Introduction

Prostate cancer (PC) is the most commonly diagnosed non-skin cancer and the third most lethal cancer among men in Europe with an estimated 89,000 annual deaths in 2008 [1]. Suspicion of PC is generally raised upon an elevated serum prostate-specific antigen (PSA) level and/or an abnormal digital rectal examination (DRE). Elevated PSA, however, is not specific for PC, and although digital rectal examinations can indicate PC, cancer lesions are frequently missed [2]. Thus, routine tests for PC are imprecise and insufficient; nonetheless, the number of PSA tests is rising. If suspicion of PC is raised by these initial tests and/or symptoms, patients are generally referred to transrectal ultrasound-guided prostate biopsy. Subsequent PC diagnosis is based on histopathological evaluation of prostate needle biopsies. Whereas the initial tests (PSA/DRE) for PC are simple, prostate biopsy is associated with considerable stress and discomfort, as well as risk of infection and other side effects [3]. Furthermore, it has been estimated that 10–20 % of PCs are not detected in the initial set of biopsies, leading to frequent repeat biopsy [4]. Thus, more specific and less invasive methods for detection of PC, e.g., in blood or urine samples, are urgently needed.

Whereas localized PC is curable through radical prostatectomy or radiation therapy, metastatic PC is incurable. Thus, early detection of PC is pivotal for complete recovery. However, many PCs do not form metastases in the lifetime of the patient even without treatment. Thus, treatment, which includes severe side effects such as incontinence and impotence, should be limited to aggressive cancers. At the time of diagnosis, PC aggressiveness is currently assessed primarily by the Gleason score and T-stage determined in prostate biopsies. Men with low-risk PC may enter a program of active surveillance, which can include periodic re-biopsy to determine the possible progression status of the primary tumor [5]. Minimally invasive markers for PC prognosis could potentially aid in determination of aggressiveness following initial PC diagnosis as well as in monitoring disease progression of men in active surveillance programs.

Recently, circulating miRNAs have been discovered as potential biomarkers for both cancer diagnosis and prognosis [6, 7]. miRNAs in circulation originate from blood cells as well as from solid tissues [8] and may be released actively or passively into the blood stream [9]. Active secretion of miRNAs can take place as an integrated part of cellular signaling via different pathways. Thus, cells produce miRNA-containing micro-vesicles (up to 1 μm) that bud from the cell surface as well as miRNA-containing exosomes (50–100 nm) that are released from larger endocytic vesicles [10–13]. It has also been reported that cells may actively release miRNAs as parts of protein complexes including, e.g., Argonaute 2 (Ago2), nucleophosmin 1, and high-density lipoprotein [14–16]. Passive release of miRNAs in complex with Ago2 is thought to occur during apoptosis or necrosis, where cellular contents are leaking [9].

The cellular origin of de-regulated miRNAs in the circulation of cancer patients has not been fully established. Thus, changes observed in circulating miRNA levels in cancer patients could derive directly from cancer cells that actively or passively release these miRNAs. Alternatively, changes in circulating miRNA levels may reflect a host response to the tumor, e.g., by surrounding tissue or by immune cells [17].

The association of miRNAs with micro-vesicles/exosomes or protein complexes renders them highly stable and able to resist degradation when exposed to, e.g., repeated freeze–thaw cycles and enzymatic degradation [8, 18]. Furthermore, miRNAs are easily extracted from serum or plasma using standard RNA purification kits and can be quantified by RT-qPCR as well as a range of microarray-, and more recently, next-generation sequencing-based methods. Thus, circulating miRNAs possess many features that make them a highly attractive source for biomarker discovery.

A number of studies have evaluated the association of miRNAs in blood with PC, and many circulating miRNAs have been suggested as potential markers for PC [19, 20]. However, few overlaps in the identified de-regulated miRNAs exist. Here, we perform a profiling study of serum miRNAs from 13 benign prostatic hyperplasia (BPH) control patients and 31 PC patients, using Exiqon’s miRNA human panel I and II assaying 732 miRNAs, to identify potential serum miRNA biomarkers for PC. Furthermore, we carefully reviewed the existing literature and compared our results to those of previously published studies. In summary, we confirmed the de-regulation of two of the most well-documented miRNA markers for PC in literature (miR-141 and miR-375) and identified new potential highly specific serum miRNA markers for both PC and disseminated PC. Furthermore, we develop three novel candidate miRNA panels, which with 100 % specificity, identified 84 % of PC patients (miR-562/miR-210/miR-501-3p/miR-375/miR-551b), 80 % of patients with disseminated PC when compared to BPH patients (let-7a*/miR-210/miR-562/miR-616), and 75 % of disseminated PC patients when compared patients with localized PC (miR-375/miR-708/miR-1203/miR-200a).

Materials and methods

Patients and samples

Blood samples from 13 BPH patients (controls), 11 patients with localized PC (LPC), 9 patients with lymph node metastases (N1) or distant metastases (M1), and 11 patients with castration-resistant PC (CRPC) collected at Department of Urology, Aarhus University Hospital, Denmark, in the period of 2001–2008 were included in the study (Online Resource Table 1). Blood samples from BPH, M1, and CRPC patients were collected prior to transurethral resection of the prostate, blood samples from LPC patients were collected prior to radical prostatectomy, and blood samples from N1 patients were collected prior to intended radical prostatectomy, which was cancelled due to the discovery of lymph node metastases. BPH, LPC, and N1/M1 patients were untreated for PC at the time of sample collection. CRPC patients had undergone palliative endocrine treatment. Serum was isolated immediately after blood draw using centrifugation in serum-gel tubes and stored at −80° C.

Ethical approvals

The study was approved by the local ethical committee, and written informed consent was obtained from all patients.

RNA extraction

Briefly, 1 ml of QIAzol Lysis Reagent (Qiagen) was added to 200 μl of serum. To ensure complete dissociation of nucleoprotein complexes, the mixtures were incubated 5 min at room temperature after mixing by vortexing or pipetting. Then, 10 μl of 1 nM synthetic cel-miRNA-39 from Caenorhabditis elegans was added as a spike-in control for purification efficiency. The mix was shaken vigorously and allowed to stand for 2–3 min at room temperature following the addition of 200 μl of chloroform. After 15 min of centrifugation at 12,000×g at 4° C, the upper aqueous phase was transferred to a new collection tube, and 1.5 volumes of 100 % ethanol was added and thoroughly mixed by pipetting up and down several times. Purification of extracted total RNA was performed with miRNeasy (Qiagen) according to the manufacturer’s instructions. RNA was eluted in a final volume of 30 μl of RNase-free water.

miRNA analysis

miRNAs were reverse transcribed using the miRCURY LNA™ microRNA PCR, polyadenylation, and cDNA synthesis kit II from Exiqon according to the manual. Relative miRNA levels were analyzed using Exiqon’s microRNA Ready-to-Use PCR, Human panel I + II, V2.M, assaying 732 miRNAs in two 384-well PCR plates, which were run on Applied Biosystem’s 7900HT real-time thermal cycler. miRNAs for which all C t values exceeded 35 in all samples were excluded from further analysis (n = 276). We used miR-320a, as identified by the NormFinder algorithm [21], for normalization of individual miRNA levels. C t values were converted to relative levels using the ∆C t method (2^(C t(miR-320a) − C t(target miR))). To control for inter-plate variations, all plates were normalized to the internal plate control UniSp3.

Statistical analysis

Two-sided Student’s t tests were used to identify significantly (P < 0.05) de-regulated miRNAs. Comparisons were made between serum samples from BPH patients and all PC patients, between BPH patients and LPC patients, between LPC patients and all patients with disseminated disease (N1/M1 and CRPC), and between patients with non-castration resistant disease (LPC and N1/M1) and CRPC patients. Receiver operating characteristic (ROC) analysis was conducted for each potential miRNA serum marker. To generate models comprising multiple miRNAs, we dichotomized individual miRNA levels using a cutoff maintaining 100 % specificity as determined by ROC analysis. For miRNA models comprising multiple miRNAs, patients were included in the high-level group if at least one miRNA was annotated as high-level by ROC analysis.

Taylor dataset

The complete original dataset consisted of miRNA expression data from 28 adjacent normal (ADJN) prostate tissue samples, 99 primary prostate cancers, and 14 metastases [22]. We used data only from ADJN tissue samples (n = 28) and from primary PC tissue samples obtained from patients not having received neo-adjuvant hormonal- or chemotherapy (n = 96).

Results and discussion

Screening of serum miRNAs from PC patients and controls; development of miRNA panels

To identify new PC biomarker candidates, we analyzed miRNA levels in serum samples from 13 patients with BPH, 11 patients with LPC, 9 patients with local or distant metastasis (N1/M1), and 11 patients with CRPC (Online Resource Table 1). The LPC patients generally had low T-stage (pT2a–b) and low Gleason score (5–7), whereas patients with disseminated PC had higher Gleason scores (7–10). Serum miRNA levels were assayed with Exiqon’s microRNA Ready-to-Use PCR panels, interrogating 732 miRNAs.

First, for investigation of cancer-specific alterations in serum miRNA levels, we used Student’s t tests to identify significantly de-regulated miRNAs in serum samples from all PC patients (n = 31) compared to serum samples from BPH control patients (n = 13). To limit the list of candidate miRNAs to those with the highest potential as detection markers for PC, we excluded all miRNAs with a fold change of less than two. This yielded eight miRNAs, which were all up-regulated in serum samples from PC patients (Table 1; Fig. 1a, b). Each of these miRNAs identified only a subset of PC patients, with sensitivities ranging from 16 % (miR-17*) to 45 % (miR-562) at a fixed specificity of 100 % (Table 1). We performed Spearman correlation analyses of the miRNA levels in PC samples to identify miRNA pairs, which could potentially supplement each other in a diagnostic test, and thus increase sensitivity when compared to individual miRNAs (Online Resource Table 2). While miR-562 had the highest sensitivity (45 %) as a single marker, miR-210 displayed the highest anti-correlation (Spearman rho = −0.415) to miR-562. We therefore combined miR-562 and miR-210, which increased the sensitivity to 65 % when fixing specificity at 100 % (Table 1; Fig. 1c). Next, we sequentially added miR-501-3p, miR-375, and miR-551b (also anti-correlated to miR-562) to the model and gained in sensitivity for each added miRNA concluding at a sensitivity of 84 % (Table 1; Fig. 1c) and 100 % specificity.

Table 1 miRNAs more than two-fold de-regulated in serum samples from PC patients compared to BPH
Full size table
Fig. 1
figure 1

miRNAs more than two-fold de-regulated in serum samples from PC patients. Serum samples from BPH patients were compared to serum samples from PC patients. a Column charts. BPH serum samples from patients with benign prostatic hyperplasia (n = 13), LPC serum samples from patients with localized PC (n = 11), N1/M1 serum samples from patients with lymph node or distant metastases (n = 9), CRPC serum samples from patients with castration resistant PC (n = 11), Rel Exp relative expression levels of miRNAs normalized to miR-320a. Black line mean miRNA level for BPH patients and for PC patients, respectively. t test p value given for comparisons. b Receiver operating characteristic (ROC) analysis of miRNAs in BPH samples vs all PC samples. AUC area under the curve. c ROC analysis of combinations of dichotomized miRNAs. miRNAs levels were classified as high or low based on ROC analysis. Cut-points were selected to maintain specificity at 100 %. AUC area under the curve of ROC analysis

Full size image

The high specificity of the miR-562/miR-210/miR-501-3p/miR-375/miR-551b model for PC suggests a promising potential for inclusion of these miRNAs into a serum-based molecular diagnostic test for PC detection. It is possible that combination with additional molecular markers, including but not limited to serum miRNAs, may improve sensitivity even further without significantly compromising specificity. Future studies are needed to clarify this. We noted that the levels of the eight miRNAs (Table 1) varied considerably in the serum samples from PC patients (Fig. 1a), possibly reflecting the heterogeneity of the patient sample set, which represented all stages of PC, from LPC to CRPC. Hence, to identify miRNAs with more consistent changes in serum, we proceeded with analyses focusing on subgroups of PC patients.

miRNAs de-regulated in serum samples from LPC patients compared to BPH patients

To identify miRNAs that could potentially aid in early detection of organ-confined PC, we compared serum from BPH control patients to serum from LPC patients. This analysis identified 12 miRNAs (miR-103-2*, miR-1179, miR-149*, miR-154, miR-181a*, miR-188-5p, miR-31, miR-329, miR-376c, miR-450a, miR-508-5p, and miR-556-5p) as more than 2-fold de-regulated between the two groups (Online Resource Table 3, Online Resource Fig. 1). However, except for miR-103-2*, all of the remaining 11 miRNAs were down-regulated in serum from LPC patients, and for each of these, one or more patients with disseminated disease (N1/M1/CRPC) displayed serum levels comparable to those seen in the BPH control group (Online Resource Fig. 1). Thus, these 11 miRNAs may be suboptimal for use in a diagnostic test, as high serum levels do not distinguish between BPH patients and PC patients with advanced disease. miR-103-2* was the only up-regulated miRNA in LPC compared to BPH and also distinguished CRPC patients from BPH patients. Thus, miR-103-2* may have potential as a serum biomarker for early as well as advanced PC (Online Resource Fig. 1), although further studies are required to confirm this. The association of only one up-regulated miRNA with LPC in this analysis could be a consequence of the low Gleason score and T-stage cancers of the LPC patients (Table 1), as small and relatively well-differentiated cancers may not give rise to large perturbations in serum miRNA levels. In agreement with our findings, there is little experimental support in the literature for the utility of circulating miRNAs as very early detection markers for localized prostate cancer [27, 28].

miRNAs de-regulated in serum from patients with disseminated PC compared to BPH patients; development of miRNA panels

To identify miRNAs with potential as markers for advanced PC, we compared serum samples from BPH patients to serum samples from patients with disseminated disease (N1/M1 and CRPC). We identified eight miRNAs de-regulated more than two-fold, all of which were up-regulated in samples from cancer patients (Table 2; Fig. 2a, b). The sensitivities of these miRNAs for advanced PC ranged from 20 to 50 %, at 100 % (fixed) specificity. Spearman correlation analysis suggested that the most sensitive single miRNA (let-7a*, sensitivity 50 %) was anti-correlated (rho = −0.204) with miR-210 in patients with disseminated PC (Online Resource Table 4). Accordingly, using a combination of let-7a* and miR-210 sensitivity increased to 70 % (Table 2). We further noted that let-7a* was not inversely correlated to any of the other de-regulated miRNAs, whereas miR-210 showed inverse correlation to miR-562 (rho = −0.466), miR-616 (rho = −0.442), and miR-297 (rho = −0.351). We sequentially added each of these miRNAs to the let-7a*/miR-210 model. By addition of miR-562, the sensitivity increased to 80 % (at 100 % fixed specificity), but the addition of more miRNAs did not improve the model further (Table 2; Fig. 2c). Thus, let-7a*, miR-210, and miR-562 seem to have promising potential as new serum biomarker candidates for advanced (aggressive) PC. It would be highly interesting to test these candidate markers in serum samples from patients with high-risk LPC to investigate their potential value as biomarkers for localized PC with high risk of dissemination.

Table 2 miRNAs more than two-fold de-regulated in serum samples from PC patients with disseminated PC when compared to BPH control patients
Full size table
Fig. 2
figure 2

miRNAs more than two-fold de-regulated in serum samples from patients with disseminated PC. Serum samples from BPH patients were compared to serum samples from patients with disseminated PC. a Column charts. BPH serum samples from patients with benign prostatic hyperplasia (n = 13), LPC serum samples from patients with localized PC (n = 11), N1/M1 serum samples from patients with lymph node or distant metastases (n = 9), CRPC serum samples from patients with castration resistant PC (n = 11), Rel Exp relative expression levels of miRNAs normalized to miR-320a. Black line mean miRNA level for BPH samples and for samples from patients with disseminated PC, respectively. t test p value given for comparisons. b Receiver operating characteristic (ROC) analysis of miRNAs in BPH samples vs samples from patients with disseminated PC. AUC area under the curve. c ROC analysis of combinations of dichotomized miRNAs. miRNAs levels were classified as high or low based on ROC analysis. Cut-points were selected to maintain specificity at 100 %. AUC area under the curve of ROC analysis

Full size image

miRNAs de-regulated in serum samples from patients with disseminated PC compared to localized PC; development of miRNA panels

To further characterize serum miRNAs associated with aggressive disease, we compared serum samples from LPC patients to serum samples from patients with disseminated disease (N1/M1 and CRPC). miRNAs identified in this comparison could potentially aid in TNM staging of PCs in order to guide treatment decisions at the time of diagnosis. We identified 26 miRNAs more than two-fold de-regulated in serum samples from patients with disseminated PC vs LPC patients (Online Resource Table 5). To reduce the number of miRNAs for further analysis, we focused only on miRNAs more than three-fold de-regulated (Table 3, Online Resource Fig. 2). Ten miRNAs were more than three-fold up-regulated in patients with disseminated PC and displayed sensitivities ranging from 20 to 45 % when specificity was fixed at 100 % (Table 3, Online Resource Fig. 2). Spearman analysis was conducted to identify anti-correlated miRNA pairs, which could supplement each other in a test comprising multiple miRNAs (Online Resource Table 6). Three miRNAs each identified disseminated PC patients with a sensitivity of 45 at 100 % specificity (miR-149*, miR-200a, and miR-329). None of the de-regulated miRNAs were inversely correlated with miR-149* or miR-329 (Online Resource Table 6), but miR-200a displayed an inverse correlation to miR-1203 (rho = −0.242). When combined, miR-200a/miR-1203 identified 60 % of disseminated PC patients at 100 % specificity (Table 3, Online Resource Fig. 3). Closer inspection of Spearman correlation results revealed that only miR-375 and miR-708 were inversely correlated (rho = −0.474) to a greater extent than miR-200a and miR-1203 (Online Resource Table 6). A combination of miR-375/miR-708 identified 60 % of disseminated PC patients (Table 3, Online Resource Fig. 3). We further combined the two models and reached a sensitivity of 75 %, when maintaining specificity at 100 %. Thus, we have identified a panel of four miRNAs (miR-200a/miR-1203/miR-375/miR-708) which identified 75 % of disseminated cancer cases in our sample set and thus potentially could be of high value in classification of PC as localized or disseminated PC (Table 3, Online Resource Fig. 3). Of note, for some of these miRNAs (miR-1203, miR-129-5p, miR-149*, and miR-329), we also found elevated levels in serum samples from BPH control patients (Online Resource Fig. 2a). Thus, a final test for disseminated PC based on these candidate markers may need to be used in combination with markers that can accurately distinguish BPH (and other healthy controls) from PC samples.

Table 3 miRNAs more than three-fold de-regulated in serum samples from PC patients with disseminated PC when compared to LPC patients
Full size table

Comparison to previously published studies of circulating miRNAs in PC

To compare our results with previously published studies, we generated a comprehensive list of miRNAs that have been reported as circulating biomarker candidates for PC. Studies comparing the levels of circulating miRNAs in serum and plasma have suggested that either of these fluids can be used with similar results for both protein- and vesicle-associated miRNAs [15, 18]. We therefore reviewed both serum- and plasma-based studies for PC and searched the PubMed database (http://www.ncbi.nlm.nih.gov/pubmed/) using the phrases “miRNA or microRNA,” “circulation, serum or plasma,” and “prostate cancer.” Since 2008, eight miRNA profiling studies [18, 23, 25–27, 30–32] and six studies focusing on smaller subsets of miRNAs [23, 29–33] in serum or plasma from PC patients have been published. Thus, prior to the present study, a total of 104 miRNAs have been suggested as circulating biomarker candidates for PC. Strikingly, only nine of these miRNAs have been associated with PC in more than one study (Table 4). The limited overlap between these studies is likely a result of relatively small sample sizes combined with differences in starting materials (serum, plasma, or purified micro-vesicles), purification methods (e.g., different miRNA extraction kits or total RNA extraction, various methods of micro-vesicle purification), miRNA detection/screening methodologies (e.g., array-based methods or RT-qPCR), and/or marker discovery strategies including statistical analysis methods [19, 36]. However, given the heterogeneity of study designs, the miRNAs that have been repeatedly identified may have particularly high potential applicability as markers for PC, as they appear to remain associated with PC under varying experimental conditions. The nine miRNA candidate markers identified in two or more studies are miR-107, miR-141, miR-21, miR-200b, miR-221, miR-30c, miR-346, miR-375, and miR-574-3p (Table 4).

Table 4 Candidate circulating miRNA biomarkers for PC
Full size table

In the present study, we also detected increased levels of miR-141 and miR-375 in serum samples from patients with disseminated PC (Tables 2 and 3). Consistent with previous reports, miR-141 levels were highly variable in our PC patient sample set [29], and the levels of miR-141 and miR-375 correlated directly (Spearman’s rho = 0.742) across PC patient samples (Online Resource Fig. 2a, Online Resource Table 6) [23]. Thus, for miR-141 and miR-375, our data confirmed previously published findings, but the remaining seven recurrently identified miRNAs were not significantly de-regulated in our sample set (data not shown).

We next compared our lists of de-regulated serum miRNAs in PC patients (Tables 1, 2, and 3 and Online Resource Tables 3 and 5) to all of the 104 miRNAs presented in previous studies. miR-141, miR-375, and miR-609, all associated with disseminated PC in the present study, have previously been identified as up-regulated in serum or plasma samples from patients with disseminated/aggressive PC [18, 23–26, 29, 30]. miR-17*, which was identified in the BPH/PC comparison (Table 1), as well as miR-326 and miR-92b, which were identified in the LPC/advanced PC comparison (Online Resource Table 5), have also been previously reported as PC biomarker candidates, but have not been associated with the same stage of PC as suggested here. Thus, miR-17* has been reported to be up-regulated in plasma-associated micro-vesicles in serum from advanced PC patients compared to patients with localized PC [23], whereas miR-326 and miR-92b have been reported to be up-regulated in plasma-associated micro-vesicles and in serum from PC patients when compared to normal controls [23, 26].

A main difference between the present study and the majority of previously published studies is our use of the Exiqon RT-qPCR-based platform for screening/detection of miRNAs in serum. This platform has only been used in one other screening study, where miRNAs associated with plasma-derived micro-vesicles were profiled [23]. The method of miRNA measurement likely affects which miRNAs are identified as top candidate biomarkers, e.g., by the selection of specific miRNAs that are assayed by the given method as well as by assay specificity and efficiency. Thus, this difference in methodology is likely, at least in part, to explain the low frequency of recurrently detected miRNAs between our study and previously published studies. Nonetheless, the present study as well as earlier studies provides proof of principle for the potential application of serum miRNAs as markers for PC. Future studies using large well-defined patient cohorts combined with (yet to be developed) standardized procedures for sample collection, sample handling, and miRNA quantification will likely facilitate the development of circulating miRNA biomarkers for PC and ensure reproducibility and consistency between studies.

Origin of de-regulated miRNAs in serum from PC patients

Up-regulation of specific miRNAs in the circulation of PC patients has been shown to correlate with increased tumor tissue miRNA levels [18, 24, 25, 30]. However, inverse correlations between tumor and circulating levels of miRNAs have also been reported [25, 30]. By investigating miRNA expression levels in the publicly available Taylor et al. [22] prostate (cancer) tissue miRNA expression dataset, we looked into the potential origin of the miRNAs de-regulated in serum samples of our PC patient set when compared to BPH control patients (Tables 1 and 2). miR-375, miR-17*, miR-200b*, and miR-501-3p were up-regulated in both serum samples from PC patients and in tumor tissue samples (Table 1), as also previously reported for miR-375 [24, 25, 30], suggesting that prostate tumors could be directly responsible for the observed increased serum levels of these miRNAs. In contrast, miR-616 was up-regulated in serum samples from PC patients but down-regulated in PC tumor tissue (Table 2). Non-tumor tissues influenced by PC cells (e.g., adjacent prostatic tissues or immune cells) could be responsible for the observed serum increase in miR-616, but a model in which tumor cells actively secrete miR-616, thus reducing cellular levels and increasing extracellular levels, is also conceivable. miR-210, miR-551b, and let-7a* were all up-regulated in serum samples, with no evident alteration in PC tissue levels (Tables 1 and 2); thus, these miRNAs may be derived from other tissues than the primary tumor. Our observations resemble those of others [37, 38]; thus, some serum miRNAs likely originate from PC cells, whereas others could originate from other tissues, e.g., as part of a host response.

Conclusions

Here, we profiled serum samples from BPH and PC patients and identified several novel potential circulating miRNA biomarkers for PC. A thorough review of previously published studies, reporting on the potential of circulating miRNAs as biomarkers for PC, revealed that a very large number of miRNAs have been proposed as serum or plasma markers for PC, but few have been identified in more than one study. Paralleling this observation, our own analysis of de-regulated miRNAs showed some, but limited, overlap with already published miRNAs. It is highly likely that different starting materials, different purification methods, different analysis platforms, and different statistical methods influence the results of the individual studies. Furthermore, miRNAs often display highly fluctuating levels in serum and plasma samples from PC patients, complicating the task of finding single miRNAs that detect PC across many patient samples. Nonetheless, we and others consistently identified miR-141 and miR-375 as up-regulated in samples from patients with disseminated PC. Furthermore, as proof of principle, we demonstrated that by combining sets of inversely correlated miRNAs, we could generate models that predicted PC with high sensitivity while maintaining 100 % specificity. Thus, a panel of five miRNAs (miR-562/miR-210/miR-501-3p/miR-375/miR-551b) identified 84 % of all PC patients in our sample set, a panel of four miRNAs (let-7a*/miR-210/miR-562/miR-616) identified 80 % of patients with disseminated PC when compared to BPH patients, and another panel of four miRNAs (miR-375/miR-708/miR-1203/miR-200a) identified 75 % of disseminated PC patients when compared to localized PC patients. Thus, circulating miRNAs, including these panels of miRNAs, seem to hold great promise for future use in biomarker applications. However, to move circulating miRNAs closer to actual clinical utility, standardization of sample collection, RNA extraction procedures, and miRNA quantification methods is needed. Furthermore, assessment of the true biomarker potential of circulation miRNAs must be done in large-scale multi-center studies encompassing well-defined patient cohorts and clearly defined clinical endpoints and requires independent validation.