The ‘Pokemon’ (ZBTB7) Gene: No Evidence of Association with Sporadic Breast Cancer

It has been proposed that the excess of familiar risk associated with breast cancer could be explained by the cumulative effect of multiple weakly predisposing alleles. The transcriptional repressor FBI1, also known as Pokemon, has recently been identified as a critical factor in oncogenesis. This protein is encoded by the ZBTB7 gene. Here we aimed to determine whether polymorphisms in ZBTB7 are associated with breast cancer risk in a sample of cases and controls collected in hospitals from North and Central Spanish patients. We genotyped 15 SNPs in ZBTB7, including the flanking regions, with an average coverage of 1 SNP/2.4 Kb, in 360 sporadic breast cancer cases and 402 controls. Comparison of allele, genotype and haplotype frequencies between cases and controls did not reveal associations using Pearson’s chi-square test and a permutation procedure to correct for multiple test. In this, the first study of the ZBTB7 gene in relation to, sporadic breast cancer, we found no evidence of an association.


Introduction
It has been suggested that breast cancer, together with prostate and colorectal, are the cancers with the highest heritable components. A substantial proportion of familiar breast cancer (~25%) is explained by mutations in the BRCA1 and BRCA2 genes [1,2]. By contrast, the excess of familial risk associated with sporadic breast cancer (as well as the unexplained genetic risk in familial breast cancer) may be better explained by the effect of multiple weakly predisposing alleles [3,4]. The identifi cation of common alleles conferring modest susceptibility to cancer (as opposed to the known high penetrance BRCA1/2 genes) is a fi eld of growing interest, especially with the development of new genotyping techniques and SNP database facilities [5].
Hence, there is much interest in the search for low penetrance gene/variants for breast cancer, which could exist with relatively high prevalence in the general population. Many polymorphisms have been proposed as candidates for susceptibility to sporadic breast cancer but reported positive associations have rarely been replicated in independent studies [6][7][8][9].
Recently, Maeda et al. [10] identifi ed the transcriptional repressor FBI1, namely Pokemon (POK erythroid myeloid ontogenic factor), as a critical factor in oncogenesis. This protein is encoded by the ZBTB7 gene ("zing fi nger and BTB domain containing 7"; Gene ID: 51341). Mouse embyronic fi broblasts lacking ZBTB7 are completely refractory to oncogene-mediated cellular transformation. Conversely, FBI1 over-expression led to overt oncogenic transformation both in vitro and in vivo in transgenic mice. FBI1 can specifi cally repress the transcription of the tumor suppressor gene ARF (600160). Maeda et al. [10] found that FBI1 is aberrantly over-expressed in human cancers, and its expression levels predict biologic behaviour and clinical outcome. On the other hand, tissue microarray (TMA) analysis in breast carcinomas has revealed high levels of Pokemon expression in a subset of these tumours. In addition, the genomic region where the ZBTB7 gene resides (19p13.3) is a hotspot for chromosomal translocations (The Cancer Genome Anatomy Project; http:// cgap.nci.nih.gov/). ZBTB7 is therefore a good candidate low penetrance breast cancer susceptibility gene.
Here we aimed to study the potential implications of common ZBTB7 variants in sporadic breast cancer in a sample of cases and controls from Spain. To do this, we selected a set of 19 SNPs covering the whole extension of ZBTB7 and fl anking regions at high density.

Study subjects and DNA extraction
Cases were 360 Spanish women with breast cancer and mean age at diagnosis of 59 years (range 25 to 85 years), recruited between 2000 and 2004 (48% of cases were recruited within one year of their diagnosis and 79% within fi ve years). All cases were collected from a consecutive series recruited via three public Spanish hospitals: Hospital La Paz (20%), Fundación Jiménez Díaz (50%) and Hospital Monte Naranco (30%). Our samples contain prevalently invasive cases of breast cancer, 96%; while only 4% of in situ breast cancer. Controls were 402 Spanish women free of breast cancer at ages ranging from 24 to 85 years (mean = 53 years) and recruited between 2000 and 2005, via the Menopause Research Centre at the Instituto Palacios (50%), the Colegio de Abogados (31%) and the Centro Nacional de Transfusiones (19%), all in Madrid. While data was not available to calculate response rates, our experience is that response rates are very high for cases (~90%).
Genomic DNA was isolated from peripheral blood lymphocytes using automatic DNA extraction (Magnapure, Roche) according to the manufacturer's recommended protocols. DNA was quantifi ed using picogreen and diluted to a fi nal concentration of 50 ng/ul for genotyping. Informed consent was obtained from all participants and the study was approved by the institutional review boards of Hospital Clínico Universitario (Santiago de Compostela, Galicia, Spain) and Hospital La Paz, Madrid.  (Table 1).

SNP genotyping
Genotyping was performed using the MassARRAY SNP genotyping system (Sequenom Inc., San Diego, CA) located at the Universidad de Santiago de Compostela node of the Spanish National Genotyping Center (Centro Nacional de Genotipado; http://www. cegen.org), following the manufacturer's instructions. This typing assay uses the extension of a single primer that binds to the sequence fl anking the mutation site. Base-specifi c primer extension products are created 1-4 bases long depending on the substitution present. The different primer extension products are then differentiated by mass. Multiple sites can be typed simultaneously by multiplexing the extension reaction. Detection uses matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry with samples automatically genotyped from each mass spectrum produced. The assays were designed using Spectro DESIGNER software. Case and control samples were genotyped using 384-well plates and automated protocols. The allele-calling of all possible SNPs in each DNA sample was performed automatically using SpectroTYPER-RT software. Positive and negative controls were incorporated in each genotyping plate in order to assess genotyping quality. We estimate a genotyping error rate below 0.001%.

Statistical analysis
We tested for differences in allele frequencies between cases and controls using Pearson's Chisquared test (the best model is provided in Table 2). We adjusted for age in categories Ͻ45, 44-49, 50-54, 55-59, and Ն60 via logistic regression using Stata v8. Disequilibrium coeffi cients (D') for adjacent SNPs were calculated using Haploview v3.11 [11]. We used Gold software [12] to graphically summarized patterns of linkage disequilibrium in ZBTB7 because it is well suited to the analysis of dense genetic maps. Assuming a minimum allele frequency (MAF) of 3% (the average MAF of our SNP set) and a genetic effect of 2, the a priori power to detect association under a dominant model is above 70%.
The Cocaphased program of the Unphased software package [13] was used to check for single SNP and haplotype associations. We tested all two-, three-, four-, and fi ve-SNP haplotypes for association in a sliding window across the gene. The option 'drop rare haplotypes' was used in order to restrict the analysis to the haplotypes with a frequency Ͼ1%. We followed the permutation test procedure implemented in Unphased which provides P-values corrected for the multiple haplotypes tested. The EM algorithm was used to impute missing data.
Evaluation of stratification was carried out based on the genotyping of 28 neutral SNPs, as previously described in a separate study that targeted a different set of low penetrance breast cancer genes in overlapping samples [14].

Results and Discussion
Three out of the 22 SNPs selected failed genotyping. Four out of the 19 remaining SNPs (namely, rs10405522, rs895330, rs350840, and rs350832) were successfully genotyped in less than 75% of the samples and were therefore excluded from association analyses. The average call rate for   these 15 SNPs was 95% (see also preliminary results in [15][16][17]) and none gave evidence of deviation from Hardy Weinberg equilibrium. Table 1 summarizes their location and allele frequencies.
We computed D' values between all 19 markers, and detected moderate levels of LD (Fig. 1). However, under the 'four gamete rule' model (see Haploview for more information) we identifi ed a haplotype blocks nearly covering the entire extension of the gene (Fig. 2). This characterization of LD along the ZTBT7 region could be useful for future association study designs in cancer.
In order to measure the percentage of variability captured by the our selected SNPs, we first collected the HapMap data from the CEPH subset (http://www.hapmap.org) and the same chromosome range explored in the present study (chromosome 19: positions 3990056-4025697). Then, we estimated the number of SNPs un-captured in the CEPH-HapMap using our SNP selection under an r 2 threshold of 0.8 and a model of 'aggressive tagging'. Only one SNP in the HapMap dataset would remain untagged by our selected SNPs, indicating that our set of SNPs covers well the whole gene region under analysis. No statistically signifi cant differences between cases and controls were observed for individual SNPs based on comparisons of allele frequencies (see Table 2 for the best fi tting models) whether or not age was adjusted for. Four-and three-SNPs haplotypes carrying markers rs350842 and rs350841 had associated P-values below 0.05 but were not signifi cant after correction for multiple testing. Note also that these adjusted P-values overestimates the real value since the software employed (cocaphase) does not correct for the multiple hypothesis tested running different sliding windows.
To our knowledge, this is the fi rst time that ZBTB7 has been evaluated as a candidate sporadic breast cancer susceptibility gene. We have not found evidence of an association for ZBTB7 SNPs nor haplotypes with breast cancer risk. It should be mentioned that most of the ZBTB7 variants studied are rare in our sample. We are aware that the main drawback in detecting positive associations of rare variants (or haplotypes) is the need for large sample sizes. Therefore, the present result needs further validation in future studies of independent case-control series before a role for ZBTB7 in breast cancer can be completely ruled out.