DYX1C1, a gene originally associated with dyslexia [9
] and neuronal migration, is expressed in several tissues, e.g., testis, ovary and brain [9
]. Two recent studies have suggested a role for DYX1C1 as a cancer biomarker [7
]. In the present study, we report that DYX1C1 is expressed on the mRNA level in breast tumors and is associated with several clinicopathological variables such as ERα, PR, and lymph node status. We also show that the DYX1C1 protein is expressed in these tumors and is connected to the overall survival of the patients. This further points towards the potential of DYX1C1 as a breast cancer biomarker of predictive and prognostic significance.
There are several proliferative diseases of the mammary gland, both benign and malignant. The pathological mechanisms leading to malignant transformation of normal mammary epithelium are not fully understood, however mutations of tumor suppressor gene TP53
and amplification of ERBB2
are common genetic alterations. Fibroadenoma is the most common benign hyperplasia in the mammary gland, especially during adolescence. Estrogen is considered a putative agent for the development of benign breast tumors and important for growth of malignant tumors of the breast. However, females with a history of fibroadenoma do not have a higher lifetime risk of breast cancer and there is little evidence that fibroadenoma is a precancerous stage (reviewed in [23
Development of malignant invasive breast carcinoma is considered as a continuous process from normal epithelium, flat epithelial atypia (FEA), atypical ductal hyperplasia (ADH), ductal carcinoma in situ
(DCIS) to invasive ductal carcinoma. Along the process, there are both loss and amplification of several genomic regions [24
]. DCIS is defined as a neoplastic proliferation of epithelial cells within the duct with intact basement membrane. DCIS is generally considered as a premalignant lesion, and women with DCIS are at higher risk of developing invasive ductal carcinoma. However the precise mechanism of progression from DCIS to invasive cancer is unknown [25
]. It has been shown that most expressional changes seen in invasive carcinoma are present already in DCIS [26
]. An interesting example is the overexpression of CXCL12 in DCIS as well as invasive cancer, a chemokine that has been shown to increase the expression of DYX1C1 in a prostate epithelial cell line [27
]. The role and expression of DYX1C1 in DCIS, premalignant and benign proliferative disorders of the mammary gland is not known and needs further study.
In this study we show by using qRT-PCR that the mean expression of DYX1C1 is 3.2 times higher in tumors that are classified as ERα-positive. We were able to replicate this association in two different independent data sets where microarray analysis of mRNA expression had been performed. Here the difference in expression was less pronounced, most likely due to the higher sensitivity of the qRT-PCR method compared with microarrays. Although several mechanisms of DYX1C1 actions have been proposed for dyslexia, the function of the gene in other diseases is still poorly understood. DYX1C1 has, however, been shown to interact with protein chaperone Hsp70, which is often up-regulated in several neoplasms and is considered a future pharmaceutical target [28
]. Furthermore, when overexpressed in the neuroblastoma cell line SH-SY5Y, DYX1C1 has been shown to interact with and regulate the levels of ERα, as well as affect the migration properties of the cells [12
]. As DYX1C1 has been shown to interact as a chaperone with ERα and also regulate its expression, higher levels of DYX1C1 protein, which could be a result of increased transcription of DYX1C1, could perhaps influence the levels of ERα and, consequently, affect the response to the anti-estrogen treatment of the patient. DYX1C1 expression is also higher in PR-positive tumors. As PR is a target gene for ERα and is considered as a marker for endocrine sensitivity, this finding is not surprising (reviewed in [30
]). Both ER and PR status was independently correlated with DYX1C1 status when included into a multivariate linear regression model, further indicating that the correlation to DYX1C1 expression was a result of the receptor status independently.
Tumors classified as basal subtype, associated with the worst survival prognosis, were also the group that, on average, expressed the lowest levels of DYX1C1 mRNA. The tumors of the basal subtype lack amplification of ERBB2
gene (HER2) and are both ERα- and PR-negative, which could explain this association. DYX1C1 expression was on the other hand highest in the Luminal A group, which is characterized by expression of ERα, PR, low expression of proliferative genes and no HER2 amplification [24
When correlating DYX1C1 mRNA level to the Elston grade of the tumor, the grade 3 tumors, shown to have the worst prognosis [31
], had significantly lower levels of DYX1C1 compared with grades 1 and 2 combined. However, this finding was only seen in the Uppsala patient cohort and not in CAHRES. This discrepancy could be due to the enrichment of invasive tumors within CAHRES, where only 3 out of 61 patients were graded as Elston grade 1, render in inadequate power. Also, in our multivariate linear model, Elston grade was not a significant independent factor. Taken together, these results indicate that DYX1C1 mRNA is lost in high-grade aggressive tumor subtype. Our results are opposite to the published data from Chen et al.
showing reduced DYX1C1 expression in benign tumors and normal tumor adjacent tissue compared with cancer [8
]. The reason for this difference is not clear to us. It should be pointed out that we have been using different methods in detecting mRNA expression. While we were using a TaqMan probe together with qRT-PCR, Chen et al.
used semi-quantitative PCR with designed probes. Also, a basic local alignment search tool (BLAST) analysis of the primers used by these authors revealed that their primer pairs were unable to detect one of the RefSeq isoforms of DYX1C1. It is therefore likely that they were not able to measure all the DYX1C1 mRNA.
Although patients retaining DYX1C1 expression seem to have less aggressive tumors, they do, on the other hand, have an increased tendency of lymph node metastasis which seems contradictory. As DYX1C1 has previously been shown to have a potential role in the migration of neuronal cells [29
], perhaps DYX1C1 has a similar role in breast cancer cells. Reelin, another protein important in neuronal migration in the brain, was recently shown to be epigenetically silenced in breast cancer compared with normal mammary tissue [32
]. Loss of Reelin protein expression was associated with decreased survival [33
]. Furthermore Roundabout homolog 1 (ROBO1), also implemented in familiar dyslexia, has been identified in the progression of several cancers. ROBO1 is a membrane bound receptor that interacts with members of the SLIT family of secreted proteins important in the migration of neurons [34
]. The binding of SLIT to ROBO-receptors has been shown to increase the migration of breast cancer cells and also in selection of metastasis to the brain [35
]. ERα-positive tumors are considered to be less prone to metastasize, suggesting DYX1C1 as a possible marker of metastasis risk within the group of ERα-positive tumors. However, it must be noted that we were unable to replicate the association of DYX1C1 in lymph node metastasis in any of the microarray datasets.
We stained 10 healthy donors using IHC with DYX1C1 antibodies raised against DYX1C1. As DYX1C1 was seen strongly expressed in all donors it seems unlikely that the levels of DYX1C1 protein is low in normal mammary tissue. The staining was, on the contrary, usually strong. We also stained 221 tumor samples using the same antibody. This indicated that 88.7% of tumors expressed DYX1C1 protein at some level, but the expression was lost in 11.3% of the tumors. The reason for this loss is unknown to us, but perhaps it could be a regulation through the ubiquination ligase CHIP as seen in cell lines [10
Patients with tumors classified as negative DYX1C1 protein expression by IHC had significantly higher mortality when controlling for several possible confounding factors, indicating that lost DYX1C1 expression is a poor prognostic factor. However, although DYX1C1 mRNA expression was associated with several clinicopathological variables we could not show similar associations on DYX1C1 protein level. This could be a result of a post-translational modification of DYX1C1 causing a non-linear correlation of mRNA and protein. This is not an uncommon observation and is seen with many genes (reviewed in [36
On the other hand, the lack of concurrence between DYX1C1 protein and mRNA results could also be a result of not choosing the appropriate cut-off for defining a positively expressing tumor. When we instead used the median of the Allred score as a cut-off for positively or negatively DYX1C1 expressing tumors we saw a correlation with the lymph node status of the patients, but not with overall survival (data not shown).