MicroRNA Expression for Early Prediction of Late Occurring Hematologic Acute Radiation Syndrome in Baboons

Abstract

For effective medical management of radiation-exposed persons after a radiological/nuclear event, blood-based screening measures in the first few days that could predict hematologic acute radiation syndrome (HARS) are needed. For HARS severity prediction, we used microRNA (miRNA) expression changes measured on days one and two after irradiation in a baboon model. Eighteen baboons underwent different patterns of partial or total body irradiation, corresponding to an equivalent dose of 2.5 or 5 Gy. According to changes in blood cell counts (BCC) the surviving baboons (n = 17) exhibited mild (H1-2, n = 4) or more severe (H2-3, n = 13) HARS. In a two Stage study design we screened 667 miRNAs using a quantitative real-time polymerase chain reaction (qRT-PCR) platform. In Stage II we validated candidates where miRNAs had to show a similar regulation (up- or down-regulated) and a significant 2-fold miRNA expression difference over H0. Seventy-two candidate miRNAs (42 for H1-2 and 30 for H2-3) were forwarded for validation. Forty-two of the H1-2 miRNA candidates from the screening phase entered the validation step and 20 of them showed a statistically significant 2-4 fold up-regulation relative to the unexposed reference (H0). Fifteen of the 30 H2-3 miRNAs were validated in Stage II. All miRNAs appeared 2-3 fold down-regulated over H0 and allowed an almost complete separation of HARS categories; the strongest candidate, miR-342-3p, showed a sustained and 10-fold down-regulation on both days 1 and 2. In summary, our data support the medical decision making of the HARS even within the first two days after exposure where diagnostic tools for early medical decision are required but so far missing. The miRNA species identified and in particular miR-342-3p add to the previously identified mRNAs and complete the portfolio of identified mRNA and miRNA transcripts for HARS prediction and medical management.

Fig 1. Screening and validation for samples eligible in Stages I and II of the two-Stage split study design.

Shown are the numbers of miRNAs included at each step of the eligibility process. Abbreviation: FC, fold-change. miRNA gene expression values within the linear-dynamic range of the methodology are called expressed. Preselected candidate miRNAs (detected miRNAs) had to fulfill certain criteria which became more defined for the selected candidates forwarded to Stage II.

Fig 2. The Figure summarizes qRT-PCR results (normalized threshold cycle [CT]-values) from Stage I (screening) comparing H1-2° severity of the hematological acute radiation syndrome (HARS) with H0° (unexposed, reference) at days 1 and 2 after radiation exposure.

Descriptive statistics reflect the distribution and the fold-change (FC) difference among miRNAs differentially expressed at H1-2°, with H0° as the reference. Mean values of both HARS groups were compared with the Students t-test. Concordance was calculated using logistic regression analysis. In the case of complete separation of data sets (concordance equals 100%), missing p-values using logistic regression analysis were replaced by p-values (Chi-Square) generated by frequency distribution comparisons. Samples showing the same direction in the FC on both days were merged into one category and analyzed together (last part of the figure). Genes with p-values <0.0025 survived Bonferroni correction for multiple comparisons (p = 0.05/20 miRNAs/hypothesis ~ 0.0025) and are marked with an asterisk.

Fig 3. Summarized qRT-PCR results (normalized threshold cycle [CT]-values) from Stage I (screening), Stage II (validation) and a combined analysis on all samples (Stage I and Stage II samples together).

MiRNAs values measured before irradiation (H0°) were compared with corresponding miRNA values measured at days 1 and 2 after irradiation in baboons developing a more aggravated H2-3° HARS. Descriptive statistics reflect the distribution and the fold-change (FC) difference among miRNAs differentially expressed at H2-3° with H0° as the reference. Mean values of both HARS groups were compared employing the Students t-test. Concordance was calculated using logistic regression analysis. In the case of complete separation of data sets (concordance equals 100%), missing p-values using logistic regression analysis were replaced by p-values (Chi-Square) generated by frequency distribution comparisons. Samples showing the same direction in the FC on both days were merged into one category and analyzed combined (last part of the figure). Genes with p-values <0.0033 survived Bonferroni correction for multiple comparisons (p = 0.05/15 miRNAs/hypothesis ~ 0.0033) and are marked with an asterisk.

Fig 4. Plots of fold changes (FC) for miR-342-3p and miR-146a relative to the mean H0°-values are shown separately for each blood sample taken on day 1 and day 2 after exposure.

The plots correspond with data shown in Fig 2 for these miRNAs. FCs generated on the samples for the screening (left side), the independent validation (middle) and all samples combined (right side) are presented. A jitter plot was chosen to avoid overlap of similar data. Circles, triangles and squares represent H0°, H1-2° and H2-3° HARS, respectively.

Fig 5. Graphical summary of predicted miRNA-mRNA interactions.

The miRNAs and mRNAs, represented by circles and squares, respectively, are grouped according to their expression levels with down-regulated mRNAs depicted in the top row, miRNAs in the middle row and down-regulated genes at the bottom. Visibility (thickness) of the lines connecting miRNAs and mRNAs is defined by the corresponding cumulative weighted context++ score. Context++ scores indicate how effective the miRNA might regulate the targeted mRNA. Visualization was done by means of Cytoscape. The upper panel depicts all 211 interactions, 29 mRNA (from a previous study) and 33 miRNAs from the present study. Predicted interactions of the most promising miR342-3p miRNA are depicted in the middle panel and expected reciprocal and stronger miRNA (miR185-3p)–mRNA interactions are shown in the lower panel.