Many RNA species are modified to contain non-canonical bases, a process known as RNA editing. The most prevalent type is adenosine-to-inosine editing, wherein adenosine deaminases (e.g., ADARs) enzymatically modify an adenosine base to an inosine base. This effectively changes the base pairing from thymine to cytosine and can influence a variety of things including secondary structure (formation of a bulge in the middle of a hairpin) and alter the sequence of a coded protein. Disruptions of A-to-I editing can lead to defects in hematopoiesis and neurological function, and global elevation of editing has been observed in most cancers.
It has been speculated that adenosine-to-inosine RNA editing influences subcellular localization patterns, such as nuclear retention, trafficking to nuclear paraspeckles, and presence at the transcription site. However, a lack of visualization tools has left this and other hypotheses untested.
We wanted to develop a system that would: 1. Provide single-nucleotide resolution of edited and unedited transcripts, 2. Visualize the subcellular localization of each of these transcripts, and 3. Allow us to characterize trends in RNA editing behavior on the single-cell level.
Discriminating between edited and unedited RNA via RNA fluorescence in situ hybridization (RNA FISH) is difficult because it relies on the hybridization of oligonucleotide probes to visualize the target of interest. Short oligonucleotides bind nonspecifically while long oligonucleotides cannot discriminate single-base differences. We thus used a ‘toehold probe’ strategy to reduce the initial hybridization region of our detection probes in order to confer selectivity based on single-nucleotide differences. We use two detection probes that compete to target the unedited, adenosine-bearing sequence using a thymine, and the edited, inosine-bearing sequence using a cytosine. Upon specific binding, the “mask” sequence is released by strand displacement to stabilize hybridization. However, single oligonucleotides are still prone to nonspecific binding, so we simultaneously used smFISH (the “mRNA guide” probe coupled to a unique fluorophore) to target a constant region of mRNA in the vicinity of the edited base.
As a control to confirm that detection probes did not colocalize with guide probes by random chance, we measured the rate of random colocalization by computationally shifting guide spots by 5 pixels in both the X and Y direction (“Pixel-shift”), thereby moving them outside the range of any true colocalization events.
As an additional control, we check for dye-specific effects by swapping the fluorophores on the detection probes.
To test inoFISH we chose the canonical, well-studied example of the Glutamate receptor 2 transcript (GRIA2). Combining four biological replicates, 10.53% of mRNA guides uniquely colocalized with adenosine or inosine detection probes, with 5.25% and 5.28% of GRIA2 guides colocalizing with the adenosine-detection and inosine-detection probes respectively). The estimated mean editing level for GRIA2 was 57.3% (95% confidence interval: 45.1%, 69.5%). Pixel-shift analysis reduced colocalization to 1.83% and 1.16% for adenosine and inosine, respectively, showing that most of colocalization events were specific.
GRIA2 mRNA is primarily edited by the enzyme ADAR2, so we used siRNA to knock down ADAR2 mRNA levels by 60% in SH-SY5Y cells. We observed a concomitant reduction in mean estimated GRIA2 editing level from 65% to 14% (Parametric bootstrap p = 0.0004)
We also chemically modified inosine bases with acrylonitrile on the N1 position to prevent base pairing to cytosine, reducing observed editing level from 52.1% to 13.5% (Parametric bootstrap p = 0.0006).
We can use inoFISH for subcellular localization analysis of editing targets. For example, here we’re using inoFISH to test whether the edited forms of three transcripts of interest (one editing site per transcript considered) in cells of interest are found more or less often in the nucleus than in the cytoplasm, or if instead there’s no detectable difference between editing levels in the two compartments. For the target in cell type pairs GRIA2 in SH-SY5Y, EIF2AK2 in U-87 MG, and NUP43 in U-87 MG we found no detectable differences in editing level between nuclear and cytoplasmic compartments.
We can use inoFISH for other types of subcellular localization analysis, as well. Here we checked whether the edited form of GRIA2 is preferentially found associating with paraspeckles in SH-SY5Y cells as compared with the unedited form of GRIA2. We captured 125 paraspeckle-associated GRIA2 transcripts, of which 21 were labelled with an editing status (either edited or unedited). The fraction of those labelled as edited at paraspeckles was not significantly different from the fraction of overall mean estimated editing level. Therefore, we don’t see an enrichment of edited GRIA2 transcripts at paraspeckles.
A third type of subcellular localization analysis made possible by inoFISH is visualizing the editing status of transcripts at their transcription sites. Here we show NUP43 transcription sites in U-87 MG cells. Since this is not a particularly highly expressed gene, the frequency of observing a transcription site was rather low. We observed 17 NUP43 transcription sites, at five of which we could label at least one transcript with editing status. All five were labelled as unedited. However, since these numbers are low, we cannot rule out the possibility that some NUP43 transcripts are co-transcriptionally edited.
We can simultaneously use inoFISH to understand single-cell trends in editing levels of targets of interest. Here we show different modeled datasets with the same population-wide editing level. Bulk analyses of RNA editing, such as bulk RNA-seq or other RT-PCR-based methods cannot resolve differences between these two scenarios, but inoFISH can. On the left, all cells have the same effective editing level; the number of observed edited transcripts per cell is drawn from a binomial distribution. On the right, some cells have near-total editing while other cells have near-zero editing.
Here we show transcript-specific single-cell editing level trends. On the left: GRIA2 editing levels vary from cell to cell more than expected under the binomial single-population model from the previous slide. In the bottom part of the left panel are the observed GRIA2 inoFISH results in single cells. In the top part of the left panel we present a density plot of simulated data likelihood values (log-transformed and negative) under the binomial model for GRIA2. The red line labelled “Obs.” is the observed -log(likelihood) for GRIA2 inoFISH data; the observed data are unlikely to come from the binomial model. On the right is a similar analysis for NUP43 editing in single cells. However, the binomial model does appear to fit the observed NUP43 data.