Detection of KRAS and BRAF Mutations in Colorectal Carcinoma

Abstract

KRAS and BRAF mutations predict the resistance of colorectal carcinomas to therapy targeted to the epidermal growth factor receptor, but their detection can be challenging because of high testing volume, frequently low tumor content, and the spectrum of rarer mutations in these genes. To address these issues, we evaluated a locked nucleic acid (LNA)-PCR sequencing assay to detect low levels of mutant DNA, and we also evaluated a mass spectrometry genotyping assay (Sequenom, San Diego, CA) that is suitable for broad mutation screening. Clinical cases (n = 308) previously tested for KRAS and BRAF by standard sequencing were retested by LNA-PCR sequencing incorporating an LNA oligonucleotide to suppress amplification of nonmutant DNA, and by a Sequenom assay panel targeting common mutations in both genes. Standard sequencing detected 121 KRAS (39%) and 10 BRAF mutations; retesting with the LNA-based method and the Sequenom assay detected 19 (140/308, 45%) and 6 (127/308, 41%) additional KRAS mutants, respectively. One additional BRAF mutant was detected by the Sequenom assay. The analytical sensitivities were 0.3% for both KRAS and BRAF by LNA-PCR and from 1% to 10% for the Sequenom assays, depending on the specific mutation. Given these results, standard sequencing is suboptimal for mutation detection in metastatic and treated lesions even with predissection for tumor enrichment. High-sensitivity LNA-PCR sequencing detects significantly more mutations, whereas the Sequenom platform shows intermediate sensitivity but offers significant advantages for broader mutation screening.

Colorectal carcinoma (CRC) is one of the leading causes of cancer-related death worldwide, with a 5-year survival rate of less than 10% for patients with metastatic disease.¹ Advances in the understanding of the biology of CRC have led to the introduction of molecular-targeted agents.^2–3 In practice, however, the use of these agents has uncovered a significant variability in clinical response, which highlights the molecular heterogeneity of colorectal tumors and underscores the need to develop reliable biomarkers to predict therapeutic efficacy.

The epidermal growth factor receptor (EGFR)-signaling pathway is commonly up-regulated in colorectal cancer.^1,4–6 Signals from the activated cell surface receptor activate multiple downstream molecules that are involved in the processes leading to tumor cell proliferation and survival. Given its central role in tumor development, the EGFR pathway has been intensively pursued as an ideal substrate for the treatment of colorectal cancer. A notable success resulting from these efforts is the development of the EGFR-targeting monoclonal antibodies cetuximab and panitumumab. Unfortunately, their broad use has shown a significant clinical benefit in only 10% to 20% of patients.^7–9 The remaining majority never demonstrate a response but may experience undesirable side effects and assume the economic burden of ineffective therapy.^10–12

The search for molecular markers that could adequately predict the response to EGFR inhibitors initially focused on the most obvious target, the EGFR itself. Receptor expression by immunohistochemistry and increased copy number by fluorescence in situ hybridization or other methods have been explored in multiple studies which, taken together, fail to establish a solid correlation with therapy effect.^13–20 Similarly, the search for EGFR mutations in colorectal cancer has been disappointing.^21–22

Two of the many signaling molecules downstream of EGFR are the small G protein RAS and the protein kinase RAF. Mutations in KRAS and BRAF occur commonly in colorectal carcinoma and, as initially postulated, their presence is associated with primary resistance to EGFR inhibitors.^23–37 KRAS mutations, reported in 30% to 40% of patients with CRC, are the most common and currently the best-documented predictor of resistance to EGFR targeting agents. A substantial body of evidence collected from multiple single-arm studies and large randomized controlled clinical trials shows a consistent correlation between mutant KRAS and the lack of response to anti-EGFR monoclonal antibodies.^{23–27,29–30,33,35–38} These studies have served as the basis for the most recently published guidelines by the American Society of Clinical Oncology, which recommend mutation testing in all patients being considered for this treatment modality.³⁹ The vast majority of KRAS mutations occur in codons 12 and 13 of exon 2; the remainder are located predominantly in codons 61 and 146. Recently, a similar resistance association has been reported in patients with BRAF mutant CRC.^28,32,34 Mutations in KRAS and BRAF are mutually exclusive,²⁸ supporting sequential testing as a rational algorithm.

Based on these studies, and further stimulated by the American Society of Clinical Oncology 2008 guidelines, there is currently intense interest in rapid, reliable, and accurate methods of KRAS and BRAF mutation screening. A wide spectrum of technical approaches can be applied to the routine detection of point mutations in KRAS and BRAF,^40–44 each with inherent advantages and disadvantages, depending on individual detection limits, labor requirements, turnaround time, expense of reagents and equipment, and potential for automation and multiplexing. Sanger sequencing of PCR-amplified DNA is the classic and most widely used method of mutation detection.^22,45 However, it has several drawbacks in this particular setting, primarily because of suboptimal sensitivity and its relatively labor-intensive, low-throughput nature. Although the problem of sensitivity can be overcome by microdissection in most cases, it remains a major challenge in the context of recurrent CRC, when testing is commonly based on post-treatment specimens in which scant tumor cells are intimately admixed with abundant non-neoplastic cells. Not infrequently, these tests yield false-negative results despite attempts at predissecting areas richer in tumor. Thus, many institutions and commercial laboratories reject samples in which the tumor cell percentage does not exceed a preset standard. For those institutions that continue routinely to predissect, this process can be time consuming. Furthermore, as the testing volume increases and additional specific markers of response and targetable pathways continue to be discovered, the low throughput inherent in standard sequencing is becoming a major issue. We need optimal testing strategies that are reliable, appropriately sensitive, and can adequately address the testing volume expected from the high prevalence of CRC.

To address these issues, we developed and evaluated two mutation genotyping assays: a modified PCR-sequencing assay using a locked nucleic acid (LNA) probe for mutant enrichment and a mass spectrometry-based genotyping assay (Sequenom, San Diego, CA) suitable for large-scale mutation screening. Herein we describe our experience with these two assays and compare them with traditional Sanger sequencing for routine mutation testing of CRC specimens. We demonstrate that both techniques have a higher analytical sensitivity for the detection of mutant DNA and, at different levels, both techniques are less labor intensive.

Materials and Methods

Tumor Samples

Metastatic CRC samples (n = 334) received for routine clinical KRAS and BRAF testing at Memorial Sloan-Kettering Cancer Center between January and May 2009 were selected for the study. H&E-stained sections of formalin-fixed paraffin-embedded tissue were reviewed for each sample to identify and circle the areas of highest tumor density, ensuring at least 50% tumor content. Macrodissection was performed using the tip of a blade to scrape off the selected tumor areas on corresponding unstained sections. Genomic DNA was extracted using the DNeasy Tissue kit (Qiagen, Valencia, CA), following the manufacturer's standard protocol.

KRAS and BRAF Mutation Analysis by Standard Sanger Sequencing

The entire coding region of exon 2 of KRAS was amplified using the following forward and reverse intronic primers (custom oligos, Proligo, Boulder, CO): KRAS-F, 5′-GTGTGACATGTTCTAATATAGTCA-3′; and KRAS-R, 5′-CTGTATCAAAGAATGGTCCTGCAC-3′. All samples were tested in duplicate. Each PCR reaction was performed in a 50-μL volume mixture containing 100 ng of genomic DNA; forward and reverse primers (20 pmol each); 50 μmol/L each of dATP, dCTP, dGTP; 400 μmol/L of dUTP; 1.5 mmol/L MgCl₂; 1X Qiagen PCR buffer containing 1.5 mmol/L MgCl₂ and 2.5 units of HotStarTaq DNA polymerase (Qiagen). AmpErase uracil N-glycosylase (Applied Biosystems, Foster City, CA) (0.5 unit) was added to each reaction mixture for carryover prevention, and they were incubated at room temperature for 10 minutes. The PCR amplification was carried out under the following conditions: 1 cycle at 95°C (15 minutes); 40 cycles at 94°C (30 seconds), at 56°C (30 seconds), and at 72°C (30 seconds); and a final extension step at 72°C (7 minutes). Amplified products were purified using Spin Columns (Qiagen) and sequenced in both directions using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems), according to the manufacturer's protocol, on an ABI3730 running ABI Prism DNA Sequence Analysis Software (Applied Biosystems).

BRAF mutation analysis was carried out following the same procedure described for KRAS, using the forward and reverse primers (custom oligos; Proligo, Boulder, CO); BRAF/15F, 5′-TCATAATGCTTGCTCTGATAGG-3′; and BRAF/15R, 5′-GGCCAAAAATTTAATCAGTGG-3′ to amplify the entire coding region of exon 15 of BRAF.

KRAS and BRAF Mutation Analysis Using Sequenom Genotyping Assays

For these assays we used the MassARRAY system (Sequenom), which is based on matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). This system was originally designed to analyze single-nucleotide polymorphisms (SNPs) in amplified DNA fragments,⁴⁸ but it is also well suited to the detection of known somatic point mutations. Genotyping by this method relies on the principle that mutant and wild-type alleles for a given point mutation produce single-allele base extension reaction products of a mass that is specific to the sequence of the product. Mutation calls are based on the mass differences between the wild-type product and the mutant products as resolved by MALDI-TOF MS.⁴⁹ Amplification and extension primers were designed using Sequenom Assay Designer v3.1 software to target the most common KRAS and BRAF mutations in exon 2 and exon 15, respectively. Amplification primers were designed with a 10mer tag sequence to increase their mass so that they fall outside the range of detection of the MALDI-TOF MS. The sequences of primers are shown in Supplemental Table S1 (http://jmd.amjpathol.org).

The Sequenom assays for KRAS exon 2 consisted of four separate, single-direction assays to assess the genotype at the four positions within codons 12 and 13 containing possible mutations. For BRAF V600 mutations, the assay comprised two separate reactions assessing the presence of the mutation at the same position in both the forward and the reverse directions or strands. Each assay was performed in duplicate.

The initial PCR amplification was performed in a 5-μL reaction mixture containing 10 to 20 ng of DNA/1.25x buffer; 1.625 mmol/L MgCl2; 500 μmol/L deoxynucleotide triphosphate (dNTP); 100 nmol/L from each primer; and 0.5 U HotStar TaqDNA Polymerase (Qiagen) under the following conditions: 95°C (15 minutes); 95°C (20 seconds): 56°C (30 seconds); and 72°C (60 seconds) for 45 cycles, and a final extension phase at 72°C (3 minutes). Remaining unincorporated dNTPs were dephosphorylated by adding 2 μL of a shrimp alkaline phosphatase cocktail containing 1.53 μL of water, 0.17 μL of reaction buffer (Sequenom), and 0.3 μL of shrimp alkaline phosphatase (Sequenom). The unincorporated dNTPs were then placed in a thermal cycler under the following conditions: 37°C for 40 minutes, 85°C for 5 minutes, and then held at 4°C indefinitely. A single-base extension reaction was then performed in a 2-μL TypePLEX reaction mix (Sequenom) consisting of 0.72 μL water; 0.20 μL TypePLEX 10x buffer (Sequenom); 0.10 μL TypePLEX terminator mix (Sequenom); 0.94 μL extension primer mixture; and 0.04 μL TypePLEX enzyme (Sequenom). Thermal cycling was performed under the following conditions: 94°C for 30 seconds followed by 40 cycles of (94°C for 5 seconds, 5 cycles of 52°C for 5 seconds and 80°C for 5 seconds), then at 72°C for 3 minutes, and was finally held at 4°C indefinitely. The reaction mixture was then desalted by adding 16 μL of water and 6 mg of cationic resin mixture, SpectroCLEAN (Sequenom), and placed in a rotating shaker for 30 minutes. Completed genotyping reactions were spotted in nanoliter volumes onto a matrix-arrayed silicon SpectroCHIP with 384 elements using the MassARRAY Nanodispenser (Sequenom). SpectroCHIPs were analyzed using the Autoflex MALDI-TOF MS (Bruker AXS, Madison, WI), and the spectra were processed using SpectroACQUIRE software (Sequenom) in real time. Genotype calls are automatically generated using complex mathematical algorithms according to the peak heights, noise-to-peak-height ratio, area under the curve, and anchoring of the peaks. Results are then linked to plate information created in MassARRAY Typer 4.0 software (Sequenom). Individual calls are manually reviewed and finalized.

KRAS and BRAF Mutation Analysis Using the LNA-PCR Sequencing Assay

To increase the sensitivity of the standard DNA sequencing, we modified the standard PCR sequencing assay by adding LNA probes to favor mutant DNA amplification during PCR. LNAs are nucleic acid analogs that contain a 2′O to 4′C methylene bridge that locks the ribose group into a C3′-endo conformation. The introduction of LNA monomers into oligonucleotides increases the melting temperature of DNA heteroduplexes between 1°C and 8°C per modification.^46–47 These modified oligonucleotides bind to complementary sequences with an unprecedented high affinity that depends on the final compositions of the LNAs and their overall length. We designed two LNA-containing probes consisting of 10 consecutive LNA monomers complementary to the wild-type sequences of KRAS (codons 12 to 13) and BRAF (codons 598 to 601). The addition of these probes to the PCR amplification step causes them to hybridize avidly to wild-type KRAS and BRAF, effectively suppressing the amplification of nonmutant, wild-type DNA and leading to preferential amplification of the mutant allele. PCR mixes and cycling conditions identical to those in the standard assays were used, except for the addition of 20 pmol of each LNA probe to the PCR mixture. All samples were amplified and sequenced in the forward and reverse directions, followed by confirmatory, duplicate runs of all positive and equivocal results. The sequences of the probes were as follows: KRAS LNA-F: 5′-G+C+T+G+G+T+G+G+C+G/3InvdT/−3′ and BRAF LNA-F: 5′-+G+C+T+A+C+A+G+T+G+Aaatctcgatgg/3InvdT/−3′ where the capital letters followed by the plus (+) sign designate the locked nucleotides. PCR product purification and sequencing were carried out following the same steps used in the standard assays.

Comparison of Analytical Sensitivities

To assess the analytical sensitivity of standard sequencing and the LNA-PCR sequencing, cell-line DNA mixing studies were carried out using a mixture of mutant cell-line DNA with wild-type DNA in dilutions of 50%, 25%, 12.5%, 6%, 3%, 1.5%, 0.75%, 0.38%, and 0.19% of mutant DNA. Mutant DNA was extracted from the KRAS-mutant H1734 (G13C-37G>T) and the BRAF-mutant SK-MEL (V600E-1799T>A) cell lines. The IVS-0000 (Invivoscribe, San Diego, CA) normal control was used as the source of wild-type DNA. All cell-line DNA mixtures were amplified, purified, and sequenced following the standard and LNA-PCR protocols described above. The analytical sensitivity of standard sequencing was assessed in a fashion similar to that described for LNA-PCR sequencing, but dilutions of mutant to wild-type DNA were carried out at 50%, 25%, 12.5%, 6%, and 3% for both KRAS and BRAF.

To assess the analytical sensitivity of the Sequenom assays, eight serial dilutions of mutant DNA were used (50%, 20%, 10%, 5%, 2.5%, 1%, 0.5%, and 0.25%). Mutant DNA was derived from either cell lines (CL) or tumor (T) samples as follows: KRAS (CL) H358 (G12C-34G>T), KRAS (CL) H2009 (G12A-35G>C), KRAS (T) M3205 (G13C-37G>T), KRAS (CL) HCT-15 (G13D-38G>A), and the BRAF (T) 369t (V600E–1799T>A). All CL- and T-derived DNA mixtures were amplified, purified, and genotyped following the standard protocols for the Sequenom assays described above.

Results

Standard Sequencing

Of 334 samples of CRC received for routine mutation testing of KRAS (exon 2) and BRAF (exon 15) by direct sequencing, 151 (45%) came from primaries and 183 (55%) from metastatic tumors. Of the primary colonic tumors, 13 were postneoadjuvant chemotherapy and/or radiation and had extensive therapy effect (Table 1). Most metastatic lesions were located in the liver (68%) followed by peritoneum, lung, lymph nodes, and other distant sites. Only two samples (0.6%) failed due to poor DNA quality. Of the successfully tested cases, 36% (121/332) were positive for KRAS mutations; the distribution of mutations is shown in Table 2. Of the 211 cases negative for KRAS mutations, 10 cases (5%) were positive for BRAF exon 15 mutations, including 8 V600E mutations, one K601E mutation, and one D594G mutation.

Table 1.

Specimen Profile (n = 334)

	Total	% of total
Number of specimens tested	334
Specimen type
Primary	151	45
-Untreated	138	41
-Extensive treatment effect	13	4
Metastases	183	55

Table 2.

Mutation Profile of KRAS and BRAF Based on All Methods

	AA mutation	CDS mutation	# Mutants	% Total mutants†
KRAS mutation profile	G12S	34 G>A	6	4
	G12C	34 G>T	15	10
	G12I	34_35 GG>AT	1	<1
	G12D	35 G>A	44	29
	G12A	35 G>C	15	10
	G12R	34 G>C	1	<1
	G12V	35 G>T	31	20
	G13C	37 G>T	1	<1
	G13D	38 G>A	28	18
	Q61R⁎	182 A<G	3	2
	Q61L⁎	182 A>T	3	2
	Q61H⁎	183 A>T	2	1
	A146T⁎	436 G>A	4	3
BRAF mutation profile	V600E	1799 T>A	9	75
	D594G⁎	1781 A>G	1	8
	K601E	1801 A>G	1	8
	G469A⁎	1406 G>C	1	8

^⁎Mutations exclusively tested by the Sequenom assay. Other mutations were tested by all methods..

^†Percentages do not add up to 100% because of rounding.

KRAS and BRAF Mutation Analysis Using Sequenom-Based Genotyping Assays

Of 334 cases, 331 were available for further testing by Sequenom. Of these, 327 cases were successfully tested, and 4 could not be adequately tested because the DNA concentrations (<10 ng/μL) were below the minimum recommended for the assay. These cases were considered failures (1.2%) and corresponded to three metastases and one primary lesion, all with extensive therapy effect. Conversely, two cases that had previously failed by standard sequencing were successfully genotyped by the Sequenom method.

For the KRAS exon 2 mutation assay, 99% (120/121) of the mutations detected by standard sequencing were confirmed by this assay, and seven additional mutations were identified (five in metastatic lesions, one in a primary tumor sample with extensive therapy effect, and one in a sample that had failed standard sequencing). Only one mutation detected by standard sequencing was not detected by Sequenom (Figure 1). The exon 2 KRAS mutation rate found by this method was therefore 38% (127/331) (Table 3). In addition to the exon 2 mutations, our Sequenom panel had been designed to detect low-frequency mutations in codons 61 and 146, and it detected 12 additional mutations, including four A146T, three Q61R, three Q61L, and two Q61H mutations, for an overall mutation rate of 42% (see Table 2 for summary).

Figure 1.

Venn diagram showing all exon 2 KRAS mutations detected in this study. A total of 142 mutations were detected by a combination of all methods; 121 mutations were detected by the standard sequencing method, of which 120 mutations were detected by all three methods. The LNA-PCR sequencing method detected 20 additional mutations, of which 6 were also detected by the Sequenom assay. One mutation was detected by the Sequenom assay only and it corresponded to one of the two failures generated by traditional sequencing.

Table 3.

Comparison of KRAS Mutation Rate by Type of Testing Method

	Standard sequencing	LNA-PCR sequencing	MALDI-TOF MS (sequenom)
Total cases tested	334	310	331
Total cases successfully tested	332	308	327
Failure rate	0.6%	0.6%	1.2%
Total exon 2 mutations	121 (36%)	141 (45%)⁎	127 (38%)⁎⁎
Additional cases compared to standard sequencing (% of total mutant cases)	NA	20 (14%)	7 (5%)

^⁎P = 0.01;

^⁎⁎P = n.s. (Fisher's exact test for comparison to standard sequencing).

The BRAF V600 Sequenom assay detected nine V600E mutants; eight were concordant with the results of the initial routine testing and one additional case corresponded to the second above-mentioned sample that had failed standard sequencing. The two non-V600E exon 15 mutations detected by standard sequencing, K601E and D594G, were also detected by the Sequenom assay, as were the low-frequency G469A mutation not tested by the other two methods. As the Sequenom assays test for both KRAS and BRAF mutations in all cases, they also confirmed that the two were mutually exclusive in all 151 samples that had one of these mutations.

The performance of the Sequenom assays was compromised at very low DNA concentrations, usually below 20 ng/μL. At these low levels, artifact mutant peaks are generated, making interpretation difficult. For very small samples that may not generate these concentrations, the elution of DNA in lower volumes after extraction may be necessary.

LNA-PCR Sequencing Assay

Of 334 cases, 310 were available for further testing by the LNA method. Of these, 177 were negative for KRAS and BRAF mutations by standard sequencing and were reevaluated using the LNA probe in the PCR step. Retesting of the negative cases detected 20 additional mutations, bringing the mutation prevalence in this group to 45% (141/310) (Figure 1, Table 3). A histopathology review to determine possible reasons for the detection failure by standard sequencing in these 20 cases revealed that all 20 were either metastatic lesions (14/20) or primary tumor samples resected after neoadjuvant chemotherapy or radiation or both (6/20). Of the 14 metastatic lesions, 12 had extensive fibrosis, necrosis, or inflammation and 2 had extensive mucinous components in which only a few malignant cells dissected through the intervening fibrous strands (Figure 2). In the primary tumor samples, a small number of tumor glands were embedded in the muscular wall with associated fibrosis and inflammation present (Figure 3).

Figure 2.

Example of a metastatic lesion yielding a negative result by standard sequencing but a positive result by both alternative assays. A: H&E-stained section of metastatic colorectal carcinoma. Scant tumor cells with a prominent mucinous component dissect through planes of fibroconnective tissue. B: Sequencing by the standard method shows a tracing containing only wild-type DNA. C: Electropherogram of sequencing after LNA-PCR shows a mutant-only peak (arrow) at position 35 consistent with a G12V mutation (35G>T). D: Spectrum of the Sequenom assay also shows a 35G>T mutation. F, forward; R, reverse; an asterisk marks the position where the mutant peak would be expected; +, other SNPs/mutations of the Multiplex Sequenom Assay.

Figure 3.

Example of a lesion yielding a false-negative KRAS result by standard sequencing and by the Sequenom assay. A: H&E-stained section of colon status after neoadjuvant chemotherapy and radiation. Scattered, small foci of residual carcinoma are embedded in the wall, with associated fibrosis and chronic inflammation. Normal colonic mucosa (top right) overlies the tumor. B: Electropherogram (standard sequencing) shows only WT DNA in the region of codons 12 and 13. Note that the reverse sequence shows a very small peak (**) which, by itself, is insufficient for a positive call. C: Sequencing after LNA-PCR shows a large mutant peak (arrow) consistent with a G13D mutation (38G>A). A small WT peak is also present. D: Testing by Sequenom yielded a negative result in this case. F, forward; R, reverse; an asterisk marks the position where the mutant peak would be expected.

To further compare the detection of mutations by each method, we divided 308 specimens into three groups: untreated primary lesions, metastatic lesions, and treated primary lesions with extensive treatment effect (Table 4). In untreated primary lesions, the prevalence of mutations did not vary significantly by method; it was 46% by standard sequencing and the LNA method and 47% by the Sequenom assay. In the metastatic lesions, the mutation detection rate was notably lower for standard sequencing (46% to 36%) and by the Sequenom assay (47% to 39%) but not with the LNA method (46% to 44%), although none of these individually reached statistical significance. However, the most significant discrepancies were seen in the treated primary lesions, where the detection rate dropped to 8% and 17% in standard sequencing and Sequenom, respectively, but remained high in the LNA method, at 58%. Overall, in specimens other than untreated primaries, the LNA method detected significantly more KRAS mutations than did standard sequencing (45% vs. 34%, P = 0.04). The improvement in mutation detection can also be expressed as the number of additional mutants detected over those detected by standard sequencing, using the LNA-PCR sequencing mutation count as the denominator. With the LNA method, 18% (13/74) more mutations were identified in the metastatic lesions and severalfold more mutations were detected in the treated primary lesions. The Sequenom method detected 6% (4/65) more mutants in the metastatic samples and one additional mutant in the treated primary lesions.

Table 4.

Comparison of KRAS exon 2 Mutation Rate by Specimen Type^⁎

	Standard sequencing			LNA-PCR sequencing			MALDI-TOF MS (sequenom)
	Total tested	Failures (rate)	Total positive (% tested)	Total tested	Failures (rate)	Total positive (% tested)	Total tested	Failures (rate)	Total positive (% tested)
Primary lesion (non-treated)	128	1 (0.8%)	59 (46)	128	1 (0.8%)	59 (46)	128	0	60 (47)
Metastases	168	1 (0.6%)	61 (36)	168	1 (0.6%)	74† (44)	168	2 (1.2%)	65 (39)
Primary lesion (treated)	12	0	1 (8)	12	0	7⁎ (58)	12	1 (9%)	2 (17)

^⁎Based on 308 samples and including data from all three methods.

^†P = 0.04 (Fisher's exact test for comparison to standard sequencing for detection of mutations in specimens other than untreated primaries).

Retesting of all KRAS-negative cases for BRAF mutations using the LNA method did not reveal any additional mutants. However, based on the additional KRAS-positive cases in this patient group (141; 45%), the resulting BRAF rate in the KRAS-negative subgroup is calculated to be 6% (10/166).

An important pitfall of LNA probes is that sequencing traces may sometimes show multiple (typically >2) small, poorly reproducible mutant peaks in the region bound by the LNA probe. We find that these peaks are usually generated when a genuine mutant allele is absent or extremely low (below 0.19% for KRAS and 0.75% for BRAF). These artifactual mutant peaks may originate from low-level mutations induced by formalin fixation or from errors in PCR incorporation, which are revealed by the LNA-induced suppression of the wild-type allele (see Supplemental Figure S1 at http://jmd.amjpathol.org). It should also be noted that LNA-PCR also requires the concurrent performance of standard PCR for quality control. In the absence of a mutation, the LNA may completely suppress amplification of wild-type DNA, generating no PCR product, and this effect may be indistinguishable from a PCR failure for any other reason.

Comparison of Analytical Sensitivities for Detection of KRAS Mutations

Standard Sequencing

Based on cell-line mixing studies, the detection sensitivity of Sanger sequencing after standard PCR amplification is 25%. At dilutions below this concentration, the mutant peak is no longer detected (Figure 4).

Figure 4.

KRAS sensitivity study: comparison of electropherograms in sequencing with standard (A) and LNA-PCR (B). Serial dilutions of mutant DNA with normal control DNA. Sequencing after standard PCR shows a mutant peak at position 37 (arrow) (37G>T - G13C). The mutant peak is seen clearly at 25% or higher. Below this level, the mutation is at the level of the background or is not seen (asterisk). With the introduction of the LNA probe, only the mutant peak is seen at this position at dilutions of 3% and above. At lower concentrations, the peak can be reliably seen and scored down to 0.19%. (percentages have been rounded to one decimal place). %M, percentage of mutant DNA. F, forward; R, reverse.

LNA-PCR Sequencing

The introduction of the LNA probe improved the sensitivity to 0.2% in both directions. Lower concentrations have also been tested (data not shown) but at the low concentrations, additional artifact peaks (representing low-level errors due to misincorporation of Taq polymerase or formalin fixation) are often unmasked, making the assay difficult to interpret. Concentrations of mutant DNA of 3% and above yield sequencing graphs that show only the mutant peaks. Below this level, the relative ratio of the wild-type to mutant peak increases, as shown in Figure 4.

Sequenom Assay

The sensitivity of the Sequenom assay was variable, depending on the specific mutation being tested. For KRAS mutations G12C, G1A, and G13C, the sensitivity of the assay was 2.5% and for G13D it was 10%. Cluster plots and representative spectra of selected dilutions are shown in Figure 5.

Figure 5.

Sensitivity study for sequenom KRAS 37G (37G>T, codon 13). Forward assay. A: Representative graphs of mass (x axis) versus intensity (y axis) showing the mutation at dilutions of 5%, 2.5%, and 1% mutant DNA. A peak is still present at the 1% dilution, but it is at the same level as the background and is insufficient for a positive call. Mutation location (arrow); wild-type peak (WT); peak from another assay in the Multiplexed Sequenom Assay (asterisk). B: The G versus the T mass height cluster distribution at all tested dilutions. The cluster plot also includes negative results (blue dots).

Detailed results of the sensitivity studies of KRAS by all methods are available in Supplemental Tables S2 and S3 (at http://jmd.amjpathol.org).

Comparison of Analytical Sensitivities for Detection of BRAF Mutation

Standard Sequencing

For the BRAF mutation assay, the detection sensitivity of standard PCR and sequencing is 25% mutant DNA in the forward direction and 50% in the reverse (see Supplemental Figure S2 at http://jmd.amjpathol.org).

LNA-PCR Sequencing

The addition of the forward LNA probe improved the sensitivity to 0.38% in both directions. At concentrations of mutant DNA of 3% and above, only the mutant peak is present. At mutant DNA concentrations of 1.5% and below, both mutant and wild-type peaks can be seen at peak ratios illustrated in Supplemental Figure S2 (http://jmd.amjpathol.org). At concentrations lower than 0.38%, multiple additional artifact peaks are generated at the positions bound by the LNA probe, similar to that as seen in the KRAS assay.

Sequenom Assay

Sequenom's sensitivity for the BRAF V600E mutation was 10% in the forward direction and 5% in the reverse direction (Supplemental Figure S3 at http://jmd.amjpathol.org).

Detailed results of the sensitivity studies for BRAF are available as Supplemental Tables S3 and S4 (http://jmd.amjpathol.org).

Discussion

The use of companion molecular diagnostics in solid tumor oncology is on the rise and is rapidly becoming the standard of care in several settings. In the case of metastatic CRC, KRAS mutational status has been swiftly adopted into treatment algorithms following the recognition of its dramatic impact on the effectiveness of EGFR inhibition. BRAF and other downstream effectors of the EGFR pathway (PIK3CA,^2,50–52 PTEN,^50,53 MEK^32,34) will probably follow as soon as sufficient evidence becomes available to warrant official recommendations.

Closely paralleling the emergence of these biomarkers of response is the rapid development of a wide variety of testing techniques that offer advantages in sensitivity and throughput over Sanger sequencing. Based on the present study and our prior experience, we find that the Sanger sequencing method is a reasonable approach for the routine testing of point mutations in both KRAS and BRAF, provided that testing is based on lesions with greater than 50% tumor. This is a tumor percentage that can be easily obtained from an untreated primary lesion or a well-defined metastasis without treatment or extensive inflammatory infiltrate. In general, there is a need for microdissection, but in well-defined lesions, this can be performed without much difficulty. In our group of 138 primary untreated lesions, all but one were successfully tested by standard sequencing, and their mutation profiles were confirmed by both alternative methods of higher sensitivity. Only one lesion failed testing; it corresponded to a very small biopsy of low-quality DNA. Thus, in the setting of primary untreated CRC, standard sequencing is highly reliable and has a very low failure rate (0.7%).

In contrast, the testing of treated CRC tumors and most metastatic lesions by standard sequencing is challenging and labor-intensive and runs significant risks for producing false-negative results. In the current series, 55% of all testing was based on metastatic lesions with variable degrees of inflammation and post-treatment effects that translated into extensive efforts to dissect areas rich in tumor. Careful dissection allowed the identification of 82% of the mutants among the metastatic lesions, but 18% were missed, based on the results of retesting with the LNA probe. This problem becomes even more acute in treated primary lesions, in which 86% of the mutations go undetected by standard Sanger sequencing. In the testing of these samples, we also find that the estimation of tumor content in a background of fibrosis or inflammation is very difficult because the non-neoplastic cells are much smaller than the malignant cells but still contribute the same amount of DNA. Therefore, given this potential for overestimating tumor content in treated CRC samples, setting a minimum tumor component of 40% to 50% as a cutoff for testing, as many institutions do to avoid dissection, could lead to a large number of false-negatives in these particular samples.

Unfortunately, testing of metastatic or treated primary lesions is hard to avoid. It is estimated that 20% of patients with CRC present with metastatic disease, so mutation testing based on a primary lesion is not always feasible. Although most studies have found that metastases maintain the mutation profile of the primary lesion in more than 90% of cases,^54–60 discordance may arise for technical reasons or, rarely, because of metastases from separate primaries. For this reason, it is preferable to test the metastasis rather than the primary tumor, if possible, because it is the former that is being treated by EGFR-targeted therapy.

Highly sensitive techniques such as our LNA-PCR sequencing assay show a clear advantage over standard PCR sequencing and allow informative genotyping over a wide range of specimen types. Mutant DNA can be identified even when it makes up only 0.2% of the DNA in a specimen, so the need for microdissection can be reduced or eliminated entirely. This superior analytical sensitivity exceeds the reported sensitivities of most commercially available allele-specific assays, high-resolution melting analysis, pyrosequencing, and Sequenom.

Testing is still based on standard PCR sequencing techniques, except for the suppression of the WT allele, so the specificity for a mutation remains as high as that of Sanger sequencing.

A potential limitation of our study is that some of the samples that tested positive by the LNA method could not be confirmed as being positive by any other method. These were the samples with the lowest tumor content, so after macrodissection they generated low DNA amounts that were insufficient for further testing after assessment by the three other methods. Despite this limitation, our study design, which required the confirmation of a mutation by duplicate runs and in both the forward and reverse directions, markedly reduces the possibility of false-positive results. Of note, six of the positive results generated by the LNA-PCR sequencing method were confirmed by the Sequenom assay. Further evidence demonstrating the validity of our data is the fact that our LNA-based assay detected the same high-frequency mutations within codons 12 and 13 as did standard sequencing and revealed no rare mutations that could raise the possibility of a false-positive result. Three relatively rare mutations in our group of 334 patients (G12I, G12R, and G13C) were confirmed by all methods as being positive.

It has been previously suggested that the use of highly sensitive methods of mutation testing may not be advantageous in the testing of all CRCs because they may detect physiologically insignificant mutant populations within a larger tumor. This could potentially deny patients with a predominantly nonmutant tumor the benefits of treatment with EGFR inhibitors. In our current series, we did not find this to be the case. All large tumor lesions that tested negative when the Sanger sequencing method was used had concordant negative results by both Sequenom and the LNA-PCR sequencing assay, suggesting that small mutant populations in the sensitivity range of 0.2% were not present, consistent with little or no intratumoral heterogeneity for KRAS and BRAF mutations in CRC.

The low percentage of BRAF V600E found in this study (6%) may reflect a patient population enriched for sporadic microsatellite stable colon cancers, and that is in keeping with the prevalence of BRAF mutation established by Samowitz and colleagues⁶¹ in a series of 803 microsatellite stable tumors, compared to 9.5% in CRCs overall.

Although less sensitive than LNA-PCR sequencing, the Sequenom platform addresses the need for a broader spectrum of mutation screening. Given its multiplexing capabilities, this platform facilitates screening for multiple, less prevalent oncogenic point mutations concurrently with screening for the more common KRAS and BRAF mutations. Thus, aside from the KRAS G12 and G13 mutations and the BRAF V600E mutations that formed the basis of the platform comparisons, our Sequenom assay panel also detected 12 low-frequency KRAS mutations (four A146T, three Q61R, three Q61L, and two Q61H), and one BRAF (G469A) mutation, bringing the overall mutation rates for the KRAS and BRAF+/KRAS- groups to 46% and 6.7%, respectively. Although the predictive value of KRAS Q61 and A146 mutations has not yet been clearly delineated, recent data⁶² suggest that these mutations are also associated with resistance to EGFR-targeted therapy and are biologically similar.⁶³ Aside from multiplexing, the Sequenom platform also offers a clear sensitivity advantage, having detection limits that range from 2.5% to 10% (vs. 25% for Sanger sequencing), depending on the specific mutation tested. Based on this higher sensitivity, six additional mutations. five in metastatic lesions and one in a treated lesion, were detected in the present study, and all of them were confirmed by the LNA-PCR method.

In summary, we find that standard Sanger sequencing techniques are suboptimal for the detection of somatic mutations in metastatic and treated CRC, even with conscientious predissection to enrich for tumor. LNA-PCR sequencing and the Sequenom platform are two alternative approaches that are both more sensitive than standard sequencing, and each has distinct advantages.

Supplementary data

Supplemental Figure 1

KRAS sequencing trace after LNA-PCR demonstrating multiple artifactual peaks in the region bound by the probe. These peaks are usually generated when a genuine mutant allele is absent or extremely low (below 0.2% for KRAS, below 0.8% for BRAF). These cases usually show >2 additional peaks and are poorly reproducible.

Supplemental Figure 2

BRAF sensitivity study. A: Electropherograms of standard sequencing. B: LNA-PCR sequencing. Serial dilutions of mutant DNA with normal control DNA at 50%, 25%, and 12.5% are shown. %M denotes percentage of mutant DNA. At the 25% concentration the mutant peak is seen barely above the background in the forward direction but is not seen in the reverse direction. At a concentration of 12% the mutation is no longer scorable. B: With the addition of LNA, the mutation can be seen as a single or predominant peak down to 1.5% and observed as a peak of the same size as the wild-type at 0.75% mutant DNA dilutions. At dilution levels of 0.38% and below, the mutant peak becomes smaller and difficult to score because of additional artifact peaks (X). Percentages have been rounded to one decimal place.

Supplemental Figure 3

Sensitivity study of BRAF V600E (1799T>A) forward and reverse assay. A: Representative graphs of mass (x axis) versus intensity (y axis), showing the mutation at dilutions of 50%, 20%, and 10% mutant DNA in the forward direction and 20%, 10%, and 5% in the reverse directions. The mutation location (arrow) and wild-type peak (WT) are shown, along with other SNP mutations indicated by the Multiplex Sequenom Assay . B: The G versus T mass height cluster distribution at all tested dilutions. The cluster plot also includes negative results (orange dots).

Supplemental Table 1

PCR Primers and Extension Primers

Supplemental Table 2

Summary of Sensitivity Studies: KRAS exon 2 Mutations: Standard and LNA-PCR Sequencing

Supplemental Table 3

Sensitivity Study for KRAS exon 2 and BRAF V600E Mutations: Sequenom Assay

Supplemental Table 4

Summary of Sensitivity Studies: BRAF exon 15 mutations: Standard and LNA-PCR Sequencing