A New Microarray Substrate for Ultra-Sensitive Genotyping of KRAS and BRAF Gene Variants in Colorectal Cancer

Molecular diagnostics of human cancers may increase accuracy in prognosis, facilitate the selection of the optimal therapeutic regimen, improve patient outcome, reduce costs of treatment and favour development of personalized approaches to patient care. Moreover sensitivity and specificity are fundamental characteristics of any diagnostic method. We developed a highly sensitive microarray for the detection of common KRAS and BRAF oncogenic mutations. In colorectal cancer, KRAS and BRAF mutations have been shown to identify a cluster of patients that does not respond to anti-EGFR therapies; the identification of these mutations is therefore clinically extremely important. To verify the technical characteristics of the microarray system for the correct identification of the KRAS mutational status at the two hotspot codons 12 and 13 and of the BRAFV600E mutation in colorectal tumor, we selected 75 samples previously characterized by conventional and CO-amplification at Lower Denaturation temperature-PCR (COLD-PCR) followed by High Resolution Melting analysis and direct sequencing. Among these samples, 60 were collected during surgery and immediately steeped in RNAlater while the 15 remainders were formalin-fixed and paraffin-embedded (FFPE) tissues. The detection limit of the proposed method was different for the 7 KRAS mutations tested and for the V600E BRAF mutation. In particular, the microarray system has been able to detect a minimum of about 0.01% of mutated alleles in a background of wild-type DNA. A blind validation displayed complete concordance of results. The excellent agreement of the results showed that the new microarray substrate is highly specific in assigning the correct genotype without any enrichment strategy.


Introduction
Defining the molecular signature of human cancers could be central to the development of a personalized approach to patient care. In fact the identification of appropriate biomarkers might increase accuracy in prognosis, facilitate the selection of the optimal therapeutic regimen, improve patient outcome, and reduce costs of treatment [1]. Thus, stratification of single patients based on molecular and genetic characteristics is the expected evolution of the modern clinical oncology.
Recently therapeutic agents targeting specific genetic variants and well characterized molecular pathways have been developed. This is the case of the oncogene KRAS which is part of the signaling pathway of several different molecules. Gain-of-function missense mutations are often somatically acquired in colorectal cancer, prevalently at three hot spots represented by codons 12, 13, and 61. Unfortunately, at these levels the number of substitutions is high, making their detection more complex with allele specific techniques. Constitutively, activating mutations at these hot spot sites can determine resistance to EGFR-targeted therapies which should otherwise significantly improve the survival rate and the quality of life of patients [2], [3], [4], [5]. The potential of KRAS codon 12/13 mutations as effective molecular markers for drug selection has received considerable attention leading to their use in the routine care of patients with colorectal cancer [6]. The European health authority (http://www.emea. europa.eu/pdfs/human/press/pr/ 27923508en.pdf.) as well as the American Society for Clinical Oncology [7] and the National Comprehensive Cancer Network (NCCN, http://www.nccn.org/ professionals/physician_gls/PDF/colon.pdf.) require KRAS mutational analysis on colorectal cancer prior to anti-EGFR therapy.
Another promising biomarker of anti-EGFR resistance is represented by BRAF V600E mutation, that occurs in about 10% of colorectal cancers. BRAF is the immediate downstream effector of KRAS in the Ras/Raf/MAPK signaling pathway and BRAF V600E activating mutation is mutually exclusive for KRAS mutations [6]. Despite the currently limited data, and lack of complete consensus, it is likely that BRAF mutations have a role in determining the response to anti-EGFR mAb treatment and it is associated with worse prognosis, independently from treatment [8], [9]. Furthermore, patients with tumors carrying mutant BRAF might also benefit from selective BRAF inhibitors such as PLX-4032 [10].
In the present scenario of screening strategies, the current methods of analysis (conventional sequencing, pyrosequencing, etc.) are time-consuming, expensive and lack robustness. Another emerging issue is connected to the real sensitivity of these methods that seem to detect minority mutated alleles only when present at concentrations higher than 10-20%. In previous works [11], [12], we underlined the importance of sensitivity in the detection of minority mutated alleles in biological samples and confirmed the usefulness of COLD-PCR for their enrichment, particularly in samples with low percentages of tumour cells. On average, 15% of patients initially classified as negative for KRAS or BRAF V600E variants were found positive after COLD-PCR [11], [12].
Microarrays represent an inexpensive and accurate tool for parallel genotyping of multiple markers, suitable for routine analysis in medical diagnostics [13]. Here, we report on the development of a highly sensitive microarray for the detection of KRAS and BRAF mutations. The microarray is developed using a crystalline silicon slide coated by a thermally grown silicon dioxide (SiO 2 ) layer and functionalized by adsorption of a copolymer of dimethylacrylamide (DMA), N-acryloyloxysucinimide (NAS) and meta-acryloy propyl trimethoxy silane (MAPS), copoly(DMA-NAS-MAPS), originally developed for glass DNA microarrays [14]. The backbone of the polymer consists of DMA, a monomer that adsorbs to the surface by hydrogen bonding; NAS is the chemically reactive monomer that reacts covalently with bioprobes, whereas MAPS contribute to film stability by condensing with surface silanols. The coating procedure is simple and reproducible, when compared to organo-silanization, a process that requires highly controlled conditions and suffer from poor reproducibility. This functional polymer has been widely applied in the biosensor field for the bio-functionalization of polystyrene nanobeads [15], silicon microcantilevers [16], polydimethylsiloxane [17], and nitrocellulose [18] substrates.
The choice of silicon substrate coated by a layer of SiO 2 of 100 nm thickness leads to a strong intensification of fluorescence on the surface resulting from optical constructive interference between the incident and reflected lights of the fluorescent radiation. The condition of constructive interference at the substrate surface is fulfilled in several types of glass slides coated with layers of dielectric or metal films [19]. However, the strategy involved in producing such complex multi-layer structures often suffers low reproducibility and difficult process control. The simplest configuration to achieve a fluorescent enhancement close to that provided by multi-layer slides consists of the silicon planar reflector coated with a thin film of SiO 2 [20], [21] proposed by this work.
To evaluate technical performances of the newly developed platform and verify its ability to correct genotyping of KRAS and BRAF V600E mutations in colorectal tumor samples, we selected 60 tissue samples already classified for these genetic variants by complete protocols of conventional and COLD-PCR, High Resolution Melting analysis and direct sequencing. The excellent outcome of the results will be discussed in order to show the advantages and the drawbacks of both technologies.

Ethics
Tissue samples were collected at Azienda Ospedaliera Universitaria Careggi in Florence (Italy) from 75 patients undergoing surgery for CRC in the period 1/6/2009-2/5/2011. All individuals enrolled in this study provided written informed consent. The study was approved by the review board of the University of Florence. Moreover, we used two cell lines provided by ATCC-LGC Standards Partnership: the CCRF-CEM cell line as reference for KRAS mutation p.G12D (heterozygous) and the human melanoma cell line A375 as the source of homozygous BRAF V600E DNA. The human colorectal carcinoma cell line SW620 (used as reference for KRAS mutation pG12V, homozygousand the human breast cancer cell line MCF-7 (wild-type for both KRAS and BRAF genes) were supplied by Banca Biologica e Cell Factory (IRCCS Azienda Ospedaliera Universitaria San Martino -IST Istituto Nazionale per la Ricerca sul Cancro).

Sample preparation
Sixty tissue samples were collected during surgery and immediately steeped in RNAlater (Qiagen Gmbh, Hilden, Germany) at 4uC for 24 hours and subsequently stored at 280uC until analysis. Tissues were disrupted by Tissue Lyser with Stainless Steel Beads 5 mm (Qiagen) according to the manufacturer's instructions. DNA purification was performed with QIAcube TM and QIAamp DNA mini Kit (Qiagen).
The concentration of DNA was determined with Nanodrop ND-1000 Spectrophotometer (Thermo Scientific, DE, USA). Moreover, 15 FFPE tissues were also analyzed; DNA from FFPE tissues was extracted using the FFPE Tissue kit (Qiagen) following manufacturer's instruction.
DNA samples were firstly investigated by means of conventional PCR and COLD-PCR amplification followed by HRM and direct sequencing. Subsequently they were blindly submitted to the analysis by the newly developed microarray device to asses its ability in KRAS and BRAF mutations genotyping.
Positive control samples were represented by DNA of cell lines harboring mutations in the target genes (See previous section Ethics). In particular wild-type and mutant samples were assayed separately as single samples and as mixtures, in order to get known percentage of mutated allele (from 6% to 0.01%) to be used for the determination of assay sensitivity for KRAS p.G12D and BRAF V600E variants.
Moreover, plasmidic DNA containing the wild-type sequences and alternatively all the considered variants was used to obtain reconstituted samples to prove assay sensitivity and specificity for all the other KRAS mutations.

Mutagenesis
Mutant-bearing plasmids were generated through the cloning of specific mutagenized PCR products harboring the seven mutations tested in the assay and the corresponding wild-type fragment. Mutagenized fragments were prepared using a modification of the method previously reported [22]. Briefly, mutagenesis was achieved by dividing each amplicon into two fragments. The 59 fragment was then amplified with the original forward primer and a mutagenizing reverse primer introducing a conservative transversion. In parallel, the 39 fragment was amplified with the original reverse primer and a mutagenized forward primer introducing the same nucleotide variation as above (see Table   S1). This led to the production of two partially overlapping fragments, which were each gel-eluted in a final volume of 50-100 mL of distilled water to eliminate non-incorporated primers. We mixed 2 mL of each eluted solution together; each mixture was then elongated for 15 cycles in the presence of the PCR reaction mixture containing all reagents but primers. The product of elongation reaction, resulting in a full-length centrally mutagenized fragment, was further PCR amplified for 20-30 cycles by addition of the complete PCR mixture and cloned in the plasmid vector (TOPO TA Cloning, Invitrogen, LifeTechnologies, Milan, Italy) according to manufacturer's protocol. Direct sequencing confirmed that the desired nucleotide change was introduced into the mutagenized control.

Reporter and stabilizer design
Reporters were designed in a dot-blot format with the base variation corresponding to the hotspots located internally (Table 1). A universal labelling format was used (Figure 1), based on reporters containing a sequence-specific tail that hybridizes to universal Cy3 or Cy5 labelled probes (wild-type probe: CTC AAT GTT CGG ACT CAG-Cy3; mutant probe: TGT CAA GCG ATA TAC TGC-Cy5) [23].  [24]. All reporters were designed using the antisense strand as target amplicon. All the 7 KRAS mutations were analyzed with the same stabilizer oligonucleotide reported in Table 1. The codon 600 BRAF mutation required the use of two stabilizers (Stab 1: located at 59 to the mutation; Stab 2: located at 39 to the mutation, reported in Table 1). . Copoly(DMA-NAS-MAPS) was synthesized and characterized as described [14].

Silicon slide coating and microarray preparation
Slides were finally rinsed with water and dried under vacuum at 80uC.
The PCR products for each gene, amplified from primers with a 59 primary amino-group, necessary to bind the amplicons covalently to the substrate through a reaction between the amino groups and the active esters of the polymer coating, were spotted on the microarray substrates. Three mL of the amino-modified amplicons were printed in 6 replicates using a piezoelectric spotter, SciFLEXARRAYER S5 (Scienion Germany), on coated silicon slides. An amino-modified oligonucleotide labelled with Cy3 was spotted as a positional reference in four columns in every array. Spotting conditions was carried out at as described [25]. The amplicons were coupled to the arrays by incubating in an uncovered storage box, laid in a sealed chamber, saturated with sodium chloride (40 g/100 mL H 2 O) and incubated at room temperature overnight.
After incubation, all residual reactive groups of the coating polymer were blocked by dipping the slide in pre-warmed blocking solution as reported [25]. Hybridization Immediately before hybridization, printed slides were dipped in 0.1 M NaOH for 5 min to denature the double-stranded immobilized amplicons, subsequently rinsed with water and dried. Sequences of reporters and stabilizers along with the hybridization temperatures are detailed in Table 1. In the first step, 0.5 mL of the stabilizer oligonucleotide were mixed with 49.5 mL of hybridization buffer (26 SSC, 0.1% SDS, 0.2 mg/mL BSA) up to 1 mM final concentration and spread onto the spotted area of the slide. The slides were incubated at 20uC for 30 min in the Thermomixer Comfort (Eppendorf) hybridization chamber, and then washed at room temperature in a 46 SSC buffer to remove the cover slip. This first wash step was followed by a brief wash (30 s) in a low-salt buffer (0.26SSC). Then, for the detection of G12S, G12D, G12C, G12R, G13D KRAS mutations and for the BRAF V600E mutations, the reporter for the wild-type and the mutated sequences and their corresponding universal oligonucleotides labelled with Cy3 and Cy5 respectively, were mixed together in equimolar amounts (final concentration 1 mM) and added to the hybridization buffer (26 SSC, 0.1% SDS, 0.2 mg/mL BSA) (see Figure 1 for the assay scheme). In the case of G12A and G12V KRAS mutations it was necessary to incubate the amplicons in two consecutive steps. In the first step the amplicons were hybridized with the reporter complementary to the mutated sequence together with the corresponding universal oligonucleotide labeled with Cy5; then, after a brief wash in 46 SSC to remove the cover slip, in the second one the amplicons were hybridized with the reporter complementary to the wild-type and its corresponding universal oligonucleotide labeled with Cy3. Both incubations lasted for 1 hour.
Finally the silicon slides were removed from the hybridization chamber and soaked briefly in 46SSC buffer to remove the cover slip, washed twice for 5 min in 26 SSC/0.1% SDS, pre-warmed at the specific hybridization temperature, then dipped, in sequence, in a solution 0.26 SSC and 0.16 SSC for 1 min at room temperature, dried by centrifuging at 780 rpm for 3 min and scanned.

Image scanning and data analysis
ProScanArray (Perkin Elmer, MA, USA) was used to scan the hybridized slides. In particular a green laser (l ex 543 nm/l em 570 nm) for the Cy3 dye and a red laser (l ex 633 nm/l em 670 nm) for the Cy5 dye were applied. The photomultiplier (PMT) tube gain and the laser power changed between fluorochromes and different experiments. 16-bit TIFF images were analyzed at 5 mm resolution. Data intensities were extracted with the scanner software and the data analysis was performed for each experiment as previously described [25].

Conventional and COLD-PCR amplification followed by High Resolution Melting (HRM) and direct sequencing
Conventional PCR, COLD-PCR, HRM and sequencing protocols have been already described [11], [26].

1) Conventional and COLD-PCR amplification HRM and sequencing results
Initially, all of the 75 samples were screened for the research of KRAS mutations and BRAF V600E by HRM and sequencing. In a second phase, all samples (mutated and wild-type) were resubmitted to a complete screening using COLD-PCR amplification followed by HRM and sequencing as already reported [11].
Globally, 59 out of 75 samples contained alternatively one mutation either on KRAS hotspots or the BRAF V600E variant, while 16 were classified as wild type samples and included as negative controls. It is important to note that in 9/59 mutated samples, the identification of base substitutions was only made possible through the use of fast or full COLD-PCR protocols, probably due to the lower percentages of mutated alleles in the starting samples. These samples were appositely introduced in the validation study to test the sensitivity of the proposed platform. For a detailed description of the distribution of KRAS and BRAF variants in our samples see Table 2.

2) Optimization of silicon slide analysis
Wild-type control samples and reconstituted heterozygous control samples (generated by properly mixing plasmidic DNA containing wild-type and mutated sequences of all 7 KRAS mutations and of the BRAF V600E variant) were used to test the assay specificity. Following optimization of temperature and hybridization time, a correct identification of all genotypes was achieved. An example is illustrated in Figures 2 and 3 where a hybridization experiment for the G12R KRAS and the BRAF V600E mutation is shown, respectively. In the optimized system, for all the analyzed mutations we could evidence that: i) fluorescent signals were high, ii) no cross-hybridization was obtained, even for variants that affect the same or adjacent  Figure 2D) and iii) a good reproducibility from spot to spot (shown by error bars) was found.

3) Sensitivity of the assay
Sensitivity of the assay in discriminating different proportions of mutated alleles was evaluated on serial dilutions (6%, 3%, 1,6%, 0,8%, 0.4%, 0.2%, 0.1%, 0.05%, 0.02%, 0.01% mutated/wildtype) of mutated DNA opportunely mixed with wild-type DNA. In particular we used CCRF-CEM cell line as reference for KRAS mutation p.G12D (heterozygous), the human colorectal carcinoma cell line SW620 (used as reference for KRAS mutation pG12V, homozygous), and the human melanoma cell lines A375 (used as the source of homozygous BRAF V600E DNA). The human breast cancer cell line MCF-7 were used as wild-type for both KRAS and BRAF genes. For all the other KRAS mutations we used the mutant-bearing plasmids generated.
The detection limit of the proposed method was different for the 7 KRAS mutations tested and for the V600E BRAF mutation. In particular, the microarray system has been able to detect a minimum of about 0.01% of mutated alleles in a background of wild-type DNA for the G13D mutation, about 0.025% of mutated allele for the G12R mutation, 0.05% for the G12C mutation, 0.1% for the G12D mutation, 0.2% for the V600E BRAF mutation, 0.4% for the G12A and the G12S mutations and 0.8% for the G12V respectively. In Figure 4 two examples of the results obtained for the G13D KRAS mutation and for the BRAF V600E mutation are shown.

4) Blind validation
Assay validation was performed in a blinded fashion by analyzing 75 samples from subjects either positive or wild-type for KRAS and BRAF mutations, previously characterized by HRM and direct sequencing. As an example, the microarray results for the G12C KRAS mutation and for BRAF V600E mutation for frozen tissues are shown in Figures 5 and 6, respectively. Moreover in Figure 7 and 8 the microarray results for the G12R KRAS mutation and for BRAF V600E mutation in FFPE samples are shown, respectively. The system was highly specific in assigning the correct genotype to all samples. Blind validation displayed complete concordance of results (Table 2) with the  Table 2).

Discussion
Sensitivity and specificity are fundamental characteristics of any diagnostic method. Highly sensitive assays, able to detect small quantities of the substance under investigation, may open new opportunities in several clinical applications. In particular, the identification of minority mutated alleles within an excess of wild type alleles that represents a common problem in the analysis of cancer DNA, is technically very challenging, and requires the use of highly sensitive techniques.
Within this context, we aimed at developing innovative methodologies to improve the detection of very low proportions of cancer mutations. In this work, rapid detection of KRAS and BRAF mutations was achieved using a microarray support based on high-sensitivity silicon slides. Nevertheless, the availability of a slide with optimized optical properties is just one step toward the improvement of assay sensitivity. In order to fully take advantage of its characteristics, the surface must have high binding capacity and high hybridization yield of the immobilized molecules with their targets in solution. Hence, it has to be pointed out that the approach followed for the functionalization of the slides to allow the immobilization of the amino-modified amplicons is crucial for the success of the assay. Thus, taking into account these requirements, we have employed, for amplicon immobilization on silicon oxide surfaces, a coating procedure that our group has introduced to functionalize glass, silicon, and other kind of slides [27]. The combination of surface chemistry and optical slides properties leads to strong sensitivity increase with an enhancement of fluorescence signals from four to six times [20] with respect to glass. The low non-specific background provided by the polymer coating is an essential requirement to take advantage of fluorescence intensification as the signal enhancement is not specific for spot fluorescence. Moreover, the coating procedure is simple and reproducible and on the surface the polymer forms a thin film of few nanometers the morphology of which does not alter the optical properties of Si/SiO 2 substrates. By applying this system, our data indicated the possibility of identifying at least 0.01% of mutated alleles in a background of wild-type DNA, based on a dilution curve obtained by serial dilutions of mutated cancer cell lines and wild-type DNA. This theoretical sensitivity seems well suited for detecting even a limited percentage of mutated alleles in a heterogeneous sample, as obtained from CRC random tissue biopsies. Notably, our previous experience based on the comparison of conventional PCR versus COLD-PCR amplification [11] indicates that HRM and direct sequencing can identify up to 6.2% and 3.1% of mutated allele if preceded respectively by conventional and COLD-PCR amplification [12]. It is relevant to notice how just the use of COLD-PCR amplification could be determinant for the identification of the presence of BRAF V600E or KRAS mutations at levels below the sensitivity of conventional methods [11], [12]. Nevertheless, other qPCR assays recently introduced on the market (TheraScreenH: K-RAS Mutation Kit, Elucigene KRAS.BRAF and Exigon) declare sensitivity levels of 1%, that represent a definite improvement compared to conventional PCR and sequencing methods, but they might be less sensitive that the method presented in the paper. Our approach could allow for genetic variants detection down to the level of 0.8-0.01% mutant allele (depending on each mutation), which is approximately 10-100 fold more sensitive than most of the other available assays.
To test the assay specificity, the system was initially validated on artificial heterozygous control samples containing wild-type and mutated sequences of all 7 KRAS mutations and BRAF V600E mutation. The optimized slide system provided high fluorescence signals, good reproducibility and allowed correct identification of all genotypes. The drawbacks of this system are that the hybridization temperature must be carefully optimized for each mutation to obtain a specific assignment of the genotype and that the hybridization temperature is different for each mutation requiring 7 different microarrays for the 7 KRAS mutations. Nevertheless the great advantage of the array is that on the same slide you can spot several samples collected from different patients; it means that it is possible to screen disease-related mutations of up to one hundred of individuals on one slide. In this way, hundreds of genomic samples can be scored in a single experiment, making it particularly useful for screening large populations for important markers, such as those implicated in disease susceptibility. Subsequent blind validation on 75 previously characterized samples from subjects either positive or wild-type for KRAS and BRAF mutations indicated 100% agreement for both FFPE and frozen tissue. The microarray's performance obtained in the analysis of FFPE samples is the same shown for the analysis of DNA extracted from tissue immerse in RNA later where the DNA integrity is excellent. This condition makes the array useful also in a clinical setting where the available tumour samples are often FFPE tissues.
All these data strongly support the application of the newly developed microarray assay for the scanning of unknown DNA samples.
KRAS and BRAF mutations are mutually exclusive. In the current study all 45 KRAS positive samples were wild-type for BRAF V600E mutation and all 12 BRAF positive samples were wildtype for KRAS, supporting the mutual exclusiveness of the two genes in the tumour process.
The development of target therapies has created a clinical need for a fast and accurate molecular characterisation of tumours.
Understanding which gene is involved in the cancer origin is mandatory for good clinical practice and allows the physician to offer their patients better therapeutic choices. Finally, the method described in this paper represents the first attempt to generate a platform for the simultaneous and sensitive detection of a panel of somatic mutations starting from a limited DNA amount. Obviously, this prototypical approach can be implemented to reduce cumbersome procedures and introduce semi-automatic steps.
Moreover, sensitivity is not to be considered a secondary aspect in view of a possible application to detect mutated alleles in biofluids, i.e. blood from cancer patients.

Supporting Information
Table S1 Gene variant, primer sequences, and base substitution for PCR-mediated site-directed mutagenesis. (DOC)