ATBF1 and NQO1 as candidate targets for allelic loss at chromosome arm 16q in breast cancer: Absence of somatic ATBF1 mutations and no role for the C609T NQO1 polymorphism

Chromosome arm 16q is one of the regions most frequently involved in loss of heterozygosity (LOH) in breast cancer [1]. Detailed mapping shows that at least two separate regions are targeted, implying the presence of more than one tumor suppressor gene (TSG) [2]. The CDH1 gene, encoding the homotypic adhesion molecule E-cadherin and located at 16q22.1, was identified as the target of LOH at 16q, but gene-truncating mutations were identified only in the relatively infrequent lobular histological subtype [3]. CDH1 shows no mutations in tumors with LOH at 16q of the more common ductal subtype. A few candidate tumor suppressor genes have been suggested, in particular CTCF, a gene located at 16q22.1 that encodes an insulator. Mutations in CTCF have been identified in breast cancer, but only in a very small subset of cases [4, 5]. Mutations in other candidate genes at chromosome 16q could not be identified, although decreased expression has been reported for some cases, suggesting that haploinsufficiency could be a mechanism of tumorigenesis [6, 7]. Thus, the major TSGs at chromosome 16q in ductal breast cancer remain to be identified [8].

Recently, the gene for AT motif-binding factor 1 (ATBF1) was reported as a TSG at chromosome arm 16q in prostate cancer. This designation was based on the fact that ATBF1 is located in the smallest region of overlap affected by loss of heterozygosity (LOH), and mutations were detected in prostate cancer cell lines, -xenografts, and -tumors [9]. ATBF1 encodes a transcription factor with four homeobox domains and 23 Zn-fingers and was first identified as a suppressor of alpha-fetoprotein transcription [10]. ATBF1 was shown to repress expression of the c-Myb oncogene [11], and it may activate the cell cycle inhibitor p21 [12].

With the goal of identifying new candidate genes for the 16q TSG, we constructed a cDNA microarray containing all known and predicted genes at chromosome arm 16q, and we used it to screen breast tumor RNA for genes that are down-regulated in tumors with LOH at 16q compared to tumors without LOH. One of the down-regulated genes is NQO1, which encodes NAD(P)H dehydrogenase quinone 1, a phase II 2-electron reductase that detoxifies metabolites of benzene and protects the p53 tumor suppressor protein from degradation [13, 14].

In this paper, we report the identification of genes at chromosome arm 16q that are significantly down-regulated in breast tumors with LOH at chromosome 16 using a cDNA microarray enriched for genes on the long arm of this chromosome. NQO1 and ATBF1 were identified as the most promising candidate TSGs. We examined in detail a polymorphism that affects the stability of the NQO1 protein in relation to LOH at 16q, but we failed to demonstrate preferential loss of one specific allele in tumors showing LOH at 16q. We did observe an association between NQO1 protein expression and histological grade.

Furthermore, we performed mutational analysis of the entire open reading frame of the ATBF1 sequence in 48 breast tumors with known LOH status. We found the same types of sequence variations as previously reported for prostate cancer [9], but we observed that these variations are not tumor-specific and that the wild-type allele is not a specific target of LOH at 16q. This suggests that some of the previously reported mutations in prostate cancer may have been erroneously reported as pathogenic.

The frequent occurrence of loss of heterozygosity (LOH) at the long arm of chromosome 16 in breast cancer has led many investigators to search for the tumor suppressor genes that may be affected by LOH in this chromosomal region. Many genes have been proposed as putative candidates. In particular, FBXO31 was recently identified not by deletion mapping, but by identification of its ability to induce senescence in tumor cells [7]. However, none of these genes follow the classical two-hit model of Knudson [30]. In an attempt to assess all known genes at chromosome 16q simultaneously within a single experiment, we designed a chromosome 16q-specific cDNA microarray and hybridized the array with RNAs from a series of well-characterized breast tumors. By comparing tumors with and without LOH at the 16q arm, we identified a small set of genes that met our strict statistical standards.

In the present work, we were unable to stratify the groups based on histological grade or the smallest region of overlap involved in LOH, since larger cohorts of patients are needed for this analysis.

Two genes that showed lower mRNA expression in breast tumors with LOH at chromosome arm 16q were of particular interest for breast cancer and were therefore subjected to more detailed investigation. The most significantly down-regulated gene was NQO1 encoding NAD(P)H dehydrogenase quinone 1. This gene has been implicated in carcinogenesis because of its role in the reduction and detoxification of quinones and their derivatives, thereby protecting cells from oxidative damage [14]. Moreover, NQO1 has been shown to stabilize the p53 tumor suppressor protein by inhibiting its degradation through a direct protein-protein interaction [13, 31]. The possible role of NQO1 in carcinogenesis has been studied extensively, especially the occurrence of a frequent polymorphism, 609C>T, which results in a proline to serine substitution at amino acid 187 [32]. This substitution results in a variant with poor enzymatic activity and no detectable protein, as shown in individuals homozygous for the T-allele [33]. The prevalence of the 609T allele of NQO1 has been studied in tumors of lung, bladder, and colorectal cancer as well as leukaemia. However, the results are inconsistent [14] and no correlation between NQO1 allele frequency and breast cancer has been observed [32, 34]. This is consistent with our findings. To our knowledge there are no studies examining the preferential loss of heterozygosity of the active 609C allele. A possible role for NQO1 in breast cancer would be confirmed if it could be shown that LOH at 16q is targeted at the active allele. However, we could not show a prevalence of 609C loss and, in fact, LOH was higher at the 609T allele. It is therefore unlikely that NQO1 is the target tumor suppressor gene at 16q. Nevertheless, the high incidence of LOH (70%) at this locus suggests that the gene lies near another target gene.

We also investigated NQO1 protein expression in 354 breast tumors. Normal breast epithelium showed no or only weak expression of this protein, whereas 20% of breast tumors showed strong staining. This is not consistent with a tumor suppressor function for NQO1. Remarkably, intermediate and poorly differentiated tumors (grade 2 and 3) showed overexpression of NQO1 protein more often than those showing low differentiation (grade 1). We have previously shown that grade 1 tumors have LOH at 16q through physical loss, whereas poorly differentiated breast tumors show 16q LOH through mitotic recombination [35]. Therefore, the absence of NQO1 expression may be attributed to physical loss of 16q and LOH at 16q may not be specifically targeted at the NQO1 gene. In addition, the expression of HER2 and NQO1 showed significant correlation. Like NQO1, also HER2 overexpression was more prevalent in poorly differentiated breast tumors, which could explain this correlation.

ATBF1 was an other candidate gene identified by our microarray analysis showing decreased mRNA expression in tumors with LOH at 16q. This gene was recently reported to contain somatic mutations in prostate cancer, a tumor type in which frequent LOH at 16q is also a hallmark [9]. ATBF1 is one of the largest transcription factors known, containing four homeodomains and 23 zinc fingers [10]. Nevertheless, little is known about its function. The ATBF1 protein binds to the AT motif of the alpha-fetoprotein (AFP) gene, thereby inhibiting its transcription. A possible role in tumorigenesis has been found in AFP-producing gastric cancer, an aggressive tumor type that lacks ATBF1 [36]. Interestingly, ATBF1 was found to bind to Myb oncoprotein, as well as to inhibit transcription of the MYB gene [11]. Combined with the identification of somatic mutations in prostate cancer [9] these features are consistent with a tumor suppressor function for ATBF1. It has been suggested that in breast cancer, higher mRNA expression of ATBF1 is associated with a better prognosis, i.e. absence of tumor-positive lymph nodes [37]. In a previous study of 712 breast tumors we analyzed a possible association between positive lymph nodes and LOH status at 16q [8], however, we failed to demonstrate a significant correlation. Since a decrease in mRNA is associated with 16q LOH, but lymph node status is not, our current study cannot support the observation by Zhang et al. that ATBF1 expression is correlated with better prognosis in breast cancer [37].

Screening of the entire ATBF1 open reading frame of 43 breast tumor DNAs by direct sequencing revealed no somatic mutations, only previously described single nucleotide polymorphisms (SNP) and 7 seven new SNPs. All of these SNPs were also present in matched normal DNA. Although we cannot exclude that these variants are pathogenic, this seems quite unlikely because linkage of chromosome arm 16q and hereditary breast cancer has never been demonstrated [38].

We confirmed the same in-frame deletions in contiguous stretches of glutamic acid codons in exons 9 and 10 as described by Sun et al. [9] in prostate cancer. However, whereas those authors reported that these mutations were somatic, we have shown here that they were also present in matched normal DNA from peripheral blood cells, confirming these variants as germline. In addition we identified similar germline variations at two different locations in the gene that affect the amino acid repeat length of stretches of glutamine and glutamic acid residues. As in the case of the NQO1 polymorphism we could not show neither preferential LOH of these repeats nor a possible role in predisposing towards breast carcinogenesis. According to our results these repeat length variations should be considered as polymorphic variations. The effect of these variations should be further investigated using functional assays. A dramatic effect is not very likely given the fact that the variations are bidirectional: both insertions and deletions were observed in three of four variants. Recently Xu and co-workers [39] reported a 21- or 24-nucleotide deletion at position 3381 in ATBF1 in the germline of prostate cancer patients and concluded that these variants are associated with prostate cancer risk. We also found deletions at the same glutamine tract, but they were only three or six base pairs in length and they showed no significant difference in frequency between breast cancer patients and normal controls.

The large number of variations found in ATBF1 can be explained by the size of the gene, which is exceptionally large, with over 11 kb of coding sequence. The presence of simple sequence repeats coding for amino acid repeats ranging from 8 to 26 residues may explain the high frequency of length variations. Repeat length variations can lead to pathogenic traits, often neurological in character. However, these variations are almost invariably expansions that increase over time [40]. All repeats detected in ATBF1 showed the same length in tumor and normal DNA, and were in some cases even shorter than the most frequent allele (Table 3). A recently published study described the sequencing of the ATBF1 open reading frame in 32 breast cancer cell lines [41]. Two possible mutations were identified, one as an amino acid substitution at codon 2622 and one undefined change in a poly(T) tract. Whether these mutations were pathogenic was not verified.

Using a chromosome 16q-specific cDNA array to identify tumor suppressor genes we have found five candidates, of which the two most interesting, NQO1 and ATBF1, were excluded as classical candidate genes. Of special note are our results indicating that ATBF1 variants showing differences in the length of amino acid repeat sequences are most likely non-pathogenic and instead are normally occurring polymorphisms. This calls into question the conclusions of Sun et al. [9], who found these variants to be somatic mutations involved in prostate cancer.

New views on the role of loss of heterozygosity in cancer are currently of importance for their relevance to the targeting of multiple genes [42, 43] and haploinsuffiency [44]. These views promise to provide better working hypotheses for elucidating somatic cancer genetics than Knudson's two-hit model.

Primer sequences have been submitted to the RTPrimerDB: http://medgen.ugent.be/rtprimerdb/

UCSC Genome Browser: http://genome.ucsc.edu/