Mammography is currently the modality of choice for screening for early BC and possesses a sufficient sensitivity and specificity. Specificity of screening mammography is over 95%, but sensitivity ranges between 67% and 95% and is strongly dependent on several factors like age, breast density and professional experience of the examiner 
. In women with very dense breast tissue sensitivity of screening mammograms can be as low as 40 to 50% 
. Magnetic resonance imaging (MRI) of the breast and breast ultrasound could improve cancer detection in these cases, but these imaging techniques are not routinely used for BC screening 
. Therefore, intensive research is currently carried out to identify new, non-invasive BC detection methods. In the discovery setting of our study we found 13 significantly up-regulated miRNAs and 46 significantly down-regulated miRNAs in our microarray panel of 1100 miRNAs and miRNA star sequences and validated our results in an independent BC cohort with RT-qPCR. A set of 240 miRNAs yielded a specificity of 78.8%, and a sensitivity of 92.5% in this very early stage BC cohort. We deliberately used frozen EDTA-blood to test and establish a miRNA-detection method which could possibly be used in larger multicentric trials collecting frozen whole blood. The stability of miRNAs in EDTA-blood and the possibility of miRNA-profiling from non-frozen EDTA-blood have been shown previously by others 
. After miRNA-microarray-profiling miR-202, and miR-718 were chosen for RT-qPCR validation based on the significant expression changes on the microarray. Moreover, the identification of biochemical pathways that are enriched with respect to the miRNA target genes revealed miR-202 as the miRNA with the highest number of significant biological pathways and Gene Ontology categories of all known miRNAs 
. This target pathway analysis indicates that miR-202, which we found to be significantly up-regulated in our early stage BC blood samples of the microarray and the RT-qPCR cohort, is influencing a plethora of cancer relevant biological pathways and might be important for BC development. Considering that miR-202 belongs to the let-7 family of miRNAs, which are known to be involved in self-renewal and tumorigenicity of BC cells, functional data are available regarding the role of miR-202 and other members of this familiy in carcinogenesis 
. The let-7 family regulates estrogen receptor alpha signaling in estrogen receptor positive BC and is highly conserved across species in sequence and function 
. The miRNAs let-7e and miR-202 target the same seed sequence and a recent study has shown that these miRNAs target the proto-oncogene MYCN in vitro 
. Further targets of the let-7 family of miRNAs are RAS and HMG A2 (high mobility group A2) 
. The two associated miRNA families, let-7 and miR-200, have also been identified as regulators of epithelial-to-mesenchymal transition (EMT) and dedifferentiation of cancer cells to stem cells 
Several miRNAs are known to be aberrantly expressed in human BC tissue and have been correlated with clinical stage and clinico-pathological variables like hormone receptor status, tumor subtypes, as well as clinical variables like metastatic potential, progression free survival and overall survival 
. Tissue-based miRNA expression profiling of the inflammatory breast cancer (IBC) subtype observed some miRNAs to be independently associated with the difference between IBC and non-IBC. Among those miRNAs with increased expression in the IBC subtype was also miR-335, which was also significantly up-regulated in our blood-based microarray discovery study 
. The expression of miR-335 was also up-regulated (3.9 fold) in colonic cancer tissues compared to para-cancerous control tissue 
. On the other hand, miR-335 has also been reported as robust inhibitor of tumor reinitiation and was found to suppress migration, invasion, and metastatic colonization in vivo 
. Another study analyzing tissues from malignant pleural mesothelioma found miR-193-3p to be over expressed in formalin-fixed, paraffin-embedded malignant pleural mesothelioma tissue compared to carcinoma tissue 
. In line with these findings, our microarray study identified miR-193-3p as one of the most significantly down-regulated miRNAs in whole blood of BC patients (2.16-fold; P
Blood-based miRNA-profiling is still far behind the improvements in tissue-based miRNA-profiling, but offers the potential for early, non-invasive, sensitive and specific BC detection and screening. First reports using serum or plasma for RT-PCR or microarray based miRNA-profiling were promising. Recently, a serum based study using next-generation sequencing of miRNAs for BC detection has also been reported and several miRNAs have been identified as potential serum/plasma biomarkers in different cancer types like lung, prostate, colon and liver cancer 
. Furthermore, strategies with whole blood have been established and likewise show favourable results in the non-invasive detection of cancer and other diseases 
One potential advantage of the whole blood approach could be the higher miRNA-content and the chance to measure not only tumor secreted oncogenic miRNAs, but also the changes in the miRNA profile following the “host-reaction” in the body of the patient 
. On the other hand, the main concern about using whole blood is a reduction of the testing accuracy due to the measurement of a miRNA profile which represents only an unspecific secondary response of blood cells during tumorigenesis 
. Secondly, the high protein content of whole blood could be a problem for RNA-extraction. Results from previous studies of our and other groups indicated that the changes in the miRNA profile of blood cells of patients with cancer also reflect tumor-specific host-reactions which might be measurable in whole blood 
. We would therefore expect that the whole blood approach offers the potential to diagnose cancer at a very early stage when the concentration of tumor-secreted miRNAs is still small, but the reaction of the immune system in response to cancer can already be detected by miRNA profiling in whole blood. Comparing serum and blood cells from the same healthy individual an almost identical miRNA profile can be found, but in cancer patients the profiles differ 
. In contrast to miRNA-studies using plasma or serum we found a contrary trend in whole blood regarding the previous published plasma levels of let-7d*, let-7c, miR-425* and miR-589 
. Although none of these miRNAs was significantly differentially expressed in our microarray analyses the median fold changes between BC cases and controls were completely contrary to the plasma expression changes published by Zhao et al. for these four miRNAs in Caucasian Americans (n
. For example, the whole blood expression levels of let-7d* in our microarray cohort (n
48) were almost 2-fold higher in BC cases compared to controls with a P-value of 0.009 in the unadjusted t-test and 0.098 in the adjusted t-test.
The comparison of miRNA-profiles from whole blood and plasma/serum is also addressed in the work of Zhao et al. 
. The authors of this plasma based miRNA-profiling study in BC were not able to reproduce the data from Heneghan et al. showing a significantly higher expression of let-7a and miR-195 in whole blood of BC cases compared to controls 
. In our study we found only a trend towards an up-regulation of miR-195 in whole blood of BC cases compared to controls in the miRNA-microarray analyses (P
0.055). Possible reasons for this discrepance are differences in sample handling, detection methods and patient selection (clinical stages). Moreover, a recent study showed significant differences between cell-free and cellular blood miRNA profiles. Using different plasma fractionation procedures for plasma, the authors showed varying degrees of efficacy in the removal of red and white blood cells and as a consequence different miRNA profiles 
. In addition to different detection techniques, these differences could be partially responsible for the discrepancies between blood based miRNA profiling studies. Based on this data whole blood miRNA profiling seems reasonable to achieve integrated, standardized analyses of circulating disease specific miRNA signatures; and is able to measure the disease specific over-expression of hematopoietically derived miRNAs and circulating cell-free miRNAs.
The median fold expression changes of deregulated miRNAs in our microarray analyses were in the range of 2 to 4 fold which is in line with previous miRNA microarray profiling studies, but fare less than previously reported in qRT-PCR based studies. These differences were expected and are probably due to the different detection and analysis techniques. The relative fold changes found in our RT-qPCR validation analyses were higher, with miR-202 and miR-718 showing a relative fold change of about 20 and over 5 between cases and age-matched controls (). The RT-qPCR validation cohort was small with only 24 age matched pairs and we were able to show statistically significant differences in expression values only for one of the two analysed miRNAs (miR-202), but in agreement with the microarray results the relative fold changes showed the same trend for both miRNAs. This is probably due to the small sample size, but could also indicate that a set of miRNAs rather than a single miRNA is needed for a reliable differentiation of cancer cases and controls.
The clinicopathological characteristics of the two cohorts are slightly different with smaller and less aggressive tumors with a higher rate of HER2-negativity in the RT-qPCR cohort compared to the microarray cohort. This is due to the different recruitment strategy (hospital based versus screening cohort) and the consecutive patient recruitment.
We used miRNA microarray technology to analyze the miRNA expression profiles of early stage BC patients compared to healthy controls from frozen EDTA-whole-blood and validated the results with RT-qPCR in an independent early stage BC cohort. Research of circulating miRNAs as blood based biomarkers is still in its infancy. However, this study as well as other recent studies indicate that miRNA-analyses have diagnostic and prognostic potential and could improve early stage BC detection in the future. There are several possible applications of miRNA profiling conceivable in the future. Firstly, miRNA profiles could help to reduce unnecessary breast biopsies if miRNA sets could be identified which reliably identify BC free individuals. Secondly, miRNA profiling could be used as a pre-screening method for example by general practitioners to identify women with an urgent need for breast diagnostics. Thirdly, in younger patients with dense breast tissue a future miRNA-based BC screening could possibly provide better sensitivity and specificity than the mammography even without radiation exposure.
Our study detected several significant deregulated miRNAs in frozen whole blood of early stage BC patients, which should be analyzed further with regard to their function in breast cancer development and progression. Moreover, large prospective clinical studies are clearly warranted to confirm our preliminary results and further explore the existing potential of circulating miRNAs in serum, plasma or whole blood as diagnostic and therapeutic BC biomarkers.