Absolute Measurement of Cardiac Injury-Induced microRNAs in Biofluids across Multiple Test Sites

Abstract

Extracellular microRNAs (miRNAs) represent a promising new source of toxicity biomarkers that are sensitive indicators of site of tissue injury. In order to establish reliable approaches for use in biomarker validation studies, the HESI technical committee on genomics initiated a multi-site study to assess sources of variance associated with quantitating levels of cardiac injury induced miRNAs in biofluids using RT-qPCR. Samples were generated at a central site using a model of acute cardiac injury induced in male Wistar rats by 0.5 mg/kg isoproterenol. Biofluid samples were sent to 11 sites for measurement of 3 cardiac enriched miRNAs (miR-1-3p, miR-208a-3p, and miR-499-5p) and 1 miRNA abundant in blood (miR-16-5p) or urine (miR-192-5p) by absolute quantification using calibration curves of synthetic miRNAs. The samples included serum and plasma prepared from blood collected at 4 h, urine collected from 6 to 24 h, and plasma prepared from blood collected at 24 h post subcutaneous injection. A 3 parameter logistic model was utilized to fit the calibration curve data and estimate levels of miRNAs in biofluid samples by inverse prediction. Most sites observed increased circulating levels of miR-1-3p and miR-208a-3p at 4 and 24 h after isoproterenol treatment, with no difference seen between serum and plasma. The biological differences in miRNA levels and sample type dominated as sources of variance, along with outlying performance by a few sites. The standard protocol established in this study was successfully implemented across multiple sites and provides a benchmark method for further improvements in quantitative assays for circulating miRNAs.

Keywords: biomarker; interlaboratory; microRNA; variance.

FIG. 1

Outline of standard protocol for sample processing and analysis.

FIG. 2

Plasma cTnI levels after sc administration of isoproterenol or saline. Results are shown from Phase 2 and 3 study samples collected 4 h after dosing with saline control (n = 12) or 0.5 mg base/kg of isoproterenol (treated) (n = 13) and 24 h after dosing with saline control (n = 5) or 0.5 mg base/kg of isoproterenol (n = 6). Levels lower than the limit of quantitation (LOQ) (0.1 ng/ml) were graphed at the LOQ. The mean and SD are shown for treated groups.

FIG. 3

Sources of variability attributable to specific effects within datasets. The inverse predictions from the Phase 2 (A) and Phase 3 (B) datasets, weighted by an estimate of precision, were fit to a variance components model to estimate the proportion of total variance attributable to included effects and interactions between effects.

FIG. 4

Heat maps of miRNA*Site least squares means. Hierarchical clustering of least squares means by site (coded by letter) and miRNA was performed for (A) Phase 2 plasma and serum samples, (B) Phase 3 plasma samples, and (C) Phase 3 urine samples. The heatmap greyscale denotes the range of least squares means across all three panels. The dendrogram heights indicate Euclidean distance.

FIG. 5

Levels of miR-1-3p and miR-208a-3p in control and isoproterenol-treated rat plasma and serum samples. Data is combined across Phase 2 and Phase 3 studies for miR-1-3p (A) and miR-208a-3p (B) from the sites indicated in the figure legend for (A). Estimated values in log₁₀ copies per µL biofluid are shown for 3 day-to-day replicates per site. Data points below the lowest calibrator concentration, with no computed value in 3PL models, or with an “undetermined” C_T value in qPCR assays were categorized as “below assay limits.” The 24-h time point samples were derived from calibration curves where carrier RNA was included in the diluent.

FIG. 6

Levels of miR-16-5p and miR-192-5p in control and isoproterenol-treated rat plasma and serum samples. Data for miR-16-5p (A) is from the Phase 2 study and the data for miR-192-5p (B) is from the Phase 3 study for the sites indicated in the figure legend for (A). Estimated values (log₁₀ copies per µL biofluid) are shown for 3 day-to-day replicates per site. The 24 h time point samples were derived from calibration curves where carrier RNA was included in the diluent.