A transcriptome-wide study on the microRNA- and the Argonaute 1-enriched small RNA-mediated regulatory networks involved in plant leaf senescence

Data sources

The cDNA sequences of SAGs in Arabidopsis and rice were retrieved from the FTP site (ftp://ftp.cbi.pku.edu.cn/pub/database/LSD_2.0/) of the Leaf Senescence Database (LSD 2.0; Liu et al. 2011; Li et al. 2014). The mature miRNA sequences of the two plant species were obtained from miRBase (release 20; http://www.mirbase.org/; Griffiths-Jones et al. 2008). The gene annotations were retrieved from TAIR (The Arabidopsis Information Resource; release 10; Huala et al. 2001) and MSU Rice Genome Annotation Project (release 7; Yuan et al. 2003). The high-throughput sequencing (HTS) datasets utilised to search for AGO1-enriched sRNAs were downloaded from GEO (Gene Expression Omnibus; http://www.ncbi.nlm.nih.gov/geo/; Barrett et al. 2009). Briefly, in Arabidopsis, the HTS datasets were divided into six groups: (i) GSM149079 (sRNAs from whole extracts) and GSM149080 (sRNA from AGO1 complexes); (ii) GSM642335 (AGO1-associated sRNAs_control) and GSM642336 (AGO1-associated sRNAs); (iii) GSM707678 (flower_control) and GSM707682 (flower_AGO1-associated sRNAs_two-step purification); (iv) GSM707679 (leaf_control) and GSM707683 (leaf_AGO1-associated sRNAs_two-step purification); (iv) GSM707680 (root_control) and GSM707684 (root_AGO1-associated sRNAs_two-step purification); (vi) GSM707681 (seedling_control) and GSM707685 (seedling_AGO1-associated sRNAs_two-step purification). In rice, there are three groups: (i) GSM455965 (seedling_total extracts) and GSM455962 (seedling_AGO1a-associated sRNAs); (ii) GSM455965 (seedling_total extracts) and GSM455963 (seedling_AGO1b-associated sRNAs); (iii) GSM455965 (seedling_total extracts) and GSM455964 (seedling_AGO1c-associated sRNAs). For expression analysis of the Arabidopsis regulatory miRNAs and sRNAs in rice, GSM278532, GSM278534, GSM278535, GSM409318, GSM722128 and GSM943194 were retrieved from GEO, and ShCn2D, ShK2D, ShKNa1I, ShKNa2D, ShN2D, ShP2D and ShS2D were downloaded from Next-Gen Sequence Databases (http://mpss.udel.edu/common/web/library_info.php?SITE=rice_sRNA2&showAll=true; Nakano et al. 2006).

Prediction and validation of the sRNA targets

Target prediction was performed using the miRU algorithm (Zhang 2005; Dai & Zhao 2011) with default parameters. The degradome sequencing data were utilised to validate the predicted sRNA–target pairs. First, in order to allow cross-library comparison, the normalised read count (in RPM, reads per million) of a short sequence from a specific degradome library was calculated by dividing the raw count of this sequence by the total counts of the library, and then multiplied by 10⁶. Second, all the degradome signatures were mapped onto the predicted target transcripts. Then, the previously proposed criteria (Shao et al. 2013) were applied to extract the potential cleavage sites. In summary, (i) ‘Average_Read count_Cleavage site’ is the averaged read count (in RPM) of all the degradome signatures (belonging to one library) with their 5′-ends mapped to a potential cleavage site; ‘Average_Read count_Surrounding’ is the averaged read count of all the degradome signatures (also belonging to this library) that mapped to the regions surrounding the cleavage site; ‘Average_Read count_Cleavage site’ should be five times or more than ‘Average_Read count_Surrounding’. (ii) Also for this degradome library, among the degradome signatures mapped to a potential cleavage site, the most abundant tag should be amongst the top 12 most-abundant degradome signatures that perfectly mapped to the corresponding transcript. (iii) The cleavage site should reside within the 9th to 10th, 10th to 11th or 11th to 12th nucleotide region of the regulatory sRNA. For any degradome library, if the three rules were fulfilled, the potential slicing sites were retained. Finally, both global and local target plots were drawn to perform manual screening, referring to our previous study (Meng et al. 2011). Only the transcripts with cleavage signals easily recognised were extracted as potential sRNA–target pairs.

Search for the AGO1-enriched sRNAs in Arabidopsis and rice

To do a comprehensive search for the AGO1-enriched sRNAs, HTS data prepared from the AGO1-associated sRNAs were downloaded from GEO (Barrett et al. 2009). Then, the HTS datasets were classified into six groups in Arabidopsis and three groups in rice (Fig. 1; see also 2 for details). We adopted the following criterion to extract the AGO1-enriched sRNAs: for each group, the accumulation levels of an AGO1-enriched sRNA should be ten RPM or higher in the ‘AGO1’ dataset, and should be undetectable in the ‘control’ dataset. As a result, 1267, 62, 516, 171, 3175 and 157 AGO1-enriched sRNAs were discovered in the six groups of Arabidopsis, and 77, 830 and 545 AGO1-enriched sRNAs were identified in the three groups of rice. Then, for each plant, the AGO1-enriched sRNAs extracted from different groups were combined. The redundant ones and the miRBase-registered miRNAs (release 20; Griffiths-Jones et al. 2008) were removed. For the following analysis, sRNAs shorter than 31 nt were retained (four sRNAs were removed in Arabidopsis and none in rice). As a result, a total of 5,123 and 1,399 AGO1-enriched sRNAs were identified in Arabidopsis and rice, respectively (Data S1 and S2).

Next, the sequence characteristics of the 5123 and 1399 AGO1-enriched sRNAs were analysed separately. Quite consistently, in both Arabidopsis and rice, large portions of the sRNAs are 21 nt in length [1243 (24.26%) in Arabidopsis and 655 (46.82%) in rice], and start with 5′ U [2250 (43.92%) in Arabidopsis and 914 (65.33%) in rice; Fig. 2). This observation correlates well with the previous report on the sequence features of the sRNAs enriched in the AGO1-associated silencing complexes (Mi et al. 2008). To some extent, it also indicates that the criterion for the identification of AGO1-enriched sRNAs is effective and reliable.

Several SAGs are targeted by AGO1-enriched sRNAs in addition to miRNAs in Arabidopsis

To address the question whether the miRNAs and the above-identified AGO1-enriched sRNAs are implicated in leaf senescence, the cDNAs of the SAGs were obtained from LSD 2.0 (Liu et al. 2011; Li et al. 2014). Then, all of the miRNAs and the AGO1-enriched sRNAs were included for target prediction using the miRU algorithm with default parameters (Zhang 2005; Dai & Zhao 2011). The predicted targets were validated using degradome sequencing data according to the previously reported approach (Meng et al. 2011; Shao et al. 2013). As a result, 200 regulatory pairs involving 120 AGO1-enriched sRNAs and 53 transcripts encoded by 40 SAGs, and 266 regulatory pairs involving 64 miRNAs and 54 transcripts belonging to 42 SAGs were discovered in Arabidopsis (a total of 61 SAGs were targeted by the AGO1-enriched sRNAs or miRNAs, or both). Notably, evident cleavage signals resided within the 9th to 11th nucleotide of the sRNAs or the miRNAs were detected for most of the regulatory pairs (Fig. 3A, B, D; Figures S1 and S2). However, in rice, only one transcript (LOC_Os03g53230.1) was targeted by miR395, and one transcript (LOC_Os06g12870.1) targeted by three AGO1-enriched sRNAs (OS_AGO1_sRNA302, OS_AGO1_sRNA304 and OS_AGO1_sRNA495; Fig. 3C, E; Figures S3 and S4).

Although the SAGs retrieved from LSD 2.0 were considered as involved in leaf senescence according to reports from Luo's group (Liu et al. 2011; Li et al. 2014), we performed a GO (Gene Ontology) enrichment analysis to interrogate the functional correlation of the target genes with leaf senescence in Arabidopsis. A web toolkit agriGO (Du et al. 2010) was employed for this analysis by using the following parameters: (i) ‘Singular Enrichment Analysis (SEA)’ was selected; (ii) the species was A. thaliana; and (iii) using suggested background ‘Arabidopsis gene model (TAIR9)’. The 53 transcripts targeted by the 120 AGO1-enriched sRNAs and the 54 transcripts targeted by 64 miRNAs were analysed separately. The results showed that for both groups of targets, the GO terms ‘organ development’, ‘shoot development’ and ‘leaf development’ were highly enriched compared to the background (Figures S5 and S6). Detailed annotations of the 61 SAGs were obtained from TAIR (release 10; Huala et al. 2001) for further investigation. In support, 23 out of the 61 SAGs are annotated as involved in ‘leaf development/morphogenesis/differentiation/senescence’ (AT2G31070, AT3G15030, AT4G18390, AT5G05700, AT5G39610 and AT5G62000), ‘shoot apical meristem formation’ (AT1G56010 and AT2G45160), ‘programmed cell death’ (AT3G11440 and AT5G48380), ‘anthocyanin-containing compound biosynthetic process’ (AT5G06510), or located/expressed in ‘chloroplast’ (AT2G41220, AT3G11960, AT3G59950, AT4G13400, AT5G01600, AT5G03560, AT5G04360, AT5G18100 and AT5G41620) or ‘guard cell’ (AT1G15125, AT3G01080, AT4G13400, AT4G27310, AT5G06510; Table 1). For the two rice SAGs targeted by miR395 and three AGO1-enriched sRNAs, LOC_Os06 g12870 encodes a leaf senescence-related protein according to the annotation of MSU Rice Genome Annotation Project (release 7; Yuan et al. 2003). Besides, a previous study indicated that the other SAG (LOC_Os03 g53230: bifunctional 3-phosphoadenosine 5-phosphosulfate synthetase) might be involved in the early senescence process of rice flag leaves (Liu et al. 2008). Taken together, our functional analyses demonstrate that the target genes, at least a large portion of them, are strong SAG candidates, and the corresponding regulatory miRNAs and AGO1-enriched sRNAs play potential roles in leaf senescence or other related developmental processes.

Construction of AGO1-enriched sRNA-mediated networks involving target interactions at the protein level

Based on the sRNA/miRNA–target list identified above, a regulatory network involved in leaf senescence could be constructed in Arabidopsis (Figure S7). However, we recognised that many factors had not been included in this network, suggesting a far more a complex network in planta. One of the important factors not considered is protein–protein interactions within the network. First, for the 61 target genes identified in Arabidopsis, are there any candidates that could interact with them at the protein level? Are these interacting candidates targeted by the miRNAs or the other AGO1-enriched sRNAs in Arabidopsis? To address these questions, we set out to predict candidate genes interacting with the 61 target genes at the protein level. PAIR (Predicted Arabidopsis Interactome Resource) was used for this prediction (Lin et al. 2011). As a result, a total of 459 interacting candidates was discovered (Data S3). Then, target identification was performed to investigate whether these candidate genes could be targeted by the miRNAs or the AGO1-enriched sRNAs in Arabidopsis. Fortunately, based on the degradome sequencing data, 16 transcripts encoded by eight interacting candidate genes were validated as regulated by 17 AGO1-enriched sRNAs in Arabidopsis (Figure S8), and 11 transcripts of five interacting candidate genes were targeted by ten miRNAs in Arabidopsis (Figure S9; a total of ten interacting candidate genes were targeted by the AGO1-enriched sRNAs or the miRNAs, or both). Notably, according to the TAIR annotations, six out of the ten target genes are either located in ‘chloroplast’ (AT1G04140, AT1G12520, AT2G28190, AT5G21930 and AT5G53480) or involved in ‘programmed cell death’ (AT5G06100) (Table 1). In this regard, the ten interacting candidate genes, along with the regulatory miRNAs and the AGO1-enriched sRNAs, were integrated into the leaf senescence-associated network (Figure S7).

Search for conserved sub-networks in rice

If the regulatory network identified in Arabidopsis is indeed important for the modulation of leaf senescence-associated processes, we speculated that some of the sub-networks should be highly conserved among diverse plant species. In this respect, all of the regulatory miRNAs and the AGO1-enriched sRNAs within the above established network of Arabidopsis were included to search for conserved sRNAs in rice. For the regulatory miRNAs of Arabidopsis, the families miR156, miR159, miR160, miR164, miR169, miR171, miR319, miR393, miR396, miR397 and miR414 are conserved in rice. Notably, most of these families are widespread among diverse plant species according to the miRBase registries (release 20; Griffiths-Jones et al. 2008), indicating the indispensible roles of these miRNAs in leaf senescence, which needs further experimental validation. For the remaining miRNA families including miR157, miR163, miR170, miR782, miR854, miR858, miR5021, miR5643 and miR5658, no conserved partner is reported in rice. Although not registered in the current release of miRBase, we observed that the sRNA sequence identical to miR157a, miR157b and miR157c of Arabidopsis was highly accumulated in the tricellular pollen of rice. Moreover, the sRNA sequence identical to ath-miR163 was abundant in the grains and roots of rice (Table 2). Both sRNAs were weakly expressed in the leaf of rice. Thus, whether these rice sRNAs possess species-specific roles, such as reproduction and root development, needs further investigation.

Five sRNAs whose sequences were identical to AT_AGO1_sRNA3814, AT_AGO1_sRNA3923, AT_AGO1_sRNA4033, AT_AGO1_sRNA5067 and AT_AGO1_sRNA5074 were identified as highly abundant in the AGO1-associated silencing complexes of rice. Thus, the five sRNAs were named AT_AGO1_sRNA3814_conserved_in_rice (81.97 RPM in AGO1c versus 1.49 RPM in control set), AT_AGO1_sRNA3923_conserved_in_rice (12.33 RPM in AGO1c versus 0.99 RPM in control), AT_AGO1_sRNA4033_conserved_in_rice (13.19 RPM in AGO1a versus 5.21 RPM in control), AT_AGO1_sRNA5067_conserved_in_rice (7.19 RPM in AGO1c versus 0 RPM in control) and AT_AGO1_sRNA5074_conserved_in_rice (7.71 RPM in AGO1c versus 0.25 RPM in control), respectively. More interestingly, four out of the five sRNAs are highly accumulated (> 5 RPM) in the leaves or shoots of rice. This expression-based evidence points to the possibility that the five sRNAs might have leaf senescence-associated functions through cleavage-based regulation of certain downstream targets. In this consideration, degradome sequencing databased identification of the target genes was performed. As a result, five transcripts encoded by five genes were targeted by AT_AGO1_sRNA3814_conserved_in_rice, four transcripts encoded by four genes were targeted by AT_AGO1_sRNA3923_conserved_in_rice, 18 transcripts encoded by ten genes were targeted by AT_AGO1_sRNA4033_conserved_in_rice, one transcript LOC_Os01g73760.1 was regulated by AT_AGO1_sRNA5067_conserved_in_rice, and eight transcripts encoded by four genes were regulated by AT_AGO1_sRNA5074_conserved_in_rice (Fig. 4; Figure S10). Based on annotations from the MSU Rice Genome Annotation Project (release 7; Yuan et al. 2003), certain sRNA–target pairs were identified as conserved between Arabidopsis and rice. For examples, in Arabidopsis, AT_AGO1_sRNA3814 regulates AT1G56010, AT5G07680, AT5G39610 and AT5G61430 encoding NAC domain-containing proteins. AT_AGO1_sRNA3814_conserved_in_rice also targets LOC_Os06g23650, LOC_Os06g46270, LOC_Os08g10080 and LOC_Os12g41680 encoding no apical meristem proteins belonging to the NAC family (Fig. 4A). Based on previous studies, the NAC gene family members play highly conserved roles in modulating the biological processes of leaf senescence in diverse plant species, such as Arabidopsis (Lin & Wu 2004; Guo & Gan 2006; Kim et al. 2009; Balazadeh et al. 2010; Yang et al. 2011; Lee et al. 2012; Zhang & Gan 2012; Hickman et al. 2013), rice (Liang et al. 2014), cotton (Gossypium hirsutum) (Fan et al. 2015) and wheat (Triticum aestivum) (Zhao et al. 2015). In Arabidopsis, it has been reported that miR164-mediated regulation of ORE1, a NAC transcription factor, is essential for age-dependent cell death during leaf senescence (Kim et al. 2009). In our study, AGO1_sRNA3814 was conserved in Arabidopsis and rice, and had the conserved target genes belonging to the NAC families in both plants. Thus, in addition to miR164, whether AGO1_sRNA3814 or some other sRNAs is indeed involved in leaf senescence, and whether such regulation is widespread across diverse plant species need to be further investigated. AT_AGO1_sRNA5074 targets AT2G28350 encoding an Auxin Response Factor (ARF), and AT_AGO1_sRNA5074_conserved_in_rice regulates LOC_Os02g41800, LOC_Os04g43910, LOC_Os06g47150 and LOC_Os10g33940 also encoding ARFs (Fig. 4B). According to previous reports, ARF1 and ARF2 positively regulate leaf senescence in Arabidopsis (Lim et al. 2010; Ellis et al. 2005). In rice, a genome-wide analysis of miRNAs and the target genes related to leaf senescence revealed that several ARF genes targeted by miR160 or miR167 were potentially implicated in miRNA-mediated leaf senescence (Xu et al. 2014). Here, we identified that AGO1_sRNA5074 also targeted ARF genes, and was related to leaf senescence in both Arabidopsis and rice. Thus, it will be interesting to thoroughly clarify the miRNA- and AGO1-enriched sRNA-mediated auxin signalling pathways related to plant leaf senescence. AT_AGO1_sRNA3923_conserved_in_rice targets LOC_Os02g44360, LOC_Os02g44370, LOC_Os04g46860 and LOC_Os10g40390 encoding SCARECROW family proteins, and AT_AGO1_sRNA3923 targets AT2G45160 encoding HAIRY MERISTEM 1 (HAM1) homologous to SCARECROW proteins (Fig. 4C). Some reports pointed to involvement of SCARECROW genes in leaf or shoot development (Wysocka-Diller et al. 2000; Dhondt et al. 2010); however, to our knowledge, the SCARECROW gene family has not been clearly related to leaf senescence. In this regard, whether the AGO1_sRNA3923–SCARECROW regulatory pathway plays an important role in leaf senescence needs to be investigated. Taken together, these conserved sub-networks mediated by the three sRNAs (AGO1_sRNA3814, AGO1_sRNA3923 and AGO1_sRNA5074; Fig. 4A, C) might play conserved roles in leaf senescence in Arabidopsis and rice. Notably, some novel target genes were identified in rice. In addition to the NAC genes, AT_AGO1_sRNA3814_conserved_in_rice also regulates LOC_Os11 g44940 encoding a seed maturation protein, PM27 (Fig. 4A), indicating its involvement in seed development in rice. In Arabidopsis, AT_AGO1_sRNA4033 targets AT1G53910 [Related to AP2.12 (RAP2.12)], and AT5G43060 [Responsive to Dehydration 21B (RD21B)] is targeted by AT_AGO1_sRNA4033 and AT_AGO1_sRNA5067. However, in rice, AT_AGO1_sRNA4033_conserved_in_rice targets nine genes encoding growth-regulating factors, and LOC_Os01g73760 (IPP transferase) is targeted by AT_AGO1_sRNA4033_conserved_in_rice and AT_AGO1_sRNA5067_conserved_in_rice (Fig. 4D). These regulatory pairs indicate that divergent biological roles of the three sRNA sequences (AGO1_sRNA3814, AGO1_sRNA4033 and AGO1_sRNA5067) might exist between Arabidopsis and rice.

In summary, a comprehensive study of the involvement of miRNAs and other AGO1-enriched sRNAs in the regulation of SAGs was conducted. The obtained results could advance our understanding of the sRNA-mediated molecular processes modulating leaf senescence, and could serve as a basis for further experimental investigations.