-- Kaplan-Meier survival curves for king adenocarcinoma and adenosquamous carcinoma patient~. (a) Data ['or all patients with low-traction K-ras mutations {n = 10). high-traction K-ras mutations (n = 42) and wild-type K-ras (n = 64). ~b) Data ft>r stage 1 patients only with low-traction K-ras mutations (n = 8). high-traction K-ras mutations (n = 20) and wild-type K-ras (n = 30).

K-ras MUTATIONS contained at least a subpopulation with a K-ras mutation in codon 12, which is in agreement with the 18% frequency reported by Rosell et aL (1996), but higher than those reported by other studies (Mitsudomi et al., 1991: Rosell et ill., 1993; Wachtenheim et al., 1995; Keohavong et aL, 1996) and absence of such mutation (Rodenhuis and Slebos. 1992) among these lung tumor subtypes, using less sensitive methods of detection. Altogether, the total occurrence of high- and low-fraction mutations in codon 12 of the K-ras gene in lung adenocarcinomas identified in both our earlier study (Keohavong et aL, 1996) and in the present study was 43.8%. This frequency was close to that reported by Mills et aL (1995a), who used a PCR-based sensitive technique to analyze lung adenocarcinomas, and found that 48% of them contained a mutation in codons 12, 13 and 61 of the K-ras gene, half of which were detected at a mutant fraction of 1% or less in lung tumors. These results suggest that one reason for the low-frequency K-ras mutations observed in earlier studies may be that a significant number of these mutations failed to be identified using less sensitive techniques (Rodenhuis et aL, 1988; Kobayashi et aL, 1990; Suzuki et aL, 1990; Rodenhuis and Slebos, 1992: Sugio et al., 1992). The technical ability exists to detect these low-fraction muta- tions, but they may not be indicators of poor prognosis. The data presented here along with those of Mills etal. (1995b) suggests that a significant number of lung tumors contain K-ras mutations only in a small population. This implies that mutation at this locus is a late event in many cases. The increase in this mutant population could be associated with progression, but if it is present in only a small fraction of the tumor at the time of surgery, this may indicate that progression has been limited. Although the number of patients with low-traction mutations in our study is small, our data suggest that adenocarcinomas with low-fraction mutations may occur in younger, early-stage patients, who have a better outcome than the general lung cancer population. Most of the patients in whom these low-fraction mutations were detected were diagnosed with lung cancer as an incidental finding. Because they were younger, their tumors may have been at an early stage of progression in which occult local or regional metastases had not yet formed. While it is possible that less sensitive methods of mutation detection may cause some false negatives, especially if the tumor contains a large amount of normal compon, ents, it is also possible that use of highly sensitive techniques will obscure differences between tumors in which the K-rqs mutation is present in most of the cells, and tumors IN LUNG CANCER 169 in which it is only a small subpopulation. The biology of these 2 groups may be different. Studies showing evidence for K-ras mutation as both as an early event ILl et aL, 1994; Mao et al., 1994; Mills et al., 1995b) and a late event (Sugio et aL, 1994: Rosell et aL, 1996) in lung cancer development have been presented. Based on the theory of tumor formation through a clonal expansion from an initial mutated cell. one would expect all tumor cells or a high fraction of them to contain the same mutation if the event is early. On the other hand, the presence of only a small percentage of mutated cells in the tumor suggests the mutation occurred later, perhaps during progres- sion. It is also possible that due to tumor remodeling the subpopula- tion, although formed late. could become the dominant population over time. Our results show that the percentage of cells with a K-ras mutation can vary. from less than 1% to up to 50%. This observation suggests that while in some cases K-ras mutations may occur early, in other tumors they occur later, probably as a result of continued exposure to tobacco smoke. This may be compounded by loss of DNA repair in an unstable, malignant nucleus. These low-fraction mutations probably do not contribute to clinical outcome. The method described herein can be applied to screen for potential infrequent mutations at specific codons in other cancer genes in lung or other tumor types. It allows analysis of several samples simultaneously and is also relatively simple because DGGE allows detection of each base-pair substitution in a unique and recognizable pattern of bands (Keohavong et al.. 1996). Therefore, once a distinguishable pattern is identified by DGGE there is no need for further characterization analysis of each individual mutant. Furthermore, the sensitivity may be further improved, for instance, by using a DNA polymerase with a fidelity higher than that of Taq and/or by using more than one step of restriction enzyme digestion, as described by Mills et al. (1995a), before DGGE analysis of the uncut fragments. ACKNOWLEDGEMENTS This work was supported by NIH grant I-RO3-CA71609-01, American Cancer Society Institutional Research grant IRG-58-32. NCI Lung SPORE grant P20 CA-58235 to the University of Pittsburgh Cancer Institute and NCI contract NO1-CN-15393-02 for the Early Detection Research Network. REFERENCES BARBACID, M., ras genes. Ann. Re~: Biochem.. 56, 779-827 (1987). BOS, J.L., ras oncogenes in human cancer: a review. Cancer Res., 49, 4682--4689 (1989). CARIELLO, N.F., KEOHAVONG, P., KAT, A.G. and THILLY, W.G., Molecular analysis of complex human cell population: mutational spectra of MNNG and ICR-19 I. Mutation Res., 231, 165-176 (1990). FISCHER, S.H. and LERMAt~, L.S., DNA fragments differing by single base-pair substitutions separated in denaturing gradient gels: correspon- dence with melting theory. Proc. nat. Acad. Sci. (Wash.), 80, 1579-1583 ~ 1983). JIANG, W.. KAHN. S.M.. GUILLEM, J.G., LU, S. and WEINSTEIN. R.B.. Rapid detection of ras oncogenes in human tumors: application to colon,. esophageal, and gastric cancer. Oncogene, 4, 923-928 (1989). KAPLAN. E.L. and MEIER. P., Nonparametric estimation for incomplete observations. J. Amer. Star. Assoc., 53, 457---481 (1958). KEOHAVONG. P., DE MICHELE. M.-A.A., MELACRINOS. A.C.. LANDRENEAU. R.J., WEYANT. R..J'. and SIEGFRIED, J.M.. Detection of K-ras mutations in lung carcinomas: relationship to prognosis. Clin. Cancer Res., 2, 41 I--418 (1996). KEOHAVONG. P. and TltlLLY. W.G.. Fidelity of DNA polymerases in DNA amplification. Proc. nat. Acad. Sci. (W~t.~h. ~. 86, 9253-9257 (1989). KEOHAVONG, P. and TtIILLY. W.G., Mutational spectrometry: a general approach for hotspot point mutations in selectable genes. Proc. nat. Aca~L Sc'i. (I¼tsh.), 89, 4623--4627 (I 992L KOBAYASHI, T., TSUDA, I'[.. NOGUCHL M.0 HIROHASHL S.. SHhMOSATO. Y., GOYA, Z. and HAYATA. Y.. Association of point mutation in oncogene in lung adenocarcinoma with particular reference to cytologic subtypes. Cancer, 66, 289-294 (1990). KUMAR, R. and DUNN, L., Designed diagnostic restriction fragment length polymorphism for the detection of point mutations in ras oncogenes. Oncogene Res.. 1,235-241 t 1989). LI. Z.-H., ZHENG, J.. WEISS. L.M. and SHIBATA, D., c-K-ras and t,53 mutations occur very early in adenocarcinoma of the lung. Amet: d. Pathol., 144, 303-309 (1994). LOEB, L.A. and KUNKEL. T.A.. Fidelity of DNA synthesis..4mr. Re~: Biochem., 52, 429--457 (1982). MAO, L., HRUBAN, R.H., BOYLE. J.O., TOCKMAN, M. and SIDRANSKYo D., Detection of oncogene mutations in sputum precedes diagno,~is of ltmg cancer. CancerRes.. 54. 1634-1637 (1994). MILLS, N.E., FISHMAN. C.L.. Ro,',t. W.N., DUBIN, N. and JACOBSON, D.R., Increased prevalence of K-tits oncogene mutations in lung adenocarci- noma. CancerRes.. ~, 1444-1447 (1995a). MILLS, N.E., FISHMAN, C.L.. SCHOLES, J., ANDERSON. S.E.. ROM. W.N.. DUBIN. N. and JACOBSON. D.R., Detection of K-ras oncogene mutations in bronchoalveolar lavage fluid for tung cancer diagnosis. J. nat. Cancer Inst.. 87, I056-1060 ( 1995bL MITSUDOMI. T.. VI,\LLET. J.. MULSHINE. J.L., LINNOILA. R.I.. NIINNA, J.D. and GAZDAR, A.F.. Mntations of ~x~s genes distinguish a subset or"

170 non-small-cell lung cancer cell lines from small-cell lung cancer cell lines. Oncogene. 6, 1353-1362 ( 1991 I. MULLtS. K.B. and FALOONA. F.A.. Specific synthesis of DNA hz vitro via a catalyzed polymerase chain reaction. Methods EnzymoL. 155, 335-350 119871. RODENHUIS. S. and SLEUOS. R.J.. Clinical significance of ras oncogene activation in human lung cancer. Cancer Res.. 52, 2665s-2669s ( 19921. RODENHrdlS. S., SLE8OS. R.J.. BOOT. A.J.M.. EVERS. S.G.. Moot. W.J.. WAGENAA~. S.S.. VAN BODEGOM. P.C. and Bos. J.L., Incidence and possible clinical significance of K-ras oncogene activation in adenocarcinoma of the human lung. Cancer Res.. 48, 5738-5741 ( 19881. ROSELL, R., L), S., SKACEL. Z., MATE. J.L., MAESTRA. J.. CANELA. TOLOSA. E.. ARMENGOL. P.. BARNADAS. A. and ARIZA, A., Prognostic impact of mutated K-ras gene in surgically resected non-small cell lung cancer patients. Oncogene. 8, 2407-2412 ( 19931. ROSELL. R.. MONZO. M.. PIFARRE, A.. ARIZA. A., SANCHEZ. J.J., MORENO. I.. MAUREL. J., LO~'Ez. M.P., AaAO. A. and DE ANTA. J.M.. Molecular staging of non-small cell lung cancer according to K-ras genotypes. Clin. Cttncer Res.o 2, 1083-1086 ( 19961. SAIKI. R.K., GELFAND. D.H.. STOFFEL. S.. SCHARF. S.G.. HIGUCHI. R.. HORN. G.T.. MULLIS, K.B. and ERL~CH, H.A.. Primer-directed enzymatic amplifica- KEOHAVONG ETAL. tion of DNA with a thermostable DNA polymerase. Science. 2~9, 487--.,t91 (1988). SHEFFmLD. V.C.. Cox, D.R.. LERMAN. L.S. and MYERS. R.M.. Attachment of a 40-base-pair G+C rich sequence to genomic DNA fragments by polymerase chain reaction results in improved detection of single base changes. Proc. nat. Acat~ SoL (Wash.). 86, 232-~36 ( 1991 ). SUGIO. K.. ISHIDA. T.. YO~<.OYA,',~A. H.. INOUE. T.. SUGIMACHI. K. and SASAZUKI. W.. rtts gene mutation as a prognostic marker in adenocarcinoma of the human lung without lymph node metastasis. Cancer Res.. 52, 2903-2906 ( 19921. SL'G~O. K.. K~S~MOTO, Y.. VmMANt. A.K.. HUNG. J.Y. and G.-~.DAR. A.E. K-ras mutations are a relatively late event in the pathogenesis of lung carcinoma. Cancer Res.. 54, 5811-5815 (1994). Suzuki. Y.. ORtTA, M.. SHmAtSH~. M.. HAYASm. K. and SE~<I'~'A. T.. Detection of ras gene mutations in human lung cancers by single-strand conformation polymorphism analysis of polymerase chain reaction prod- ucts. Oncogene. 5, 1037-1043 (19901. ~VACHTENHEIM. J.. HORAKOVA. l.. NOVOTNA. H.. OPALKA. P. and ROUBKOVA. H.. Mutations of the ras-oncogene and absence of H-ras mutations in squamous cell carcinoma of the lung. Clin. CancerRes.. 1,359-365 ( 19951. li i ,! ( [ ( I i I I I