Background

MicroRNAs (miRNA) are highly conserved small non-protein-coding RNAs (21-22 nt) that regulate post-transcriptionally gene expression and modulate fundamental biological processes like development and cell homeostasis ( Lagos-Quintana et al., 2001; Fabian et al., 2010; Bartel, 2018), including as well several pathologies where miRNA expression correlates with tumor progression and aggressiveness (Lu et al., 2005; Di Leva and Croce, 2013; Krishnan et al., 2015; Bertoli et al., 2015 and 2016). The study of miRNAs has boosted the evolution of numerous biochemical and computational techniques that unravel new mechanisms and networks applicable to clinical scenarios. These high-throughput procedures include profiling methods (microarrays and NGS), in vivo target validation and single-molecule imaging (Steinkraus et al., 2016). In contrast, functional reporters that directly assess miRNA-target repression have been kept simple and with low resolution.

Deep sequencing and mass spectrometry have increased the possibilities to assess miRNAs and targets expression (Ender et al., 2008; Yang et al., 2009; Brameier et al., 2011; Schramedei et al., 2011; Bai et al., 2014a and 2014b; Muller et al., 2015; Yu et al., 2015), since quantifying the number of RNA molecules and running statistical analyses for the differential expression (miRNAs or mRNA) can generate hypotheses that need further experimental evaluation, using, for example, miRNA exogenous expression or inhibition. In brief mRNA analysis on cDNA level (qPCR or next-generation sequencing) estimates RNAs degradation, mass spectrometry defines the affected coding targets (Yang et al., 2009), and ribosome profiling determines which miRNAs targets are translationally repressed (Bazzini et al., 2012).

Moreover, to narrow down the analysis to the Argonaute bound miRNA captured-based profiling has been used, these profiles rely on RNA isolation from ribonucleoproteins, cDNA synthesis and its quantification (qPCR, microarrays or NGS). Crosslink-immunoprecipitation (CLIP) together with high-throughput sequencing increase enormously our understanding of miRNA regulation (Hafner et al., 2010), to the extent that including an RNA-RNA ligation step (in the capture) allows detecting miRNA-mRNA hybrids called chimeras (Helwak et al., 2013; Grosswendt et al., 2014; Moore et al., 2015).

FRET-based assays in vitro have calculated kinetic parameters for AGO-miRNA binding and RISC formation, increasing the structural basis of miRNA target recognition and suggesting that AGO-miRNA behaves more like RNA binding domain rather than RNA-RNA solely interaction (Wee et al., 2012; Salomon et al., 2015). Using this vast amount of information computational scientist developed algorithms to understand miRNA as an intricate network (Vera et al., 2013; Lai et al., 2016), expanding further with these integrative models the concepts of miRNA as fine-tuners and switches of gene expression (Bartel and Chen, 2004).

The increased resolution of sequencing methods did not boost the development of better in vivo reporters. Usually, studies involved in exhaustive biochemical characterizations (Bait profiling) tested functionally miRNA-mRNA interaction using luciferase reporter assays (Helwak et al., 2013; Hasler et al., 2016; Steinkraus et al., 2016). These experiments require two proteins (a reference protein and another with a miRNA response element) to create quantitative results (ratios) for miRNA regulation, dismissing the intricate functional network behind miRNAs; creating a misbalance between the high-resolution biochemical outputs (NGS, CLIP-Seq, FRET, etc.) and low-resolution reporters.

For that reason, we implemented a system for analytical flow cytometry (Denzler et al., 2016), using a single plasmid reporter system and validated it for miRNAs (Lemus-Diaz et al., 2017). To increase resolution, we used fluorescent proteins instead of luciferases, analyzed single cells by flow cytometry, and processed the data using the R environment. Furthermore, we adopted a titration model for miRNAs regulation, tested and validated its prediction, creating three categorical variables that integrate miRNA binding and expression (Mukherji et al., 2011; Lemus-Diaz et al., 2017), which can be used to estimate miRNA proficiency (Garcia et al., 2011).

Here, we provide a detailed protocol that includes plasmid generation, transfection, data acquisition, data handling and plotting using a simplified code. The reporter described here expresses two fluorescent proteins (CFP and YFP) under control of two constitutive promoters (Figure 1A), while one protein is the reference (YFP); the other has a miRNA response element (CFP). In the empty plasmid (CFP w/o miRNA target site) the two fluorescent proteins are expressed proportionally (Figure 1B), while in a plasmid with a miRNA response element this proportionality is shifted (Figure 1C).

To tidy up the raw events and analyze them using a threshold model for miRNA regulation, we process the data into a transfer function (called here Analytical function) by binning the reference protein intensities and calculating the mean for sensor protein (Bosson et al., 2014; Denzler et al., 2016). For our construct, we transfer the raw data in FCS 2.0 format into the R environment using FlowCore Bioconductor package (Gentleman et al., 2004; Hahne et al., 2009; Huber et al., 2015), then we logarithmically transform the YFP relative intensities, bin them at 0.05 intervals, and calculate the average of the log CFP intensities of each range.


Figure 1. Dual fluorescence-reporter analytical flow cytometry. A. The plasmid contains two fluorescent proteins (YFP and CFP) with two constitutive promoters and a miRNA response element. B and C. To generate analytical functions using raw cytometry data (Grey dots), the YFP relative intensities are binned at 0.05 intervals, and the average CFP intensity per bin is calculated (Green dots). Transfection of HEK 293 with two constructs: (B) Non-cognate and (C) miR-27-3p targeted insert.