Formalin-fixed paraffin-embedded (FFPE) human CRC samples (N = 40) were obtained from the following procurement service providers: Asterand plc (Detroit, MI), Ardais Corp (Lexington, MA), and the National Disease Research Interchange (Philadelphia, PA). Fourteen (35%) samples were from men, 20 (50%) were from women, and 6 (15%) were unassigned (Table 1). The median age was 67 years (range, 35-94 y); age data were not available for 7 patients. The samples were primary resections from colon adenocarcinoma (n = 36), rectum adenocarcinoma (n = 3), and rectum carcinoma (n = 1) with a range of poorly to well-differentiated tumors of different stages with variable tumor, normal, stromal, and necrotic content. The aim was to select samples that were representative of samples expected in clinical trials. All study procedures were conducted in accordance with the Declaration of Helsinki.

Mutational analysis of KRAS sequences was performed by the Amgen DNA Sequencing Laboratory and 5 independent laboratories that provide diagnostic services for academic/clinical research laboratories and/or clinical trials. Each laboratory was provided with 10-μm tissue sections from all 40 samples in 2007 except Agencourt. Agencourt was provided with extracted genomic DNA from the remaining 35 specimens in 2009. DNA was extracted using the QIAamp FFPE Tissue kit (Qiagen Inc, Carlsbad, CA) according to the manufacturer's instructions, with the addition of a 16-hour proteinase K lysis step.

Exon 2 of KRAS was amplified from isolated genomic DNA using the Roche Expand Long Template PCR System (Roche Applied Science, Indianapolis, IN). The forward primer sequence was 5'-AAGGTACTGGTGGAGTATTTG-3', and the reverse was 5'-GTACTCATGAAAATGGTCAGAG-3', resulting in a predicted amplicon length of 295 bp. Cycling conditions were as follows: 93°C, 3 minutes; 40-46 cycles at 93°C, 15 seconds; 62°C, 30 seconds; 72°C, 30 seconds; and 72°C, 4 minutes. Polymerase chain reaction (PCR) products were directly sequenced in triplicate (3730xl DNA Analyzers; Applied Biosystems). Sequences were analyzed using Sequencher^® software (Gene Codes Corporation, Ann Arbor, MI). Results were reported when all 3 parallel PCR products generated 4 acceptable sequences, resulting in a total of 12 sequences for each sample.

Five commercial services were contracted to analyze KRAS mutational status. HistoGeneX (Antwerp, Belgium) used the DxS K-RAS Mutation Test Kit (DxS Ltd, Manchester, UK) that interrogates the 7 most common somatic mutations of codons 12 and 13 (G12A, G12D, G12R, G12V, G12C, G12S, and G13D) using allele-specific PCR amplification with an amplification-refractory mutation system technology in combination with Scorpion^® detection to measure amplification products by real-time PCR 5 6 . The DxS K-RAS Mutation Test Kit is capable of detecting approximately 1% of mutated KRAS from background wild-type DNA.

Gentris Clinical Genetics, Inc. (Morrisville, NC, USA) assessed KRAS mutational status by direct DNA sequencing of codons 12 and 13. Gentris used a pathology service (Integrated Laboratory Systems, Inc.) to enrich for tumor tissue by removing normal tissue from the slides and then extracting genomic DNA from the remaining material. Isolated genomic DNA was then amplified with primers that spanned codons 12 and 13 of KRAS, and the resulting PCR products were subjected to cycle sequencing in reverse reactions.

Genzyme Corporation (Cambridge, MA) used an allele-specific primer extension technique and reported results for any mutation of codons 12 and 13.

Invitek GmbH (Berlin, Germany) used their Invisorb^® Spin Tissue Mini Kit for DNA extraction and evaluated KRAS mutational status using their Invigene^® K-ras Genotyping Kit, which employs allele-specific oligonucleotide hybridization. Results were reported for 6 possible mutations at each of codons 12 and 13.

Agencourt Bioscience Corporation conducted picotiter plate pyrosequencing using the Roche 454 system described previously 7 8 9 . Briefly, unique barcode-containing primers were used to amplify KRAS exon 2. All the uniquely tagged samples were pooled at equimolar concentrations and adaptors were ligated to allow bead binding. Optimized mixing conditions were then used to ensure that each bead-bound DNA molecule was within an oil droplet that contained PCR reagents. Emulsion PCR was performed using primers that included biotin to enable selective capture of those beads where successful amplification took place. The emulsion was subsequently broken, the biotin-containing beads captured, and each bead loaded by centrifugation into 1 well of a picotiter plate. Traditional pyrosequencing was performed, yielding a library of sequences linked to unique barcodes. Results were reported as the sequences derived from flowgrams, together with the 10-bp sequences employed for each sample to allow deconvolution of the data at Amgen. The number of mutant and wild-type sequences for each sample was reported.

KRAS genotype was classified as either wild type or mutant. Techniques were considered to be in agreement if both identified wild type or a mutant, even if the allele reported was different. Agreement was assessed by κ statistics 10 using SAS/STAT^® software version 9.1 (SAS Institute, Cary, NC).

Twenty (50%) patients had KRAS exon 2 mutations in sequencing performed by Amgen: G13D (n = 10), G12D (n = 5), G12V (n = 5), G12A (n = 2), and G12S (n = 1) (Table 2). Three of the G13D mutations occurred in combination with a codon 12 mutation.

Variation was observed in the comparability of direct sequencing by Amgen and the results of commercial analyses as shown by the tabulation of the reported results (Table 2) and the comparison of these results using the κ statistic (Table 3). The results of allele-specific testing by HistoGeneX using the DxS kit (κ = 0.95; 95% CI, 0.83—1.00) and by Genzyme (κ = 0.94; 95% CI, 0.85—1.00) closely matched the results of direct sequencing by Amgen, differing on 2/40 samples and 1/35 samples, respectively. The results of picotiter plate sequencing by Agencourt were also almost in perfect agreement with the results of direct sequencing by Amgen (κ = 0.94; 95% CI 0.52—0.98; 32 results reported from 35 samples). Results from the 2 techniques differed in only 1 sample when making a determination of KRAS wild-type versus mutant sequence. Amgen direct sequencing identified sample 9 as wild type, whereas Agencourt picotiter plate sequencing detected a G12V mutation. The results of direct sequencing by Gentris were in substantial agreement with the results of direct sequencing by Amgen, with disagreement on 4 samples (κ = 0.75; 95% CI, 0.52—0.98; 32 results reported from 40 samples). The results of mutational analysis by allele-specific oligonucleotide hybridization by Invitek (as conducted in 2007) were in slight agreement with the results of direct sequencing by Amgen (κ = 0.13; 95% CI -0.15—0.42; 27 results reported from 40 samples). KRAS analysis by Invitek identified only 3 wild-type samples, 2 of which were in agreement with direct sequencing by Amgen.

The quantitative nature of the Agencourt picotiter plate sequencing assay, which generates counts of mutant and wild-type sequences, provided additional data that, in many instances, allowed clarification of the results from other laboratories (Table 2). For example, 3 of the 5 other laboratories detected the G12S mutation in sample 7, whereas picotiter plate sequencing identified the mutant allele in 6% of the DNA. In sample 21, direct sequencing by Amgen identified a G13D mutation, HistoGeneX detected a G12D mutation, and Invitek and Genzyme detected both G13D and G12D mutants. Picotiter plate sequencing identified G13D and G12D mutations in 26% and 12% of the DNA, respectively. In sample 9, picotiter plate sequencing identified a G12V mutation in 41% of sample DNA, whereas the other assays identified a wild-type genotype. It is unclear whether this is due to tumor heterogeneity or other factors.