Clonal and Parallel Evolution of Primary Lung Cancers and Their Metastases Revealed by Molecular Dissection of Cancer Cells

Metastasis is a principal event leading to death in individuals with cancer. However, the molecular basis of metastasis is still unclear. A generally accepted model for tumor progression is the “clonal evolution” model. This model is well illustrated in colorectal carcinogenesis (1) and holds that more malignant cells with additional genetic alterations predominate in a tumor cell population. In this model, metastasis represents the end stage of evolution, and the presence of genetic alterations responsible for the metastatic ability of tumor cells is predicted. The finding that primary tumor cell populations are comprised of cells with different metastatic abilities supports this model (2). If tumor cells with such genetic alterations consist of a small subpopulation among the primary tumor cells, these alterations can be detected only in metastatic tumors but not (or hardly) detected in the corresponding primary tumor (3–5). In fact, we and others identified several genetic alterations that were detected only in metastatic tumors but not in the corresponding primary tumors (6–12), supporting the authenticity of this model.

Recently, other models for tumor progression and metastasis were also proposed. One is “the parallel evolution model,” which proposes early occurrence of metastasis and parallel evolution of primary and metastatic tumors (13). This was based on the report of Schmidt-Kittler et al. that genetic alterations in breast cancer cells that disseminated into the bone marrow of patients generally do not resemble those in the corresponding primary tumors (14). The parallel evolution model has been very applicable to the metastatic pattern of several solid epithelial tumors (15–17). In addition, information provided by global gene expression profiling of primary tumor cells led to the proposal of another model, “the same-gene model” by Bernards and Weinberg (18). This model holds that genetic alterations acquired early in carcinogenesis confer not only a selected replicative advantage but also a proclivity to metastasize on cancer cells. This was based on a report of van't Veer et al. that prognosis of patients with breast cancer can be predicted by gene expression profiles of primary tumors (19). Ramaswamy et al. then showed that a subset of primary tumors resembled metastatic tumors with respect to gene expression signature (20). In this model, it was hypothesized that there are no genes and genetic changes specifically and exclusively involved in orchestrating the process of metastasis.

In the present study, we investigated the authenticity of the tumor progression models above by comparison of primary and corresponding metastatic tumors obtained from eight patients with lung cancer. We focused on genetic alterations rather than expression profiles because genetic alterations are stable and irreversible and, thus, can be used as a “molecular footprint” of cancer progression (4). There have been several studies using this strategy (6, 7, 9–12), including the one by Schmidt-Kittler et al. as described above (14). However, in most of these studies, only a limited number of genetic loci and/or genes were examined. Recent progress in array technology has enabled us to accomplish not only high-resolution analysis of expression status but also that of genetic status. Thus, in this study, we used a high-resolution single nucleotide polymorphism (SNP) array, mapping 10k, which covers 11,560 loci throughout the whole human genome in 210-kb mean intervals, for the allelic imbalance scanning to obtain comprehensively information on allelic status. Furthermore, to obtain more precise data, the laser capture microdissection method was used to enrich cancer cell components in five non–small cell cancer (NSCLC) cases because various fractions of noncancerous cells are often contaminated in macrodissected NSCLC samples. In addition to allelic imbalance scanning, mutation analysis of the p53, EGFR, and KRAS genes was done because these genes are frequently mutated in lung cancer.

Patients and tissues. Seven primary lung tumors, their 12 corresponding metastases (seven brain, one liver, one pleural, and three lymph node), and seven corresponding normal lung tissues were obtained at surgery or autopsy from seven patients who were treated at the National Cancer Center Hospital, Tokyo, Japan. One primary lung tumor, two corresponding metastases (one pulmonary and one liver), and a corresponding normal lung tissue were also obtained by autopsy at Tokushima University Hospital, Tokushima, Japan (case 7 in Table 1). In total, 30 DNA samples from the eight cases were subjected to DNA array analysis (Table 1). These cases were histologically classified as three adenocarcinomas, one squamous cell carcinoma, one large cell carcinoma, and three small cell carcinomas (SCLC) according to the WHO criteria (21). In five NSCLC cases (cases 1-5), normal lung tissues were obtained at surgery from regions >5 cm distance from tumors and showing macroscopically normal morphology in the resected lobes of the lung. Tumors and normal lung tissues were then separately fixed with methanol and embedded in paraffin. Cancerous and noncancerous cells of these five cases were obtained by the laser capture microdissection method using the Pixcell Laser Capture Microdissection system (Arcturus Engineering, Mountain View, CA). Noncancerous cells were isolated from a normal lung tissue section but not from a region surrounding the tumor of a tumor tissue section under microscopic observation. Genomic DNAs were extracted as described previously (22). In an SCLC case (case 6), normal lung tissue was also obtained at surgery from a region >5 cm distance from the primary tumor and showing macroscopically normal morphology in the resected lobe. In two other SCLC cases (cases 7 and 8), normal lung tissues were obtained at autopsy from lobes without tumors macroscopically. In these three SCLC cases, tumors and normal lung tissues were stored at −80°C without fixation until DNA extraction. Cancerous and noncancerous cells of the three SCLC cases (cases 6-8) were macrodissected, and genomic DNAs were prepared as described previously (23, 24). This study was undertaken according to the guidelines for medical research in Japan. All the samples were analyzed after declaring anonymity to keep the privacy of individuals.

SNP array analysis. High-resolution SNP array, mapping array 10k (Affymetrix, Santa Clara, CA), was used according to the manufacturer's protocol with some modifications as described below. In the original protocol, PCR of 35 cycles is undertaken against 250 ng DNA to obtain 20 μg of whole-genome amplicon. However, due to the small number of cells obtained by laser capture microdissection, 10 to 50 ng DNAs were subjected to PCR of 35 to 45 cycles to obtain 20 μg of the amplicon. Twenty micrograms of the amplicon were purified, labeled, and hybridized to the array, and genotype calls were obtained as described previously (25).

The accuracy of the genotype calls obtained by the modified protocol was estimated as follows. A noncancerous lung tissue of another large cell carcinoma patient was subjected to macrodissection and laser capture microdissection, and genomic DNAs were extracted from both materials. Two hundred fifty nanograms of DNA from the macrodissected material were subjected to DNA array analysis according to the standard protocol (i.e., 35 cycles of PCR). Five or 50 ng of DNA from the laser capture microdissection material were subjected to DNA array analysis according to the modified protocol (i.e., 35, 40, and 45 cycles of PCR). The concordance of genotype calls among these preparations was >99%.

Statistical analyses for allelic imbalance in primary and metastatic tumors. The fraction of allelic imbalance for each tumor sample was calculated as the fraction of SNP probes for which noncancerous cell DNA was called as heterozygous and cancer cell DNA was called as homozygous. The fraction of error calls for each tumor sample was calculated as the fraction of SNP probes for which noncancerous cell DNA was called as homozygous and for which cancer cell DNA was called as heterozygous. The fraction of different calls between primary and metastatic tumors was calculated as the fraction of SNP probes for which primary tumor cell DNA was called as homozygous and heterozygous and for which the metastatic tumor cell DNA was called as heterozygous and homozygous, respectively. The statistical significance for the excess of fraction of allelic imbalance in primary and metastatic tumors and of different calls between primary and metastatic tumors over the fraction of error calls was calculated by the χ² test. A level of P < 0.05 was considered statistically significant.

Definition of regions of allelic imbalance. When a locus was called “homozygous” in tumor DNA and “heterozygous” in the corresponding normal tissue DNA, such a locus was judged as being an “allelic imbalance” in the tumor. On the other hand, when a locus was called “heterozygous” both in the tumor and corresponding normal tissue DNA, such a locus was judged as being “not allelic imbalance” in the tumor. By taking the call error in the SNP array analysis into account, regions containing at least six consecutive “allelic imbalance” (or “not allelic imbalance”) loci were defined as the ones of allelic imbalance (or not allelic imbalance). If the region of allelic imbalance in a primary tumor overlapped that in the corresponding metastatic tumor(s) and the overlapping region contained more than six consecutive allelic imbalance loci both in the primary and metastatic tumors, such a region was judged as a common region of allelic imbalance. In contrast, if a region judged as allelic imbalance in a primary tumor (or metastatic tumors) contained more than six consecutive “not allelic imbalance” loci in metastatic tumors (or primary tumor), such a region was judged as a unique region of allelic imbalance in either primary or metastatic tumor. If allelic imbalance at the same region was due to a gain or loss of different alleles between primary and metastatic tumors, such a region was also judged as being a unique region of allelic imbalance in respective primary and metastatic tumors. If there was a common region between primary and metastatic tumors, but the region in the metastatic tumor was wider (or narrower) than that in the corresponding primary tumor, allelic imbalance in the metastatic tumor was judged to have occurred independently of allelic imbalance in the primary tumor.

Microsatellite analysis. Microsatellite markers were chosen based on the chromosomal locations mapped in the human genome-wide screening set version 9 (Research Genetics, Inc., Huntsville, AL) or the Japan Biological Information Research Center genome database (http://www.jbirc.aist.go.jp/gdbs/). Two hundred picograms to 1 ng DNA was used for PCR of 40 cycles with a set of primers labeled with FAM or TET. PCR products were run through an ABI Prism 310 DNA Sequencer (Applied Biosystems, Foster City, CA) and analyzed by the ABI PRISM GeneScan and Genotyper software. A reduction >75% of an allele in tumor was determined as allelic imbalance.

Mutation analysis of the p53, EGFR, and KRAS genes. All eight cases subjected to the SNP array analysis were examined for mutations in exons 4 to 8 of the p53 gene. Three adenocarcinoma cases were previously examined for mutations in exons 1 to 2 of the KRAS gene and exons 18 to 21 of the EGFR gene (26). Two hundred picograms to 1 ng of DNA was subjected to PCR amplification followed by sequencing as described previously (27).

Detection of allelic imbalance in primary and metastatic tumors. Eight primary lung tumors, eight corresponding normal lung tissues, and 14 metastases were subjected to the SNP array analysis (Table 1). Genotype calls were obtained in 81.9% to 96.4% (average = 89.7%) of the 11,560 SNP sites on the array; 1,439 to 3,001 loci were informative (i.e., heterozygous in noncancerous tissues) for detection of allelic imbalance in the tumors, and fractions of allelic imbalance ranged from 19.8% to 73.3% (Table 1). Fractions of error calls for the tumors were estimated as being 0.3% to 26.3% (see Materials and Methods; Table 1) and were significantly lower than the fractions of allelic imbalance (P < 0.05). Therefore, it was indicated that all tumors analyzed had allelic imbalances.

Fractions of different calls between primary and metastatic tumors ranged from 2.1% to 25.2% (Table 1) and were much smaller than fractions of allelic imbalance in primary tumors in all cases. Therefore, it was indicated that the majority of genetic alterations are common between primary and metastatic tumors. However, in 8 of the 14 metastatic tumors, fractions of different calls against primary tumors were significantly higher than fractions of error calls (P < 0.05), indicating the presence of allelic imbalances that differentially occurred between primary and metastatic tumors. In the remaining six metastatic tumors, fractions of different calls were not significantly higher but lower than fractions of error calls. Thus, the allelic status of these metastatic tumors might not have been different from that of respective primary tumors. Alternatively, differences in the allelic status between them might have been masked by the error calls.

Common genetic alterations among eight cases. Regions of allelic imbalance were next searched for along the genomes of the 8 primary and 14 metastatic tumors. Fractions of error calls in the present SNP array analysis were estimated as being up to 26.3%. Therefore, by taking the call error into account, regions containing at least six consecutive allelic imbalance loci were defined as allelic imbalance regions because the appearance of such loci by the call error was <1 even for case 4 for which the highest probability of call error (26.3%) was inferred at a SNP locus [i.e., 1,849 informative loci × (0.263)⁶ = 0.61]. The number of allelic imbalance regions defined by this criterion ranged from 5 to 20 among 22 tumors analyzed. This was not the same between primary and metastatic tumors in six of the eight cases (cases 1, 2, and 5-8) and was the same in the remaining two cases (cases 3 and 4). The number of allelic imbalance regions in metastases was higher than that in primary tumors in five of the six cases (cases 1, 2, 5, 6, and 7), whereas it was lower in the remaining one case (case 8). These results indicate the presence of allelic imbalance only in either metastases or primary tumors. Chromosomes showing allelic imbalance and p53/EGFR mutations in each tumor are indicated in gray in Table 2. There were various regions showing allelic imbalance commonly both in primary and metastatic tumors in all eight cases, and such regions were distributed among all chromosomes. p53 mutations were detected in seven of the eight cases, whereas EGFR mutations were detected in all three adenocarcinoma cases (cases 1-3). No KRAS mutations were detected in them. All the p53 and EGFR mutations were detected in primary tumors, and the same types of mutations were detected in their corresponding metastatic tumors. The chromosome 17p11-p13 region was the most common region of allelic imbalance detected both in primary and metastatic tumors (8 of 8, 100%). Allelic imbalances of chromosome regions 3p11-p26, 10q23-q26, and 13q12-q31 were the next common ones (7 of 8, 88%). All allelic imbalances defined as common were gains or losses of the same allele between primary and metastatic tumors. The allelic status of several loci that showed allelic imbalance in tumors by SNP array analysis was confirmed by microsatellite analysis. All the loci examined showed allelic imbalance as indicated by the SNP array analysis (data not shown).

In these eight cases, there were also various chromosomal regions showing allelic imbalance only in metastases but not in primary tumors (indicated in red in Table 2) or those only in primary tumors but not in metastases (indicated in blue in Table 2). Seven of the eight cases, except case 3, had chromosomal regions with allelic imbalance only in metastases, and four of the eight cases (cases 2, 5, 6, and 8) had regions with allelic imbalance only in primary tumors. Such regions were widely distributed among diverse chromosomes among the cases, but several regions were common in multiple cases. Chromosomal regions showing allelic imbalance only in primary tumors or in metastases are listed in Table 3. Allelic imbalance of chromosomal regions 11p15 and 11p11-p13 was detected only in metastases in four of the eight cases (cases 2, 5, 6, and 7). These regions were the most common ones showing allelic imbalance only in metastases in this study. 11q11-q13 was the next common region (three cases), and 4q11-q12, 5p15, 12q11-q22, and 18p11 showed allelic imbalance only in metastases in two cases, respectively. On the other hand, allelic imbalance of the 7p14-p22 region was observed only in primary tumors but not in metastases in two cases (cases 2 and 5). Allelic imbalance of other chromosomal regions detected only in primary tumors or metastases was observed in a single case.

Similarity of genetic alterations between primary and metastatic tumors. The numbers of regions with allelic imbalance detected in each tumor are summarized in Tables 1 and 2. In case 1, 13 regions were defined as allelic imbalance in the primary tumor, and 14 regions were defined as allelic imbalance in the corresponding metastatic tumor. In this case, 13 regions were the same between primary and metastatic tumors (13 of 14, 93%), and one region on chromosome 1 (1p31-p35) showed allelic imbalance only in the metastatic tumor (1 of 14, 7%). The metastatic tumor of case 4 displayed 14 regions with allelic imbalance, and all of these regions were common between primary and metastatic tumors (14 of 14, 100%). However, among them, three regions were largely extended in the metastatic tumor (number of extended region is indicated in parenthesis in Table 1). Fractions of common allelic imbalance regions (number of common allelic imbalance regions / number of allelic imbalance regions) in metastatic tumors were 13 of 16 (81%) in case 5 and 8 of 12 (67%) in case 6. Likewise, fractions of common allelic imbalance regions in multiple metastatic tumors were 17 of 19 (89%) and 18 of 20 (90%) in case 2, 13 of 14 (93%) and 13 of 16 (81%) in case 7, and 9 of 9 (100%), 8 of 8 (100%), 4 of 5 (80%), and 6 of 7 (86%) in case 8. Interestingly, regions of allelic imbalance in case 3 were completely the same among the primary tumor and two metastatic tumors. These results further indicated that the majority of genetic alterations are common between primary and metastatic tumors in lung cancer. Furthermore, a considerable fraction of allelic imbalance (15 of 38, 40%) detected only in primary tumor or metastasis was caused by extension of allelic imbalance regions in paired tumors (indicated as extended allelic imbalance in Table 1).

Natural history of cancer progression deduced from the genotype differences between primary and metastatic tumors. Single metastatic tumors were available in four cases (cases 1, 4, 5, and 6; Fig. 1A). Our previous study showed a PTEN mutation only in the metastatic tumor but not in the primary tumor of case 6 (28). In cases 1 and 4, all genetic alterations detected in primary tumors were also detected in the respective metastatic tumors, and the metastatic tumors had additional allelic imbalances that were not observed in the primary tumors. On the other hand, in cases 5 and 6, there were regions with allelic imbalance that were detected only in primary tumors, and there were also regions with allelic imbalance that were observed only in metastatic tumors. Allelic imbalance of chromosomes 15 and 22 detected in case 6 were a loss or gain of different alleles between primary and metastatic tumors.

Two metastases were available in three cases (cases 2, 3, and 7; Fig. 1B). In case 3, no differences in the status of allelic imbalance were detected among the primary tumor and two metastases. On the other hand, two metastases in case 7 carried all allelic imbalances detected in the primary tumor and had additional allelic imbalances that were not detected in the primary tumor. Interestingly, allelic imbalances detected only in two metastases in case 7 did not overlap each other. In case 2, there were regions with allelic imbalances that were detected only in the primary tumor, in addition to regions with allelic imbalances that were observed only in metastatic tumors. Furthermore, two brain metastases had common regions of allelic imbalance in addition to unique regions of allelic imbalance in one of two metastases.

Four metastases were available in case 8 (Fig. 1C). The primary tumor of this case was previously shown to be composed of two different areas, and the N-myc gene was heterogeneously amplified in these areas (29). Genetic alterations detected in all tumors were a p53 mutation and allelic imbalances of chromosomes 3p, 11p, 13q, 17p, and 17q. Primary tumor, liver metastasis, and paraaortic lymph node metastasis had unique genetic alterations, whereas all genetic alterations detected in pleural metastasis and hilar lymph node metastasis were also detected in the primary tumor. Because common genetic alterations in the primary tumor were fewer in liver and paraaortic lymph node metastases than in pleural and hilar lymph node metastases, it is likely that metastases to the liver and paraaortic lymph node occurred earlier than those to pleural and hilar lymph nodes.

Common regions of allelic imbalance on 11p in lung cancer. Frequent occurrence of allelic imbalance on chromosome 11p in metastatic tumors prompted us to examine the allelic status of this chromosome arm in primary and metastatic tumors by microsatellite analysis (Fig. 2). Two metastatic brain tumors of case 2 showed allelic imbalance at the D11S0814i (11p15) and D11S0586i (11p13) loci, whereas the corresponding primary tumors did not show allelic imbalance at these loci. Because the D11S0149i, D11S0368i, and D11S4101 loci between the D11S0814i and D11S0586i loci did not show allelic imbalance in these metastases, the regions of allelic imbalance accumulated only in metastases of case 2 were determined as two different ones on chromosome 11p (Figs. 2 and 3). Cases 5, 6, and 7 also showed allelic imbalance of chromosome 11p only in metastases, and allelic imbalance was observed at all loci examined in these metastases, suggesting the occurrence of whole chromosome arm deletions only in metastases. Three other cases showed allelic imbalance at 11p15 both in primary and metastatic tumors (cases 1, 3, and 8). In two of them (cases 1 and 3), allelic imbalance was extended to the 11p11-p13 region both in primary and metastatic tumors. Thus, the 11p15 and 11p11-p13 regions were confirmed as being common for allelic imbalance in metastases. As indicated in Figs. 2 and 3, the result of the SNP array analysis was concordant with the result of microsatellite analysis.

The present study indicates a high similarity of genetic alterations between primary and metastatic tumors. That is, metastatic tumors carried the majority of the genetic alterations present in the corresponding primary tumors. All the p53 and EGFR mutations were detected in primary tumors and were retained in their corresponding metastatic tumors. This indicates that metastases have occurred at a late stage of lung cancer progression. The fact that genetic alterations in primary and metastatic tumors resembled each other went along well with previous results that expression profiles of metastatic tumors were similar to those of the corresponding primary tumors (19, 20). However, in seven of the eight cases, there were allelic imbalances detected only in metastases but not in primary tumors. To explain this result, we should consider two possibilities. The first possibility is that there were few cells carrying these additional genetic alterations in primary tumors; thus, we could not detect these alterations in the analysis of primary tumors (3–5). However, those few cells with these alterations selectively metastasized to distant organs or lymph nodes; thus, we were able to detect them in metastases. If this assumption is correct, the results imply that cancer cells in primary tumors are heterogeneous for accumulated genetic alterations. Because the analysis in this study looked at primary tumors in aggregate, we could not detect the alterations present in a few cells. Thus, taking the heterogeneity of cancer cells in primary tumors into consideration, the results of this study are consistent with those of a previous study (2) and match the clonal evolution model (1). From the viewpoint of this model, genetic alterations detected only in metastases might be responsible for controlling metastatic ability of cancer cells, such as detachment, invasion, survival in circulation, attachment extravasation, proliferation, induction of neovasculature, and evasion of host defenses (5, 30). The other possibility is that these additional alterations have occurred after metastasis, and these alterations conferred the growth advantage on the cells in metastatic sites. This is in agreement with the same-gene model in which genetic alterations specifically involved in metastasis do not exist (18). The results of case 3, in which all tumors showed completely the same genetic alterations, also match this model. The results can also explain the result of previous reports that analysis of primary lung and breast tumors can predict metastasis (19, 20). From the viewpoint of this model, additional alterations might confer some growth advantage on the cells in a metastatic site but not confer metastatic ability on the cells in the primary site.

In the present study, four of the eight cases showed the genetic alterations detected only in primary tumors but not in corresponding metastases. This is consistent with the concept of parallel evolution because cancer cells in primary tumors in the four cases, as well as breast cancer cases reported by Schmidt-Kittler et al., have acquired additional genetic alterations after the occurrence of metastasis (13, 14). Genetic alterations specific in primary tumors were detected by the analysis in primary tumors in aggregate; thus, cells with these alterations might comprise the majority of cells in primary sites, suggesting that these alterations conferred the cells in primary sites some growth advantage.

In a case of SCLC (case 8), we were able to analyze both lymph node metastases and distant metastases. The result indicates that liver metastasis occurred before lymph node metastasis at an early stage in SCLC progression. Thus, as indicated in breast cancer progression, it is possible that SCLC cells bypass the lymph nodes and disseminate directly through the blood to distant organs. Accordingly, the cascade model with hematogenous dissemination can be applicable to the process of SCLC progression (15). Because SCLC is the most aggressive type of lung cancer with early and wide dissemination, such aggressiveness would be well explained by this model.

As described above, the process of metastasis in each lung cancer is diverse among cases. All three progression models were applicable in lung tumor progression. Genetic alterations detected specific to primary tumors or metastases might have some biological significance, such as growth advantage and/or metastatic ability. Thus, allelic imbalances at chromosome 11p15 and 11p11-p13 regions, the most frequent alterations detected only in metastases, should be further analyzed in association with the biological behavior of lung cancer cells. According to the clonal and parallel evolution models, a metastasis suppressor gene(s) was predicted to be present in these chromosomal regions. It was previously reported that complementation with chromosome 11 induced growth inhibition in lung cancer cells (31, 32). The possible involvement of allelic imbalance on 11p in the progression of lung cancer was also indicated by the loss of heterozygosity analysis (33, 34). The 11p15 region contains a candidate tumor suppressor gene p57^KIP2, and the 11p11-p13 region contains a candidate metastasis suppressor gene KAI1. p57^KIP2 is a member of the cyclin-dependent kinase inhibitor family, and overexpression of p57^KIP2 causes a cell cycle arrest of SAOS-2 and mink lung epithelial cells in G₁ phase (35, 36). p57^KIP2 is known to be imprinted, and the occurrence of selective loss of the expressed allele for the p57^KIP2 gene in lung cancer cells has been reported (37). KAI1 encodes a protein with an ability to suppress metastasis of several types of cancer cells (38, 39). In fact, KAI1 protein expression was reported as being preferentially decreased in metastatic lung tumors rather than primary tumors (40). Because RNA and protein of the metastatic tumors analyzed in this study were not available, we could not examine the expression of these genes. Biological analyses of these genes in relation to metastatic ability and/or growth advantage of lung cancer cells as well as genetic analyses of 11p alterations in additional sets of primary and metastatic lung tumors will be necessary to elucidate how genes on 11p are involved in lung cancer progression and metastasis.

Based on the genotype data of primary and metastatic lung tumors obtained from eight patients, the natural history of each case was deduced and candidate metastasis suppressor loci were determined. However, in this study, only a small number of paired samples were analyzed; tumors of different histologic types with and without chemotherapy were compared together, and lymph node metastases and distant metastases were compared together. Therefore, further analyses with larger sets of paired primary and metastatic tumors will give us more comprehensive information on genetic alterations involved in lung cancer progression and metastasis.

We thank Kazuhiro Nagayama, Masamitsu Sato, and Reika Iwakawa for considerable contributions to DNA array analysis and DNA sequencing.