Analyses of <i>Marsilea vestita</i> genome and transcriptomes do not support widespread intron retention during spermatogenesis

Abstract

Intron retention (IR), a form of alternative splicing, plays a key role in post-transcriptional gene regulation. The fate of intron-retaining transcripts varies; for example, they could trigger nonsense-mediated decay or lead to alternative protein isoforms. Of particular interest are ‘detained introns’, which remain in the nucleus and can be productively spliced in a regulated way (Boutz et al., 2015; Fig. 1). In plants, while IR has been well-documented (Ner-Gaon et al., 2004; Campbell et al., 2006; Jia et al., 2020), its biological function remains poorly understood. The heterosporous fern, Marsilea vestita, appeared to present a rare functional example of pervasive post-transcriptional regulation by removal of detained introns (Boothby et al., 2013). The development of M. vestita male gametophytes is extremely rapid – from the rehydration of spores to the release of 32 sperm takes < 11 h. This process proceeds normally in the presence of transcriptional inhibitors (Hart & Wolniak, 1998, 1999) and is therefore presumed to be controlled post-transcriptionally (Wolniak et al., 2011, 2015; Boothby et al., 2013). By comparing de novo transcriptome assemblies from a developmental series, Boothby et al. (2013) discovered many intron-retaining transcripts that were present early in development, with the corresponding fully spliced isoforms appearing only later. They concluded that M. vestita microspores store a pool of incompletely spliced transcripts, and the removal of detained introns enables rapid spermatogenesis. However, because no Marsilea genome was available at the time, Boothby et al.'s (2013) study was based almost entirely on transcriptomic data, making it difficult to thoroughly characterize the intron landscape. Here, we generated a chromosome-level genome assembly for M. vestita and revisited IR and splicing during spermatogenesis. Mapping the original reads from Boothby et al. (2013) to this genome, we found minimal IR, not even in the genes highlighted by Boothby et al. (2013). Thus, our results, obtained using the new genome, suggest that rapid spermatogenesis in Marsilea is not driven by splicing of detained intron and the basis of which is open to new investigation. To sequence the M. vestita genome, we used the same experimental strain as in Boothby et al. (2013). The genome was sequenced on both Illumina and Nanopore MinION platforms (Supporting Information Table S1), and assembled using the MaSuRCA hybrid assembler (Zimin et al., 2017). To further scaffold the genome, a Hi-C library was prepared and sequenced through Phase Genomics (Seattle, WA, USA). All the sequencing reads were deposited in NCBI SRA under BioProject PRJNA849927. After Hi-C scaffolding by Juicer (Durand et al., 2016) and 3d-dna (Dudchenko et al., 2017) followed by gap-closing by TGS-GapCloser (Xu et al., 2020), we obtained a final assembly of 1.0 Gb with 20 pseudochromosomes and a contig N50 of 29.0 Mb (Table S2). To our knowledge, this assembly has the highest contig continuity of the sequenced fern genomes to date and is the first at the chromosomal scale for Salviniales (Li et al., 2018; Fang et al., 2022; Huang et al., 2022; Marchant et al., 2022). We annotated and masked the repetitive elements following the procedure from Li et al. (2018). For gene prediction, we used protein evidence from Azolla filiculoides and Salvinia cucullata (Li et al., 2018), and transcript evidence from the Boothby et al.'s (2013) RNA-seq data and a newly generated leaf transcriptome. Augustus and SNAP ab initio gene predictors were trained by Braker2 (Brůna et al., 2021) and Maker2 (Holt & Yandell, 2011), respectively. Maker2 was used to consolidate all the gene models to derive a consensus gene set. The final M. vestita genome annotation includes 22 541 gene models and has 96.5% complete Busco (Benchmarking Universal Single-Copy Orthologs; Simão et al., 2015) genes against the Eukaryota database (Table S3), which is one of the highest among all published fern genomes (Li et al., 2018; Fang et al., 2022; Huang et al., 2022; Marchant et al., 2022). Lastly, we used eggNOG v.5.0 (Cantalapiedra et al., 2021) for functional annotation. The genome assembly and annotation are available at www.fernbase.org. To quantify IR, we first mapped the RNA-seq data from Boothby et al. (2013) to our M. vestita genome using Hisat2 (Kim et al., 2019; Fig. 2). These RNA-seq data came from three spermatogenesis time points, at 1–2 h, 3–5 h, and 6–8 h postrehydration of microspores; each time point has three replicates (see Methods S1 for details). All the replicates have high read-mapping rates to the genome (> 97%) and are clustered by time points on a principal component analysis plot (Fig. 3a), indicative of high data quality. We used IRFinder (Middleton et al., 2017) to calculate intron retention ratio (IR ratio), which is the abundance of intronic reads divided by intronic + normally spliced reads. An intron with an IR ratio of 0.5 means that this intron is retained in 50% of the transcripts. We found that the vast majority of M. vestita introns had low IR ratios (< 0.1; Fig. 2a), suggesting IR is rare. Importantly, if intron-retaining transcripts were spliced during spermatogenesis, we would expect to see a drop in IR ratio over time. However, we observed the opposite pattern – the first time point had the fewest introns with IR ratio > 0.1 (Fig. 2a). Next, we examined the 17 genes that were highlighted in the Boothby et al.'s (2013) study – each of these has at least one detained intron in an isoform assembled by Trinity (Grabherr et al., 2011) and was described as showing programmed removal of one or more introns. Our RNA-seq coverage plots clearly showed that there were minimal number of reads (if not none) that span intron–exon junctions (Figs 2b, S1), and all the introns in these 17 genes had low IR ratios across the three time points. While two introns of Mvestita_S5g09449.1 (flagellar-associated protein 71-like) were identified having differential retention (Fig. 2c green lines) according to our DEXSeq analysis (Middleton et al., 2017), this gene was highly upregulated in the last time point (> 100-fold), contradicting with the scenario described by Boothby et al. (2013). Overall, we found more upregulated genes than downregulated ones across the three time points (Fig. 3b), based on our differential gene expression analyses using DESeq2 (Anders & Huber, 2010). Gene ontology (GO) enrichment using TBtools (Chen et al., 2020) found that upregulated genes are highly enriched in ‘cilium’- and ‘membrane’-related terms (Fig. 3c; Table S4), which are consistent with the spermatogenesis process. Taken together, it is clear that new transcription, not splicing of detained introns, is the primary mechanism for gene regulation in M. vestita spermatogenesis. This finding, however, runs counter to earlier results with transcriptional inhibitors (Hart & Wolniak, 1998, 1999), in which spermatogenesis was able to proceed normally in the presence of α-amanitin (an RNA polymerase inhibitor). One possibility is that RNA in the early stages might be sequestered in a state that is inaccessible to the extraction methods used. Alternatively, resistance to α-amanitin is not uncommon and has been shown in carrots to be developmentally regulated depending on tyrosinase activities (Pitto et al., 1985). It is thus possible that α-amanitin resistance likewise occurs in M. vestita spermatogenesis and could reconcile our results. With regard to IR, several factors could explain the different conclusions reached by Boothby et al. (2013) and our study. First and foremost, we have a high-quality, chromosome-level genome as the reference. Boothby et al. (2013), by contrast, had to rely only on de novo transcriptome assemblies, which contained a total of 187 159 transcripts, a number that is substantially higher than what we annotated from the genome. Many of their transcripts are likely assembly artifacts or incorrect isoforms that could lead to overestimation of IR. Boothby et al. (2013) did attempt to confirm their results by selecting 17 genes for RT-PCR. While intron-retaining transcripts could be successfully amplified, the possibility of genomic DNA contamination cannot be ruled out because one of the primers had to sit on intronic regions. Indeed, bands from supposedly intron-retaining transcripts were always present even in the later time points, and the decrease in their band intensities could be due to competition with the bona fide transcripts. It should be stressed that our results do not completely discount the presence of detained introns or stored mRNA in M. vestita spermatogenesis. Instead, we found that these processes are not the primary mechanism for gene regulation as previously believed. In summary, here, we present a genome assembly for M. vestita at the chromosomal level, first for the heterosporous ferns. This assembly will be an important reference for future comparisons with the growing number of homosporous fern genomes (Szövényi et al., 2021; Fang et al., 2022; Huang et al., 2022; Marchant et al., 2022) and will help elucidate the homospory–heterospory transitions (Kinosian et al., 2022). With this new genome, we reanalyzed the RNA-seq dataset from Boothby et al. (2013). We find no support for the notion that the removal of detained introns regulates rapid spermatogenesis in M. vestita, despite it being cited as a classic example of the biological significance of IR in plants. We would like to thank Larry Wu for help with RiboPlotR, Junhui Zhou for help with DNA isolation, Corine van der Weele, Vincent Klink, and Li laboratory members for discussions, as well as the reviewers and Editor for their feedback. This project was made possible in part by the contributors to the UMD Marsilea vestita genome crowdfunding project. None declared. SMW, CFD, SM and F-WL planned and designed the research. NR, L-YK, JK, EH, LL, ML, SS and RZ performed experiments. NR, SM and F-WL analyzed data. F-WL wrote the manuscript. Fig. S1 RNA-seq coverage plots of the 17 focal genes. Methods S1 Sampling of microspores for time-course RNA-seq. Table S1 Nanopore sequencing statistics. Table S2 Assembly statistics of Marsilea vestita. Table S3 Comparison of assembly and annotation quality of the published fern genomes. Table S4 Enriched gene ontology terms for the upregulated genes. Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Topics & Concepts