Cooperation of alternative splicing and microRNA targeting in the gene regulation network of Arabidopsis thaliana

MicroRNA (miRNA) is a typical class of small RNAs that could modulate gene expression in trans at the post-transcriptional level. miRNAs bind to the miRNA binding sites (MBSs) in target mRNAs by sequence complementarity. Alternative splicing (AS) is another commonly occurred process in pre-mRNAs that changes the isoforms of a gene. It is hypothesized that there should be an interaction for gene regulation that involves both AS and miRNA targeting. Studies have verified this hypothesis in the model organism Arabidopsis thaliana. High-throughput sequencing data suggested that in Arabidopsis a considerably large fraction of MBSs are affected by AS events. The overlapping between MBS and AS exceeds the randomly simulated number. Functional experiments have indicated that the AS events are required for the gene expression changes of miRNA targets. Therefore, AS and MBS are mutually favored. The observed expression changes caused by miRNAs could also be contributed by AS events. In the present perspective article, we propose that the AS analysis should be incorporated in the differential-expression analysis of miRNA studies. When defining a differentially-expressed gene, it should be clarified whether the change in gene expression is caused by AS events or solely by miRNA targeting.

even higher when the poorly expressed genes in the samples were excluded by conducting high-throughput RNA sequencing (RNA-Seq) (Filichkin et al., 2010).
Nevertheless, it is difficult to quantify the amount of each transcript solely from RNA-Seq data. The reason is: many transcripts have overlapped exons.

Plant microRNAs
Similar to the post-transcriptional AS events, the microRNA (miRNA) system appears as a typical class of small RNA that regulates gene expression in trans (Bartel, 2009). The miRNA precursors are finally processed into 20-24 nt mature miRNAs (Bartel, 2009). In plants, an endonuclease named DCL1 (Dicer-like 1) is responsible for processing miRNAs (Papp et al., 2003). The mature miRNAs would be transported to the cytosol to form the RISC (RNA-induced silencing complex) with other factors.
RISC facilitates the miRNA to find its target genes by sequence complementarity (Bartel, 2009). The mRNA sequences targeted by miRNAs are the miRNA binding sites (MBSs). The miRNA cleaves its target transcripts and represses both the gene expression (Llave et al., 2002;Reinhart et al., 2002) and translation (Brodersen et al., 2008). So far, databases have collected and stored hundreds of miRNA species in a wide range of plants (Kozomara and Griffiths-Jones, 2011). It is not surprising that various miRNAs are experimentally verified to play crucial roles in many plant development stages and stress responses (Jones-Rhoades et al., 2006;Voinnet, 2009).
Researches (including our owns) have revealed the continuously emerging of species-specific miRNAs during evolution (Fahlgren et al., 2007;Molnar et al., 2007;Rajagopalan et al., 2006;Yang et al., 2012;Yang et al., 2011;Zhu, 2008). Of course, when a newly-emerged miRNA obtains MBSs in target genes, it would participate in the global miRNA regulation network and expand this network by regulating its own targets (Fahlgren et al., 2007;Rajagopalan et al., 2006). Since sequence complementarity is the rule of miRNA targeting, the conserved miRNAs across plant lineages should have similar or even identical MBSs in homologous plant genes (Axtell and Bowman, 2008). Conceivably, mutations in MBSs might cause the loss of MBSs so that they could have larger impact than mutations in other regions. As a result, the MBSs are maintained by purifying selection (Guo et al., 2008). Indeed, it is observed that the mutations leading to beneficial miRNA-target interaction have been selectively maintained while those deleterious mutations for miRNA-target interaction undergo genetic drift and purifying selection in humans (Chen and Rajewsky, 2006) or plants (Axtell and Bowman, 2008), respectively.

Results and Discussions miRNA profile in Arabidopsis
To accomplish the task of miRNAs, the MBSs in target genes must be properly maintained during the splicing process. If the MBS-containing region is spliced out, then the so-called target gene is no longer affected by the miRNA. Interestingly, software has been developed to quantify the differentially-expressed genes (DEG) between two samples or among multiple samples (Love et al., 2014), which is commonly used in the analysis where miRNA-treated samples are compared to control samples. Meanwhile, software on differential splicing was developed to quantify the isoform switch and relative isoform abundance (Shen et al., 2014). It is completely possible that the miRNA effect and the splicing switches both contribute to the observed expression changes. Thus, the crosstalk and cooperation of AS and MBS should be taken into account.

Alternative splicing in Arabidopsis
In Arabidopsis thaliana, the mRNA splicing changes could generate transcripts that differ in the MBS. In other words, AS events cause the gain or loss of MBS ( Figure   1A). This phenomenon influences a substantial fraction of genes and MBSs, and all the five types of AS patterns are involved in particular genes to alter the presence or absence of MBSs ( Figure 1A) (Yang et al., 2012). Note that these observations are based on the fact that the number of AS events is under-estimated due to limited types of tissues, limited physiological states, and insufficient sequencing coverage.
The numbers of AS events affecting the MBSs might be much more, and there should be a larger proportion of overlapping between AS and MBS ( Figure 1B).

miRNAs and alternative splicing in Arabidopsis
Question comes that is the alternative splicing of MBS regions a coincidence of the two molecular reactions, or are there any intrinsic mechanisms leading to the enrichment of these two biological processes? We admit that there might be some random AS events which alter the MBSs. However, simulation results show that the frequency of AS events near MBS regions is significantly higher than randomness (Yang et al., 2012), which means that the actually observed overlapping between AS and MBS exceeds the expected value ( Figure 1C). Even within each individual gene, the occurrence of AS is elevated around the MBS. This series of results collectively indicate that AS and MBS are not two independent biological features. Instead, MBS changes the local sequences to increase the chance of AS, or AS is designed to alter the MBS and regulates gene expression.

Functional implication of miRNAs and alternative splicing
Given the fact that the two biological features, AS and MBS, are mutually favored, there should be molecular interactions between the RISC and spliceosome. As mentioned above, the differential expression analysis is prevalent with the advent of high-throughput sequencing technology. In the samples involving miRNA over-expression, it is possible that the observed expression changes are not only caused by miRNAs but also could be contributed by AS events. We intuitively think that the AS should be taken into account in the differential expression analysis. For example, the expression level of genes is usually based on RNA reads count of the exons. If isoform A is longer than isoform B and the two isoforms dominate in condition1 and condition2 respectively (Figure 2), then even the actual number of molecule is the same, the reads count would be substantially different. When performing differential expression analysis, this gene would be inevitably considered as a differentially-expressed gene DEG (Figure 2). If the spliced region between isoform A and B contains any MBSs, then the reads count would give a misleading indication that the miRNA has changed the gene expression.

Analyzing miRNA targeting and alternative splicing separately
There seems to be a dilemma about how to define the terminology "gene expression level". In the case shown in Figure 2, both isoform A and isoform B have 5 molecules but the reads count is substantially different. It is known that the transcription factors have the power to decide the number of molecules to be transcribed. If isoform A and isoform B both have 5 molecules in the two conditions respectively, then this suggests that the power of transcription is unchanged under these two conditions ( Figure 2). There is no reason to claim that this gene is down regulated under condition2 solely based on the reads count. The only acceptable definition of down regulation is when 3 out of the 5 molecules of isoform B are removed under condition2.
Thus, AS and miRNA targeting cooperate to regulate gene expression. We appeal that when defining a DEG, it should be stated whether the change in gene expression is caused by AS events or solely by miRNA targeting. Mixing the two features together would produce ambiguous and misleading results. To avoid the mixture of alternative splicing and differential expression, the researchers should analyze both the splicing changes and isoform expression in different conditions. This also means that the sequencing data have been fully taken advantage of as the information of splicing change and expression change is already included in the sequencing reads. Distinguishing these two biological processes would help clarify the underlying molecular mechanisms as well as understand how they cooperate to regulate gene expression.

Funding
This work was financially supported by the National Science Foundation.

Conflict of interest
The authors declare they have no conflict of interest.  . Differential expression analysis based on reads count. Isoform A is longer than isoform B. In two conditions 1 and 2, these two isoforms have 5 molecules, which indicates that they have the same expression level. However, based on reads count, isoform A would have higher expression level than isoform B, bur actually the expression is unchanged in these two conditions.

RNA-Seq
More reads Less reads