The circadian clock shapes the Arabidopsis transcriptome by regulating alternative splicing and alternative polyadenylation

Abstract

The circadian clock in plants temporally coordinates biological processes throughout the day, synchronizing gene expression with diurnal environmental changes. Circadian oscillator proteins are known to regulate the expression of clock-controlled plant genes by controlling their transcription. Here, using a high-throughput RNA-Seq approach, we examined genome-wide circadian and diurnal control of the Arabidopsis transcriptome, finding that the oscillation patterns of different transcripts of multitranscript genes can exhibit substantial differences and demonstrating that the circadian clock affects posttranscriptional regulation. In parallel, we found that two major posttranscriptional mechanisms, alternative splicing (AS; especially intron retention) and alternative polyadenylation (APA), display circadian rhythmicity resulting from oscillation in the genes involved in AS and APA. Moreover, AS-related genes exhibited rhythmic AS and APA regulation, adding another layer of complexity to circadian regulation of gene expression. We conclude that the Arabidopsis circadian clock not only controls transcription of genes but also affects their posttranscriptional regulation by influencing alternative splicing and alternative polyadenylation.

Plant 3’ Regulatory Regions From mRNA-Encoding Genes and Their Uses to Modulate Expression

“…Although we have only just begun to unravel the mechanism, we know that APA has temporal (cell and developmental cycle) and tissue specificity, in addition to being involved in different biological processes, such as embryogenesis, gametogenesis, morphogenesis, control of flowering time in plants and control of oncogenes expression in animals ( Xing and Li, 2011 ; Deridder et al., 2012 ; Tian and Manley, 2016 ; Ji et al., 2018 ). Moreover, APA events are also involved with growth and development ( Hong et al., 2018 ), circadian rhythm ( Yang et al., 2020 ), cell signaling ( Chakrabarti and Hunt, 2015 ; Li et al., 2017 ; Conesa et al., 2020 ), immunity ( Lyons et al., 2013 ; Ye et al., 2019 ) and stress response in plants ( Zheng et al., 2018 ; Conesa et al., 2020 ).…”

Interpreting machine learning models to investigate circadian regulation and facilitate exploration of clock function

“…Therefore, secondly, we prioritized selection of transcripts using model interpretation in the form of feature selection to make the frequency distribution across the modules more uniform (see Methods; Glossary). Optimizing performance based on the validation dataset, our best performing model overall used a final subset of 15 transcripts (Table S14) and had a MAE of 21 minutes on the training data, 56 minutes on the [9] validation data and 46 minutes on the test data from [10]. Figure S6b and S6c also highlight that after such feature selection there was a decrease in the generalisation error on average across the [10] test dataset with the improvements in MAE decreasing as the number of genes increased.…”

“…As such, we applied ZeitZeiger to our datasets [8,9,10], to compare directly with our model. To reflect our previous approach, firstly, dataset [8] was used to fit ZeitZeiger, with predictions then being generated on the validation [9] and testing [10] datasets to compare with the predictions generated by our method. Our approach significantly outperformed ZeitZeiger on the test dataset (MAE of 46 compared to 143 minutes, Figure S7) demonstrating our efficacy at generating highly accurate predictions for circadian time.…”

“…This has yielded a plethora of public multi-omic resources [5,6,7] that can be re-analysed to give new insights into the roles and functions of complex regulatory networks. In this study, we use newly generated datasets, published temporal datasets [8,9,10] (Table S1) and…”

“…This has yielded a plethora of public multi-omic resources [5,6,7] that can be re-analysed to give new insights into the roles and functions of complex regulatory networks. In this study, we use newly generated datasets, published temporal datasets [8,9,10] (Table S1) and Arabidopsis genomes, in combination with machine learning (ML) approaches (see Glossary for definitions of terms), to make predictions about circadian gene regulation and expression patterns. Critically, we advance existing approaches using explainable AI algorithms and interpretation of our models (Glossary), such methods help us to understand the predictions made by ML models.…”