DNA Microarray to Monitor the Expression of MAGE-A Genes

Transcription

1 Clinical Chemistry 48: (2002) Molecular Diagnostics and Genetics DNA Microarray to Monitor the Expression of MAGE-A Genes Nathalie Zammatteo, 1* Laurence Lockman, 1 Francis Brasseur, 2 Etienne De Plaen, 2 Christophe Lurquin, 2 Pierre-Emmanuel Lobert, 1 Sandrine Hamels, 1 Thierry Boon, 2 and José Remacle 1 Background: The MAGE-A genes encode antigens that are of particular interest for antitumor immunotherapy because they are strictly tumor specific and are shared by many tumors. We developed a rapid method to identify the MAGE-A genes expressed in tumors. Methods: A low-density DNA microarray was designed to discriminate between the 12 MAGE-A cdnas amplified by PCR with only one pair of consensus primers. The assay involved reverse transcription of total RNA with oligo(dt) primer, followed by PCR amplification and hybridization on a microarray. Amplification in the presence of Biotin-16-dUTP allowed subsequent detection of the amplicons on the microarray carrying 12 capture probes, each being specific for a MAGE-A gene. Probe amplicon hybrids were detected by a streptavidin-based method. Results: PCR conditions were optimized for low detection limits and comparable amplification efficiencies among all MAGE-A nucleotide sequences. The microarray assay was validated with a panel of 32 samples, by comparison with well-established reverse transcription- PCR assays relying on amplification with primers specific for each gene. Virtually identical results were obtained with both methods, except for MAGE-A3 and MAGE-A5. Detection of MAGE-A5 was more sensitive with the microarray assay. Detection of MAGE-A3 was hampered by the presence of MAGE-A6, which is 98% identical: the MAGE-A3 capture probe cross-hybridized with MAGE-A6 amplicons because these sequences differed by only a single base. 1 Laboratoire de Biochimie Cellulaire, Facultés Universitaires Notre-Dame de la Paix, B-5000 Namur, Belgium. 2 Ludwig Institute for Cancer Research, B-1200 Brussels, Belgium. *Address correspondence to this author at: Laboratoire de Biochimie Cellulaire, Facultés Universitaires Notre-Dame de la Paix, 61 Rue de Bruxelles, B-5000 Namur, Belgium. Fax ; nathalie.zammatteo@ fundp.ac.be. Received June 12, 2001; accepted October 15, Conclusions: This post-pcr microarray assay could be useful to evaluate MAGE expression in tumors before therapeutic vaccinations with MAGE-A gene products American Association for Clinical Chemistry The MAGE genes were initially identified because they encode tumor antigens that can be recognized by cytolytic T lymphocytes derived from blood lymphocytes of cancer patients (1). The MAGE gene family is composed of 23 related genes divided into four clusters located in regions q28 (MAGE-A genes), p21.3 (MAGE-B genes), q26 (MAGE-C genes), and p11 (MAGE-D genes) of the X chromosome (2 9). MAGE-A, -B, and -C genes are expressed in various types of tumors, but not in healthy tissues, with the exception of testis and placenta. Unlike the other MAGE genes, MAGE-D genes are expressed ubiquitously. The MAGE-A subfamily comprises 12 genes that share 60 98% identity in their coding sequence, located entirely in the last exon. They encode proteins of amino acids (2). Seven MAGE-A genes (MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12) were found to be highly transcribed in a large proportion of tumors of various histologic origins [for a review, see Ref. (10)]. The five other MAGE-A genes are either not expressed in tumors (such as the MAGE-A7 pseudogene) or are expressed in very small amounts. However, MAGE-A8 and -A11 were found to be highly expressed in a small number of tumors (2, 11). Many antigenic peptides presented to CD8 cytolytic T lymphocytes by HLA class I molecules have been identified in proteins MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 [for a review, see Ref. (10)]. MAGE-A1 and -A3 peptides were also reported to be recognized on HLA class II molecules by CD4 T lymphocytes (12 15). The MAGE-A antigens are of particular interest for antitumor immunotherapy because they are strictly tumor specific and are shared by many tumors. The lack of expression of the encoding genes in healthy tissues should ensure strict 25

2 26 Zammatteo et al.: MAGE-A DNA Microarray tumoral specificity of immune responses obtained after vaccination of patients with these antigens. No adverse side effects are expected in testis because HLA molecules are not present at the surface of germline cells, the subset of testis cells expressing MAGE-A genes (16, 17). These antigens are good candidates for immunization of a large number of patients because the encoding genes are expressed in a wide range of tumors. Clinical trials have been initiated with MAGE-A-derived immunogens. In one of these trials, 25 tumorbearing HLA-A1 melanoma patients were immunized with a MAGE-A3 peptide presented by HLA-A1; objective regressions of metastases were observed in 7 patients. Three of these regressions were complete (18). The patients eligible for immunization with a defined MAGE-A antigen are those who have a tumor that expresses the relevant MAGE-A gene along with the appropriate HLA specificity. Whether these conditions are met can be tested readily by HLA typing and by reverse transcription-pcr (RT-PCR) on RNA extracted from tumor samples. Each MAGE-A gene can be identified by specific amplification with pairs of primers unique for each sequence, which generates amplicons of different sizes that are visualized in agarose gels (2). This method is robust but labor-intensive because separate PCR assays using different pairs of primers and annealing temperatures are required to detect the presence of the various MAGE-A mrnas. Recently, a RT-PCR electrochemiluminescence assay was developed to detect MAGE-A-positive metastatic cancers in tissues and blood (19). This assay seems to be very sensitive, but it does not allow identification of the various MAGE-A members that are expressed in the sample. The emerging technology of DNA microarrays is especially useful for multiparametric detection, such as discrimination between the 12 MAGE-A sequences. The array is composed of a series of discrete regions bearing capture nucleotide probes that are able to hybridize complementary target nucleotide sequences. If the latter are labeled, a signal can be detected and measured at the hybridization site; its intensity should be proportional to the amount of target sequences present in the sample. However, such a multiparametric detection system may be difficult to optimize because the hybridization rate and the stability of hybrids in washing conditions are dependent on the sizes and nucleotide sequences of the various targets and probes that are incubated together. Homology between target sequences, as is the case with the MAGE-A genes, increases the difficulty because several targets may compete for the same capture probes. In this study, we designed a post-pcr low-density microarray assay to analyze the expression of the 12 MAGE-A genes. The assay involved reverse transcription of total RNA with oligo(dt) primer, PCR amplification in the presence of Biotin-16-dUTP, and hybridization of the resulting PCR products on a DNA microarray (Fig. 1). We assessed the specificity and the detection limits of the assay and performed a method comparison study on 32 RNA samples obtained from healthy tissues, tumor tissues, and tumor cell lines. Materials and Methods cdna clones, tumor cell lines, and tissues cdna clones of MAGE-A1 (20); MAGE-A2 (21); MAGE-A3 (22); MAGE-A4 (23); MAGE-A5, -A8, -A9, -A10, -A11 (11); and MAGE-A6 and -A12 (unpublished results) as well as tumor cell lines AVL3-MEL, BB90-MEL, LB265- MEL, LB373-MEL, LB583-MEL.6, LB1448-MEL.2, LB1751- MEL, LB1781-MEL, MZ2-MEL.3.0 (melanomas), BB49- HNC (squamous cell head and neck carcinoma), LB11/ OC1 (small-cell lung carcinoma), LB905-BLC (bladder carcinoma), and LB1828-RCC (renal cell carcinoma) were from the Brussels branch of the Ludwig Institute. Pleural mesothelioma line DUN was obtained from Dr. L. Saint- Etienne (Henri Mondor Hospital, Créteil, France) and medullary thyroid carcinoma line TT (CRL-1803) from ATCC. Tumor tissue samples (three melanomas, three non-small-cell lung carcinomas, two colorectal carcinomas, two squamous cell head and neck carcinomas, and one hepatocarcinoma) were collected at resection. They were frozen in liquid nitrogen immediately and stored at 80 C until RNA extraction. Samples of total RNA from healthy adult tissues (liver, brain, lung, colon, skeletal muscle, and testis) were purchased from Clontech. detection of mage-a cdna sequences with specific primers RNA extraction and RT-PCR amplifications were performed as described previously (2) with slight modifications. Briefly, total cellular RNA from melanoma tissues and cell lines was extracted by the guanidine isothiocyanate cesium chloride procedure (24), whereas samples from nonmelanoma tissues and cell lines were prepared with TriPure reagent (Roche). cdna was synthesized from 2 g of total RNA by extension with oligo(dt) primer and 200 U of Moloney murine leukemia virus reverse transcriptase (Life Technologies) at 42 C for 90 min. PCR amplification was performed on 2.5 L (1/40) of the cdna solution with U of Taq DNA polymerase (Takara), 0.4 mol/l each primer, and 100 mol/l each dntp (Takara) in a final volume of 25 L. Pairs of specific primers and amplification programs for each of the MAGE-A genes were as described by De Plaen et al. (2) (for MAGE-A5, -A7, -A8, -A9, -A11) and van Baren et al. (25). These primers were chosen in different exons to avoid the occasional false positives caused by DNA contamination of the RNA preparations. Thirty cycles of amplification were performed, except for MAGE-A5, -A7, -A8, -A9, -A11, and -A12 cdna sequences, which were amplified for 32 cycles. After amplification, PCR products (10 L) were electrophoresed in 1.3% agarose gels and visualized by ethidium bromide fluorescence.

3 Clinical Chemistry 48, No. 1, Fig. 1. Outline of the microarray assay to identify MAGE-A mrnas. (A), cdnas obtained by reverse transcription of total RNA were amplified by PCR with consensus primers DPSCONS2 and DPASCONB4 in the presence of Biotin-16-dUTP. The labeled amplicons were hybridized on a DNA microarray containing 12 capture probes, each specific for a MAGE-A gene. Probe amplicon hybrids were detected by fluorescence or colorimetry with a streptavidin-based method. The diagram shows the result for a sample expressing only MAGE-A4. (B), alignment of the sequences of the consensus primers with the corresponding sequences for each of the 12 MAGE-A genes. Identities are indicated by black boxes and discrepancies by white boxes. Nucleotide numbering is relative to the ATG codon of each MAGE-A gene. detection of mage-a cdna sequences with consensus primers and hybridization on microarray cdna was synthesized as described above. Primers for PCR amplification were DPSCONS2 (5 -GGGCTCCAG- CAGCCAAGAAGAGGA-3, sense primer) and DPAS- CONB4 (5 -CGGTACTCCAGGTAGTTTTCCTGC-3, antisense primer), which were derived from a consensus sequence for the last exon of the MAGE-A genes. They were used together with specific sense primers DPS- MAGE1 (5 -GGGTTCCAGCAGCCGTGAAGAGGA-3 ), DPSMAGE8 (5 -GGGTTCCAGCAGCAATGAAGAGGA- 3 ), and DPSMAGE12 (5 -GGGCTCCAGCAACGAA- GAACAGGA-3 ) to improve the amplification efficiencies of the MAGE-A1, -A8, and -A12 cdnas. The size of the PCR products was 540 bp. Amplification was performed in a 25- L reaction volume containing 2.5 L of cdna; 1 PCR buffer [75 mmol/l Tris, ph 9, 50 mmol/l KCl, 20 mmol/l (NH 4 ) 2 SO 4, 2.5 mmol/l MgCl 2 ]; 200 mol/l each of datp, dctp, and dgtp; 150 mol/l dttp; 50 mol/l Biotin-16-dUTP (Roche); 1 mol/l antisense consensus primer; 0.25 mol/l each sense primer, including the consensus sense primer DPSCONS2 (Eurogentec); U of DNA polymerase (Biotools); and 0.2 U of Uracil-DNA Glycosylase (Roche) to prevent carryover contamination (26). Cycling was performed in a PTC-200 thermocycler (Biozym). PCR reactions were heated to 95 C for 5 min. Amplification was then carried out for 30 cycles (30 s at 94 C,30sat55 C,30sat72 C), and a final extension was performed at 72 C for 10 min. These conditions were chosen after an optimization study using recombinant plasmids containing MAGE-A1, -A2,

4 28 Zammatteo et al.: MAGE-A DNA Microarray Table 1. Characteristics of the MAGE-A capture probes. Probe Probe sequence (5 33 ) Length, bases T m, a C Location b MAGE-A1 GCACCTCTTGTATCCTGGAGTCCTTGT MAGE-A2 GAATGTTTCCCGACCTGGAGTCCGAGT MAGE-A3 AAGACAGTATCTTGGGGGATCCCAAGA MAGE-A4 TCGCCTGACGCAGAGTCCTTGTTCCGA MAGE-A5 CCCTGACCCAGAGTCTGTGTTCCGAGC MAGE-A6 CAAGACAGGCTTCCTGATAATCATCCT MAGE-A7 AGACACACCCCGCTCACCTGGCGTCCT MAGE-A8 CCCCGGACCCAGCTCACCTGGAGTCCC MAGE-A9 CGGTCGACCCAGCTCAGCTGGAGTTCA MAGE-A10 TCCTGCCAGACAGTGAGTCTTTACCCA MAGE-A11 TGACCTGATAGACCCTGAGTCCTTTTC MAGE-A12 ACCTGGAGACGAGCTTCCAAGTAGCAC a Calculated by Oligo-4 software (National Biosciences, Inc.) using the nearest-neighbor method. b Nucleotide numbering is relative to the ATG codon of each MAGE-A gene. -A3, -A4, -A5, -A6, -A8, -A9, -A10, -A11, or-a12 cdna as templates and comparing different PCR volumes (25, 50, and 100 L), annealing temperatures (55, 57, 59, 61, 63, and 65 C), numbers of cycles (30, 35, and 40), and DNA polymerases from different suppliers. After amplification, PCR products (5 L) were electrophoresed on a 2% agarose gel and visualized by ethidium bromide staining. MAGE DNA microarrays used the technology developed by Advanced Array Technology (AAT). They contained capture probes specific for each of the 12 MAGE-A sequences. Each probe was spotted on the array in quadruplicate. The spots were 400 m in diameter, and the distance between two adjacent spots was 600 m. The characteristics of the 12 capture probes are given in Tables 1 and 2. Each microarray had two types of control spots. Detection control spots were biotinylated DNA to check the reliability of the detection step. Positive and negative hybridization control spots were capture probes that were allowed to react with biotinylated complementary or noncomplementary target sequences to check the specificity of the hybridization. The array included an additional negative control, which was the MAGE-A7 capture probe. Indeed, because MAGE-A7 is a pseudogene, no signal should be observed for that probe unless the mrna has been contaminated by genomic DNA. Microarrays were used as recommended by the manufacturer. The completed PCR reaction (15 L) was mixed with 55 L of hybridization solution (AAT) containing 5 nmol/l biotinylated DNA control; the mixture was then loaded on the array framed by an hybridization chamber (MJ Research Inc.). The chamber was closed with a coverslip. Slides were incubated at 65 C for 2 h and then washed four times for 2 min each with washing buffer (AAT). Detection was by either fluorescence or colorimetry. Both detection methods have the same detection limits (27). For fluorescence detection, slides were incubated for 45 min at room temperature with streptavidin labeled with cyanin-5 (Sigma) at 2 mg/l in blocking buffer (AAT). After incubation, slides were washed five times (2 Table 2. Sequence comparison of the MAGE-A capture probes and the MAGE-A target sequences. a MAGE-A target DNA Capture probes A1 A2 A3 A4a A5a A6 A7 A8 A9 A10 A11 A12 MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A MAGE-A a The percentages of nucleotide identity are given.

5 Clinical Chemistry 48, No. 1, min each) with washing buffer, rinsed twice with water, dried at 50 C for 15 min, and read with a GMS 418 array scanner (Genetic Microsystem). For colorimetric detection, slides were incubated for 45 min at room temperature with streptavidin labeled with colloidal gold (Sigma) diluted 1:1000 in blocking buffer. After incubation, slides were washed five times (2 min each) with washing buffer, rinsed twice with water, incubated for 15 min at room temperature with Silver Blue revelation solution (AAT), rinsed with water, dried, and read with a microarray colorimetric reader (AAT). The image of each array was analyzed with software from AAT. Briefly, this software uses three detection control spots as a reference and assigns a correlation coefficient between 0 and 1 to each of the other spots in the array. All analyses were done with the detection cutoff fixed to a correlation coefficient of Signals below this value were considered as negative. Results specificity of the assay To check the specificity of the assay, recombinant plasmids containing cdna of MAGE-A1, -A2, -A3, -A4, -A5, -A6, -A8, -A9, -A10, -A11, or-a12 were subjected to PCR amplification with consensus primers DPSCONS2 and DPASCONB4. Analysis of PCR products by gel electrophoresis showed that the optimized PCR conditions described in Materials and Methods allowed us to amplify each of the 11 cdna templates with high efficiency. The PCR products were hybridized on microarrays. Hybridization was specific except for the MAGE-A6 amplicons, which cross-hybridized on the MAGE-A3 capture probe (Fig. 2 and data not shown). The reason for this crosshybridization was the 96% identity shared by these sequences (Table 2). Different hybridization conditions were then tested by varying the amplicon concentration, the incubation temperature, and the buffer composition. Fig. 2. Evaluation of the specificity of the microarray assay. Fluorescence images were obtained on arrays hybridized with 5 L of PCR products ( 17.5 ng) obtained by amplification with consensus primers DPSCONS2 and DPASCONB4 from MAGE-A1, -A2, -A3, -A4, -A5, -A6, -A8, -A9, -A10, -A11, or-a12 cdna cloned in plasmids. After 2hofincubation at 65 C, detection was performed using a streptavidin cyanin 5 conjugate and a confocal scanner reader.

6 30 Zammatteo et al.: MAGE-A DNA Microarray Although still present, the nonspecific signal from the MAGE-A3 capture probe was considerably reduced when a5- L volume of biotinylated amplicons at 3.5 mg/l was incubated at 65 C for 2 h with the capture probes in the buffer provided by AAT. As shown in Fig. 2, these conditions were used to hybridize PCR products amplified from 5 ng of recombinant plasmid containing MAGE- A1, -A2, -A3, -A4, -A5, -A6, -A8, -A9, -A10, -A11, or-a12 cdna on microarrays carrying four replicates of the 12 capture probes. Each hybridization was specific, except for a slight hybridization of the MAGE-A6 amplicons with the MAGE-A3 capture probe. This cross-reaction represented only 5% of the specific signal obtained with MAGE-A6 amplicons on the MAGE-A6 capture probe (data not shown). No cross-reaction was observed between the MAGE-A3 amplicon and the MAGE-A6 capture probe, whose sequences share only 88% identity (Table 2). detection limit of the assay To estimate the detection limit of the microarray assay, various amounts of total RNA extracted from MZ2-MEL cells were tested. This melanoma line expresses MAGE- A1, -A2, -A3, -A6, and -A10 and weakly expresses MAGE-A5 and -A12 (2, 11). RNA (2.0, 0.2, or 0.02 g) was reverse-transcribed with oligo(dt), and 1/40 of each cdna reaction was PCR-amplified for 30 cycles with consensus primers DPSCONS2 and DPASCONB4. Portions of the PCR reactions ranging from 25 to 5 L were hybridized on microarrays under optimal conditions. When we used 2 g of RNA and either 25 or 15 L ofthe amplification reaction, the assay detected the expected MAGE-A genes, except MAGE-A12 (Table 3). With 15 L of the amplification reaction or 2 g of RNA, some of the MAGE-A genes expressed in MZ2-MEL were not detected. Total RNA, g Table 3. Estimation of the detection limit of the microarray assay. Volume of amplicons, L MAGE-A cdnas detected in MZ2-MEL cells a A1 A2 A3 A5 A6 A10 A b a As determined by RT-PCR with specific primers and analysis of PCR products in agarose gels. b Expression of MAGE-A genes determined by RT-PCR with consensus primers and hybridization on DNA microarray:, expression detected;, no expression detected (see text). The failure of the microarray assay to detect MAGE- A12 expression in MZ2-MEL cells was probably the consequence of inefficient PCR amplification because of the three mismatches present in the consensus primer DPSCONS2 vs the sequence of MAGE-A12, including a destabilizing mismatch close to the 3 end (Fig. 1). The MAGE-A genes whose expression in MZ2-MEL cells was readily detected with the microarray have only one (MAGE-A2 and -A5), two (MAGE-A3, -A6, and -A10), or three mismatches (MAGE-A1) that are less destabilizing. Thus, it was likely that the various MAGE-A cdnas were not amplified with the same efficiency in the PCR performed with the selected consensus primers. We therefore supplemented the PCR mixture with specific sense primers corresponding to the less-matching MAGE-A sequences. By including sense primers specific for MAGE- A1, MAGE-A8, and MAGE-A12, the microarray assay could detect the expected MAGE-A genes in MZ2-MEL, including MAGE-A12 (data not shown). This method was also tested on melanoma cell line LB1751-MEL, which expresses MAGE-A1, -A2, -A3, -A4, -A6, and -A10 and weakly expresses MAGE-A5, -A8, -A11, and -A12. Two parallel assays were run: one without and the other with the three additional sense primers. With these specific primers added, hybridization of the resulting PCR products on the microarray allowed the detection of all expected MAGE-A genes except MAGE-A11. In the absence of specific primers (i.e., 1 mol/l each consensus primer), MAGE-A8 and -A12 were not detected, nor was MAGE-A11 (data not shown). The lack of detection of MAGE-A11 by the microarray was probably a consequence of inefficient PCR amplification caused by three mismatches in the sequence of consensus primer DPS- CONS2 vs the sequence of MAGE-A11 (Fig. 1). method comparison studies with cell lines and tissues To validate the assay, a series of 32 oligo(dt)-primed cdna samples from different tumor cell lines, tumor tissues, and healthy tissues was analyzed. Four independent PCRs were performed on each sample, and the resulting products were hybridized on separate microarrays. The pattern of MAGE-A gene expression in each sample was deduced from the signals detected in 4 4 replicates of the 12 capture probes, as explained in the legend of Fig. 3. Table 4 summarizes the results obtained with the microarray assay and with the semiquantitative PCR assays performed with specific primers generating amplicons of different sizes that were visualized in agarose gels. Virtually identical results were obtained with both methods, except for MAGE-A3 and MAGE-A5. MAGE-A3 was detected in sample LB1781 by the microarray assay, but not by PCR with specific primers. The cloned PCR products obtained from this sample were sequenced, and MAGE-A6 sequences, but no MAGE-A3 sequences, were found. This indicated that false positives in the microarray assay resulted from a cross-hybridiza-

7 Clinical Chemistry 48, No. 1, Fig. 3. Microarray display and data analysis for sample LB1751. The assay conditions were as described in Materials and Methods. The products of four independent PCRs were hybridized on separate arrays. Colorimetric detection was used. After scanning of the arrays, image analysis software was used to detect spots with signals above the threshold. The detected spots are framed ( ). An array was considered positive for a MAGE-A gene when at least three spots were detected and negative when fewer than three spots were detected. Gene expression was scored according to the number of positive arrays:, strong or moderate (four or three positive arrays); ( ), weak (two positive arrays);, very weak (one positive array) or no expression detected (no positive array). tion of MAGE-A6 amplicons on the MAGE-A3 capture probe, as was observed previously with amplicons obtained from a MAGE-A6 cdna cloned in a plasmid (see Fig. 2). To confirm these results, we analyzed seven additional samples that had been found positive for MAGE-A6 and negative for MAGE-A3 by PCR with specific primers; these samples were also positive for MAGE-A3 in the microarray assay. Therefore, when positive signals are observed simultaneously on the MAGE-A3 and MAGE-A6 probes, the expression of MAGE-A3 must be confirmed by PCR with specific primers. The microarray assay detected MAGE-A5 in eight samples in which expression of MAGE-A5 was not detected by PCR with specific primers. The sequence of the MAGE-A5 capture probe shares at most 85% identity with the other MAGE-A sequences (Table 2). It was therefore unlikely that positive MAGE-A5 signals on the microarrays resulted from cross-hybridizations. A more probable explanation for this discrepancy was a lower detection limit of the hybridization assay for MAGE-A5 amplicons. In line with this explanation, three of the eight samples positive for MAGE-A5 in the arrays but negative by PCR with specific primers were found positive when the PCR was carried out for 35 cycles instead of 32 (Table 4). For the other MAGE-A genes, discrepancies between the results of the microarray assay and those of the comparison method were observed only in samples showing signals close to the detection limit with one or the other method. In 10 of 31 cases, the microarray assay failed to detect low MAGE-A expression (i.e., 10% of the expression found in the reference cell line, as assessed by PCR with specific primers; Table 4). Conversely, in four cases, low expression of MAGE-A2, -A9, or -A12 was detected with the microarray assay but not by PCR with specific primers. These discrepancies seem to be independent of the origin of the tested sample (tumor cell line or tissue). Six tissue samples from healthy adults were analyzed. In agreement with previous reports (2, 10), each MAGE-A gene except the MAGE-A7 pseudogene was found to be expressed in testis, and none of these genes was expressed in the other healthy tissues tested (Table 4). Discussion The assay presented here can be considered an easy screening test to identify the MAGE-A genes expressed in a tumor sample to determine which MAGE antigens could be used for immunization. The test is very fast and avoids the use of hazardous compounds such as ethidium bromide. Because a single capture probe is used for each MAGE-A gene, the cost of the assay is reduced and the interpretation of the data is straightforward, unlike highdensity microarrays, which rely on a pattern of hybridization to identify one target (28). The whole assay, including PCR and hybridization on a microarray, was repeated four times for each of 32 cdna samples from different tumor cell lines, tumor tissues, and healthy tissues to ascertain its reproducibility. A good correlation was found with the preexisting, well-established method of detection, which used one pair of primers for each MAGE-A sequence. The capture probes for discrimination between the 12 MAGE-A cdnas were chosen to have the same length (27 bases) and the greatest possible number of mismatches between the sequence of each MAGE-A gene and the sequences of the other genes of the family. Hybrids of the selected probes (27 bases) and the complementary se-

8 32 Zammatteo et al.: MAGE-A DNA Microarray Table 4. Expression of MAGE-A genes determined by RT-PCR with specific primers or by RT-PCR with consensus primers followed by hybridization on DNA microarrays. MAGE-A genes A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 Tumor cell lines a and tissues b LB1751-MEL c / d / / / ( )/( ) / / / / / ( )/ ( )/ TT / / / / / e / / / / / / / LB373-MEL / / / / / f / / / ( )/ / / / AVL3-MEL / / / / / / / / / / ( )/ ( )/ MZ2-MEL.3.0 / / / / ( )/( ) / / / / / / ( )/ OC1 / / / / / / / / /( ) / ( )/ / 3445 / ( )/ / / / ( )/ / / / / / ( )/ 1330 / / / / / / / / / / / ( )/ LB1448-MEL.2 / / / ( )/( ) /( ) f / / / / / / / 3214 / ( )/ / / / / / / / / ( )/( ) / 1363 / /( ) / / / ( )/ / / / / / / LB583-MEL.6 / / / / / f / / / ( )/ / / / LB905-BLC / / / / / / / / / / ( )/ / BB49-HNC / / / / / / / / /( ) / / /( ) 1439 / / / / / / / / / / / / LB1781-MEL / / / / / / / / / / / / 2764 / / / / /( ) / / / ( )/ / / / 2724 / / / / / / / / / ( )/ / / 679 / / / / ( )/ / / / / / / / 1920 / / / / ( )/( ) / / / / / ( )/( ) / 2189 / ( )/( ) / / /( ) / / ( )/ ( )/ / ( )/( ) / 2756 / / / / /( ) / / / ( )/ / ( )/( ) ( )/ DUN / / /( ) / / / / / / / / / LB265-MEL / / / / / / / / / / / / LB1828-RCC / / / / / / / / / / / / BB90-MEL / / / / / / / / / / / / Healthy tissues Liver / / / / / / / / / / / / Brain / / / / / / / / / / / / Lung / / / / / / / / / / / / Colon / / / / / / / / / / / / Skeletal muscle / / / / / / / / / / / / Testis / / / / / / / / / / / / a Tumor cell lines: MEL, melanoma; BLC, bladder carcinoma; HNC, head and neck carcinoma; RCC, renal cell carcinoma; TT, medullary thyroid carcinoma; OC1, small cell lung carcinoma; DUN, pleural mesothelioma. b Tumor tissue samples: melanomas (679, 1920, and 2724), non-small-cell lung carcinomas (2189, 2756, and 2764), colorectal carcinomas (1330 and 3214), squamous cell head and neck carcinomas (1363 and 1439), and hepatocarcinoma (3445). c Expression of MAGE-A genes determined by semiquantitative RT-PCR with specific primers. Gene expression was scored according to band intensity of PCR products fractionated in agarose gels:, 10% to 100%; ( ), 10%;, no gene expression detected. These symbols represent expression as a percentage of the expression found in the reference cell lines, which were MZ2-MEL.3.0 for MAGE-A1, -A2, -A3, -A5, -A6, and -A10; LB23-SARC for MAGE-A4; LB373-MEL for MAGE-A9, -A11, and -A12; and TT for MAGE-A8. d Expression of MAGE-A genes determined by RT-PCR with consensus primers and hybridization on DNA microarrays. Experiments were performed in quadruplicate. Gene expression was scored according to the number of positive microarrays as described in the legend for Fig. 3:, strong or moderate; ( ), weak;, very weak, or no expression detected. e Discrepancies / are framed. f Sample was found to express MAGE-A5 when PCR with specific primers was carried out for 35 cycles instead of 32. quence are quite stable because their melting temperatures (T m s) are between 72 and 87 C. These differences in T m do not greatly influence the yield of the hybridization performed at 65 C. This temperature is within the limits of 5 25 C below the T m s of the various probes, thus ensuring maximal hybridization rates (29). The probes are fixed on the substrate by their ends through covalent bonds that are compatible with the high salt and temperature conditions used in the hybridization step; fixation by their ends allows better control of the length and sequence of the capture probe available for hybridization (30). Probes of identical lengths have similar access to

9 Clinical Chemistry 48, No. 1, their complementary sequences fixed on the substrate. Because the probes are present in a large excess in comparison with the target molecules, reassociation kinetics are largely determined by the concentration of target molecules. These pseudo-first-order reaction conditions greatly improve quantitative evaluation by minimizing the effects of minor differences in capture probe concentration (30). Although selected in regions that are highly conserved in the 12 MAGE-A sequences, the consensus primers were not able to amplify the various MAGE-A cdnas with the same efficiency. It was necessary to add three specific primers to ensure the efficient amplification of each of the 12 MAGE-A sequences. The microarray assay could detect MAGE-A5, which is expressed in very small amounts in a tumor cell line such as MZ2-MEL, where the relative abundance of MAGE-A5 mrna is as low as 1/ (11). For detection of the other MAGE-A genes, the microarray assay appeared to have higher detection limits than the comparison method. The latter allowed the detection of expression corresponding to 1% of the expression found in the reference cell line (data not shown), whereas for some samples the microarray assay failed to detect expression 10% of the reference cell lines (Table 4). An increased number of PCR cycles (e.g., 35 cycles instead of 30) would certainly improve the detection limit of the microarray assay, but as discussed below, this would also increase the risk of false positives resulting from amplification of contaminating genomic DNA sequences. Another possibility would be to modify the detection method by use of anti-biotin antibodies instead of streptavidin conjugates. Preliminary results indicate that this modification is likely to improve the detection limit of the microarray assay 10- to 100-fold. A 10-fold improvement should be sufficient because patients whose tumors express 3% of the amount found in the reference cell line are not included in protocols of immunization with MAGE antigens. A drawback of the microarray assay is its inability to discriminate between MAGE-A6 and MAGE-A3 sequences, which differ by only one nucleotide. One possibility to avoid this cross-hybridization would be to dilute the amplicons, but this alternative would lead to weaker signals for other MAGE-A genes (data not shown). Another possibility would be to use shorter capture probes of bases, which are very discriminating because, in that case, the presence of only one mismatch lowers the melting temperature of the probe by as much as 5 C (31). Short capture probes are used to detect single-nucleotide polymorphisms (32, 33). Another original approach for the detection of single-nucleotide polymorphisms is the insertion of nucleotide analogs known as locked nucleic acid monomers in the probe to increase the thermal stability of the perfect matches vs mismatches (34). The applicability of these approaches to improve discrimination between MAGE-A3 and -A6 needs to be investigated. The microarray assay does not distinguish between PCR products amplified from mrna and those amplified from genomic DNA because primers were chosen in the same exon of the MAGE-A genes. This may lead to false-positive results if genomic DNA sequences present in the sample as contaminants are amplified. However, because MAGE-A7 is a pseudogene, no signal should be observed with the MAGE-A7 capture probe unless genomic DNA sequences were amplified. In this study, no signal for that capture probe was observed in any sample (Table 4). To confirm the absence of DNA contamination, a mock RT-PCR was performed on each of the 32 RNA samples (omitting the reverse transcriptase in the reverse transcription step), and PCR products were hybridized on microarrays carrying the 12 MAGE-A capture probes. No signal was detected in any sample (data not shown). In conclusion, we have shown that low-density microarrays can be a good discriminating tool for detecting homologous sequences amplified by consensus primers. The microarray assay described here should facilitate tumor diagnosis related to therapeutic vaccinations involving MAGE-A gene products. This approach is also suitable for sensitive identification of different Staphylococcus species (35). We thank Madeleine Swinarska, Muriel Art, Nathalie Chandelier, and Anne-France Dabee for technical assistance. This work was supported financially by the Région Wallonne (Belgium). References 1. Van den Eynde B, van der Bruggen P. T cell-defined tumor antigens. Curr Opin Immunol 1997;9: De Plaen E, Arden K, Traversari C, Gaforio JJ, Szikora J-P, De Smet C, et al. Structure, chromosomal localization and expression of twelve genes of the MAGE family. Immunogenetics 1994;40: Lucas S, De Smet C, Arden KC, Viars CS, Lethé B, Lurquin C, et al. Identification of a new MAGE gene with tumor-specific expression by representational difference analysis. Cancer Res 1998;58: Lucas S, Brasseur F, Boon T. A new MAGE gene with ubiquitous expression does not code for known MAGE antigens recognized by T cells. Cancer Res 1999;59: Lucas S, De Plaen E, Boon T. MAGE-B5, MAGE-B6, MAGE-C2 and MAGE-C3: four new members of the MAGE family with tumorspecific expression. Int J Cancer 2000;87: Lurquin C, De Smet C, Brasseur F, Muscatelli F, Martelange V, De Plaen E, et al. Two members of the human MAGEB gene family located in Xp.21.3 are expressed in tumors of various histological origins. Genomics 1997;46: Muscatelli F, Walker AP, De Plaen E, Stafford AN, Monaco AP. Isolation and characterization of a new MAGE gene family in the Xp21.3 region. Proc Natl Acad Sci USA1995;92: Pold M, Zhou J, Chen GL, Hall JM, Vescio RA, Berenson JR. Identification of a new, unorthodox member of the MAGE gene family. Genomics 1999;59: Rogner UC, Wilke K, Steck E, Korn B, Poustka A. The melanoma

10 34 Zammatteo et al.: MAGE-A DNA Microarray antigen gene (MAGE) family is clustered in the chromosomal band Xq28. Genomics 1995;29: Boon T, Van den Eynde B. Cancer vaccines: cancer antigens, shared tumor-specific antigens. In: Rosenberg SA, ed. Principles and practice of the biologic therapy of cancer, 3rd ed. Philadelphia: Lippincott Williams & Wilkins, 2000: Serrano A, Lethé B, Delroisse J-M, Lurquin C, De Plaen E, Brasseur F, et al. Quantitative evaluation of the expression of MAGE genes in tumors by limiting dilution of cdna libraries. Int J Cancer 1999;83: Chaux P, Vantomme V, Stroobant V., Thielemans K., Corthals J, Luiten R, et al. Identification of MAGE-3 epitopes presented by HLA-DR molecules to CD4 T lymphocytes. J Exp Med 1999;189: Chaux P, Lethé B, Van Snick J, Corthals J, Schultz ES, Cambiaso CL, et al. A MAGE-1 peptide recognized on HLA-DR15 by CD4 T cells. Eur J Immunol 2001;31: Manici S, Sturniolo T, Imro MA, Hammer J, Sinigaglia F, Noppen C, et al. Melanoma cells present a MAGE-3 epitope to CD4( ) cytotoxic T cells in association with histocompatibility leukocyte antigen DR11. J Exp Med 1999;189: Schultz ES, Lethé B, Cambiaso CL, Van Snick J, Chaux P, Corthals J, et al. A MAGE-A3 peptide presented by HLA-DP4 is recognized on tumor cells by CD4 cytolytic T lymphocytes. Cancer Res 2000;60: Haas GG Jr, D Cruz OJ, De Bault LE. Distribution of human leukocyte antigen-abc and -D/DR antigens in the unfixed human testis. Am J Reprod Immunol Microbiol 1988;18: Takahashi K, Shichijo S, Noguchi M, Hirohata M, Itoh K. Identification of MAGE-1 and MAGE-4 proteins in spermatogonia and primary spermatocytes of testis. Cancer Res 1995;55: Marchand M, van Baren N, Weynants P, Brichard V, Dréno B, Tessier M-H, et al. Tumor regressions observed in patients with metastatic melanoma treated with an antigenic peptide encoded by gene MAGE-3 and presented by HLA-A1. Int J Cancer 1999;80: Miyashiro I, Kuo C, Huynh K, Iida A, Morton D, Bilchik A, et al. Molecular strategy for detecting metastatic cancers with use of multiple tumor-specific MAGE-A genes. Clin Chem 2001;47: van der Bruggen P, Traversari C, Chomez P, Lurquin C, De Plaen E, Van den Eynde B, et al. A gene encoding an antigen recognized by cytolytic T lymphocytes on a human melanoma. Science 1991; 254: De Smet C, Lurquin C, van der Bruggen P, De Plaen E, Brasseur F, Boon T. Sequence and expression pattern of the human MAGE2 gene. Immunogenetics 1994;39: Gaugler B, Van den Eynde B, van der Bruggen P, Romero P, Gaforio JJ, De Plaen E, et al. Human gene MAGE-3 codes for an antigen recognized on a melanoma by autologous cytolytic T lymphocytes. J Exp Med 1994;179: De Plaen E, Naerhuyzen B, De Smet C, Szikora J-P, Boon T. Alternative promoters of gene MAGE4a. Genomics 1997;40: Davis LG, Dibner MD, Battey JF. Guanidine isothiocyanate preparation of total RNA. In: Basic methods in molecular biology. New York: Elsevier, 1986: van Baren N, Brasseur F, Godelaine D, Hames G, Ferrant A, Lehmann F, et al. Genes encoding tumor-specific antigens are expressed in human myeloma cells. Blood 1999;94: Longo MC, Berninger MS, Hartley JL. Use of uracil DNA glycosylase to control carry-over contamination in polymerase chain reactions. Gene 1990;93: Alexandre I, Hamels S, Dufour S, Collet J, Zammatteo N, De Longueville F, et al. Colorimetric silver detection of DNA microarrays. Anal Biochem 2001;295: Lipshutz RJ, Fodor SP, Gingeras TR, Lockhart DJ. High density synthetic oligonucleotide arrays. Nat Genet 1999;21: Cox KH, DeLeon DV, Angerer LM, Angerer RC. Detection of mrnas in sea urchin embryos by in situ hybridization using asymmetric RNA probes. Dev Biol 1984;101: Zammatteo N, Jeanmart L, Hamels S, Courtois S, Louette P, Hevesi L, et al. Comparison between different strategies of covalent attachment of DNA to glass surfaces to build DNA microarrays. Anal Biochem 2000;280: Conner BJ, Reyes AA, Morin C, Itakura K, Teplitz RL, Wallace RB. Detection of sickle cell S-globin allele by hybridization with synthetic oligonucleotides. Proc Natl Acad Sci U S A 1983;80: Gilles PN, Wu DJ, Foster CB, Dillon PJ, Chanock SJ. Single nucleotide polymorphic discrimination by an electronic dot blot assay on semiconductor microchips. Nat Biotechnol 1999;17: Kozal MJ, Shah N, Shen N, Yang R, Fucini R, Merigan TC, et al. Extensive polymorphisms observed in HIV-1 clade B protease gene using high-density oligonucleotide arrays. Nat Med 1996;2: Orum H, Jakobsen MH, Koch T, Vuust J, Borre MB. Detection of the factor V Leiden mutation by direct allele-specific hybridization of PCR amplicons to photoimmobilized locked nucleic acids. Clin Chem 1999;45: Hamels S, Gala J-L, Dufour S, Vannuffel P, Zammatteo N, Leclerc G, et al. DNA for simultaneous diagnosis of genus Staphylococcus, staphylococcal species and methicillin-resistance. Biotechniques, in press.