In all eukaryotes, the basic principles controlling cell division appear to be conserved (Nurse, 2000). Thus, the cell cycle is composed of four phases: in G1 (gap phase 1), cells increase their number of organelles, during S phase DNA replication occurs, in G2 (gap phase 2), cells still increase their size by extensive protein synthesis, and in M phase (mitosis), chromosomes segregate into two nuclei, followed by cytokinesis, during which cells are divided into two daughter cells. The orchestration of the cell cycle, and specially the transition from G1 to S phase as well as the progression and exit from mitosis, requires multiple levels of control. In particular, cyclin-dependent kinases (CDKs) that are specifically activated by cyclins are key players in this process (Malumbres and Barbacid, 2005;De Veylder et al., 2007). Several other kinases and phosphatases, as well as additional regulatory proteins, such as cyclindependent kinase inhibitors, also regulate progression through the cell cycle (Boutros et al., 2006;Fisher, 2012;Starostina and Kipreos, 2012). In Arabidopsis thaliana, and plants in general, most core cell cycle genes are conserved (Vandepoele et al., 2002), although several are present in multiple copies. Thus, while in mammals three Cip/Kip members inhibit a broad spectrum of CDK-cyclin complexes (Denicourt and Dowdy, 2004), the Arabidopsis genome encodes at least 21 related Cip/Kip proteins (Kumar et al., 2015). Likewise, multigenic gene families also encode several plant cyclin families, including a unique class of CDKs (B-type) for which no counterpart is found in other organisms (Boudolf et al., 2004;Vandepoele et al., 2002).
The control of each step of the cell cycle therefore requires the dynamic accumulation of specific mRNAs and proteins, achieved through transcriptional, post-transcriptional and post-translational controls. The overaccumulation of cyclins, the absence of CDK inhibitors, or the misexpression of Retinoblastoma proteins (pRB) are commonly found in human cancers, indicating that the proper protein level of cell cycle regulators is key for cell division. The ubiquitin proteasome system actively participates in this process by promoting irreversible and timely proteolysis of numerous regulatory proteins absolutely required for transitions in the cell cycle phase (Mocciaro and Rape, 2012;Genschik et al., 2014). A failure to degrade some cell cycle regulatory proteins leads to abnormal cell division and chromosomal instability.
Beside selective proteolysis, microRNAs (miRNAs) have a demonstrated role in controlling the levels of multiple cell cycle regulators in mammals (Bueno and Malumbres, 2011). miRNAs are derived from processed, single-stranded RNAs, functioning to control the stability and translation of protein-coding mRNAs (Krol et al., 2010;Axtell, 2013). This is achieved through their incorporation into a protein complex called the RNA-induced silencing complex (RISC) (Czech and Hannon, 2011). Central components of RISC are ARGONAUTE (AGO) proteins that bind the mature miRNA and orient it for interaction with target mRNAs (Meister, 2013;Poulsen et al., 2013). In animal cells, a single miRNA can have a large number of different mRNA targets; in humans, the >1000 annotated miRNAs may control up to 60% of the genes (Friedman et al., 2009;Leung and Sharp, 2010). miRNAs have been implicated in numerous developmental and physiological processes and are involved in a number of diseases, including cell proliferative diseases such as cancer (Williams, 2008). miRNAs target dozens of cell cycle genes in mammals and thus play functions in the regulation of the G1/S transition and the entry and progression through mitosis (Bueno and Malumbres, 2011). For instance, the mir-15a-16-1 cluster could induce cell cycle arrest at G1 by targeting CDKs and cyclins (Linsley et al., 2007;Liu et al., 2008). miRNAs also target negative regulators of the cell cycle such as pRB (Volinia et al., 2006) and cell cycle inhibitors of the INK4 or Cip/Kip families (Visone et al., 2007;Kim et al., 2009;Wang et al., 2009), thus promoting cell cycle entry. AGO proteins also bind other classes of small RNAs (sRNAs), such as small interfering RNAs (siRNAs). In plants, siRNAs that derive from a double-stranded RNA precursor include heterochromatic siRNAs (hc-siRNAs) and phased siRNAs (phasiRNAs) (Borges and Martienssen, 2015). In addition, recent work showed that AGO proteins bind sRNAs that are tRNA fragments (tRFs), which act like miRNAs to regulate cellular functions, including cell proliferation and in genome protection against retrotransposons (Lee et al., 2009;Haussecker et al., 2010;Shigematsu and Kirino, 2015;Martinez et al., 2017).
Unlike the situation in animal cells, little is known in plants about post-transcriptional regulation of gene expression during the cell cycle and the role and regulation of AGO proteins in this process. Mutations in genes encoding miRNA pathway components cause pattern formation defects leading, in the worse cases, to embryonic lethality (Nodine and Bartel, 2010;Vashisht and Nodine, 2014). Similarly, aberrant patterns of cell division and cell expansion during embryogenesis have been identified in mutants of Arabidopsis SERRATE (SE) and HYPONASTIC LEAVES 1 (HYL1); both genes encode proteins that cooperate with DICER-LIKE 1 (DCL1) for miRNA biogenesis (Grigg et al., 2009;Armenta-Medina et al., 2017) pointing to an important role of plant miRNAs in the regulation and or maintenance of cell division. Here, we investigated the regulation of AGO1 during the cell cycle and examined the repertoire of total sRNA and AGO1-bound sRNAs at different phases of the cell cycle.
Arabidopsis ago1 null mutants exhibit a severe morphological phenotype affecting leaf shape and polarity, along with defects in meristem identity and function (Bohmert et al., 1998;Kidner and Martienssen, 2005;Morel et al., 2002). Analysis of primary root growth of ago1 mutants revealed a clear reduction in the root length of the hypomorphic ago1-27 allele, while this phenotype was severely compromised in the strong ago1-36 mutant (Supplemental Figure 1A). Moreover, the highly organized structure of root apical meristem was altered in a significant proportion of ago1-27 roots and even lost in ago1-36 mutant roots (Supplemental Figure 1B), precluding the quantification of their meristematic cells.
Because these defects in root meristem activity might originate during embryogenesis, we aimed to deplete AGO1 post-embryonically. For this, we used the β-estradiol-inducible (XVE:P0) line to express the F-box protein P0 from Turnip Yellows Virus (TuYV), which induces the degradation of plant AGO proteins (Derrien et al., 2018). Upon P0 induction, both the root length and the root apical meristem size were significantly reduced (Figure 1A-B) and thus recapitulated the phenotype observed in strong ago1 mutant alleles. Because P0 triggers the degradation of several, if not all plant AGOs (Baumberger et al., 2007;Derrien et al., 2018), we also induced P0 expression in the ago1-57 genetic background, expressing a mutated form of AGO1 resistant to P0-mediated degradation (Derrien et al., 2018). The strong effect of P0 on meristem size and cell division activity was significantly suppressed in ago1-57 (Figure 1A-B), indicating that this phenotype is mainly dependent on AGO1. The fact that the phenotype was not fully suppressed might be attributed to the ago1-57 mutation affecting some siRNA pathways (see (Derrien et al., 2018)) or the involvement, thought minor, of some other AGO proteins.
We next wondered whether AGO1 degradation would block cell division at a specific phase of the cell cycle. To this end, we transformed the XVE:P0 and XVE:P0 (ago1-57) lines with the pHTR2:CDT1a (C3)-GFP (Yin et al., 2014) and pCYCB1;2:CYCB1;2(Dbox)-GFP constructs as markers of cells in S/G2 and G2/M phases, respectively. Roots of double homozygous plants were visualised by confocal microscopy before and upon P0 induction. In the wild type background, the number of cells expressing either the S/G2 or G2/M markers dramatically decreased upon P0-mediated AGO1 degradation, while this was not the case in ago1-57 expressing the stable version of the AGO1 protein (Figure 2). From these experiments, we conclude that AGO1 activity is required to maintain normal cell proliferation in the Arabidopsis root meristem, but that its depletion does not lead to an arrest specific to the S-, G2-or M-cell cycle phases. Note that we cannot exclude the possibility that AGO1 depletion might also affect the timing of a specific cell cycle phase.
To address more specifically the question of the regulation and function of AGO1 during the cell cycle, we moved to the tobacco (Nicotiana tabacum) BY2 cell suspension (Nagata et al., 1992), which offers optimal synchronization efficiencies and also the possibility to biochemically monitor AGO1 protein steady state levels at different cell cycle stages. We established clonal cell suspension cultures with the pAGO1:GFP-AGO1 construct expressing a functional GFP-AGO1 protein under the control of its own promoter (Derrien et al., 2012).
We observed that at a high expression level of the GFP-AGO1 transgene, progression of the cell cycle was significantly delayed (Supplemental Figure 2). We therefore selected clonal BY2 cell suspensions with moderate GFP-AGO1 protein expression for synchronization experiments. BY2 cells were synchronized by a 24 h treatment with aphidicolin (an inhibitor of DNA polymerase), as previously reported (Criqui et al., 2000), and total RNAs and proteins were extracted at different time points after removal of the drug (Figure 3). Cell cycle progression was monitored by flow cytometry measurements of DNA contents, by RT-qPCR monitoring of cell cycle gene expression, and by determination of the mitotic index at the different time points (Figure 3A-B). From S-phase (1-2 h after aphidicolin removal) to G2 (5-6 h after aphidicolin removal), the GFP-AGO1 protein was detected at all time points of the cell cycle without substantial fluctuations (Figure 3C). By contrast and as expected, cyclin B1 protein levels only accumulated during G2/M, rapidly decreased after the time of the peak in the mitotic index (6 h) and reached background levels at the exit of mitosis (9 h). Similar to the AGO1 protein level, the accumulation of the AGO1 transcript (Figure 3B) did not dramatically change during the cell cycle. A second synchronization experiment is presented in Supplemental Figure 3 showing similar results. Thus, we conclude that the AGO1 expression level remains overall constant throughout the cell cycle.
We next imaged the localization of GFP-AGO1 in our transgenic cell lines. In asynchronized BY2 cells, the GFP-AGO1 fusion protein was found predominantly in the cytosol, as previously reported in Arabidopsis root cells (Derrien et al., 2012), but was also present in the nuclear compartment (Figure 4A). Note that the nuclear localization of AGO1 has been previously documented in Arabidopsis (Dolata et al., 2016;Bologna et al., 2018;Liu et al., 2018). We next synchronized BY2 cells and imaged GFP-AGO1 at the different cell cycle phases (Figure 4B-H). During S-phase, the GFP-AGO1 signal was mainly enriched in the cytosol (Figure 4C-D). Interestingly, during G2, we observed the GFP-AGO1 signal also in larger nuclear bodies (of ~2 to 6 μm) distinct from nucleoli (Figure 4E). Quantitative analysis showed that the number of cells that contain these nuclear bodies increased from S to G2 and dropped significantly when the mitotic index started to increase, suggesting a cell cycle-dependent periodic accumulation of GFP-AGO1 nuclear foci (Figure 4B). In mitotic cells, the fluorescence signal was excluded from condensed chromosomes (Figure 4F-G).
To confirm this observation, we transformed the GFP-AGO1 clonal cell suspensions with a construct expressing H2B fused to tdTomato, which once incorporated into chromatin allows visualization of nuclear events. Hence, the dual localization of both fluorescent proteins in mitotic cells indicated no substantial overlap in condensed chromatin (Figure 4I). Finally, after mitotic exit, cells in G1 showed again a predominant cytosolic distribution of the GFP-AGO1 protein (Figure 4H). From these observations, we conclude that AGO1 exhibits a complex subcellular localization pattern with foci in both the nucleus and cytosol, supporting the existence of pools of RISCs that may play distinct functions along the cell cycle.
To achieve a global view of gene expression during the cell cycle, we performed five independent synchronization experiments (called Replicates 1 to 5) using a clonal BY2 cell suspension (131) expressing GFP-AGO1; the aim was to analyze total RNA, sRNAs and AGO1-associated sRNAs at key points of the cell cycle. According to flow cytometry measurements of DNA contents and the monitoring of mitotic indexes, all five synchronization experiments were highly reproducible in the timing of the different cell cycle phases and their level of synchrony (Supplemental Figure 4). For each experiment, we collected RNA and protein samples for S, G2, M and G1 phases (at 1, 3, 7 and 10 hours after aphidicolin removal, respectively).
To focus on genes regulated during the cell cycle, we performed RNA-seq with three biological replicates to identify differentially expressed genes, comparing S, G2, M and G1 phases (Supplemental Figure 5A; Supplemental Table 1). Out of the 69,500 genes identified in the tobacco genome (Edwards et al., 2017) we were able to detect 66,752 (96%). Among those, considering all timepoints together, we identified a total of 23,060 differentially expressed (DE) transcripts corresponding to 13,623 genes with a functional annotation (Supplemental Data Set 1A,B). However, considering each comparison separately, the number of DE genes was variable, ranging from 6,737 to 14,771 DE genes in M/G1 and S/M comparisons, respectively (Figure 5A). However, the ratio of down-and up-regulated genes was maintained in each phase of the cell cycle, always around 50%. Considering only the DE genes with a fold change of 5 or higher, we observed a change in the expression trend, i.e. a high percentage of the DE genes were up-regulated in M and G1 phases, and down-regulated in S and G2 (Figure 5B and Supplemental Figure 5B). The genes highly expressed (minimal CPM>10) and showing high fold change (FC>5) in expression mainly encode histone-related proteins, and proteins involved in cytoskeleton and cytokinesis (Supplemental Figure 5C, green dots and Supplemental Data Set 2A,B). As expected, the expression of the histonerelated genes peaked in S-phase and they were weakly expressed in M-phase, while cytokinesis-related genes peaked in M-phase. In the group of genes showing high fold changes, we identified multiple cell cycle regulators such as cyclins. These genes are not highly abundant in all phases, but instead have expression levels dropping to values close to 0 in at least one cell cycle phase (Supplemental Figure 5C, blue dots and Supplemental Data Set 2C, D).
For a better understanding of the RNA-seq data structure, we performed a soft clustering analysis using Mfuzz package. Soft clustering groups together genes that have a similar expression pattern, providing insights in gene function and networks. We divided our DE dataset, containing more than 23,000 genes (q-value < 0.05), into 16 different clusters (Figure 5C and Supplemental Figure 6). We observed distinct expression patters, mostly due to the highly dynamic transcription. The two most-abundant clusters were clusters 11 and 6, which represent a single peak of expression during phases M and S, respectively (Figure 5C). Genes included in cluster 6 are involved in chromatin assembly and organization, receptormediated endocytosis, and phosphorylation, while genes included in cluster 11 were enriched in cellular component movement, cytoskeleton organization and intercellular protein transport (Supplemental Data Set 1).
Next, we conducted a Gene Ontology (GO) analysis on the DE genes by using the PANTHER software (see Methods). As expected, gene categories related to cell cycle functions such as DNA replication, chromosome segregation and mitotic processes including cytokinesis were well represented (Figure 5D). In addition, genes involved in DNA repair and epigenetic regulation of gene expression were also greatly represented. Finally, we searched specifically for genes known to play regulatory functions during the cell cycle and found that most of them are DE in our dataset (Supplemental Data Set 2E). Moreover, for many of them, we could identify their homologs in Arabidopsis, for which expression or functional data are already available.
As hundreds of different transcripts showed highly dynamic patterns of accumulation, we questioned whether sRNAs might be involved in their regulation. We performed deepsequencing analyses on total sRNAs and GFP-AGO1-associated sRNAs in the different cell cycle phases (Supplemental Table 2). AGO1-loaded sRNAs were isolated by immunoprecipitation with a GFP antibody (Supplemental Figure 7). As expected, we observed a size distribution of sRNAs in total RNA samples typical of Solanaceae, that is, one in which 24-mers are the most abundant class of sRNA, followed by 22-and 21-mers (Supplemental Figure 8A). In contrast, the AGO1-associated sRNAs were strongly enriched in 21 and 22 nt sRNAs (Supplemental Figure 8B), consistent with published AGO1 IP data (Mallory and Vaucheret, 2010), and the specific 22-nt peak observed in Solanaceae species, namely tomato (Solanum lycopersicum, Tomato and Consortium, 2012).
To investigate the distribution of sRNAs, we quantified the accumulation of miRNAs, tasiRNAs, hc-siRNAs and tRFs along the cell cycle (Figure 6A). hc-siRNAs were consistently the most abundant category in total RNA samples, while miRNAs represented the most abundant category in AGO1-IP samples. We observed no differences in tasiRNAs or hc-siRNAs in our data, nor were there significant differences between the cell cycle phases for the proportion of different sRNA classes (Supplemental Data Set 3). Therefore, we focused on miRNAs (this section) and tRFs (the subsequent section).
For miRNA analysis, we used the workflow described in Supplemental Figure 9. A total of 312 different mature sequences (including those from 3' and 5' arms, i.e. 3p and 5p miRNAs), belonging to 192 miRNA precursors and 103 different miRNA families (Supplemental Data Set 4A-C), were identified in this study, including the already-annotated miRNAs deposited in miRBase as well as newly-predicted miRNAs. This combined set totalled 221 distinct sequences. Most miRNAs (215 of 221) were present in both samples, total RNA and AGO1-IP. We found 10 miRNAs differentially accumulated ("DA") in total RNA and three in AGO1-IP (Figure 6B). Most of the DA miRNAs had a phase-specific accumulation. Six miRNAs showed a higher accumulation during the G2 phase, while only a single miRNA, miR168a-5p, was specific to M phase (Figure 6C). Two newly identified miRNAs (miR-33-3p and miR-12-5p) and two known miRNAs (miR6147a-3p and miR479) were more abundant in the G1 phase. Additionally, miR390-3p and miR482-5p levels peaked during S phase. In the AGO1-IP samples, we only found DA miRNAs when comparing opposite phases of the cell cycle (S/M and G2/G1). Thus, only a few miRNAs showed DA when comparing contiguous cell cycle phases.
We next asked whether some miRNAs identified in our analysis could target cell cycleregulated transcripts. To address this question, we generated Parallel Analysis of RNA Ends (PARE) libraries for three biological replicates of each phase of the cell cycle (Supplemental Table 3). To minimize sample variability, we focused on miRNA-target pairs present in all biological replicates and supported with robust evidence of cleavage (i.e. a sPARTA-defined score ≤ 3, see Methods) (Supplemental Data Set 4D). Thus, we found 769 unique miRNAtarget pairs considering all samples, corresponding to 124 different miRNAs and 168 different genes (Figure 6D). Among those, we found a core subset of miRNAs and their respective target genes that were present in all phases of the cell cycle. This includes 179 miRNA/target pairs, involving 51 miRNAs from 24 families, and 31 target genes (Figure 6D). When considering only genes that are DE during the cell cycle, we identified 48 target genes, representing about 30% of the total miRNA-target pairs, but only a small proportion of all the DE genes (Figure 6E; Supplemental Data Set 4E-H and Supplemental Figure 10).
The majority of the miRNA targets encode transcription factors (TFs) from AP2/ERF, MYB, SCARECROW and GRF TF families (Supplemental Data Set 4D). The elevated number of TFs targeted by miRNAs in all phases of the cell cycle suggests two layers of transcriptional control during the cell cycle -TFs and miRNAs. The second biggest group of target genes were disease resistance proteins (Supplemental Data Set 4D), suggesting that miRNAs may dampen their expression in proliferating cells. Finally, we observed several phase-specific miRNA-target pairs, that is, miRNA-target pairs that were only found in one phase of the cell cycle (Figure 6D; Supplemental Data Set 4).
Knowing that tRFs can form a complex with AGO proteins with important cellular functions (Shigematsu and Kirino, 2015;Martinez et al., 2017), we investigated if there were any DA tRFs in our samples. None of the tRFs derived from the 1,529 tRNAs analysed showed DA in total RNA samples. However, tRFs originating from 67 different tRNAs were DA in the AGO1-IP samples (Supplemental Data Set 5A-C). These tRFs were derived from 14 different types of tRNAs (Figure 7A). Interestingly, each DA tRF-producing tRNA was associated with a different phase of the cell cycle, with the exception of G1, which did not accumulate high levels of DA tRFs (Figure 7B). To further characterize the role of the DA tRFs during the cell cycle, we analysed their size, sequence, and targets. These DA tRFs had a size distribution pattern different from all tRFs (Figure 7C), with a higher abundance at 19 nt. We observed a conserved sequence among the DA 19-mers (Figure 7D), and they lacked the CCA or polyU signatures typical of the 3' end of the tRNA. We confirmed this 5' origination by analysis of their position of origin in tRNAs; they were thus classified as 5'-end tRFs.
Taking advantage of our PARE data, we looked for putative targets for the identified tRFs. As described previously (Martinez et al., 2017), tRFs have the ability to target transposable elements. Most of the targets of the tRFs that we could validate by PARE analysis were retrotransposons (Supplemental Data Set 5D; Figure 7E). Among these TEs, the majority was from the LTR-Gypsy superfamily (Figure 7F -left panel). However, when normalizing to the total number of TEs of each superfamily present in the tobacco genome, Short Interspersed Nuclear Elements (SINE) elements (a category of non-LTR retrotransposons with an internal region that originated from tRNAs) were enriched among the tRF targets (Figure 7F -right panel). We next analysed the expression of the TEs identified as tRF targets in our RNA-seq dataset. We identified 381,452 expressed TEs in our RNA-seq data out of 1,980,189 (~20% expressed). Of these 18 (~0.005% of the identified TEs) were differentially expressed (Supplemental Data Set 5) in one or more cell cycle comparisons, and only one of these was targeted by a tRF (~0.0003% of identified TEs).
In mammalian cells, the role of AGO proteins and miRNAs in the regulation of gene expression during the cell cycle has been well established (Bueno and Malumbres, 2011). However, in plants, this question remained unclear despite the fact that defects in miRNA biogenesis or in PTGS impact cell proliferation during embryonic and post-embryonic development. For instance, mutation of DCL1 alters embryogenesis at the globular stage, showing abnormal cell divisions throughout the suspensor and hypophysis (Schwartz et al., 1994;Nodine and Bartel, 2010). Similarly, aberrant patterns of cell division and cell expansion during embryogenesis have also been identified in mutants of Arabidopsis proteins that act together with DCL1, SE and HYL1 (Grigg et al., 2009;Armenta-Medina et al., 2017). The molecular basis of altered embryogenesis, pattern formation and abnormal cell divisions is however not yet well understood. Although dcl1 mutants exhibit altered auxin responses (Seefried et al., 2014;Grigg et al., 2009;Nodine and Bartel, 2010) and several miRNAs target AUXIN RESPONSE FACTORs (ARFs) (Jones-Rhoades and Bartel, 2004;Armenta-Medina et al., 2017), a causal link between auxin signaling and defects of embryogenesis in these mutants has not been established. To investigate whether miRNAs are required to maintain cell proliferations in meristems at a post-embryonic stage, we depleted AGO1 by targeted degradation using the viral encoded F-box protein P0. Under these conditions, depletion of AGO1 impaired meristem size and cell division activity in the root apical meristem. According to our cell cycle reporters, cells did however not arrest in the S or G2/M phases of the cell cycle, suggesting that they remained in G1.
To investigate the repertoire of miRNAs that are expressed at specific cell cycle phases and identify their targets, we chose to work with the tobacco BY2 cell suspension culture (Nagata et al., 1992), offering optimal synchronization efficiencies. The analysis of total and AGO1-associated miRNAs revealed that only a few miRNAs showed a differential accumulation pattern during the cell cycle. This situation is very different from mammals for which the transcription of some miRNAs is cell cycle regulated. For instance, some miRNAs are differentially expressed through direct activation by E2F and other transcription factors (Sylvestre et al., 2007;Woods et al., 2007;Bueno et al., 2010) and degraded in a cell cycledependent manner by specific endonucleases (Elbarbary et al., 2017). In addition, while mammalian miRNAs target numerous cell cycle transcripts at different cell cycle stages, this is not the case for plant miRNAs. Our analysis shows that miRNAs target only a subset of genes DE during the cell cycle and none of them correspond to core cell cycle genes (Supplemental Data Set 4; Supplemental Figure 10). A likely explanation of such a difference is that the mode of action of RISC in gene silencing differs between plant and animal cells.
Even though plant AGO1 is able to perform translational repression (Brodersen et al., 2008;Li et al., 2013), most of its activity is in slicing target transcripts. As this process requires a perfect or near-perfect complementarity between a miRNA and its mRNA target, it might preclude the possibility that a unique miRNA targets a wide spectrum of genes, as is the case in mammals (Friedman et al., 2009). Nonetheless, according to our PARE data requiring cleavage to be detected, we presently cannot exclude the possibility that some plant miRNAs might also control the translation of DE genes during the cell cycle.
Yet, our data support that plant miRNAs indirectly contribute to the control of cell proliferation by targeting some important regulators. Indeed, we found that most of miRNA targets are transcription factors, some being expressed in a cell cycle dependent manner (Supplemental Data Set 4; Supplemental Figure 10). For instance, we found that miR164 and miR160 target the NAC transcription factor CUC2 (CUP-SHAPED COTYLEDON2) and ARFs (AUXIN-RESPONSE FACTORs), respectively, in a cell cycle-dependent manner. CUC2 promotes the generation of auxin response maxima and therefore could release ARF repression, triggering downstream auxin signalling components such as cyclins and CDKs. Also interesting are some miRNAs that target F-box domain containing transcripts, including AUXIN SIGNALLING F-BOX 2. Two F-box encoding genes are targeted by members of the miR6149 family and will deserve particular attention as their expression is enriched in S-phase.
Though they belong to a different subclade than the Arabidopsis F-box FBL17, a key player of the G1/S transition (Noir et al., 2015), they could be involved in the turnover of important cell cycle regulatory proteins.
The molecular links between the cell cycle and defense responses in plants are still not well understood, but surprising connections have already been pointed out (Zebell and Dong, 2015). Microarray data from synchronized Arabidopsis cell suspension revealed that some resistance genes are differentially expressed during the cell cycle with a peak of expression either in S or M phase (Menges et al., 2005) and we also identified 100 of those DE resistance genes from our RNA-seq libraries. Interestingly, it has been shown that cell cycle misregulation can lead to the activation of disease resistance genes (Bao et al., 2013). For instance, overexpression of OSD1 (OMISSION OF SECOND DIVISION 1) or UVI4 (UV-B-INSENSITIVE 4), two negative regulators of the APC/C (anaphase-promoting complex or cyclosome), results in the expression of the defense marker gene PR1 and confers resistance to virulent pathogens. Conversely, loss of these two proteins, as well as the APC10 subunit or the CCS52 APC/C coactivator, led to enhanced susceptibility to bacterial infection (Bao et al., 2013). In addition, the loss of the resistance gene SNC1 (SUPPRESSOR OF npr1-1 CONSTITUTIVE 1) abolished pathogen resistance conferred by the osd1 mutation. On the other hand, plant pathogens and beneficial symbionts can also modulate the cell cycle activity by inducing or suppressing cell cycle gene expression (Favery et al., 2002). Taken together, these data suggest that the balance between cell proliferation and defense against pathogens implies complex regulatory circuits.
What our study showed is that genes encoding disease resistance proteins represent the second main group of miRNA targets. While many of these miRNA targets are constitutively expressed, some are DE during the cell cycle (Supplemental Data Set 4; Supplemental Figure 10). Therefore, one could assume that miRNAs in proliferating cells or even at specific cell cycle stages repress the expression of numerous resistance genes and are thus key for the antagonistic regulation of cell cycle and defense gene expression programs.
By contrast to other sRNAs, tRF abundances were surprisingly different between total sRNAs and AGO1-associated datasets. This suggests that tRF biogenesis is most likely constant during cell cycle, while loading in AGO1 of at least some tRFs must be cell cycle-regulated. It is already known that some plant tRFs accumulate in a wide range of stresses including oxidative stresses (Thompson et al., 2008) and UV stresses (Gebetsberger et al., 2012) that induce deleterious effects such as DNA damage. Interestingly, some tRFs that differentially accumulated in our libraries originate from the tRNAs that are also processed upon oxidative stresses, particularly tRNA-Arg (CCT) and tRNA-Glu (CTC) (Figure 7; Supplemental Data Set 5). In addition, 11 tRFs among the 14 DE tRFs appear to be enriched in S and G2, suggesting that they could be linked to DNA replication and/or G2 checkpoint processes.
Notably, most of the 19-mer tRFs associated with AGO1 have a G or a C at the 5' position (Figure 7; Supplemental Data Set 5), while it is established that AGO1 has a preference for a 5' U or A (Mi et al., 2008). This might explain why tRFs represent such a small subset of AGO1-associated small RNAs. Moreover, the presence of G/C at the tRFs 5' end might also affect the stability of the RISC complex and therefore further explain why we have detected high fold changes for AGO1-associated tRFs while other sRNAs remains overall unchanged during cell cycle. This raises the question of the biological significance of these tRFs differentially accumulated, as well as their mode of action.
To examine whether tRFs are able to trigger mRNA slicing while associated with RISC complex, we investigated the PARE signatures that match with tRF sequences. We found that Gypsy elements are the most abundant target for tRFs (Figure 7; Supplemental Data Set 5); this corresponds to what was found in previous studies in Arabidopsis (Martinez et al., 2017).
On the other hand, SINE elements were also abundant targets of tRFs. This might be due to their highly conserved internal regions that originated from tRNAs.
Besides slicing functions, some tRFs have also been shown to induce translation inhibition of some transcripts. For example, Val-tRF from the archaea Haloferax volcanii associates to the small ribosomal subunit and is able to inhibit protein synthesis by interfering with peptide bond formation, presumably to fine-tune overall protein synthesis in response to several stresses (Gebetsberger et al., 2012). A similar mechanism of tRF-mediated translational inhibition in HeLa cells has also been described (Sobala and Hutvagner, 2013). The 3' "GG" dinucleotide of these tRFs is essential for their function. Interestingly, this motif is also represented in our AGO1-associated tRFs sequences (Figure 7D), but whether they are involved in translational repression in plants, will require further investigation.
For the pAGO1:GFP-AGO1 construct, the pAGO1:GFP-AGO1 described in Derrien et al. (GGGGACCACTTTGTACAAGAAAGCTGGGTACTTAGGTGACATCGCTACTTCCTT) and used for recombination in the pDONR221 vector (Invitrogen). pCYCB1;2:CYCB1;2(Dbox) and pHTR2:CDT1a (C3) (Yin et al., 2014) entry clones were recombined with pGWB650 binary vector (RIKEN, Japan) to create a C-terminal fusion to the eGFP.
Arabidopsis thaliana missense mutant alleles ago1-36 (Baumberger and Baulcombe, 2005) and ago1-27 (Morel et al., 2002) have been described previously. The XVE:P0 and the XVE:P0 (ago1-57) transgenic lines are described (Derrien et al., 2018). Both lines were transformed with the pHTR2:CDT1a (C3)-GFP and pCYCB1;2:CYCB1.2(Dbox)-GFP constructs and double homozygous plants for both transgenes were obtained and used for further experiments.
For in vitro culture conditions, Arabidopsis seeds were surface-sterilized using ethanol and plated on growth medium (MES-buffered MS salts medium [Duchefa], 1% sucrose, and 0.8% or 1% agar, pH 5.7) in the presence of selective agent when appropriate. Seeds were stratified for 2 days at 4°C in the dark and then transferred in 16h-light/8h-dark (20,5/17°C,70% humidity) growth chamber, under fluorescent light (Osram Biolux 58W/965). For P0 induction, seedlings were grown as indicated above on medium supplemented with 10µM -oestradiol or supplemented with an equal volume of DMSO for the mock treatment.
A rapidly growing suspension culture of tobacco BY2 cells (Nicotiana tabacum cv Bright Yellow 2) was maintained by weekly dilution (1.5:80) of cells into fresh medium according to (Nagata et al., 1992) and cultured at 27°C with shaking at 130 r.p.m. in the dark. The pAGO1:GFP-AGO1 (Derrien et al., 2012) and H2B-tdTomato constructs (Adachi et al., 2011) were introduced by electroporation into the disarmed Agrobacterium tumefaciens strain LBA4404 and the resulting Agrobacterium strains were used to transform BY2 cells. Four ml of a 3-dayold BY2 culture were co-cultivated with 100 μl of an overnight culture of Agrobacterium in Petri dishes in the dark for 2 days at 27°C. Cells were then collected and washed three times with BY2 culture medium by centrifugation and were plated on solid medium supplemented with carbenicillin (500 μg ml-1), cefotaxime (500 μg ml-1), (hygromycin (30 μg ml-1). In order to establish clonal cell cultures for both pAGO1:GFP-AGO1 and pAGO1:GFP-AGO1; pRPS5a:H2B-tdTomato, at least 4-6 weeks were required before calli could be recovered.
Individual calli were subcultured into liquid medium containing carbenicillin (500 μg ml-1), cefotaxime (500 μg ml-1) and hygromycin (30 μg ml-1).
Tobacco BY2 cells were synchronized according to Nagata et al. (1992). Clonal BY2 cell suspensions 131 and 408 were used for Figure 3 and Supplemental Figure 3, respectively.
Note that the cell suspension 408 expresses slightly more GFP-AGO1 protein than 131, likely explaining the delay in synchronization. For determination of the mitotic index, cells were stained with 0.2 mg ml-1 4',6-diamidino-2-phenylindole (Sigma) in the presence of 0.2% Triton X-100; interphase, prophase, metaphase, anaphase and telophase figures were determined for at least 600 cells by using UV light microscopy. Flow cytometry analysis of the tobacco BY2 cell suspension was performed as described previously (Noir et al., 2015). Measurements with at least two biological replicates were performed.
After 6 to 12 d of growth in vitro in the vertical orientation, the plates were scanned and the root length of Arabidopsis wild type, mutants and transgenic lines was measured with ImageJ.
For root meristem measurements and the imaging of the root tip, roots were mounted on a microscopic slide in a 75 mg/mL propidium iodide solution. Root tips were imaged under the confocal microscope (Leica TCS SP8) in the plane of the quiescent centre. GFP was detected using argon laser excitation at 488 nm and through 505-550 nm emission filter-set. Propidium iodide was detected using DSSP laser excitation at 561 nm and through 600-630 nm emission filter-set. Imaging was performed using HCX APO CS (20x -0.7) objective lens. The meristem size was measured in ImageJ as the distance between the quiescent centre and the last dividing cell of the cortex. To assess the statistical significance, ANOVA tests were performed using the R software.
To image GFP-AGO1 in transformed BY2 cells (clonal cell suspensions 304), confocal images were obtained using detection settings described above. H2B-tdTomato was detected with DSSP laser excitation at 561 nm and though 575-650 nm emission filter-set. Imaging was performed using HCX APO CS (20x -0.7) air, HC PL APO CS2 (40x -1.1) water and HC PL APO CS2 (63x -1.4) oil objective lens. Images are presented as single sections. Differential interference contrast (Nomarski) was used for transmission light images.
RNA extraction was performed on dried material vacuum-filtered from 2 ml of BY-2 cell suspension using Tri-Reagent (Sigma-Aldrich, USA) according to the manufacturer's instructions. For quantitative RT-PCR, total RNA was extracted from BY-2 cells using Tri-Reagent (see above). In all assays, 2 μg of total RNA treated with DNAse I (ThermoFisher Scientific, USA) was reverse transcribed using High-Capacity cDNA reverse transcription kit (Thermo Fisher Scientific, USA). Quantitative PCR reactions were performed in a total volume of 10 μl of LightCycler 480 SYBR GREEN 1 mix (Roche, Switzerland) on a Lightcycler LC480 apparatus (Roche, Switzerland) according to the manufacturer's instructions. The mean values of at least three biological replicates were normalized using N. tabacum ubiquitinconjugating enzyme E2 (AB026056), protein phosphatase 2A (X97913) and L25 Ribosomal protein (L18908) genes as internal controls (Schmidt and Delaney, 2010).
Total proteins were extracted from tobacco BY2 cells and Arabidopsis seedlings using a phenol extraction protocol.
Frozen material was ground and resuspended in extraction buffer (0.7 M saccharose, 500 mM Tris HCl pH 8.5, 5 mM EDTA, pH 8, 100 mM NaCl, 2% βmercaptoethanol, complete mini protease inhibitor-Roche, Switzerland). The lysate was incubated with an equal volume of biophenol pH 8 under agitation. Proteins were precipitated from the phenol phase with five volumes of cold methanol/100mM ammonium acetate. Pellets were washed twice with cold methanol and resuspended in a buffer (10% glycerol, 3% SDS, 62 mM Tris HCl pH 8). Prior loading, protein samples were supplemented with 4X loading buffer (40% glycerol, 16% SDS, 250mM Tris HCl pH 8, 20% β-mercaptoethanol) to 1X final
concentration and incubated at 95°C for 2 minutes. 10 to 30 µg of total protein extracts were separated on SDS-PAGE gels and blotted onto Immobilon-P membrane (Millipore, USA).
GFP-AGO1 was detected using anti-GFP antibody (Mouse monoclonal JL-8, Clontech, USA) diluted to 1:2,000, Nta AGO1 was detected using anti-N. benthamiana AGO1 antibody (Rabbit polyclonal; (Csorba et al., 2010)) diluted to 1:10,000 (v:v); Nta CYCLIN B1 was detected using anti-N. tabacum CyclinB1 antibody (Rabbit polyclonal; (Criqui et al., 2000)) diluted to 1:1,000 (v:v); ACTIN protein was detected using anti-ACTIN antibody (Mouse monoclonal -AS163141 -Agrisera, Germany) diluted 1:20,000 (v:v). Arabidopsis AGO1 and AGO2 were detected using anti-AGO1 (Rabbit polyclonal AS09527 -Agrisera, Germany) diluted to 1:10,000 (v:v) and anti-AGO2 (Rabbit polyclonal AS132682 -Agrisera, Germany) diluted to 1:2,000 (v:v) , respectively.
1 g of frozen BY-2 cells was ground and resuspended in 3 volumes of extraction buffer (50 mM Tris HCl pH 7.5, 300 mM NaCl, 10% Glycerol, 5 mM MgCl 2 , 0.5% Triton X-100, 5 mM DTT, compete mini protease inhibitor; Roche, Switzerland). The lysate was incubated with 50 µl of GFP-trap magnetic beads (Chromotek, Germany) for 1 hour at 4°C on a rotary shaker.
Beads were washed 4 times with extraction buffer on a magnetic stand. RNA was recovered from IP fractions using Tri-Reagent (see above for RNA extraction) and resuspended in ultrapure water (Sigma-Aldrich, USA). For proteins, 1/10 th of the magnetic beads were resuspended in 1X loading buffer (10% Glycerol, 4% SDS, 50 mM Tris HCl pH 8, 5% βmercaptoethanol) and incubated at 95°C for 2 minutes.
For this study, a total of 40 small RNA libraries (only the 32 with at least 10 M mapped reads were ultimately used), 12 RNA libraries and 12 Parallel Analysis of RNA End (PARE) libraries were constructed. Small RNA libraries were constructed using total RNA and AGO1 IP RNAs, using TruSeq Small RNA Sample Preparation Kit (Illumina, USA) according to manufacturer's manual. RNA libraries were constructed using TruSeq Stranded mRNA Preparation kit (Illumina, USA). These libraries were sequenced using Illumina HiSeq 2500 and Illumina HiSeq 4000, respectively, by Fasteris SA (Switzerland). PARE libraries were constructed using total RNA following the protocol described in (Zhai et al., 2014) and sequenced using
Illumina HiSeq 2500 at the Delaware Biotechnology Institute (USA).
Sequencing data quality was assessed using FastQC. Reads were processed by removing the adaptor sequences using Trimmomatic (Bolger et al., 2014) and mapped to the Nicotiana tabacum genome version 4 ((Edwards et al., 2017) - https://solgenomics.net/organism/Nicotiana_tabacum/genome) using Bowtie (Langmead and Salzberg, 2012). For sRNA analysis, a modified version of miREAP was used as described in Arikit et al. (2014) for miRNA prediction. Only miRNA precursors with reads present in both arms and mature miRNAs 21 nt or 22 nt long were selected. New miRNA candidates were then used for homology search using miRBase v22 (Kozomara and Griffiths-Jones, 2014), with a maximum difference of four nucleotides. To analyze hc-siRNAs, we first ran RepeatMasker (http://www.repeatmasker.org) using the Viridiplantae RepBase database
(http://www.girinst.org/repbase) as library. For RNA-seq, analyses were performed using the HISAT2 and StrigtTie as described in (Pertea et al., 2016). Clustering analysis was done using MFuzz v3.6. Gene Ontology (GO) analyses were done using the Arabidopsis orthologs of tobacco genes and the software PANTHER v11 (Mi et al., 2017) with default parameters.
PARE libraries were analyzed using sPARTA (Kakrana et al., 2014) and CleaveLand v4 (Addo-Quaye et al., 2009). Three biological replicates of each cell cycle phase were used as starting material. Only miRNA-target pairs present in all biological replicates, with a score of three or lower, were considered for further analysis. All statistical analysis were performed with DESeq2 (Love et al., 2014) using default parameters and raw counts, establishing the significant threshold at a q-value < 0.05. (A) Root length measurements at 6 and 12 days after stratification (DAS) of the indicated genotypes under mock (-) or β-estradiol (10 μM) to induce P0 (+). At least 30 seedlings per line, per treatment were measured and ANOVA was performed to assess significant differences. *** highlights comparisons for which p<0.001. (B) Root-meristem size of wild-type seedlings compared to the indicated genotypes. Cortex meristematic cells showing no sign of differentiation were counted. Values are meristem length of 6 DAS seedlings germinated with (+) or without (-) β-estradiol (10 μM). ANOVA was performed to assess significant differences. * highlights comparison for which 0.01<p<0.05. Right panels show primary root tips of XVE:P0 and XVE:P0 (ago1-57) after 12 days of P0 induction. Roots were counterstained with propidium iodide. The arrow indicates the length of the meristematic zone (from the QC to the first elongating cell of the cortex). Bars = 50 μm. (A) Confocal laser scanning images of primary root tips of XVE:P0 and XVE:P0 (ago1-57) lines expressing the indicated cell cycle markers after 6 and 10 days of P0 induction with β-estradiol (10 μ M), or mock treatment. Bars = 50 µM. The timing of expression of both cell cycle markers is illustrated on the top panel. (B) Immunoblot analysis of AGO1 content in mock (-) or P0 induced (+) 6-days-old seedlings of the same genotypes as in (A), with or without β-estradiol. Probing with the ACTIN antibody and Coomassie blue (CB) staining were used as loading controls. Quantification of AGO1 and AGO2 signal normalized (E) Transposable element distribution of the tRFs targets identified using PARE data. The two main categories are represented here, DNA and retrotransposons.
(F) Retrotransposable elements targeted by tRFs identified using PARE data, divided into superfamilies.
Represented here are the percentages of total number of targeted retrotransposons (left panel) and the normalized number of retrotransposons (right panel). (A) Root length measurements at 6 and 12 days after stratification (DAS) of the indicated genotypes under mock (-) or β-estradiol (10 μM) to induce P0 (+). At least 30 seedlings per line, per treatment were measured and ANOVA was performed to assess significant differences. *** highlights comparisons for which p<0.001. (B) Root-meristem size of wild-type seedlings compared to the indicated genotypes. Cortex meristematic cells showing no sign of differentiation were counted. Values are meristem length of 6 DAS seedlings germinated with (+) or without (-) β-estradiol (10 μM). ANOVA was performed to assess significant differences. * highlights comparison for which 0.01<p<0.05. Right panels show primary root tips of XVE:P0 and XVE:P0 (ago1-57) after 12 days of P0 induction. Roots were counterstained with propidium iodide. The arrow indicates the length of the meristematic zone (from the QC to the first elongating cell of the cortex). Bars = 50 µm. @AGO2 @ACTIN @ACTIN @AGO1 CB1 CB2 XVE:P0 pCYCB1.2:CYCB1.2 (dBox)-GFP XVE:P0/ ago1-57 XVE:P0 XVE:P0/ ago1-57 pHTR2:CDT1a(C3)-GFP 1.0 0.5 1.0 0.9 1.0 0.3 1.0 1.0 1.0 0.6 1.0 0.1 1.0 0.3 1.0 0.2 kDa 130 50 50 130 B Figure 2 0 20000.0 40000.0 60000.0 80000.0 100000.0 Relative P0 expression level XVE:P0 pCYCB1.2:CYCB1.2 (dBox)-GFP XVE:P0/ ago1-57 XVE:P0 XVE:P0/ ago1-57 pHTR2:CDT1a(C3) -GFP C D 0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 Relative expression level AGO1 AGO2 XVE:P0 pCYCB1.2:CYCB1.2 (dBox)-GFP XVE:P0/ ago1-57 XVE:P0 XVE:P0/ ago1-57 pHTR2:CDT1a(C3)-GFP 10 DAS +P0 6 DAS Mock 6 DAS +P0 XVE:P0 pHTR2:CDT1a(C3) -GFP pCycB1.2: CycB1.2(dbox)-GFP XVE:P0/ago1-57 pHTR2:CDT1a(C3) -GFP pCycB1.2: CycB1.2(dbox)-GFP A C D T 1 a ( C 3 ) -G F P C Y C B 1 .2 -G F P G2 G1 S M 10 DAS +P0 6 DAS Mock 6 DAS +P0 -Estradiol:
-Estradiol:
-Estradiol: (A) Confocal laser scanning images of primary root tips of XVE:P0 and XVE:P0 (ago1-57) lines expressing the indicated cell cycle markers after 6 and 10 days of P0 induction with β-estradiol (10 μ M), or mock treatment. Bars = 50 µM. The timing of expression of both cell cycle markers is illustrated on the top panel. (B) Immunoblot analysis of AGO1 content in mock (-) or P0 induced (+) 6-days-old seedlings of the same genotypes as in (A), with or without β-estradiol. Probing with the ACTIN antibody and Coomassie blue (CB) staining were used as loading controls. Quantification of AGO1 and AGO2 signal normalized to actin is indicated on top of their respective panels (C) RT-qPCR analysis of P0 (upper panel) and AGO1 and AGO2 (bottom panel) mRNA levels in the same samples as in (B). 0.5 0 1.0 10h 4C 2C 4C 4C 2C 2C 4C 2C 4C 2C 4C 2C 4C 2C @ GFP 50 CB1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 kDa @ ACTIN @ Nta AGO1 @ Nta CYCLIN B1 CB2 130 50 130 Time after aphidicolin removal (h) Time after aphidicolin removal (h) 0 0.5 1.0 1.5 2.0 2.5 0 1 3 5 7 9 11 13 14 Relative protein accumulation Time after aphidicolin removal (h) 0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 0 1 2 3 4 5 6 7 8 9 10 11 12 Mitotic index (%) Time after aphidicolin removal (h) cells with dots (%) MI GFP-AGO1+H2B-tdTomato 20 µm B GFP-AGO1 20 µm D 50 µm 1h30 20min C 50 µm 50 µm DIC WT WT 50 µm A 50 µm GFP-AGO1 10 µm 5h 50 µm E DIC 10 µm 9h 10 µm G H 50 µm 11h F F 8h 10 µm GFP-AGO1 50 µm I GFP-AGO1+H2B-tdTomato 50 µm (A) Confocal laser imaging of GFP-AGO1 expressing BY2 cells (left panel) and wild type (WT) BY2 cells (middle panel) with its corresponding differential interference contrast (DIC) reference image to highlight the background signal. The fluorescence signal was acquired in the 500-550nm band (excitation: 480nm) for GFP detection. (B-H) Subcellular localisation of GFP-AGO1 in synchronized cell culture. (B) Mitotic index (MI) of the synchronization experiment used for imaging. The green line shows the percentage of cells exhibiting GFP-AGO1 nuclear foci. (C-H) Confocal acquisition images of GFP-AGO1-expressing synchronized BY2 cells at the indicated time points (after aphidicolin removal). Images correspond to cells in S-phase (C-D), G2 phase (E), metaphase and anaphase (respectively F and G) and cells in G1 (H). Arrows highlight the presence of nuclear foci in G2 cells (E), quantified in (B). (I) Confocal laser imaging of asynchronous cells expressing GFP-AGO1 and the chromatin marker pRPS5:H2B-tdTomato. The two panels on the left show the GFP-AGO1 signal only or both the GFP-AGO1 and H2B-tdTomato signals. The two panels on the right are close-up views of the area delimited by the dashed squares on the left panels. Arrows highlight the presence of nuclear foci. 0 50.0 100.0 150.0 200.0 250.0 300.0 c e l l c y c l e v e s i c l e -m e d i a t e d t r a n s p o r t D N A m e t a b o l i c p r o c e s s D N A r e p a i r m i t o s i s R N A s p l i c i n g e n d o c y t o s i s r R N A m e t a b o l i c p r o c e s s D N A r e p l i c a t i o n e x o c y t o s i s p r o t e i n f o l d i n g c y t o s k e l e t o n o r g a n i z a t i o n c e l l u l a r c o m p o n e n t m o v e m e n t p r o t e i n p h o s p h o r y l a t i o n n u c l e a r t r a n s p o r t p h o s p h o l i p i d m e t a b o l i c p r o c e s s f a t t y a c i d m e t a b o l i c p r o c e s s m o n o s a c c h a r i d e m e t a b o l i c p r o c e s s c h r o m o s o m e s e g r e g a t i o n D N A r e c o m b i n a t i o n m e i o s i s p r o t e i n l i p i d a t i o n Number of genes GO Slim category Expected genes DE genes A B SvsG2 G2vsM MvsG1 G1vsS SvsM G2vsG1 C D Figure 5 0 2000.0 4000.0 6000.0 8000.0 10000.0 12000.0 14000.0 16000.0 S v s G 2 G 2 v s M M v s G 1 G 1 v s S S v s M G 2 v s G 1 Cell cycle phase Down Up -1 0 1 Average expression Color Key Phase S Color Key Phase G2 Phase M Phase G1 13 16 14 15 12 11 10 I II III IV V VI -4 -2 0 2 4 log 2 FC Color Key Figure 5. RNA-seq analysis of differentially expressed genes during cell cycle. (A) Summary of differentially expressed genes upon pairwise comparison of 2 cell cycle phases. Downand up-regulated genes are represented in purple and orange, respectively. (B) Heatmap representation of differentially regulated genes with a fold change bigger or equal to 5. (C) Heatmap representing the expression pattern of all clusters. Clusters are divided into 6 groups (I to VI) following their expression in each phase of the cell cycle. (D) GO Slim categories representing the biological function of differentially expressed genes. Here are only represented GO categories with more than 1.5x representation fold change. B C Figure 6 A miR8036-5p miR6025-3p miR166-3p miR166-5p miR6149b-3p miR-16-3p S G 2 G 1 M miR168a-5p miR390-3p miR482-5p miR6147a-3p miR479-5p miR-33-3p miR-12-5p D RNA se q DE PARE seq 13574 48 (0.4 % of RNA seq DE / 30% of PARE seq) E 120 S G2 M G1 m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s 0 5x10 5 1x10 5 Reads per million (RPM) S G2 M G1 m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s m i R N A s t a s i R N A s h c s i R N A s t R F s 0 2x10 5 4x10 5 6x10 5 Reads per million (RPM) SvsG2 G2vsM MvsG1 G1vsS SvsM G2vsG1 ** * * AGO1IP miR479-5p miR168a-5p miR390b-3p miR6149b-3p miR166d-3p miR-33-3p miR-12-5p miR6147a-3p miR16-3p miR6025c-3p miR482a-5p miR166b-5p miR8036-5p (E) Transposable element distribution of the tRFs targets identified using PARE data. The two main categories are represented here, DNA and retrotransposons.
(F) Retrotransposable elements targeted by tRFs identified using PARE data, divided into superfamilies.
Represented here are the percentages of total number of targeted retrotransposons (left panel) and the normalized number of retrotransposons (right panel).
Pubmed: Author and Title Google Scholar: Author Only Title Only Author and Title