Oligonucleotides used in this study are summarized in Table I. All cell lines listed in Table II were maintained in our laboratory (National Veterinary Institute, Technical University of Denmark) or kindly provided by Mogens Kruhøffer (Aarhus University Hospital, Denmark). Cell lines were classified into four groups based on the sequence of codon 12 of KRAS gene: wild-type sequence (GGT) was designated ‘WT’, GTT was designated ‘M1T1’, GAT was designated ‘M2T2’, AGT was designated ‘M3T3’ (Table II). Cell lines HT29 and CALU -1 were maintained in McCoy’s medium (HyClone); SW480, SW620 and A549 cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen); all other cell lines were grown in RPMI-1640 medium (Invitrogen). All media used for cell culture were supplemented with 10% fetal calf serum (Invitrogen). Cells were cultured at 37°C in a humidified 5% CO₂ atmosphere. Four unique tags were selected based on published data (30) after analysis with regard to cross-priming, melting temperature, percentage of G + C contain, and possible secondary structure. Four single nucleotide polymorphism (SNP)-specific forward primers specific to four genotypes were designed. One of the four tags was linked to the 5′-end of each of these primers, respectively, to form forward chimeric tagged-primers (Table I and Fig. 1A). Four oligonucleotide probes were modified at the 5′-end with a poly (T) 10 - poly (C) 10 - probe binding tail (TC tail) to facilitate the attachment to the solid substrate, as previously described (31,32). One of the four unique tag sequences was introduced into the 3′-end of each probe (Table I). The reverse primer (Cy5-K-ras-RP2) was labeled with Cy5 fluorescent dye to facilitate detection (Table I). Oligonucleotides were synthesized at DNA Technology A/S (Arhus, Denmark).

Genomic DNA from all cell lines (Table II) was isolated using QIAamp DNA Mini kit (Qiagen, Hilden, Germany). A 111-bp fragment encompassing codon 12 of KRAS gene was amplified by PCR using primers KRASFP1 and KRASRP2 (Table I). PCR was performed in a 20 μl reaction mixture containing 10 μl HotStarTaq Plus Master Mix buffer (Qiagen), 2.5 mM MgCl₂, 20 μM forward and reverse primers and 10 ng extracted DNA. The reaction mixtures were subjected to amplification on PTC-200 Peltier Thermal Cycler (MJ Research, Watertown, NY, USA) with initial step at 95°C for 5 min, followed by 30 cycles with denaturation at 95°C for 40 sec, annealing at 55°C for 40 sec and extension at 72°C for 30 sec. Final extension was at 72°C for 10 min. The PCR products were purified by QIAquick PCR purification kit (Qiagen) and were sequenced directly to determine the altered nucleotide of KRAS gene in codon 12. Equal amounts of four forward ‘tag-primers’ (Table I) were mixed thoroughly and used for multiplex PCR. The multiplex PCR was carried out in a total volume of 20 μl containing the following: 10 μl HotStarTag Plus Master Mix, 2.5 mM MgCl₂, 50 pg purified PCR product, 0.05 pM of mixed forward primer and 2.5 pM of reverse primer Cy5-K-ras-RP2 (Table II). PCR was performed with initial step at 95°C for 5 min, followed by 30 cycles with denaturation at 95°C for 40 sec, annealing at 68°C for 40 sec and extension at 72°C for 50 sec. Final extension at 72°C for 10 min.

Topas plastic microscope slides (Microfluidic, Germany) were treated with 95% ethanol for 1 h and then rinsed with sterile water and dried at room temperature before use. Super Frost glass microscope slides were purchased from Menzel (Braunschweig, Germany). GAPS-II coated glass slides were purchased from Corning (NY, USA). Four DNA probes (Table I) modified at the 5′-end with TC tail (31) were diluted in 150 mM sodium phosphate buffer (pH 8.5) to a final concentration of 50 pM, and deposited in a volume of 10 nl/spot onto the slides using Q-Array robot spotter (Genetix, UK).

The DNA probes were linked to the solid support using a simple one-step method previously described (31,32). After UV irradiation, the slides were washed with agitation in filtered 0.2% (w/v) sodium dodecyl sulfate (SDS) for 5 min followed by 0.1X saline-sodium citrate (SSC, 300 mM NaCl, 30 mM sodium citrate, pH 7.0) for 5 min and then spin-dried.

Cy5-labeled multiplex PCR products (without any additional post-PCR manipulation steps) were mixed with PerfectHyb Plus hybridization buffer (Sigma). Single strand DNA was obtained by boiling the mixture for 5 min followed by incubation in ice-water for 2 min. The solution was transferred immediately onto the surface of the slide. The microarrays were hybridized under a glass coverslip for 1 h at 37°C in a humidified chamber. Slides were washed at room temperature in filtered 0.5 × SDS + 0.1% SSC for 10 min with agitation, rinsed in water and spin-dried. The microarrays were scanned in a LaVision scanner (LaVision BioTech, Germany) using appropriate laser power and exposure time suggested by the manufacturer. The fluorescent intensities of the spots were quantified using Fips BioAnalyzer 4F/4S software.

Genomic DNA extracted from CALU -1 (M3T3) was mixed with either Colon205 DNA (WT) or DNA extracted from feces of a healthy donor at the indicated ratios. The mixed genomic DNA was used as a template for PCR and hybridization.

Twenty-eight DNA samples from patients with tumors were used as template for PCR and hybridization as described above to determine the KRAS codon 12 mutations, which were further verified by direct DNA sequencing.

Hybridization was performed as described above using WT DNA as a target. To remove the hybridized oligonucleotide DNA target, the arrays were boiled in distilled water for 5 min. SW480 DNA (M1T1) was used for hybridization utilizing the same array. Further boiling and hybridization were performed using the same arrayed slide and A549 DNA (M3T3) as the target. Before and after hybridization, the DNA arrays were scanned for Cy5 fluorescence.

To develop a robust, affordable, high-throughput method for the detection of KRAS mutations, a tag-microarray-based genotyping protocol, as shown schematically in Fig. 1A, was developed. A 111-bp fragment encompassing the SNP region of KRAS codon 12 was amplified. The obtained PCR product was used as a template for nested multiplex PCR. In the multiplex PCR, the forward primers were chimeric tag-primers consisting of one of the four unique tags, linked at 3′-end with allele-specific sequences corresponding to WT and three mutant genotypes (M1T1, M2T2, M3T3, see Tables I and II). These primers were used in pair with a 5′-Cy5-labeled reverse primer (Cy5-K-ras-RP2, see Table I). The PCR products were hybridized to the tag probes immobilized on solid substrates (Fig. 1A).

The selection of the tag is an important step because the tag sequence significantly affects efficiency and specificity of hybridization (29,33–35). To minimize cross-hybridization, we first identified tag sequences that were predicted to have minimal cross-priming and unlikely to produce secondary structure. The selected four tags were tested experimentally for cross-hybridization by labeling all four tags with Cy5 fluorescent dye and hybridizing them respectively to a microarray containing four anti-tag probes (Table I). Strong fluorescent signal was reproducibly observed when Cy5-labeled tag target was hybridized with its complementary probe, whereas no signal was observed in random-combination groups (data not shown).

To assess the usefulness of the tag-microarray approach for detection of SNPs in genomic DNA, we examined DNA samples from various cancer cell lines carrying WT or different mutations in KRAS codon 12 (27,29,36,37). The specificity of the tag-microarray protocol described here relies largely on the performance of multiplex PCR, because the forward ‘tag-primers’ are allele-specific for the mutations of KRAS. We established an effective PCR amplification system in which no PCR amplicon was observed in gel-electrophoresis when one forward ‘tag-primer’ was omitted and its corresponding genotype DNA template was used in multiplex PCR which was accomplished based essentially on the same conditions described above (data not shown). DNA isolated from each cancer cell line (Table II) was used as template to obtain PCR amplicons containing codon 12. The PCR amplicons were purified and sequenced. The sequencing results revealed that there were five WT genotypes (HT29, PC3, LNCaP, MCF-7, Colon205), four homozygous mutants (SW480, SW620, A549, CALU -1), and two heterozygous mutants (LS174T, A427) (Table II). These results were consistent with data previously reported (2,14,22,23,27,38). A multiplex PCR was subsequently performed and the PCR products were used directly for the tag-microarray hybridization. Representative images obtained from WT samples were shown in Fig. 1B. As expected, specific signals were readily observed in all WT hybridizations in repeated experiments (Fig. 1B), indicating that DNA isolated from WT cell lines was hybridized specifically to its complementary probe immobilized on all substrates (Topas plastic, glass and GAPS-II coated slides). No cross-hybridization was detected in any WT hybridizations (Fig. 1B), demonstrating specificity of the hybridization. It should be noted that the size of the spot on plastic surface was slightly smaller than that on the surface of glass slide or GAPS-II coated slide (Fig. 1B, compare top panel and middle or bottom panel). This likely resulted from the different physical characteristics (surface tension) between plastic and glass substrates, rather than from hybridization. Relative fluorescence intensity value was calculated by dividing the value of intensity of fluorescence obtained from each spot to that obtained from background. The final relative value for each probe was determined as the average value obtained from six spots in three independent experiments. The results showed that higher averaged relative fluorescent signals were detected in normal glass slide, compared to those obtained in Topas plastic slide or GAPS-II coated slide (data not shown). However, overall, the intensity of fluorescence measured on different substrates was not significantly different.

Next, we analyzed all mutant DNAs extracted from cancer cell lines using procedures as described above. The results indicated that all mutant genotypes were assigned correctly (Fig. 2). Fluorescent signal was consistently detected at the expected genotyped sites in repeated experiments (Fig. 2), indicating that mutations in codon 12 could be distinguished from each other using our tag-microarray approach. The sequencing results showed that LS174 and A427 cells contain a WT allele in addition to the mutated genotype in KRAS gene (data not shown). Consistently, fluorescent signals were detected in both WT and M2T2 probe areas on all slides where the genomic DNAs extracted from LS174T and A427 cells were examined (Fig. 2, M2T2). The intensity of fluorescence between WT and M2T2 genotyped sites was equivalent within the margin of error for LS174T and A427 (Fig. 2), suggesting that the ratio between M2T2 and WT genotypes is 50% as expected in a heterozygous sample. The relative fluorescent intensity of all mutant DNAs hybridized to all immobilized probes was determined, and are shown in Fig. 2. Overall, the calculated values correlated well with the observed images with respect to the intensity of signal (Fig. 2). The presence of WT and mutations in different allele argued that it is of importance to include four primer sets (WT, M1T1, M2T2 and M3T3) in a multiplex PCR. Moreover, targeting both WT and all three mutation possibilities in a multiplex PCR allowed the assay to be self-controlled on a hybridization slide, as at least one set of the probes would be detected as positive, demonstrating the assay was successful. Thus, we did not test other primer mixes, e.g., with less primer sets, against a specific template. In sum, our data indicated that all genotypes of the cells tested were accurately discriminated by using the tag-microarray hybridization technique.

To quantify the detection limit of the tag-microarray method in a background of WT DNA, CALU -1 (M3T3) cell genomic DNA was mixed with varying quantities of WT genomic DNA from Colon205 cells (Fig. 3A) or fecal samples (Fig. 3B) collected from a healthy donor to generate spiked DNA samples. Fragment containing KRAS codon 12 was amplified using the mixed DNA sample as template, and hybridization was performed using normal glass slides, as described above. The results revealed that the KRAS mutant sequence was detectable when it was present only in 10% or more of the starting mixed materials (Fig. 3, proportion of DNA 90:10), demonstrating that the tag-microarray assay was sensitive. Overall, the intensity of the signal observed in both WT and mutant DNA samples was dose-dependent (Fig. 3).

To evaluate the applicability of the tag-microarray approach, 28 tumor DNAs from clinical patients were analyzed using this technique. As shown in Table III, 20 (75%) of 28 samples were detected as WT and seven (25%) of those samples were detected as mutant. All detections were confirmed by direct sequencing. Of seven detected KRAS mutations, two (28.6%) were identified as M1T1 (GGT→GTT); four (57.1%) were identified as M2T2 (GGT→GAT); one (14.3%) was determined as M3T3 (GGT→AGT) (Table III). As verified by a comparison with direct sequencing, all the WT and mutant samples were correctly identified, suggesting that the tag-microarray detection technique performed well.

Finally, we investigated the stability of immobilization of probes on the solid surfaces and examined whether the normal glass slides were reusable. We performed hybridization-boiling-hybridization cycles. Following a successful hybridization with significant fluorescence read, the 5-min boiling led to a background signal (data not shown), indicating that the hybridized targets were completely removed by the boiling. Re-hybridization of the boiled slides with HT29 target re-gained a specific signal, indicating the probes are stably attached and recognized by its target sequence after hybridization-boiling cycle. Similar results were observed for SW480 and A549 targets, and that an additional 15 min of boiling did not result in significant loss of the signal (data not shown). These results suggested that the immobilized probes were stable on the solid phase and the slide could be re-used after a thermo treatment. The corresponding value of signal for each target in repeated experiments is shown in Fig. 4.