The diversity of small non-coding RNAs in the diatom Phaeodactylum tricornutum

Non-protein coding RNAs (ncRNAs) play a critical role in maintaining cellular activity and are engaged in a wide variety of molecular tasks and functions in all studied eukaryotic organisms. For instance, transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs) are essential actors of protein translation, small nuclear RNAs (snRNAs) are involved in the pre-mRNA splicing, nucleolar RNAs (snoRNAs) direct the processing and modification of ribosomal RNAs. In addition, a plethora of other types of ncRNAs of variable size (small RNAs of 20–32 nucleotides up to heterogeneous long ncRNAs of several kilobases) has been reported in different kingdoms of life [1–7]. Although some of these ncRNA species show overlapping features in plants and animals, the information concerning their evolution and functionality is still limited [8]. Until now, the most extensively characterized endogenous ncRNAs are short interfering RNAs (siRNAs), also referred to as endo-siRNAs, and microRNAs (miRNAs), which are generated by the processing of double stranded RNAs (dsRNAs) by a Dicer-class RNase III enzyme [1, 9, 10]. These small RNAs are loaded into RNA-induced silencing complex (RISC) containing Argonaute (AGO) proteins, and guide these protein complexes onto sequence complementary mRNA targets to direct their silencing at the post-transcriptional level through mRNA degradation and/or translation inhibition. Small RNAs, and particularly siRNAs, can also guide AGO-RISC onto sequence complementary DNA targets to direct cytosine DNA methylation and/or chromatin modifications resulting in their transcriptional gene silencing [11]. Besides these small RNA species, a variety of new small RNA classes, involved in novel biological functions such as DNA elimination and DNA repair [12–15], continues to be discovered in different organisms. A class of animal small RNA species that has been recently extensively characterized are PIWI-interacting small RNAs (piRNAs) [16]. These DICER-independent small RNA species are 24-32 nt in length and are derived from long precursor transcripts that originate from distinct genomic regions referred to as piRNA clusters. Mature piRNAs bind specifically to the PIWI subfamily of AGO proteins to promote silencing of transposable elements (TEs) -but also of some protein-coding genes- at the transcriptional and post-transcriptional levels [16]. In mice, this pathway has been mostly characterized in the male germ line, where piRNAs presumably direct transcriptional silencing of TEs through DNA methylation [17, 18], a mechanism that is thus analogous to RNA-directed DNA methylation (RdDM) initially described in plants [19].

The existence of a sophisticated level of regulation of gene expression involving interactions and cross-talk between different small RNAs, generated by different RNA silencing pathways, becomes more and more apparent [1]. Furthermore, several recent discoveries (e.g., the identification of snoRNAs with microRNA-like function [20], the finding of small RNAs derived from tRNAs (tRFs) [12], the co-regulation of miRNA biogenesis and splicing from the same host mRNA [21]) shed light on even more complex and unanticipated RNA regulatory networks involving various ncRNA populations with different origins and functions.

Marine diatoms represent a valuable biological system to address questions about the evolution and diversification of gene regulatory mechanisms in eukaryotes. This hugely diverse phytoplanktonic group of the Stramenopiles derives from a secondary endosymbiosis event and shows a peculiar genetic makeup [22, 23]. Most significantly, almost half of the diatom proteins have similar alignment scores to their closest homologs in plants and animals and at least 5% of their gene repertoire has been recruited by horizontal gene transfer from bacteria [23, 24]. As a consequence, diatoms possess a unique mix of bacterial, animal and plant metabolic features. About half of the diatom genes encodes for proteins of unknown function, suggesting that many regulators of central processes in diatom biology remain to be identified. Nevertheless, the ecological success of diatoms in the contemporary oceans and their capability to grow in very different environments imply that these organisms evolved sophisticated mechanisms to efficiently regulate genome structure and expression, mechanisms that still await characterization. Comparative analyses of the diatom genomes have indicated transposable elements as key contributing factors to genome diversification [23, 25]. Moreover, the first genome-wide DNA methylation map in the model diatom Phaeodactylum tricornutum has revealed the potential impact of DNA cytosine methylation for the control of transcriptional or post-transcriptional silencing and of differential gene expression under specific growth conditions. The data indicated that around 6% of the genome displays detectable DNA methylation, with extensive methylation being localized over transposable elements [26]. A few studies have also provided an initial characterization of small RNAs in the marine diatoms Thalassiosira pseudonana and P. tricornutum[27–29]. Putative miRNAs have been predicted in these two species, but none of them has so far been experimentally validated. Broadly speaking, the potential regulatory role of small RNAs in diatoms still awaits to be demonstrated. Interestingly enough, we have shown that the expression of anti-sense or inverted repeat sequences of selected target genes can trigger efficient gene silencing in P. tricornutum[30], a clear indication for the existence of RNA interference-like processes in diatoms. Moreover, we have also described the presence of genes encoding a predicted Dicer-like protein, an Argonaute-like protein and a potential RNA-dependent RNA polymerase (RdRP) in the P. tricornutum genome [30]. RNAi-related proteins have been identified in T. pseudonana as well [27]. Although distantly related to plant or animal RNAi machinery components, these diatom proteins possess conserved structural and functional domains including key amino acid residues that support their potential role in the small RNA pathways described above.

To gain insight into the complexity of small ncRNA biogenesis and function in diatoms, we combined computational and experimental analyses to generate a comprehensive characterization of P. tricornutum small RNA populations. In particular, we carried out a high-throughput sequencing of short RNA fragments extracted from cells grown under different conditions of light and iron, two abiotic factors highly relevant for diatom growth in the oceans [31–33]. The characterization of the sRNA transcriptomes revealed a small ncRNA landscape in diatoms that is much more complex than anticipated. We identified and characterized different functional categories of small RNAs with different sizes, suggesting the presence of distinct biosynthetic pathways. Our study also questioned the occurrence of miRNAs, which were not detected. Instead, it identified highly expressed 21-25 nt small RNAs derived from longer non-coding RNAs such as tRNAs and U2 snRNA, or of unknown origin. The majority of small RNAs correspond to TEs, in particular to the Copia-like diatom-specific retrotransposon family [23, 25]. Finally, a clear correlation between the presence of small RNA/DNA methylation and the silencing of TEs and coding genes was established, suggesting that a small RNA-dependent DNA methylation pathway is operational in diatoms as in other plant and animal organisms. An unprecedented periodic pattern of small RNAs on TEs and coding genes, related to the methylation, is also unveiled by our genome-wide analysis, opening a completely novel line of investigation on the molecular determinants establishing sRNA distribution.

Although the annotation of sRNAs in almost all studied organisms is becoming more and more comprehensive, these regulatory elements have only recently been identified in unicellular and multicellular algae belonging to different lineages [53–58]. In this work, we have significantly extended the characterization of the sRNAs from a model Stramenopile species, the marine diatom P. tricornutum, providing new information regarding diversification and function of these elements in eukaryotes. The genome-wide analysis of sRNAs, obtained from cells grown under different light and nutrient stress conditions, revealed the existence of a variety of sRNA populations with characteristic lengths and specific distribution patterns on diverse genomic locations (Figure 1). The functional and in silico characterization of these elements allowed us to reach several novel conclusions.

We questioned the presence of canonical miRNAs in diatoms, previously reported in P. tricornutum[28]. We used a computational approach to look for putative miRNAs in our small RNA dataset, but none of our attempts revealed canonical pre-miRNA-like hairpins that could be substrates for a Dicer-like enzyme. Moreover, none of the sRNAs we identified share significant sequence similarity when compared with microRNA sequences from miRBAse [41] and miRNAs from Ectocarpus siliculosus, a Stramenopile multicellular brown alga that represents the only example reported for this lineage in the literature with experimentally validated microRNAs [55, 56]. These results strongly indicate that canonical miRNAs, such as those characterized in different eukaryotic organisms so far, are not major regulators of gene expression in P. tricornutum. However, at this stage, we do not rule out the possibility that canonical miRNAs could be identified in response to other environmental cues and/or through a more important small RNA sequencing coverage than the one used in the present study. Nevertheless, we were able to identify some sRNAs located on stable precursor secondary structures, which we designated as miRNA-like because they possess only a part of canonical miRNA features. In particular, an interesting polycistronic miRNA-like cluster has been identified on chromosome 10, with a RNA stem containing a predicted miRNA locus, consistently detected in all our libraries, and another miRNA-like, on another RNA stem, that is observed only in cells under iron starvation conditions. The data suggests that under different environmental conditions, primary transcripts might be differentially processed to generate different miRNA-like elements with different function. Similar examples have been already reported in human [29]. Interestingly, even if these miRNA-like are well represented and predicted to originate from stable structures, we identified 2 nucleotide 3′ overhangs at only one strand of the putative dsRNAs. This observation suggests that this type of small RNAs might be generated via a Dicer-independent mechanism [59, 60]. Alternatively, the P. tricornutum Dicer-like enzyme, which shows a partial sequence conservation with plant and animal Dicer homologues and possesses only a dsRBD domain followed by 2 RNAs III domains [30] might bind and cleave the opposite strands of a dsRNA in an atypical way to generate the miRNA-like molecules identified in this study. Therefore, further characterization of these miRNA-like elements will likely help identifying novel sRNA biogenesis pathways in diatoms.

Our dataset allowed us to identify the most abundant population of sRNAs in P. tricornutum, showing a clear distribution peak of 24-25 nt in length, localized on an intergenic region of chromosome 2 (chr2: 29.039-29.123). Northern blot analyses experimentally demonstrated the existence of sRNAs in P. tricornutum for the first time (Figure 2B). Their detection in cells grown under different conditions excludes a non-specific RNA degradation origin. Similarly to the miRNAs, these sRNAs may result from the processing of longer RNA precursors, which could successfully be detected. However, despite numerous attempts, we did not succeed to amplify and validate the primary transcript and to identify the dsRNA structure that may generate these products. We can thus hypothesize that, differently from the miRNAs, the precursor is not transcribed by the RNA polymerase II [61] or is subjected to other 5′ or 3′ end modifications (e.g., decapping and subsequent 5′-3′ degradation), that preclude their amplification using standard RACE. Interestingly, these sRNAs share sequence similarities with sRNAs matching an intergenic region of the T. pseudonana genome. This suggests that novel sRNA populations of unknown origin are conserved between different diatom species and likely evolved to play similar regulatory functions in these microalgae.

Novel results derived by next-generation sequencing experiments have provided some indications about the existence of sRNAs associated to spliceosomal snRNAs [62] and in mRNA length regulation (telescripting) [63]. The results presented in this study unveiled a previously unknown regulatory network involving U2 snRNA-derived sRNAs. Two different sRNA populations at the 5′ and 3′ ends of the U2 snRNA (the U2-5′ and U2-3′ sRNAs) were identified. We validated the expression of both U2-5′ and U2-3′ and provided functional evidences that the full length U2 RNA represents the primary precursor of the sRNA populations. Namely, the reported increase in U2-3′ content in transgenic lines over-expressing U2 snRNA robustly supports this possibility. Of particular interest, the expression of both U2-5′ and U2-3′ is strongly repressed in cells grown under iron starvation condition. Interestingly, an induction in the U2 snRNA transcript was observed in the same condition. This negative correlation supports the existence of a regulatory network balancing the reciprocal expression of the U2 snRNA and the U2-derived sRNAs. An increase in the U2 snRNA might be needed to control the cellular responses to deficiency of iron, a critical nutrient for life in the ocean [32, 33, 64, 65]. Likewise, we can also hypothesize a regulatory role played by the U2-derived sRNAs in controlling P. tricornutum’s response to iron availability. Finally, a feedback loop controlling the expression of U2 snRNA precursors by the associated sRNAs might also occur. Interestingly, highly expressed U2-3′ smallRNAs have been detected also in T. pseudonana, strongly supporting a conserved biogenesis of these smallRNA populations in diatoms. The characterization of this complex regulatory process, the identification of U2-3′ mode of action possibly involving sequence-specific targeting, the putative interaction between these sRNAs and the splicing machinery, all represent important questions to address in the future.

Similarly to what was reported in the last years in many organisms [38, 66], we identified a variety of sRNAs associated to tRNAs also in P. tricornutum and T. pseudonana (Additional file 6: Figure S3C and Additional file 7: Table S3). These sRNAs possibly derive from different processing mechanisms and endorse different cellular functions. In P. tricornutum, we found putative tRFs of around 19 nt, mostly corresponding to the tRNA 3′ ends and carrying the trinucleotide CCA at the acceptor stem, which likely derive from tRNA cleavage in or at the T loop [48]. We also detected short fragments associated to the D loop and to the anticodon of 9 different P. tricornutum tRNAs (Additional file 7: Table S3). Because they do not fit into the general classification scheme [48], we are not able to anticipate the processing events generating these short tRFs from precursor or mature tRNAs. Moreover, we detected several longer small RNAs (30-35 nt) that correspond to both 5′ and 3′ parts of full length tRNAs, and probably derive from cleavages at different positions (Figure 4). Some of these long fragments might be classified as half tRNAs (e.g., Glu^GAG tRNA). In other species, such tRNA fragments, usually referred to as ‘tRNA, stress-induced small RNAs’, derive from the tRNA cleavage in the anticodon loop, and have been shown to regulate a wide range of cellular processes (e.g., translational efficiency and mitochondrial-mediated apoptosis [48, 66–69] in response to nutritional, biological, or physicochemical stresses). Interestingly, our data also unveil an enrichment of the long tRNA fragments in P. tricornutum cells grown under iron starvation (Additional file 7: Table S3). The data suggest that cells exposed to iron deficiency may differently modulate the processing of specific tRNAs, thereby generating fragments of different sizes that may play different roles in the control of diatom responses to nutrient limitation. We validated the expression of selected sRNAs associated to tRNAs by Northern blot and in agreement with the sequencing data, we observed different populations of tRNA fragments (Figure 4). In general, the detection of the shorter tRFs appeared more problematic in Northern blot analyses as compared to the longer tRFs and to mature tRNAs, likely due to the different stability of these products.

In this study, we also dedicated particular attention to the characterization of sRNAs matching TEs and repeats, which represent almost half of the P. tricornutum small RNA repertoire. TEs are considered an important driving force for genome evolution by inserting at various genomic locations and altering gene expression and regulatory functions [70, 71]. A significant part of diatom genomes is devoted to TEs and repeat sequences, particularly in P. tricornutum[23]. Moreover, a comparative genomic analysis between P. tricornutum and the centric diatom T. pseudonana revealed that TEs have been important contributors to the genome divergence between the two diatom species, with P. tricornutum having undergone a significant amplification of TE sequences as compared to T. pseudonana, respectively 6.4% and 1.9% of the genomes [23, 25]. Interestingly, 7 to 15% of the sRNAs are matching repetitive sequences in T. pseudonana[27], a proportion significantly lower than in P. tricornutum. These observations suggest that the increased amplification of TEs in P. tricornutum is likely responsible for the over-representation of endogenous sRNAs associated to TEs in this species. Such a mechanism would be consistent with recent findings in Arabidopsis whereby gradual enhanced accumulation of siRNAs derived from the retrotransposon Evadé were also observed upon progressive amplification of this TE in the genome [72]. TE-derived sRNA species likely regulate the expression/transposition of these mutagenic sequences in normal growth conditions but may also control the transposition events under selective conditions thereby contributing to genome evolution and adaptation. Consistently with this concept, expression of some LTR retrotransposons (LTR-RTs) can be modulated by different stresses in P. tricornutum, suggesting that the mobilization of some TEs may contribute to genome rearrangements [25]. The likely essential role of TEs in genome evolution and adaptation, and the high representation of TE-derived sRNAs in P. tricornutum therefore prompted us to explore the possible regulatory role of these sRNA species. Similarly to other organisms, our data clearly indicated that a high sRNA coverage correlates with low TE expression (Figure 5), suggesting that the sRNAs indeed direct TE silencing. We gained more insights into the mechanisms by which sRNAs could direct TE silencing by comparing our small RNA profiling with the methylated regions obtained from a recent genome-wide analysis generated in the same species [26]. This actually confirmed that most of the TEs targeted by sRNAs are characterized by high methylation levels, further supporting the possibility that these ncRNAs contribute to the establishment of DNA methylation at these TEs and to their transcriptional or post-transcriptional silencing. We then extended the analysis of small RNA-associated DNA methylation to coding regions. Interestingly, as already observed by Veluchamy et al.[26], a high methylation level was observed in genes that are specifically covered by sRNAs (Figure 1, Additional file 1: Table S1 and Additional file 8: Figure S4). Our data further show that these sRNAs have sizes and distribution patterns like TE-associated sRNAs, suggesting that they are also generated via similar mechanisms. Interestingly, we already showed in a previous study [30] that de novo cytosine methylation is associated with transgene silencing induced by artificially expressing antisense and inverted-repeat RNAs in P. tricornutum. Altogether, this information suggests that a small RNA-dependent DNA methylation process is widely used by diatoms not only to control the expression of TEs but also of transgenes as well as protein-coding genes by presumably directing DNA methylation within these genetic elements. These results also open interesting novel questions on the evolution, distribution and function of such epigenetic-based regulatory pathways in different organisms. In mammals, DNA methylation occurs throughout the genome, and many genes are body methylated, although this methylation pattern has not been associated with the presence of sRNAs [73]. In wild-type Arabidopsis plants, gene body methylation has also been reported for many genes, but this methylation occurs in a small RNA-independent manner and appears to be irrelevant for silencing [74]. By contrast, loss-of-function mutations in some active Arabidopsis DNA demethylase mutants were reported to enhance DNA methylation in specific gene body regions through siRNA-mediated DNA methylation, presumably leading to their silencing [75, 76] and supporting a functional role for siRNA-directed gene body methylation in these mutant backgrounds. Therefore, we can hypothesize that different photosynthetic organisms belonging to different lineages had the potential to use the small RNA-dependent DNA methylation pathway to control expression of coding genes, although in plants this effect is solely appreciated in specific mutants but not in wild-type plants grown under standard conditions. Furthermore, the sRNA distribution and DNA methylation profile over gene bodies differs significantly between diatoms and plants (Figure 1 and Additional file 8: Figure S4) [75]. Therefore, the observations made on P. tricornutum will provide fertile ground for new exciting insights into the mechanisms involved in small RNA-directed gene body methylation and into the eventual effects of such epigenetic modification on transcription initiation and elongation.

In addition to the likely functional relevance of P. tricornutum TE-associated sRNAs in DNA methylation and transcriptional gene silencing, we found that these sRNA species were 25 to 30 nt in length and have a 5′ uridine bias (Figure 5). Interestingly, these features are shared with animal piRNAs suggesting a potential link between these silencing pathways. However, our analysis did not reveal self-complementary sequences, characteristic of a Ping-Pong amplification, and no canonical PIWI protein could be retrieved in the P. tricornutum genome (data not shown), arguing against the presence of a canonical piRNA biosynthetic pathway in this marine microalga. Future in depth characterization of such diatom small RNA species will thus be necessary to compare and contrast these sRNAs with animal piRNAs.

The periodic distribution of sRNAs over methylated regions represents a completely novel hint. This phenomenon is characterized by a 180 nt-long period (at times of 90 nt) that is sharp on coding genes and also visible with high confidence on TEs. At this stage, it is still difficult to predict the mechanisms generating such a peculiar distribution pattern or to link sRNA distribution with DNA methylation activities and transcriptional silencing. In future studies, it will be relevant to assess whether periodic sRNA distribution correlates with local chromatin properties such as nucleosome occupancy, histone post-translational modifications or histone variants, which have not been mapped in diatoms so far. More specifically, it might be interesting to test whether the 180 nt periodic enrichment of sRNAs matches inter-nucleosomal domains with about the same size, possibly reflecting the capacity of RNA Polymerases producing small RNAs to efficiently transcribe these more accessible regions in a condensed and silenced chromatin context. Reciprocally, on protein-coding genes, sRNAs might also target inter-nucleosomal domains more easily than histone-wrapped nucleosomal DNA, particularly in a non-replicative context. Interestingly, the existence of periodic cytosine methylation has recently been detected in microalgae with compact nuclei and strong gene body methylation, regardless of gene expression levels. In Ostreococcus lucimarinus and Micromonas pusilla, the periodic DNA methylation occurs specifically over nucleosome linker (inter-nucleosomal) regions, and appears to contribute to genome architecture and nucleosome positioning [77]. Based on our observations, it might be interesting to test whether sRNAs also display a periodic enrichment at the same loci in these species. Of note, regularly spaced methylation has not been detected in P. tricornutum[77]. This might reflect the different methylome features between P. tricornutum and these other algae. Alternatively, more refined DNA methylation mapping might be required to determine whether periodic DNA methylation and sRNA distribution overlap at some TEs or gene loci in P. tricornutum.

The analysis of diatom genomes reveals the presence of genes encoding several proteins that might be required for generating DNA methylation-associated sRNAs. P. tricornutum and T. pseudonana genomes both encode Dicer- and Argonaute-like proteins and several putative RNA-Dependent RNA Polymerases (RdRPs) but no PIWI protein [30]. These proteins might be responsible for generating sRNAs, for amplification mechanisms and for the possible recruitment of DNA methyltransferases (DNMTs) identified in previous studies [25, 77]. Considering the peculiar genome of diatoms, their evolutionary history, and the information on the sRNAs landscape from this study, we also expect that additional, still unknown, components participate in the small RNA-based regulation of diatom genome function. Their characterization will likely provide novel insights into the genetic bases allowing the successful proliferation of diatoms in the oceans and will help addressing relevant questions about the conservation of these processes in organisms that cover different branches of eukaryotic evolution.

This comprehensive characterization of the P. tricornutum small RNAs landscape uncovered an unexpectedly complex composition of the sRNA complement in diatoms. While miRNAs seem not to play a major regulatory role in P. tricornutum, other ncRNA types with well-described function, such as tRNAs and U2 snRNA, represent a major source of sRNAs and could play important functional roles in P. tricornutum’s responses to environmental stresses such as iron starvation. This discovery also implies that complex and interconnected networks of ncRNA pathways exist in diatom cells, and their fine-tuning is likely to be fundamental for the biology of these organisms and their extraordinary ability to proliferate worldwide. The existence of highly expressed sRNAs of unknown origin and conserved in two distantly related diatom species also indicates that diverse small RNAs sources await to be discovered in these algae. The most represented population of sRNAs match TEs and likely contributes to their transcriptional silencing through a yet-to-be characterized RNA-dependent DNA methylation mechanism. Protein-coding genes might also be subjected to similar regulatory processes for down regulating their expression at a transcriptional, and possibly also at a post-transcriptional level. In conclusion, this extensive characterization of P. tricornutum small ncRNAs can contribute to future studies aiming at understanding diatom biology, the molecular bases underlying their ecological success in the contemporary oceans, and their capacity to optimally perform in highly changing environments. The periodic pattern of sRNAs distribution associated with genic DNA methylation unveiled here, represents an unexpected discovery that opens the way to search for similar periodic features in other eukaryotic groups.