A Plasma Biomarker Panel of Four MicroRNAs for the Diagnosis of Prostate Cancer

Abstract

Prostate cancer is diagnosed in over 1 million men every year globally, yet current diagnostic modalities are inadequate for identification of significant cancer and more reliable early diagnostic biomarkers are necessary for improved clinical management of prostate cancer patients. MicroRNAs (miRNAs) modulate important cellular processes/pathways contributing to cancer and are stably present in body fluids. In this study we profiled 372 cancer-associated miRNAs in plasma collected before (~60% patients) and after/during commencement of treatment (~40% patients), from age-matched prostate cancer patients and healthy controls, and observed elevated levels of 4 miRNAs - miR-4289, miR-326, miR-152-3p and miR-98-5p, which were validated in an independent cohort. The miRNA panel was able to differentiate between prostate cancer patients and controls (AUC = 0.88). Analysis of published miRNA transcriptomic data from clinical samples demonstrated low expression of miR-152-3p in tumour compared to adjacent non-malignant tissues. Overexpression of miR-152-3p increased proliferation and migration of prostate cancer cells, suggesting a role for this miRNA in prostate cancer pathogenesis, a concept that was supported by pathway analysis of predicted miR-152-3p target genes. In summary, a four miRNA panel, including miR-152-3p which likely targets genes with key roles in prostate cancer pathogenesis, has the potential to improve early prostate cancer diagnosis.

Figure 1

Screening of plasma miRNA markers in pooled patient and control samples and validation in the discovery cohort. (a) Scatter plot of 372 cancer-associated miRNAs in a screening cohort of pooled plasma samples from prostate cancer patients and healthy controls screened using a miScript miRNA PCR array. miRNAs found to be differentially expressed between the patient and control groups are shown in black, and unchanged miRNAs are shown in grey. A fold regulation cut-off of 2.5 was selected for the analysis. (b) Scatter plot showing the 11 differentially regulated miRNAs re-analysed by qRT-PCR in patient samples in the discovery cohort (N = 61). The mean fold regulation of each miRNA across the patient and control samples was taken into account and those that were below the 2 fold regulation cut-off were excluded from further analysis. The selected miRNAs are shown in colour. (c) Relative levels of miR-4289, miR-326, miR-152-3p and miR-98-5p analysed by qRT-PCR as in (b) in patients vs healthy controls. Statistically significant differences were assessed using a Mann-Whitney U test; p values are shown after Bonferroni correction for multiple testing. Each data point represents a plasma sample, the horizontal middle line in each data set represents the mean, and the limits of the vertical lines represent the standard deviation.

Figure 2

Relative expression of miR-4289, miR-326, miR-152-3p and miR-98-5p in the validation cohort (N = 58). Statistically significant differences in miRNA expression levels between the patients and control groups were assessed using a Mann-Whitney U test; p values are shown after Bonferroni correction for multiple testing. Each data point represents a plasma sample, the horizontal middle line in each data set represents the mean, and the limits of the vertical lines represent the standard deviation.

Figure 3

ROC curve analysis in the discovery, validation and combined cohorts comparing the ability of the miRNA signature to identify prostate cancer patients. (a) A combined measure of the sensitivity and specificity of the miRNA signature in the discovery cohort (N = 61) is represented by the Area under the curve AUC = 0.82 (p < 0.0001, 95% CI = 0.72–0.93). (b) A combined measure of the sensitivity and specificity of the miRNA signature in the validation cohort (N = 58) is represented by AUC = 0.95 (p < 0.0001, 95% CI = 0.89–1.00). (c) A combined measure of the sensitivity and specificity of the miRNA signature in a combined cohort (N = 119) is represented by AUC = 0.88 (p < 0.0001, 95% CI = 0.82–0.94). The diagonal reference line reflects the performance of the diagnostic test i.e. whether a test yields the positive or negative results by chance or due to a relation with the true disease status.

Figure 4

Analysis of published miRNA transcriptomic data from clinical samples. TCGA data expression analysis of miR-152-3p, miR-98-5p and miR-326 in 52 tumour and adjacent non-malignant prostate tissues. The expression of miR-152-3p was significantly lower (p = 0.0011) in tumour compared to adjacent non-malignant prostate tissues, while the expression of miR-98-5p and miR-326 did not reach statistical significance (p = 0.6288 and p = 0.4182). TCGA data was not available for miR-4289. The differences between the paired samples were assessed using a Wilcoxon test.

Figure 5

miR-152-3p mediates cell proliferation and migration in LNCaP cells. (a) Overexpression of miR-152-3p using miRNA mimics in LNCaP cells increased their proliferative capacity (p = 0.0002) measured as an increase in percentage confluence by the IncuCyte live-cell imaging system. (b) Overexpression of miR-152-3p also increased migration in LNCaP cells (p = 0.0084) measured as an increase in percentage relative wound density. (c) An increase in proliferation was accompanied by a change in morphology in miR-152-3p treated LNCaP cells when compared to non-targeting negative control treated cells. (d) LNCaP cells were grown to form a confluent monolayer before scratches were made and wound healing was measured by the IncuCyte system. Both the functional assays were performed for a period of 72 hours and data was collected at every 2 hour time point throughout the experiments. The differences between the miR-152-3p and negative control treated cells were assessed using a Mann-Whitney U test, N = 3.

Figure 6

Ingenuity pathway analysis (IPA) of miR-152-3p-target interactions in prostate cancer signalling. Prostate cancer signalling was among the top ten canonical pathways consisting of eight deregulated miR-152-3p target genes (highlighted in purple) i.e. FGFR3, IRS1, SOS2, HSP90AA1, KRAS, CDKN1B, CCND1 and PTEN where FGFR3 and IRS1 are closely related to PI3K and hence shown as a group. Some of the targets for e.g. HSP90, KRAS and SOS exist as a complex or group of genes. The p value was represented as −log p = 5.57 for the analysis.

Figure 7

Study design and discovery and validation cohort study profiles. Flow diagram summarising the methodology and statistical approach used to identify diagnostic miRNAs followed by in silico target identification and pathway analysis, TCGA data expression analysis and functional assays.