Identification of endogenous normalizers for serum MicroRNAs by microarray profiling: U6 small nuclear RNA is not a reliable normalizer

Abstract

We read the letter from Qi et al. regarding the use of endogenous normalizers for serum microRNAs with much interest. The authors describe a panel of serum miRNAs (miR-374a, miR-374b, and let-7d) that appear to be expressed at consistent levels between healthy and chronic disease groups. These microRNAs may be superior normalizers over currently used reference species. This is a young and evolving field, and this contribution is welcomed as it may help define parameters for the most accurate quantification of human circulating microRNAs. There is no single universally accepted reference species used in circulatory microRNA studies. In fact, the use of such normalizers employed in this field has been diverse, and include a range of exogenous and endogenous species. 1-4 In response to the findings by the authors, we have now profiled endogenous human let-7d-5p and miR-374a-3p (the closest available species using the TaqMan platform) alongside U6 small nuclear RNA (snRNA) and exogenous Caenorhabditis elegans-specific lin-4 in 15 randomly selected healthy controls and acute liver injury (ALI) patients from our study. We found that let-7d has low variation within healthy controls (coefficient of variation [CV] 1.7%) and ALI patients (CV 3.0%) (Fig. 1A). However, in common with U6 snRNA, we observed a modest but statistically significant difference (P < 0.001) in let-7d levels between healthy controls and ALI patients, which supports our notion that many reference molecules may be perturbed in these very sick, acutely injured patients. The serum profile of miR-374a-3p was less promising, with high variation in controls (CV 7.58%) and ALI patients (CV 7.21%) and low mean abundance (Ct = 40.1) (Fig. 1B). Furthermore, miR-374a-3p was not detected in six ALI patients. It would be interesting to profile the related miR-374a-5p in light of the authors' findings. The exogenous species, cel-lin-4, provided the least variation in controls (CV 1.5%) and ALI patients (CV 1.9%), while exhibiting no difference between groups (Fig. 1D). Importantly, we find that miR-122 is increased in ALI patients regardless of which normalizer we use (Fig. 1E-1H). Furthermore, as we report in our study (Supporting fig. 4A), serum miR-122 is increased in ALI patients when we remove normalization (Fig. 1I). These data support our findings and suggest that let-7d-5p in particular may add value as an endogenous normalizer, alongside or instead of U6 snRNA and other reference species in circulating microRNA studies. Identification of suitable internal controls by microarray profiling. (A) Serum miRNA expression profiles. Serum miRNA expressions from 117 human samples, including healthy controls at different ages (young group [n = 25; age, 27.4 ± 10.9 years]; aging group [n = 27; age, 58.3 ± 10.5 years]) as well as different disease conditions, including autoimmune type 1 diabetes (n = 31; age, 23.5 ± 12.1 years) and malignant melanoma (n = 34; age, 60.1 ± 16.7 years), were profiled with TaqMan Human MicroRNA array card A (v2.1), performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA), using the manufacturer's recommended protocol. The cycle threshold (Ct) values were obtained with SDS 2.3 and RQ manager 1.2 software (Applied Biosystems) and then data were analyzed with Real-Time StatMiner 4.2 software (Integromics, Inc., Granada, Spain). Of 332 expressed miRNAs, 58 miRNAs were consistently expressed across all samples and were chosen to determine a global mean for further global normalization. The −ΔCt [−(Ct-global mean)] was calculated, and heatmap analysis was performed with hierarchical clustering. (B) Internal normalizers for serum miRNA analysis. Pairwise comparisons among the four groups by a nonparametric Wilcoxon test identified three miRNAs (miR-374a, miR-374b, and let-7d), showed no significant differences among the four groups (P > 0.1), small variations (SD = 0.86, 0.78, and 0.82, respectively), and was very close to the global mean expression value. (C) Expression of miR-16 and U6 showed significant differences among the four groups (P < 0.01-0.001) with a large variation (SD of 1.46 and 2.85, respectively). (D) Expression of U6 was significantly different between the healthy young and aging groups. SD, standard deviation. Philip Starkey Lewis B.Sc., M.Res.*, James Dear Ph.D., M.R.C.P.*, Vivien Platt B.Sc., M.Sc.*, Jonathan Moggs B.Sc., Ph.D.*, Chris Goldring B.Sc., Ph.D.*, B. Kevin Park B.Sc., Ph.D.*, * Department of Pharmacology and Therapeutics, School of Biomedical Sciences, University of Liverpool, Liverpool, UK.