Differences in the frequencies of K-ras c12–13 genotypes by gender and pathologic phenotypes in colorectal tumors measured using the allele discrimination method

Abstract

The frequencies of different genotypes of the K-ras oncogene in colorectal cancer (CRC) reveal complex relationships among gender, age, and tumor aggression, however, differences among these studies could also be attributed to a lack of standardization of the detection methods used. We developed the allele discrimination assay, which uses dual-color real-time polymerase chain reaction (qPCR) as a fast K-ras genotyping method, and demonstrated higher sensitivity and specificity than DNA sequencing with formalin-fixed paraffin tissues. The assay detected K-ras mutations among 83 of 204 patients with CRC (40.7%); 20.6% of these mutations were G12D (GAT) mutations, 7.4% were G13D (GAC) and G12V (GTT), and 5.3% were other types. A higher proportion of females was observed overall in tumors with K-ras mutations (60.2%, P = 0.01), codon 12 mutations (63.2%, P = 0.005), and transversions (69.6%, P = 0.02), which reflected the higher prevalence of females among the well- to moderately differentiated tumors (29% in males vs. 53% in females; interaction P = 0.03). The opposite was observed for poorly differentiated tumors (47% in males vs. 35% in females). No significant influence of age was found on the prevalence of K-ras mutation. Males with pathological changes and females with poorly differentiated tumors displayed GAT as a less common genotype compared with most other prevalence studies. In conclusion, allele discrimination, with no additional amplification step, is a fast and reliable genotyping method for detecting K-ras c12–13 mutations. Using this method, we demonstrate differences in the frequencies of K-ras genotypes by gender and pathologic phenotypes of CRC. Environ. Mol. Mutagen., 2012. © 2011 Wiley-Liss,Inc.

INTRODUCTION

Mutations in the K-ras genes play an important role in tumorigenesis, such as in the transition from adenomas to carcinomas in colorectal cancer (CRC). Clinical evidence has associated K-ras mutations with poor prognosis and resistance to anti-epidermal growth factor receptor (EGFR) therapy in CRC [Amado et al.,2008; Pajkos et al.,2000; Wong and Cunningham,2008]. Prescreening of mutations in K-ras codon 12–13 is used for early detection of CRC and is considered mandatory for the selection of therapeutic options [Lefferts et al.,2010; Souglakos et al.,2009; Wong and Cunningham,2008]. However, because evidence of direct associations between K-ras mutations and pathologic phenotypes is lacking, how the K-ras mutations predispose to CRC malignancy and metastasis remains unclear [Al-Mulla et al.,1998; Ohnishi et al.,1997; Vogelstein et al.,1989].

Single-base substitutions of the K-ras gene in the ras oncogene family are extensively investigated genetic markers in human cancer pathogenesis and therapies. More than 90% of single-base substitutions in the K-ras gene occur at glycine 12 (GGT) and 13 (GGC), whereas 5% occur at glutamine 61 (CAA) and at other codons [Marchetti and Gasparini,2009; Poehlmann et al.,2007; Toyooka et al.,2003]. Novel oncogenic mutants at other codons—including G15 (GGC) [Wang et al.,2003, 2007], L19 (TTG) [Akagi et al.,2007; Rouleau et al.,2008; Simi et al.,2008], Q22 (AAG) [Miyakura et al.,2002; Palmirotta et al.,2009; Tsukuda et al.,2000], and A146 (GCA) [Chang et al.,2009; Edkins et al.,2006]—have been detected using predominantly DNA sequencing and single-strand conformation polymorphism (SSCP) analysis. Nonetheless, the detection of the single-nucleotide variants in K-ras codon 12–13 has the most important implications for clinical prognosis and drug discovery.

Frequency of the K-ras codon 12–13 mutants may eventually become a marker for distinguishing phases of CRC (from carcinogenesis to metastasis) and predicting clinical outcomes of patients with CRC [Brink et al.,2003]. However, correlation studies of K-ras mutation frequencies to CRC pathology may be influenced by the mutation detection method used [Kobunai et al.,2010], as well as pathological features, and sample size or characteristics. In this study, the dual-color allele discrimination was used to detect single-nucleotide mutations in the K-ras gene, with the Fam-fluorescence intensity (Y) indicating the presence of mutant alleles. Individual quality controls (mixtures of mutant and wide-type plasmids) were monitored in each assay. The assay's performance was evaluated against DNA sequencing. For verification of discrepant results, the restriction fragment length polymorphism (RFLP) or denatured high-performance liquid chromatography (dHPLC) method was used. The aims of this study are to validate the allele discrimination assay, determine the K-ras status of 204 patients with CRC, and assess the prevalence of different K-ras genotypes by age, gender, tumor differentiation, and pathological changes.

CRC, colorectal cancer; dHPLC, denatured high-performance liquid chromatography; MT, mutant; NTC, no-template control; qPCR, real-time polymerase chain reaction; RFLP, restriction fragment length polymorphism; WT, wild-type.

MATERIALS AND METHODS

Genomic DNA (gDNA) Extraction and Collection of Patient Specimens

A total of 21 cell lines of colorectal and lung cancers were sequenced and tested using all assays. The genomic DNA (gDNA) was extracted with proteinase K digestion, followed by DNA extraction using DNeasy Blood & Tissue Kits (Qiagene, Valencia, CA). Tumor samples were collected from 204 patients with CRC diagnosed between 2007 and 2009. All patients gave informed consent for tissue collection and analysis. All specimens were stained with Hematoxylin and Eosin and were re-evaluated by a pathologist with the National Health Research Institutes' National Institute of Cancer Research. This study was approved by the Human Experiment and Ethics Committee of National Cheng Kung University Hospital.

Sequencing Analysis

Exon 1 of the K-ras gene was amplified using a T3 thermocycler (Biometra, Germany) using 1 μM of primers (as previously described [Poehlmann et al.,2007]), dNTPs, and VioTaq DNA polymerase (Viogene, Taiwan), with the following polymerase chain reaction (PCR) conditions: 94°C for 10 min, followed by 35 cycles of 94°C for 15 sec, 55°C for 10 sec, and 72°C for 20 sec. The PCR products were purified with Gel/PCR DNA Fragments Extraction Kits (GeneAid, Taiwan) before being amplified using a GeneAmp 9700 thermocycler (Applied Biosystems, Foster City, CA) and sequenced on an ABI 3730xL DNA analyzer (Applied Biosystems, Foster City, CA) using BigDye Terminator Cycle Sequencing Ready Reaction Kits (Applied Biosystems, Foster City, CA).

Allele Discrimination Analysis of 7 K-ras Mutations at Codon 12–13

Fam-labeled hybridization probes were designed for seven individual mutations (1A, 1T, 1C, 2A, 2T, 2C, and 5A) as follows: TTGGAGCTAGTGGC, CTACGCCACAAGCT, TTGGAGCTCGTGGC, CTGATGGCGTAGGC, ACGCCAACAGCTC, CTACGCCAGCAGCT, and CTGGTGACGTAGGCA, respectively. Each was paired with a Vic-labeled unidirectional wild-type (WT) probe (sense: TCTGGTGGCGTAGGC or anti-sense: CCTACGCCACCAGCT) for simultaneous detection of both mutant (Y) and WT (X) alleles. To each real-time PCR (qPCR) reaction of 1 × of TaqMan Genotyping Master Mix (Applied Biosystems, Foster City, CA) the following were added: 20 ng of gDNA, 3.38 pmol of paired Fam- and Vic-labeled probes, and 13.5 pmol of primers (forward: 5′-AggCCTgCTgAAAATgACTgA AT and reverse: 5′-gCTgTA TCgTCAAggCACT). An ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA) was programmed as follows: stabilization at 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 sec and 60°C for 60 sec. Mixtures of 1–5% mutant DNA plasmids in WT DNA plasmids, constructed by site directed mutagenesis, were used as positive controls. For each experiment, the no-template control (NTC) reactions were included as the negative control. For patient specimens, K-ras mutants were distinguished as those with a Fam-intensity equal or above the 5% mutant DNA plasmid.

RFLP and dHPLC Analyses of K-ras Codon 12–13 Mutations

In PCR-RFLP, a F12s-BstNI primer (5′-ACTgAATATAAACTTgTggTAgTTggACCT) and a K-ras ex2F primer (5′-gAATggTCCTgCACCAgTAA) were used to introduce a BstNI recognition site into codon 12 of K-ras WT as previously described [Ohnishi et al.,1997]. The R13as-HaeIII primer, 5′-gTATCgTCAAggCACTCTTgCCTAgg, and the K-ras ex2R primer, 5′-gTgTgACATgTTCTAATATAgTCA, were used to introduce a HaeIII site into codon 13 of K-ras WT. The PCR program was as follows: 94°C for 10 min, 35 cycles of 94°C for 15 sec, 55°C for 10 sec, and 72°C for 20 sec. The PCR products were digested with BstNI and HaeIII at 37°C overnight, and separated by 4% agarose gel.

For dHPLC, the PCR products (produced above) were denatured at 95°C for 4 min and cooled to 25°C at a rate of −0.1°C/min to form DNA duplex. The PCR products were injected into a WAVE DNA Fragment Analysis System (Transgenomic, Omaha, NE) and eluted at a flow rate of 0.9 mL/min within a linear acetonitrile gradient as previously described [Li et al.,2008]. The eluted DNA fragments were monitored by a UV detector at 260 nm; the elution temperature of dHPLC was 59.5°C.

Statistical Analyses

The distribution of K-ras mutations in patients with CRC was assessed by SAS Version 9.1 (Cary, NC). Differences in the distributions of K-ras mutations, specific mutations, as well as combined mutational groups (codons 12 and 13, transitions, and transversions) by age, sex, and histotypes were determined using χ² tests or Fisher's exact test. Transitions and transversions are different classes of bp substitutions that may reflect differences in carcinogen exposure and/or genetic pathways. Interactions between age, sex, and histotypes on the distribution of K-ras mutations were also assessed by multivariable logistic regression.

RESULTS

Validation of the Allele Discrimination Assay

Individual probes and primer sets used for the allele discrimination assay were examined with 1% and 2% mixtures of K-ras mutant plasmids. The Fam-fluorescence intensity (Y) of each mutant allele displayed responsiveness to increments of mutant percentage (diagram A, Fig. 1). Specimens were tested in triplicate to analyze the accuracy of results with intra-run precision. The assay's inter-run precision was validated with five runs of five CRC specimens including two 2A mutants (P18 and P48), one 2T mutant (P30), and two WT mutants (P02 and P04), with the corresponding quality controls (NTC, WT, and 5% mutant plasmids). The Fam (mutant) fluorescence values of each specimen among five runs (diagram B, Fig. 1) had small standard deviations (SDs), thus demonstrating acceptable reproducibility.

Figure 1. Sensitivity (A) and reproducibility (B) of the allele discrimination assay. (A) Responsive changes of Fam-fluorescence intensity (Y axis) to seven individual mixtures of 1% and 2% K-ras mutants at codon 12–13 (1A, 1T, 1C, 2A, 2T, 2C, and 5A on the X axis), with the allele discrimination. (B) Inter-run precision of the individual alleles among five runs of five CRC specimens (P02, P04, P18, P30, and P48 in solid bars) and the corresponding quality controls (in open bars) of no-template control (NTC), wild-type (WT), and 5% mutant allele-specific QC mutants. The wild-types, P02 and P04, had Fam-fluorescence intensities (Y) for all seven alleles below those of the QC mutants, whereas the respective 2A, 2T, and 2A mutants—P18, P30, and P48—had allele intensities (Y) above those of the corresponding QC mutants. Numbers displayed in parenthesis are SD.

In primary tests of 21 cancer cell lines using both allele discrimination and DNA sequencing, six were detected as K-ras codon 12 mutants (A549 and H358 at base 1; and SW480, SW 620, RPMI-8226, and H2444 at base 2), eight as codon 13 mutants (H1355, H1734, and H1755 at base 1; and DLD-1, HCT-8, HCT-116, Lovo, and SW48 at base 2), and seven as K-ras WTs. RFLP detected Lovo as a codon 12 mutant. However, this was verified as a codon 13 mutant using both allele discrimination and DNA sequencing yielding a 6.7% false positive rate among the 15 mutant cell lines; this incorrect mutant might be caused by incomplete digestion that worsened with the formalin-fixed paraffin-embedded tissues (n = 62).

Among the 62 CR tumors analyzed by RFLP, 25 were identified as K-ras mutants at codon 12 or 13. Allele discrimination and sequencing verified that 8 were false positive mutants (8/25 = 32%, Table IA), 16 were correct, and 1 had a different mutation (1/17 = 6% wrong mutant). Due to the poor specificity of PCR-RFLP for the detection of mutations in the CRC specimens (82%), RFLP was replaced with dHPLC as a third method to verify discrepancies between allele discrimination and sequencing. There was a 12.4% failure rate of sequencing during handling of the formalin-fixed tissue specimens. However, there were no problems with allele discrimination. If a paraffin section of CR tumor failed sequencing the first time, more tissue sections or DNA extracts were resubmitted to improve the results.

Table IA. RFLP in Methods Comparison of 62 CRC Specimens

	RFLP
For RFLP, sensitivity = 100% (17/17) and specificity = 82% (37/45). The RFLP method showed false positive MT = 32% (8/25) and wrong MT = 6% (1/17). Results of sequencing and allele discrimination were used as standards to confirm the RFLP results.
Sequencing/allele discrimination	MT (mutant)	WT (wild-type)
Confirmed MT	17 (1: c13 MT detected as c12 MT)	0
Confirmed WT	8	37

Discrepant results between the two methods were resolved by retesting the one that differed from the other two and ruling out incidental errors. Allele discrimination as well as dHPLC detected one mutant (5A) and three WTs that were sequenced as a WT and a 2T and 2A5A and 1C5A double mutations. Other discrepancies—including one 2C2A double mutation, and one 2T and two 2A single mutations—were detected by allele discrimination and dHPLC but sequenced as 2C, 2T5A, 2A5A, and 1A2A, respectively. The detected 2C2A double mutation (1%), sequenced as 2C only, was found in a well-differentiated female tumor with mucinous histotype (P138 in diagram A, Fig. 2). The other 1A2A double mutation (1%) in a female lung metastasized from a well-differentiated CR tumor was identified by both allele discrimination and sequencing (P192 in diagram B, Fig. 2). Therefore, sensitivity (100%) and specificity (100%) of allele discrimination were slightly better than sequencing (98.8% and 97.5%, respectively; Table IB) with verification of RFLP or dHPLC. The genotype frequencies of K-ras mutations detected among 83 of 204 CRC specimens (40.7%) were 20.6% of G12D (GAT), 7.4% of 12V (GTT), 7.4% of 13D (GAC), and 5.3% of four other mutations (12C, 12R, 12A, and 12S) combined (Table II).

Figure 2. The K-ras double mutants, 2C2A and 1A2A, detected by allele discrimination assay in a mucinous tumor (A, P138) and a metastasized lung (B, P192). (A) In the allele discrimination results of one 2C (upper left) and two 2A alleles (upper middle and right), patient's 2C allele (green-circled dot) is above the 5% mutant QC (blue-circled dot), and patient's 2A allele (green-circled dot) is above both 2% and 1% mutant QC (blue-circled dots), and the wild-types (red dot) on the bottom. Demography of DNA sequencing (lower left) displays 2C allele only. The corresponding hematoxylin and eosin (H&E) stains (40 × and 400 × magnitude in lower middle and right) demonstrate the morphology of a mucinous histotype (arrows). (B) Demography of DNA sequencing (upper left) displays the 1A2A alleles. The patient's 1A and 2A alleles both (green-circled dot in upper middle and right) are far above the individual 1% mutant QC (blue-circled dots) and the wild-types (red dots). Corresponding H&E stains demonstrate characteristics of the lung (arrows), metastasized by colorectal tumor.

Table IB. Comparison of Allele Discrimination and DNA Sequencing

	Sequencing MT	Sequencing WT
a dHPLC, the third method, was used to confirm the genotypes that were detected by allele discrimination.
Allele discrimination MT	82	1 (5A sequenced as wild-type)a
Allele discrimination WT	3a	118
	Sequencing	Allele discrimination
Sensitivity	82/83 = 98.8%	83/83 = 100%
Specificity	118/121 = 97.5%	121/121 = 100%

Table II. Frequencies and Variations of K-ras Genotypes of the 204 Patients with CRC between Using Allele Discrimination and Sequencing

Type	Amino acid	Allele discrimination	Sequencing	n	%
Two sequenced double mutations, 2A5A and 1C5A (underlined), were detected as wild-types by allele discrimination with dHPLC confirmation. Other discrepancies were three sequenced double mutations—2A5A, 2T5A, and 1A2A—which were detected as single mutations 2A, 2T, and 2A, respectively, and one sequenced single mutation, 2C, which was detected as 2C2A in a mucinous histotype. n, final counts of the same genotypes from two of the three methods with deduction of two 2As from double mutations. %, the percentage of genotype counts in a total count of 204. a Two 2As of double mutations were deducted to avoid counting twice. b A common double mutation, 1A2A, simultaneously sequenced and detected by the allele discrimination in a metastasized lung.
2G → A	G12D	44 (−2a)	44	42	20.6
2G → T	G12V	15	16	15	7.4
5G → A	G13D	15	18	15	7.4
1G → T	G12C	3	3	3	1.5
1G → C	G12R	2	3	2	1.0
2G → C	G12A	3	3	3	1.5
1G → A	G12S	3	4	3	1.5
Double mutations		1A2Ab, 2C2A	2A5A, 1C5A, 2A5A, 2T5A, and two 1A2A (oneb)	2	1.0
Sum of mutants		85 (−2a)	91 (−6a)	83	40.7
Wild-type		121 (59.3%)	119 (58.3%)

Prevalence of K-ras Mutations in Patients with CRC

Prevalence of the K-ras mutation and genotypes was assessed by sex, age, and tumor histology. A higher proportion of females among the K-ras mutants [60.2% (50/83) in mutants vs. 42.2% (51/121) in WTs, P = 0.01 (Table III)] was the result of a higher proportion of females among the tumors with transversions (P = 0.02) and transitions (P = 0.05) concentrated at the second base of codon 12 (P = 0.005). Very few female-only GCT mutations (1.5%; Table II) were found, whereas males younger than 65 years of age were found with only one transversion and no GTT mutations (Table III). The female-to-male ratios of GTT and transversions occurrences remained 2:1 or greater across differentiation statuses. However, there was no significant difference in the prevalence of K-ras genotypes between age groups (<65 vs. ≥65 years old) by gender (Table III) or by differentiation status (data not shown). The overall higher prevalence of K-ras mutations among females is mainly due to a higher proportion of females in the K-ras mutants among the well- to moderately differentiated tumors (n = 165, 80.9% of the patients with CRC): 64.2% (43/67) females versus 35.8% (24/67) males.

Table III. Summary of K-ras Mutation Genotypes in 204 Patients with CRC

	Total	Wild-type	Mutants	Specific types of mutants: n (%)
				C12	C13	Transition	Transversion	2G > T	2G > A
a P-Values were derived by comparing the distribution of sex, age, or histology between K-ras mutations and wild type using χ^2 tests or Fisher's exact tests when 25% of the cells have expected counts of less than five. Significant P-values (P < 0.05) are bolded. b Interaction P-values were calculated using multivariable logistic regression. Significant P-values (P < 0.05) are bolded. c Interaction P-values were not calculable due to a zero count in one of the cells.
Sex
Female	101 (49.5)	51 (42.2)	50 (60.2)	43 (63.2)	7 (46.7)	35 (57.4)	16 (69.6)	10 (66.7)	26 (59.1)
Male	103 (50.5)	70 (57.8)	33 (39.8)	25 (36.8)	8 (53.3)	26 (42.6)	7 (30.4)	5 (33.3)	18 (40.9)
Pa		Referent	0.01	0.005	0.74	0.05	0.02	0.07	0.05
Age
<65	98 (48.0)	61 (50.4)	37 (44.6)	33 (48.5)	4 (26.7)	27 (44.3)	10 (43.5)	6 (40.0)	22 (50.0)
≥65	106 (52.0)	60 (49.6)	46 (55.4)	35 (51.5)	11 (73.3)	34 (55.7)	13 (56.5)	9 (60.0)	22 (50.0)
Pa		Referent	0.41	0.80	0.08	0.43	0.54	0.45	0.96
Histology
Poor	39 (19.1)	23 (19.0)	16 (19.3)	13 (19.1)	3 (20.0)	12 (19.7)	4 (17.4)	3 (20.0)	9 (20.5)
Well/moderate	165 (80.9)	98 (81.0)	67 (80.7)	55 (80.9)	12 (80.0)	49 (80.3)	19 (82.6)	12 (80.0)	35 (79.5)
Pa		Referent	0.96	0.99	1.00	0.91	1.00	0.93	0.84
Age–sex
<65
Female	50 (51.0)	25 (41.0)	25 (67.6)	23 (69.7)	2 (50.0)	16 (59.3)	9 (90.0)	6 (100.0)	13 (59.1)
Male	48 (49.0)	36 (59.0)	12 (32.4)	10 (30.3)	2 (50.0)	11 (49.7)	1 (10.0)	0 (0.0)	9 (40.9)
≥65
Female	51 (48.1)	26 (43.3)	25 (54.4)	20 (57.1)	5 (45.5)	19 (55.9)	7 (53.9)	4 (44.4)	13 (59.1)
Male	55 (51.9)	34 (56.7)	21 (46.7)	15 (42.9)	6 (54.5)	15 (44.1)	6 (46.1)	5 (55.6)	9 (40.9)
Interaction Pb		Referent	0.27	0.31	0.82	0.71	0.09	—c	0.89
Histology–sex
Poor	41% = 16/39 mutation rate
Female	20 (51.3)	13 (56.5)	7 (43.8)	5 (38.5)	2 (66.7)	4 (33.3)	3 (75.0)	2 (66.7)	2 (22.2)
Male	19 (48.7)	10 (43.5)	9 (56.2)	8 (61.5)	1 (33.3)	8 (66.7)	1 (25.0)	1 (33.3)	7 (77.8)
Well/moderate	40.6% = 67/165 mutation rate
Female	81 (49.1)	38 (38.8)	43 (64.2)	38 (69.1)	5 (41.7)	31 (63.3)	13 (68.4)	8 (66.7)	24 (68.6)
Male	84 (50.9)	60 (61.2)	24 (35.8)	17 (30.9)	7 (58.3)	18 (36.7)	6 (31.6)	4 (33.3)	11 (31.4)
Interaction Pb		Referent	0.03	0.01	0.83	0.02	0.77	0.62	0.006

Despite the similar K-ras mutation rates between poorly differentiated and well- to moderately differentiated tumors [41% (16/39) vs. 40.6% (67/165) respectively; Table III], a higher proportion of males (56.2%, 9/19) was observed for the K-ras mutants of poorly differentiated tumors (n = 39). This is in contrast to the lower proportion of males (35.8%, 24/84) in the K-ras mutants of well- to moderately differentiated tumors (interaction P = 0.03). This observation was especially true for codon 12, transition, and GAT mutations (interaction P-values = 0.01, 0.02, and 0.006, respectively). Among the well- to moderately differentiated tumors, mutant genotypes frequencies were 30% GAT, 10% GTT, and 6% GAC for females, and 13% GAT, 5% GTT, and 8% GAC for males (Table IV), with GAT being the most common K-ras mutation in both genders. Among the males with poorly differentiated tumors, GAT was also the most common K-ras mutation, with 37% GAT, 5% GTT, and 5% GAC; however, the three mutations (GAT, GAC, and GTT) among the females with poorly differentiated tumors were equal in prevalence (10%).

Table IV. Comparison of K-ras Genotypes between Genders, Differentiation Statuses and for Some Pathological Changes

	2G > A (%)	2G > T (%)	5G > A (%)	Mutants (%)	Wild-type	Total N
Underlines indicate an uncommon prevalence of the three major genotypes. a One female patient had both mucinous histotype and perineural invasion.
Poor
Female	2 (10)	2 (10)	2 (10)	7 (35)	13 (65)	20
Male	7 (37)	1 (5)	1 (5)	9 (47)	10 (53)	19
Well/moderate
Female	24 (30)	8 (10)	5 (6)	43 (53)	38 (47)	81
Male	11 (13)	4 (5)	7 (8)	24 (29)	60 (71)	84
Pathological changes (total)	8 (32)	3 (12)	3 (12)	14 (56)	11 (44)	25
Mucinous/signet ring				6 (46)		13
Female	3 (2C2A included)			3	4	7
Male		2	1 (poor)	3	3	6
Metastatic/invasive				5 (63)		8
Female	1 (1A2A)	0	0	1	2	3
Male	2 (both poor)	1	1	4	1a	5
Calcification/necrosis				3 (60)		5
Female	2 (1 poor)	0	0	2	1	3
Male	0	0	1	1	1	2
Females	6 (50)	0	0	6 (50)	6 (50)	12
Males	2 (15)	3 (23)	3 (23)	8 (61.5)	5 (38.5)	13

Tumors with pathological changes (n = 25, Table IV)—including mucinous histotype (n = 13), metastatic or invasive (n = 8), and necrotic or calcified tumors (n = 5)—showed a higher K-ras prevalence, 56% (14/25), than those without changes: 38.5% (69/179). K-ras mutation rates of 46% (n = 6, one 2C2A and two 2A in females, and two 2T and one 5A in males), 62.5% (n = 5, one 1A2A in females, and two 2A, one 2T, and 5A in males), and 60% (n = 3, two 2A in females and one 5A in males), respectively, were observed in each category. In contrast to no occurrences of GTT or GAC, and 50% GAT (with double mutants, 33% without) among females, the occurrence of 23% GTT and GAC, and 15% GAT among males, suggests a gender difference for the K-ras genotypes in association with pathological changes of CRC.

DISCUSSION

In contrast to RFLP and DNA sequencing, which amplify DNA regions twice to derive sequence information, the allele discrimination assay reduces genotyping errors by synchronizing genotype detection with amplification. The improved detection rate and sensitivity during handling of formalin-fixed tissues is conferred by the shorter amplicon (<100-bp) used in the qPCR for the allele discrimination assay [Do et al.,2008]. We found that the RFLP assay, which depends on the complete digestion of DNA using restriction enzymes as well as PCR manipulations, is less specific (Table IA) and more time consuming. In contrast, DNA sequencing on formalin-fixed tissues, in spite of a failure rate of 12.4%, generated a high level of sensitivity (98.8%, Table IB) and specificity (97.5%) that is sufficient for clinical application. However, the potential introduction of errors in processing of DNA sequencing may cause difficulty in troubleshooting and subsequently delay retesting. Allele discrimination was sensitive to even 1–2% of mutant plasmid DNA (diagram A, Fig. 1), and reproducibility displayed consistent inter-run precision with five representative specimens in five genotyping runs (diagram B, Fig. 1). In every assay, quality controls including individual positive mutants and WT controls and NTC were monitored to help distinguishing the analytical errors (assay itself) from other kinds (sample and preparation).

The 40.7% K-ras mutation rate (83/204) observed in this study is within the average (20−60%) reported by studies conducted in similar geographic areas. However, the mutational genotypes were different from the results of other studies using Sanger sequencing; these differences might relate to the use of different Taq polymerases (e.g., Promega [Lee et al.1996] or Biotechnological & Medical Laboratories [Wang et al.2003, 2007]). Although DNA sequencing produces the exact identity of mutation(s), the process contains two separate PCR reactions: one amplifies the gene product from the gDNA extract, the other incorporates the fluorescent-dye-labeled nucleotides in sequencing. Thus, the chance of introducing and magnifying a biased mutation is more likely to occur using the conventional Sanger DNA sequencing. For example, in our study, more double mutations found using sequencing than allele discrimination were verified to be errors, thereby sensitivity and specificity of sequencing was lower. In earlier experiments, we also observed biased mutations that differed with the Taq polymerases used (Viogene, Qiagene, Invitrogen, and KB HotStart) during the processing of some older specimens (data not shown). Bio-degeneration or poor tissue processing results in DNA fragmentation [Gallegos Ruiz et al.2007], which can worsen the problem. Newer DNA sequencing technologies, such as pyrosequencing and next-generation sequencing, eliminate the additional amplification step and greatly improve accuracy as well as efficiency.

Double mutations at codon 12–13, which reflect clonal expansion and intratumoral genetic heterogeneity that can be distinguished through microdissection [Al-Mulla et al.,1998; Baisse et al.,2001; Losi et al.,2005], are often reported with sequencing and SSCP [Bazan et al.,2002; Span et al.,1996], but not with qPCR applications. In a metastasized lung and a mucinous histotype, the 2A (GAT) mutants of double mutations 1A2A and 2C2A were minor (Fig. 2), raising the question of how each mutational clone occurs and how one overtakes the other. An ongoing effort characterizing the prevalence of K-ras genotypes in affiliation with pathological changes—such as early [Losi et al.,2005] or advanced CRC progression [Baisse et al.,2001], and tumor differentiation—should not only have prognostic value for EGFR-targeted therapy and related combinational therapies but should also help facilitate personalized treatments and preventive strategies.

The higher proportion of females having K-ras mutants among the well- to moderately differentiated tumors (female 64.2% vs. male 35.8%; Table III) and the higher proportion of males with K-ras mutants among the poorly differentiated tumors (male 56.2% vs. female 43.8%; interaction P = 0.03) were concordant with the prevalence of GAT (12D; P = 0.006) but not with the prevalence of GAC (13D). The GAC mutation is reported to be of predictive and prognostic value for local recurrence and short-term mortality [Bazan et al.,2002; Chang et al.,2009; Pajkos et al.,2000; Samowitz et al.,2000]. Among the poorly differentiated tumors, the K-ras mutants in males were predominantly GAT (37%, Table IV [Samowitz et al.,2000]); in contrast, among the poorly differentiated tumors of females, GAT, GTT, and GAC mutations were equal in prevalence (10% each). Regardless of gender, more K-ras mutations were also reported with poorly differentiated colorectal tumors [Abubaker et al.,2009; Lee et al.,1996].

The higher prevalence of transversion and codon 12 mutations (P = 0.02 and 0.005, Table III) among females resulted from the presence of quasi female-only GCT [Breivik et al.,1994; Breivik et al.,1994] and GTT (absent in males at age <65) mutations. Studies on relapse- and death-related K-ras genotypes consistently associate GTT with tumor aggression [Al-Mulla et al.,1998; Andreyev et al.,1998; Span et al.,1996]. In this study, older males tended to have a higher K-ras mutation rate and more mutational genotypes; however, the age imbalance between genders (observed by the median ages in groups) offset the statistical significance.

Prevalence of K-ras mutations was also higher among the tumors with pathological changes, including mucinous histotype (signet ring), metastasis, and calcification and necrosis, 56% (14/25, Table IV), than those without the changes, 38.5% (69/179). Interestingly, males in this category were found with a lower occurrence of GAT (15%) than GTT and GAC (both 23%), whereas both prevalences of K-ras mutations were similar. Mucinous histotype was previously associated more with the codon 12 mutations [Bazan et al.,2002; Neumann et al.,2009].

The K-ras pathway plays a role in the etiology of many cancers by initiating processes that are capable of inducing malignant morphological phenotypes [Ohnishi et al.,1997]. In population studies of K-ras genotypes in patients with CRC, the morphological and genetic complexities of CRC etiology result in inconclusive observations for their role in cancer progression or metastatic potential. Reports on gender differences in K-ras mutations in CRC [Breivik et al.,1994; Brink et al.,2003] have postulated that the sex hormone that affects secretion of bile acid and subsequently, the microenvironment of proximal colon, prevalence of constipation, and fecal compositions, may affect occurrences of K-ras genotypes. However, obtaining direct evidence to support these theories remains a challenge. We found that GAT was not a predominant K-ras mutational genotype in either poorly differentiated tumors among females or the tumors with pathological changes among males, in contrast to being predominant in other groups of this study and in most prior reports. Regardless of possible causes or intrinsic factors, such as dietary intake, family history, or other CRC-related genetic alterations, such differences in the occurrences of K-ras mutations between females and males suggest that their susceptibility to K-ras mutations might differ with the CRC pathological phenotypes. More work on larger populations is needed to clarify this.

CONCLUSIONS

The allele discrimination assay is a rapid and reliable method for screening K-ras mutations at codon 12 and 13 in formalin-fixed CRC tissue. We found this method to be superior to sequencing, exhibiting a higher detection rate and fewer false mutations. Using this assay, we demonstrate that females harbor more codon 12 and transversion (GTT and GCT) CRC K-ras mutants than males. The overall prevalence of K-ras mutations was higher in females than in males, as well as among the well- to moderately differentiated tumors. However, the prevalence of K-ras mutations was higher in males than in females among poorly differentiated tumors and in tumors with pathological changes. We also found that neither the poorly differentiated tumors among females nor the tumors with pathological changes among males displayed GAT as the most common mutational genotype as has been suggested in prior studies. Therefore, our findings suggest that K-ras mutations may occur differently in females and males and in association with the pathological phenotypes of CRC than previously reported.

Acknowledgements

Special thanks to Dr. Jia-Yi Wang for essential input and guidance on this manuscript and to Dr. Hsu-Wei Hung who provided us with three known mutant specimens as quality controls.