In summary, the minor allele of SNP rs3803662 was associated with shorter survival in breast cancer patients with luminal A tumours, with lower mRNA expression of TOX3 in ER positive tumours as well as being more frequently observed in lobular tumours (Additional file
1: Figure S5). Also, low expression of TOX3 and the neighbouring gene LOC643714 was associated with high Ki67 and the basal tumour subtype while high mRNA expression was associated with ER status, positive lymph nodes at diagnosis and shorter survival (Additional file
1: Figure S6).
The risk allele of rs3803662 has been shown to associate with ER and PgR status
[6, 11, 20–22, 40], grade of tumours
[21, 22, 40] and diagnosis before the age of 60
. In our study cohort, that was smaller than the aforementioned studies, we did not observe a correlation between genotype at rs3803662 and age or tumour characteristics except that the risk allele was more frequent in patients with lobular tumours, which is in agreement with published data
[6, 21]. The Breast Cancer Association Consortium (BCAC) recently reported that ER positive patients, homozygous for the risk allele, had poorer overall and breast cancer specific survival than patients, homo- and heterozygous for the non-risk allele
. The BCAC study did not include information on the molecular subtypes of the tumours. In our cohort, the risk allele was significantly associated with poor overall survival only in patients with luminal A tumours. The rs3803662 genotype has been shown to associate most strongly with the ER/PgR positive and ERBB2 negative subgroups
, which is the expression pattern of tumours of the luminal A subtype, further supporting that the effect seen in the BCAC study may be connected to luminal A tumours.
A decrease in TOX3 mRNA according to the rs3803662 genotype was observed in a large Dutch study of 1401 breast tumours
 but analysis according to ER status was not reported. In our study, the allele-dependent decrease of TOX3 or LOC643714 mRNA level was only observed in ER positive tumours. The TOX3 mRNA levels were quantified by microarrays and confirmed by qRT-PCR. A significant association was not seen using the latter method, possibly due to the lower number of tumours that were available for qRT-PCR. Expression of LOC643714 was only measured by qRT-PCR and a positive correlation was found between TOX3 and LOC643714 mRNA levels. The closely located genes are potentially co-regulated. LOC643714 mRNA expression in breast tumours was very low and near the detection limit of the assay. This was previously observed by Udler et al. who reported negligible mRNA levels of LOC643714 in normal breast and breast tumours
. A larger cohort will be necessary to confirm the association between rs3803662 and LOC643714 expression.
The rs3803662 genotype may only affect expression of TOX3 and LOC643714 in certain tumour subtypes. Although we observed a reduction in gene expression in ER positive tumours, our preliminary analysis suggests that this is limited to luminal A tumours (Additional file
1: Figure S3) and another mechanism may be responsible for controlling expression of these genes in luminal B tumours. Although a higher number of tumours are required to confirm this observation, the majority of the adverse effect of high TOX3 expression in ER positive tumours was related to luminal B tumours (Additional file
1: Figure S4). Furthermore, higher TOX3 mRNA was observed in patients with luminal B tumours with metastasis (p = 0.056, data not shown) but more tumours are necessary to observe a significant association of metastasis to bone as has been reported with high TOX3 expression
. The risk allele of rs3803662 was highly correlated with a subgroup of basal tumours that expresses EGFR and CK5/6 but not ER, PgR and ERBB2
. Expression of TOX3 and LOC643714 was significantly lower in basal tumours in our cohort but we did not observe differences in expression according to genotype or a difference in allele distribution compared to the other tumour subtypes as has been reported for rs9940048, which is located within TOX3.
The rs3803662 SNP may regulate expression of genes more distantly located than TOX3 and LOC643714. It lies within an evolutionary conserved region that has an open chromatin conformation suggestive of an active transcription regulatory region
. Our preliminary analysis of the association of the rs3803662 risk allele with the expression of genes located within 1 Mb of the SNP revealed an association with the retinoblastoma-like 2 gene (RBL2) only, where an increase in expression was seen with the number of risk alleles (p = 0.02, data not shown). This result is in agreement with Udler et al., who reported a dose-dependent association between the rs3803662 risk allele with increased expression of RBL2 mRNA expression in lymphocytes; however, an association between the rs3803662 genotype and RBL2 mRNA expression in 77 breast tumours was not seen
. Expression of RBL2 was recently reported to be associated with prognosis
. It is a member of the retinoblastoma gene family
 and has been shown to be involved in cell cycle regulation, and it is frequently deleted in breast tumours
Very low expression of TOX3 and LOC643714 was observed in the basal subtype, an observation that has been reported for TOX3. The significantly lower TOX3 and LOC643714 expression in tumours expressing high Ki67 was due to the relatively high number of basal tumours expressing high Ki67. Expression correlated significantly with metastasis in ER negative tumours but not with metastasis to bone
. High expression of both TOX3 and LOC643714 mRNA was observed in tumours of lymph node positive patients but only TOX3 mRNA correlated with survival where high levels were seen in patients with shorter overall and distant metastasis free survival in the entire group and for patients with ER positive tumours. This discrepancy between TOX3 and LOC643714 may possibly be due to fewer LOC643714 tumours than for TOX3, different distribution of available samples into subtypes, or the fact that there was not a correlation between TOX3 and LOC643714 in ER positive tumours. It was recently reported that TOX3 expression was not associated with prognosis
. The study was large but the data came from 9 separate studies in publicly available databases. Thus, we suggest that the effect potentially vanished due to variability in study design and data acquisition.
The biological mechanism that potentially explains our results is the implication of TOX3 in proliferation and apoptosis. Knockdown of TOX3 expression with siRNA in the luminal A breast cancer cell line ZR-75-1 increased cellular proliferation
. The same study described a FOXA1 binding site at breast cancer SNP rs4784227, whose risk allele segregates with the risk allele of rs3803662; FOXA1 preferentially bound the site that included the rs4784227 risk allele resulting in a 5-fold decrease in TOX3 expression
. Therefore, the correlation of FOXA1 expression with luminal A tumours
 could explain lower expression of TOX3 in our study and poorer survival of carriers of the rs3803662 risk allele and luminal A tumours. In these tumours, TOX3 would appear to be a tumour suppressor. However, a different scenario has been described in neuronal cells that potentially may explain the effect of high TOX3 mRNA in ER positive tumours which we ascribe to luminal B tumours. Increased expression of TOX3 protected neuronal cells from cell death by altering expression levels of genes that affect apoptosis
. Such function in ER positive breast cancer cells, where TOX3 expression is high, could promote tumour enlargement, invasion of nearby tissues, seeding of lymph nodes which would subsequently shorten the time until death or distant metastasis. The effect of TOX3 on apoptosis in neuronal cells was mediated via its interaction with CITED1 and ERE
 and may be the case for proliferation as well. However, the TOX3 ERE-mediated transcription appeared independent of ER and oestrogen
 unlike that of CITED1, a co-activator of ER that modulates ER responsive genes
. We can only speculate that LOC643714 has a role in tumour development based on results similar to TOX3 but its mRNA is not translated and its function is unknown.
Although we assume that mRNA expression of our candidate genes may differently affect breast tumour subtypes, the study is small and the analyses according to ER status or molecular subtypes further reduced the number of tumours. As such, the study is exploratory and the results should be confirmed with a larger number of tumours. Our findings provide a suitable basis for further validation studies.