MicroRNAs (miRNAs) are short (~22 nucleotides), non-coding RNA molecules that post-transcriptionally regulate gene expression. As the miRNA field is still in its relative infancy, there is currently a lack of consensus regarding optimal methodologies for miRNA quantification, data analysis and data standardization. To investigate miRNA measurement we selected a panel of both synthetic miRNA spikes and endogenous miRNAs to evaluate assay performance, copy number estimation, and relative quantification. We compared two different miRNA quantification methodologies and also assessed the impact of short RNA enrichment on the miRNA measurement. We found that both short RNA enrichment and quantification strategy used had a significant impact on miRNA measurement. Our findings illustrate that miRNA quantification can be influenced by the choice of methodology and this must be considered when interpreting miRNA analyses. Furthermore, we show that synthetic miRNA spikes can be used as effective experimental controls for the short RNA enrichment procedure.
MicroRNAs (miRNAs) are short (~22 nucleotides), non-coding, RNA molecules that control diverse biological processes, including cell fate determination, cell proliferation, cell differentiation, and cell death (1, 2). miRNAs regulate gene expression post-transcriptionally by interacting with and down-regulating target mRNA molecules (3-6). There has been a growing interest in their use for clinical applications, largely due to their high stability and cell/tissue-specificity (7, 8). Most notably, the recent discovery of miRNAs in the circulating bloodstream led to extensive investigations into their suitability as biomarkers for diseases including cancer (8-12) and cardiovascular disease (13, 14).
If the recent advances in miRNA research are to be successfully implemented, there are still many challenges that need to be addressed, particularly those associated with the accuracy and reproducibility of miRNA quantification measurements. There is currently little consensus as to which are the optimal methodologies for sample collection, miRNA isolation, miRNA quantification, and data analysis. Furthermore, repeatability and reproducibility, which are especially important for successful clinical application of miRNAs as biomarkers, require the development of robust tools to enable the standardization of miRNA data. There are a range of techniques used to quantify miRNA expression (15-21), but RT-qPCR is currently considered to be the gold standard due to its unparalleled sensitivity and specificity (22-26). There are several commercially available miRNA RT-qPCR assays that employ diverse approaches to address the challenges of achieving accurate miRNA quantification (20, 21, 25). Although these assays have been described in previous publication, they have not been extensively compared experimentally. The effects of upstream variables, such as sample preparation, on downstream miRNA quantification have received less attention in previous investigations (27-33). For example, RNA samples are routinely prepared for miRNA analysis by enriching the short RNA (<200 nucleotides) fraction. To our knowledge the effects of enrichment on RT-qPCR measurement, however, have not yet been extensively investigated. In addition to issues affecting the accuracy of miRNA quantification, possibly one of the most important and difficult challenges that needs to be addressed by the field is the standardization of miRNA data. Typically, miRNA profiling involves a series of steps that are highly sensitive to technical manipulations; therefore, there is an urgent need for methods to standardize various procedures for within-platform or cross-platform comparisons. Considerable work has been done to facilitate data normalization through the identification of stable reference genes (34) and global mean strategies (35). Several studies have also investigated the use of external standards to control for variability in RNA extraction between samples (8, 36-38). However the use of standards to control for downstream RNA processing procedures, such as short RNA enrichment, has not previously been evaluated.
In the present study we compare two of the most prominent commercially available miRNA RT-qPCR assays — Life Technologies’ Taqman miRNA Assay (20) and Exiqon's miRCURY LNA Universal RT microRNA PCR assay — and evaluate the impact of a short RNA enrichment method on RT-qPCR measurement. Furthermore, we investigate the application of external Arabidopsis thaliana miRNAs standards for use as experimental procedure controls.
Materials and methods
Synthetic miRNA oligonucleotides used for spike-in material and standard curves were synthesized by Eurofins MWG Operon (Ebersberg, Germany) and quantified by measuring absorbance at A260 nm (Nanodrop, Thermo Scientific, Waltham, MA, USA) and purity assessed by calculation of the A260/A280 ratio. 50 ng/µl of human fetal brain total RNA (Cat. No. 540157, Agilent Technologies, Santa Clara, CA, USA) was spiked to make the following concentrations of synthetic Arabidopsis miRNAs (miR-159a: 1.2 × 107 copies/µl, miR-172a: 6.4 × 105 copies/µl, miR-394a: 1.6 × 106 copies/µl). For standard curves, synthetic miRNA molecules were spiked into yeast transfer RNA (tRNA) (10ng/µL) (Cat. No. R5636, Sigma, St. Louis, MO, USA) carrier solution at the following miRNA copy number ranges per reaction: 105 - 108 for let-7a, let-7c, miR-16, -394a, -26b, -159a; 104 – 107 for miRs-21 and -172a. All RNA was stored in RNA Storage Solution (Cat. No. AM9937, Life Technologies, Carlsbad, CA, USA) after dilution at -80°C.
Short RNA enrichment
Short RNA enrichment was performed using the miRVana miRNA isolation kit (Cat. No. AM1560, Life Technologies) according to the manufacturer's instructions and ethanol from Sigma (Cat. No. 1147662). In addition to the extraction of total RNA in which the short RNA fraction is retained and optionally enriched, the miRVana miRNA isolation kit also enables the user to enrich the short RNA fraction that was isolated by another method. The latter procedure, which we performed in this study, involves the separation of the larger (>200 nucleotides) and shorter (<200 nucleotides) RNA species in a sample. Briefly, total RNA samples were mixed with 5 volumes of Lysis/Binding buffer and 1/10 volume of miRNA homogenate additive and left on ice for 10 min. A low concentration of ethanol (25% v/v) was then added to the samples which were subsequently mixed and bound to a filter cartridge by centrifugation. The relatively low concentration of ethanol in this first treatment allows the binding of the larger RNAs to the column while the shorter, more soluble RNAs pass through and are collected. In the second step, a higher concentration of ethanol was added to the eluant (40% v/v), allowing the shorter RNAs to be immobilized to filter cartridges during centrifugation and subsequently eluted. This procedure was performed on 50 µL of total human fetal brain RNA (2.5 µg) (Cat. No. 540157, Agilent Technologies, Santa Clara, CA, USA) containing synthetic Arabidopsis miRNA spike-ins. Short RNA was eluted in 50 µL elution buffer so that the effective volume was identical before and after the enrichment.
Reverse transcription reactions for the Taqman miRNA Assays (Life Technologies) were performed using the Taqman microRNA Reverse Transcription Kit (Cat. No. 4366596, Life Technologies) with 10 ng total RNA containing Arabidopsis miRNA spike-ins, per 7.5 µL reaction. Ten µl reverse transcription reactions for the miRCURY LNA Universal RT microRNA PCR assays (Exiqon, Vedbaek, Denmark) were performed using the Universal cDNA synthesis kit (Cat. No. 203300, Exiqon) with 20 ng total RNA with Arabidopsis miRNA spike-ins per reaction. The differences in RNA used between the two kits reflect the recommended working concentrations. All reverse transcription reactions were performed in accordance with the manufacturers’ protocols (Life Technologies: Cat. No. 4364031, Rev. E, 01/2011; Exiqon: Cat. No. 203300, Version 4.1, 08/2011) with the exception that half volume reactions were used. Reverse transcription thermocycling parameters were as follows: for Life Technologies assay: 16°C for 30 min, 42°C for 30 min, 85°C for 5 min; for Exiqon assay: 42°C for 60 min, 95°C for 5 min. For short RNA enriched samples, the same volumes of RNA eluant were added to each reaction as used for the corresponding non-enriched samples. Reactions were performed on a DNA Engine Tetrad 2 Thermocycler (BioRad, Hercules, CA, USA). All reactions were performed in triplicate and included the following controls: no template (RT NTC), no reverse transcriptase enzyme (no RT) and yeast carrier tRNA (carrier only). cDNA was stored at -20°C for less than one week prior to qPCR analysis.
Quantitative real-time PCR (qPCR)
qPCR was performed in accordance with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines (Supplementary Table S1); however, primer sequences used for the different methods are not currently provided by the kit providers. Prior to qPCR reactions being performed, cDNA was diluted 1 in 5 and 1 in 80 for the Life Technologies and Exiqon assays respectively. The Taqman miRNA assays (Cat. No. 4427975, Life Technologies) and miRCURY LNA Universal RT microRNA PCR assays (Cat. No. 206999, Exiqon) were performed using the Taqman Universal PCR mastermix (Cat. No. 4304449, Life Technologies) and Sybr Green mastermix, Universal RT (Cat. No. 203450, Exiqon) respectively as per manufacturers’ instructions (see Reverse transcription for references to manufacturers’ protocols) with the exception that 10 µL reactions (half volume) were used for the Life Technologies assays, as was also used for the Exiqon assays. Supplementary Table S3 gives details of the RT-qPCR assays used. All assays were available off-the-shelf with the exception of the Exiqon Arabidopsis miRNA assays, which were custom designed. Details of the design of the Arabidopsis miRNA assays can be found at www.exiqon.com/miRNA-qPCR-primers. Reactions were performed according to the manufacturers’ instructions using a Prism 7900HT Real Time PCR system (Life Technologies). qPCR thermocycling conditions were as follows: for Life Technologies assay: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min; for Exiqon assay: 95°C for 10 min, followed by 45 cycles of 95°C for 10 s, 60°C for 1 min, melt curve analysis performed between 60–95°C for 15 min at a ramp-rate of 1.6°C/s. Quantification was performed using the standard curve method with four serial dilutions (see RNA samples for concentrations) and three replicates at each dilution. Interpolated miRNA copy number values were normalized for differences in input RNA between the two RT-qPCR technologies. PCR NTCs were run as controls on all plates in addition to the reverse transcription controls (RT NTCs, no RTs and carrier only). The SDS software v2.4 (Life Technologies) was used to calculate the quantification cycle (Cq) value, which is defined as the number of cycles at which the fluorescence signal is significantly above the threshold, which was converted to copy number using the relevant standard curve.
The data were analyzed using linear mixed effects models with maximum likelihood estimation. The R statistical programming environment (build 2.13.0) (www.r-project.org/), running under Microsoft Windows XP, was used to perform all statistical analysis, and to produce graphical output. Sample replicate and qPCR plate were modeled as random effects, while the miRNA target assay technologies used and enrichment status (total RNA or enriched) were modeled as fixed effects. The inclusion of all eight miRNAs in the same model entails pooling all the residuals; however, given the within group variances, this was justifiable and increased the number of residual degrees of freedom. The random effects were, as with linear models of this type, assumed to be normally distributed with a mean of 0 and variance representing the unit-to-unit variability (run-to-run or plate-to-plate in this case). Copy numbers were log-transformed in order to stabilize the within-group variances and to better approximate a normal distribution. Further details describing the statistical analysis can be found in the Supplementary Material.
Results and discussion
RT-qPCR is the method of choice for the accurate quantification of miRNA expression (22-26), and there are several commercially available miRNA RT-qPCR methods that employ distinct approaches to prime the miRNA for reverse transcription and then amplify the cDNA. A previous study compared two of these assays, one of which employs a universal tailing reverse transcription primer platform and another, a sequence-specific stem-loop reverse transcription primer platform (26). These two assays were compared in terms of their sensitivity and specificity, revealing that the former assay was less specific as it generated more non-specific products but was generally more sensitive as it detected more low abundance miRNAs successfully. The majority of the low abundance products detected by the universal tailing method, however, were non-specific (26). In our study, however, we do not assess the sensitivity or specificity of miRNA RT-qPCR analysis, but investigate measurement precision, accuracy, and linearity, and the effect of short RNA enrichment on these parameters, as these important performance characteristics have not been previously evaluated. We quantified miRNA levels using two popular RT-qPCR technologies: Life Technologies’ Taqman miRNA Assay (20) and Exiqon's miRCURY LNA Universal RT microRNA PCR assay, as these have not to our knowledge been previously compared. The Taqman miRNA assay uses a miRNA-specific stem-loop reverse transcription primer to generate cDNA for subsequent hydrolysis probe qPCR amplification (20). In contrast, the miRCury LNA Universal RT PCR assay involves the poly(A)tailing of mature miRNAs and the subsequent use of a poly(T) reverse transcription primer containing a 3′ degenerate anchor and 5′ universal tag to generate cDNA for amplification by SYBR Green qPCR using miRNA-specific forward and reverse primers containing locked nucleic acids (LNAs). The use of miRNA-specific forward and reverse primers offers an advantage in terms of the specificity of amplification, compared with other assays of similar design. For example, Qiagen's miScript PCR System uses a miRNA-specific forward primer but with a universal reverse primer for the qPCR, possibly leading to a comparatively lower specificity of amplification. The Exiqon system also offers the potential advantage of using LNA containing primers. LNA molecules are modified RNA nucleotides that are mixed with DNA or RNA nucleotides to increase hybridization properties (39). Levin et al. (40) demonstrated, however, that LNA containing primers can display poorer amplification efficiencies than those that do not contain LNA molecules (40).
We analyzed the expression of five endogenous human miRNAs and three Arabidopsis thaliana synthetic spikes. The human miRNAs were selected on the basis that they are well characterized and show a high abundance across a wide range of cell and tissue types. For the synthetic Arabidopsis spike-ins, well-characterized miRNAs were also chosen as opposed to synthetic or re-assorted existing miRNAs because artificial oligonucleotides may not exhibit the same physiochemical properties of natural miRNA species. Furthermore, the three Arabidopsis miRNAs were selected because they displayed no homology with any known human miRNAs. The suitability of the Arabidopsis miRNAs was confirmed by experiments that showed there was an absence or a very low-level of cross-reactivity (Cq >35) with each other, background human brain total RNA, or a yeast tRNA carrier (data not shown).
The performance of the assays, in terms of efficiency and variability of efficiency, were evaluated by measuring serial dilutions of synthetic miRNAs in three independent experiments and miRNAs quantified using both qPCR technologies (Table 1). The RT-qPCR technologies did not differ significantly in their assay efficiency (Table 1, all R2 values >0.988), but the Exiqon assay efficiency measurement appeared to be more variable between runs when compared with the Life Technologies assay (Table 1), as the latter usually showed a lower average standard deviation across the three repeats. However, these differences in standard deviation were not statistically significant.