A genetic link between epigenetic repressor AS1-AS2 and a putative small subunit processome in leaf polarity establishment of Arabidopsis

ABSTRACT Although the DEAD-box RNA helicase family is ubiquitous in eukaryotes, its developmental role remains unelucidated. Here, we report that cooperative action between the Arabidopsis nucleolar protein RH10, an ortholog of human DEAD-box RNA helicase DDX47, and the epigenetic repressor complex of ASYMMETRIC-LEAVES1 (AS1) and AS2 (AS1-AS2) is critical to repress abaxial (ventral) genes ETT/ARF3 and ARF4, which leads to adaxial (dorsal) development in leaf primordia at shoot apices. Double mutations of rh10-1 and as2 (or as1) synergistically up-regulated the abaxial genes, which generated abaxialized filamentous leaves with loss of the adaxial domain. DDX47 is part of the small subunit processome (SSUP) that mediates rRNA biogenesis. In rh10-1 we found various defects in SSUP-related events, such as: accumulation of 35S/33S rRNA precursors; reduction in the 18S/25S ratio; and nucleolar hypertrophy. Double mutants of as2 with mutations of genes that encode other candidate SSUP-related components such as nucleolin and putative rRNA methyltransferase exhibited similar synergistic defects caused by up-regulation of ETT/ARF3 and ARF4. These results suggest a tight link between putative SSUP and AS1-AS2 in repression of the abaxial-determining genes for cell fate decisions for adaxial development.


INTRODUCTION
A step of developmental events in multicellular organisms is characterized by the activation of a new genetic program and the inactivation of the previous program, which is mostly mediated by epigenetic systems. Recently, a complex of plant-specific nuclear proteins ASYMMETRIC LEAVES1 (AS1) and ASYMMETRIC LEAVES2 (AS2) (AS1-AS2) of Arabidopsis thaliana has been proposed to play a role in the epigenetic silencing involved in early leaf development from stem cells at the shoot apices (Phelps-Durr et al., 2005;Guo et al., 2008;Lodha et al., 2013;Iwasaki et al., 2013). Both nuclear proteins are localized as speckles, called AS2 bodies, to regions adjacent to and sometimes inside the nucleolus Luo et al., 2012a), suggesting a potential role of AS1 and AS2 in the gene silencing through nucleolus-associated events.
Leaf primordia appear around the shoot apical meristem (SAM: a group of stem cells at the shoot summit), and adaxial-abaxial (dorsal-ventral) leaf polarity is subsequently established by actions of various polarity-specific genes during the early development. The abaxialization (ventralization) is thought to proceed to adaxialization (dorsalization) because abaxial-determining genes were expressed earlier than that of adaxial-determining genes (Sawa et al., 1999;Eshed et al., 1999Eshed et al., , 2001 and the defect of the adaxial domains results in the generation of the filamentous leaf that retains only abaxial traits (Waites and Hudson, 1995;Li et al., 2005;Ueno et al., 2007). These abaxial and adaxial genes have genetically antagonistic relationships with each other, as expression of the abaxial genes could repress that of the adaxial genes and vice versa (Bowman and Floyd, 2008;Nakata and Okada, 2013). For example, KANADI genes ( parts of abaxial genes) directly repress AS2 ( part of adaxial genes) transcription (Wu et al., 2008).
AS1-AS2 is regarded as a key regulator for the proper confinement of the stem cell fate and the adaxial development (Waites et al., 1998;Byrne et al., 2000;Semiarti et al., 2001;Iwakawa et al., 2002Iwakawa et al., , 2007Matsumura et al., 2009). Direct repression by AS1-AS2 of the expression of at least two classes of genes is critical for leaf development, one of which is the class 1 KNOX homeobox genes such as BREVIPEDICELLUS (BP), KNAT2, and KNAT6 (Guo et al., 2008;Ikezaki et al., 2010). AS1-AS2 interacts physically with the plant homolog of the component of Polycomb-Repressive Complex 2 and recruits this complex to these KNOX genes, and is involved in maintenance of levels of trimethylation of histone H3Lys27 in the KNOX genes (Lodha et al., 2013). The other class of direct targets is the abaxial-determining gene ETTIN (ETT)/AUXIN RESPONSE FACTOR3 (ARF3) , which is directly repressed by the binding of AS1-AS2 to its promoter region. In addition, AS1-AS2 also indirectly represses expression of ETT/ARF3 and functionally redundant gene ARF4 through activating the miR390-tasiR-ARF pathway, which is essential for the adaxial development. AS1-AS2 is required for maintaining levels of CG methylation in the ETT/ARF3 coding region ; thus AS1-AS2 plays a key role in the antagonistic action by adaxial/abaxial factors and in epigenetic repression of ETT/ARF3.
Many factors are involved in leaf adaxialization in an AS1-AS2 dependent manner, in the as1 or as2 mutation background various modifier mutations have been identified that markedly enhance the defects of adaxial development to generate abaxialized filamentous leaves with loss of the adaxial domain. Causative mutations occur in genes that are involved in the biogenesis of small RNAs, chromatin modification and cell cycle progression (Williams et al., 2005;Machida et al., 2015). Mutations or disruptions of some ribosomal protein genes (Pinon et al., 2008;Yao et al., 2008;Horiguchi et al., 2011;Szakonyi and Byrne, 2011) and genes for ribosome biogenesis Huang et al., 2014) also act as enhancers of the as1 and as2 phenotypes. Collectively, many modifiers are weak alleles of the essential genes and strongly enhance the adaxial defects of as1 and as2. These findings suggest that AS1-AS2 and an enhancer protein act cooperatively on adaxial development through the epigenetic repression of ETT/ARF3 and ARF4 genes, however mechanisms of these cooperative actions are still unknown.
In the present study, we identified a new enhancer mutation in the Arabidopsis RH10 gene which potentially encodes an ortholog of yeast Rrp3 and human DDX47 that belong to the DEAD-box RNA helicase family. This helicase family has an indispensable role in gene regulation through RNA metabolism and many of them locate in the nucleolus. Rrp3 and DDX47 are parts of the nucleolar protein complex designated as the small subunit (SSU) processome that is a large ribonucleoprotein complex involved in 18S rRNA biogenesis, assembling the SSU of ribosomes (Dragon et al., 2002;Granneman et al., 2006;Phipps et al., 2011), that is ubiquitous in eukaryotes (Feng et al., 2013). Although DDX47 is necessary for maintaining the pluripotency of mouse stem cells (You et al., 2015), the subfamily of this helicase has yet to be intensively studied in relation to the development of multi-cellular organisms. Here, we report that RH10 represses cooperatively with AS1-AS2, the expression of abaxial-determining genes to promote the adaxial development of leaves. We also show that two other components that are related with the putative Arabidopsis SSU processome have similar functions in leaf development. Taken together with earlier results, the present results reveal a genetic link between AS1-AS2 and a protein complex such as the SSU processome for the polarity establishment of leaves.

RESULTS
A novel enhancer mutation of as2 that causes adaxial defects of leaves is present in the gene for DEAD-box-type RNA helicase RH10 We found the strong enhancer mutant in our library of mutagenized seeds of as2-1 Ishibashi et al., 2012), which showed filamentous and trumpet-like leaves and was named enhancer of asymmetric leaves two (east2). Molecular and genetic analyses showed that east2 was a recessive single mutation in the 9th exon in At5g60990, which is predicted to encode a protein belonging to the DEAD-box RNA helicase family called RNA HELICASE10 (RH10) (Fig. S1A-D) (Aubourg et al., 1999). In east2, the cytosine residue at nucleotide position 821 was converted to a thymine residue to replace the threonine residue at position 273, near motif IV, with an isoleucine residue (T273I). There are no RH10 knockout line and point mutants in public collections. In Arabidopsis, the amino acid sequence of RH10 showed the highest identity with that of RH36 (42.4%). RH10 had high amino acid identities with rice RH10 (63.0%), human DDX47 (60.5%), mouse DDX47 (59.8%), and budding yeast Rrp3 (50.1%) (Fig. S1D). Transcripts of RH10 were detected in all tissues and organs examined (Fig. S1E). The phenotype of east2 (see below) was complemented by the recombinant RH10 cDNA (Fig. S1F). Accordingly, this mutation was designated as rh10-1.
The rh10-1 phenotype was indistinguishable from that of wild type at 16°C (Fig. 1A), however it generated a weak phenotype of pointed leaves at 22°C which became narrower at 26°C (Fig. 1A), suggestive of a temperature-sensitive mutation in rh10-1. The level of the rh10 transcript in the mutant was similar to that of the RH10 transcript in Col-0 (Fig. S1E), which suggests that the temperaturesensitive mutation might be due to decrease in its enzymatic activity and/or protein amounts.
Although the as2-1 rh10-1 double mutant at 16°C produced leaves that were indistinguishable from those of the as2-1 single mutant (Fig. 1A), at 22°C approximately 50% of the leaves produced were short, filamentous, and trumpet-like (Fig. 1A,D). The defectiveness of as2-1 rh10-1 leaves was enhanced at 26°C as 90% of their rosette leaves were filamentous (Fig. 1A,D). Shapes of epidermal cells of the filamentous leaves were simple and rectangular, similar to those of a petiole, but different from those of flat leaves of wild-type plants (Fig. 1C). These results suggest that the adaxial-abaxial leaf polarity was more severely disrupted in as2-1 rh10-1 at 26°C, and similar abnormalities of leaf phenotypes were observed in as1-1 rh10-1 (Fig. 1B).
We observed the vascular patterns and GFP signals due to FILp: GFP (FIL promoter-fused gene for green fluorescent protein, Watanabe and Okada, 2003) as an abaxial marker. In the leaves of either single mutant grown at 26°C, the xylems were located on the adaxial sides and the phloems were on the abaxial sides, similar to those observed in the wild type ( Fig. 2A). The filamentous leaves of as2-1 rh10-1 at 26°C showed primitive or no vascular tissue without apparent xylem cells inside the bundle sheath ( Fig. 2A), suggesting defects in differentiation of xylem cells on the adaxial side. As shown in Fig. 2B, the GFP signals from FILp:GFP were detected on the abaxial side of the leaves of wild type, as2-1, and rh10-1 plants, and on the surface of the filamentous leaves and the entire young primordia of as2-1 rh10-1. These results show that as2-1 rh10-1 was defective in development of the adaxial domain of leaves, which generated abaxialized leaves. Apparently, there is a tight genetic interaction between the AS2-AS1 and RH10 for establishing leaf polarity.
RH10 is localized to the nucleolus and involved in processing pre-rRNAs We constructed the functional GFP:RH10 chimeric gene (Fig. S1F), and examined the sub-cellular localization of signals due to GFP in Arabidopsis. We used the YFP:HDT1 chimeric protein as a nucleolus marker . The fluorescent signal of GFP:RH10 was colocalized with that of YFP-HDT1 in the nucleolus (Fig. 3A), suggesting the nucleolar location of RH10.
We examined the effect of the rh10 mutation on levels of precursors of ribosomal RNAs ( pre-rRNAs) by northern blot with RNA isolated from shoot apices of plants grown at 22°C, 26°C, and 28°C, and a probe specific for the internal transcribed spacer1 (ITS1) sequence (colored red in Fig. 3B). As depicted in Fig. 3B, in Arabidopsis (Zakrzewska-Placzek et al., 2010), the 35S pre-rRNA corresponding to the primary transcript is rapidly cleaved at a specific site, named the P site (Saéz-Vasquez et al., 2004), in the 5′-external transcribed spacer (5′ETS) to generate 33S pre-rRNA, which is cleaved at either A2 or A3 sites in ITS1 to generate P-A2, P-A3 and 27S A2/A3 pre-rRNAs. Here we designated the pre-rRNA as P-A2/A3 because it was difficult to discriminate P-A2 from P-A3 by the experiment in the present study. These precursors undergo several processing steps to eventually generate mature 18S, 5.8S, and 25S rRNAs. The 35S and/or 33S (35/33S), P-A2/A3 and 27S A2/A3 pre-rRNAs could be detected with the probe.
As shown in Fig. 3C, three bands (denoted by b1, b2, and b3) that corresponded to these pre-rRNA molecules were detected. These precursors were detected as faint bands in wild-type (Col-0) and as2-1 plants grown at all the temperatures. The signals of b1 corresponding to 35/33S pre-rRNAs potentially containing the primary transcript and the cleavage product at the P site, and those of b2 corresponding to P-A2/A3, were higher in the rh10-1 and as2-1 rh10-1 than those in wild-type and as2-1 plants even at 22°C. Signal intensities of b1 and b2 in rh10-1 and as2-1 rh10-1 were markedly increased at 26°C and 28°C. Since b1 and b2 signals were also detected with the probe between the P site and the A1 site (data not shown), b1 and b2 bands contain the 5′ETS region, and b2 corresponds to the P-A2/A3 pre-rRNA, which is similar in terms of size to the 23S/22S pre-rRNA of yeast (O'Day et al., 1996). Signals of b3 corresponding to the 27S A2/A3 pre-rRNA were slightly increased at higher temperatures. These data showed that at least 33S and P-A2/A3 pre-rRNAs were significantly accumulated in rh10-1 and as2 rh10-1 at high temperatures.
These results suggest that the rh10 mutation mainly prevents 35S/ 33S and P-A2/A3 pre-rRNAs from being processed or degraded, RH10 might be involved in early stages of the processing reactions of 35/33S and P-A2/A3 pre-rRNAs. Although the accumulation of the 35S/33S pre-rRNA might be explained by up-regulation of rDNA transcription, the reduction of the ratio of 18S rRNA/25S rRNA as mentioned in the next section supports the processing impediment in rh10-1. These results are consistent with those obtained by experiments with yeast Rrp3 and human DDX47 showing that they are involved at least in cleavages at the A1 and/or A2 sites during early stages of processing (O'Day et al., 1996;Granneman et al., 2006;Sekiguchi et al., 2006).
The ratio of 18S rRNA/25S rRNA was reduced and nucleoli were enlarged in the rh10-1 mutant at high temperatures We examined by quantitative one-step RT-PCR alteration of the ratios of 18S to 25S rRNAs in wild-type and rh10-1 plants grown at various temperatures. The ratio was reduced in rh10-1 under higher temperature conditions ( Fig. 3D: 79% at 26°C and 46% at 28°C as compared to each value in the wild type). Considering the roughly constant levels of the 25S rRNAs in gel (Fig. 3C), the imbalanced ratio of 18S rRNA/25S rRNA in rh10-1 at higher temperatures might have been caused by the reduction of 18S rRNA.
We examined size differences of nucleoli in rh10-1 mutant cells. To this end we observed fluorescent signals from NUC1-fused GFP protein, as a nucleolar marker, in the young leaves of wild-type and rh10-1 plants grown at 22°C and 26°C. Fluorescence signals due to the NUC1-GFP were expanded in the rh10-1 mutant grown at 26°C as compared with those in Col-0 and rh10-1 plants grown at 22°C ( Fig. 3E), which suggests that the mutation in RH10 causes increases in the size of nucleoli. Two genes DOMINO and AtREN1, mutations of which affect the nucleolar size, have been reported (Lahmy et al., 2004;Reň ák et al., 2014); however, levels of transcripts of these genes were not affected by rh10-1 (data not shown).
Transcript levels of abaxial genes and class 1 KNOX genes were elevated in as2 rh10 at high temperatures leading to the adaxial defect We conducted microarray analysis (with the Affymetrix GeneChip) of mRNAs from shoot apices of Col-0, as2-1, rh10, and as2-1 rh10 grown at 26°C to obtain gene expression profiles in the tissues (Dataset-RH10-26). We performed cluster analysis of Dataset-RH10-26 by the knowledge-based fuzzy adaptive resonance theory (KB-FuzzyART) using Gene-list-4 (Takahashi et al., 2008 (Table S1). Constructed clusters are shown in Table S2 and Fig. S2A.
We confirmed the transcript levels of these genes in the shoot apices of plants grown at 26°C by quantitative real-time RT-PCR ( Fig. 4A; Table S6). The accumulated transcript levels of the genes involved in the abaxialization, ETT/ARF3, ARF4, KAN1, KAN2, and YAB5, were increased by two-to eight-fold in as2-1 rh10-1 as compared with those in wild-type plants (Col-0) (Fig. 4A). ETT/ ARF3 and ARF4 transcript levels were increased by two-fold in each of rh10-1 and as2-1 single mutants compared with those levels in wild-type plants. Transcript levels of the class 1 KNOX genes BP, KNAT2, KNAT6, and STM were markedly increased (by five-to eight-fold) in as2-1 rh10-1, compared with two-to four-fold increases in each single mutant. Transcript levels of adaxial genes HD-ZIP III (PHB, PHV, and REV) in the mutants did not significantly differ from those in the wild-type plants. These results showed that the as2-1 rh10-1 double mutation causes a synergistic enhancement of transcript levels of the abaxial genes and class 1 KNOX genes in shoot apices at 26°C.
We also examined transcript levels of genes for nucleolar proteins such as two NUCLEOLINs, AtFIBRILLARIN, TITAN, DOMINO and AtREN1 in as2-1 rh10-1, but synergistic enhancement of transcript levels was not observed (data not shown).
Mutations of NUC1 gene encoding nucleolin involved in pre-rRNA processing affect leaf polarity establishment in as1 and as2 In the previous section we showed that RH10 is a homolog of human DDX47 and yeast Rrp3 ( Fig. S1D; Fig. 3C); Rrp3 and DDX47 are reportedly parts of the SSU processome (Granneman et al., 2006;Phipps et al., 2011;You et al., 2015). Using information reported by Phipps et al. (2011), we conducted in silico analyses based on the similarity of protein sequences to detect orthologous proteins of the putative SSU processome in Arabidopsis. As shown in Table 1 (and Table S7), 74 protein sequences of Arabidopsis, corresponding to 54, including RH10 and two nucleolins (NUC1 and NUC2), out of the 72 proteins in yeast, were identified by these analyses. These results predict the presence of a SSU processomelike complex in Arabidopsis.
The prediction allowed us to examine the mutation in NUC1 for their involvement in leaf polarity determination since nucleolin is a major protein in the nucleolus that is involved in the processing of pre-rRNAs at various stages and the single mutant exhibit a pointed narrow shape of leaves, which is observed often in other modifier mutations including rh10 (Fig. 5) (Pontvianne et al., 2007;Kojima et al., 2007;Petricka and Nelson, 2007). In addition, human nucleolin is reported to be an assembly intermediate of the SSU processome and its candidate components (Turner et al., 2009;Phipps et al., 2011).
Mutation of the RID2 gene encoding nucleolar rRNA methyltransferase involved in pre-rRNA processing affect leaf polarity establishment in as2 We also examined the effect of a mutation in the ROOT-INITIATION-DEFECTIVE2 (RID2) gene (Ohbayashi et al., 2011) on the adaxial-abaxial organization of leaves on the as2 background. RID2 encodes an orthologous protein of yeast Bud23, which is nucleolus-localized rRNA methyltransferase that exhibits tight functional and physical interactions with some of the SSU processome components (Sardana et al., 2013(Sardana et al., , 2014Zhu et al., 2016). RID2 encodes a nucleolar protein that is involved in the processing of pre-rRNAs at various stages from early steps, and its single mutant also shows a typical pointed shape of leaves like other modifier mutants (Ohbayashi et al., 2011).
We observed GFP signals due to FILp:GFP in the filamentous leaves generated by as2-1 rid2-1. The GFP signals were detected in the epidermis of the filamentous leaves (Fig. 6B); thus they were surrounded by abaxialized epidermis, suggesting that adaxial development was blocked during leaf development of as2-1 rid2-1.

DISCUSSION
Cooperative repression by AS1-AS2 and the nucleolar proteins As we have reported here, the Arabidopsis rh10-1 mutation is a novel modifier that markedly enhances the adaxial defect of the as2 mutant to generate abaxialized filamentous leaves with the loss of the adaxial domain (Figs 1 and 2). Further molecular and genetic analyses of the as2 rh10-1 double mutant have shown that the severe adaxial defect is attributable to synergistically increased transcript levels of abaxial genes ETT/ARF3 and ARF4 in the double mutant (Fig. 4). These results suggest the involvement of a cooperative action of nucleus-localized protein complex AS1-AS2 and nucleolus-localized protein RH10 in repression of ETT/ARF3 and ARF4 in the presumptive adaxial domain of leaf primordia, which Bars indicate the s.d. among more than three biological replicates. White-bluecoloured graphs: abaxial genes; white-pink-coloured graphs: adaxial genes; white-brown-coloured graphs: class 1 KNOX genes. Significant differences from wild type were evaluated by Student's t-test and are represented by asterisks (*P<0.01). (B) Suppression of leaf phenotypes of as2-1 rh10-1 by ett-13 arf4-1. Wild-type and mutant plants with indicated mutations were grown on the soil at 26°C for 23 days. Magnified images of shoot apices of as2-1 rh10-1 and as2-1 rh10-1 ett-13 arf4-1 in the bottom panel were taken as different photographs of the corresponding plants in the middle panel. Arrowheads indicate filamentous leaves. Scale bars=10 mm. (C) Number of plants that generated filamentous leaves or did not. Wild-type and mutant plants with indicated mutations were grown on the soil at 26°C for 23 days.  (Fig. 7). A similar repression mechanism also operates on the repression of all four class 1 KNOX genes during development of leaf primordia ( Fig. 4A; Fig. 7) which are involved in proximal-distal growth, and formation of the prominent mid vein and stem cell maintenance in leaf primordia (Semiarti et al., 2001;Guo et al., 2008;Ikezaki et al., 2010). Since 54 out of 72 yeast components of the SSU processome are conserved in Arabidopsis (Table 1), we predicted the presence of an SSU processome-like complex, including RH10, in the plant. Taken together with results of experiments with nuc1 and rid2 mutants, the present results also predict that there might be a molecular link between the nucleolar complex and the AS1-AS2 complex that might cooperatively repress the abaxial genes and class 1 KNOX genes during leaf development. In addition, the observation that AS1 and AS2 are colocalized as a few speckles called AS2 bodies to regions adjacent to the nucleolus Luo et al., 2012a) implies a substantial link between these complexes. To approach the mechanism behind the link, molecular investigations should be further required with mutations of genes encoding core components of the SSU processome in Table 1.
It cannot be ruled out that the abnormal phenotypes observed in the present study might have also been due to a combinatorial effect of impairments unrelated to rh10-1 that might be induced at high temperatures.

Involvement of other nucleolar factors in the cooperative repression
With respect to nucleolar factors and leaf polarity establishment, we should mention that two genes encoding nucleolar proteins are also involved in rRNA biogenesis and leaf polarity establishment. Knockdown of the Arabidopsis genes HDT1 and HDT2 for nucleolar histone deacetylases (HDACs) enhances the leaf adaxial defects of as1 and as2 to generate the severely abaxialized filamentous leaves as seen in rh10-1 . These HDACs are responsible for establishment of nucleolar dominance, which is an epigenetic silencing of a set of rDNAs in genetic hybrids of Arabidopsis and is correlated with heterochromatic states of rDNAs in nucleolus organizing regions (NORs) (Pontes et al., 2007). These observations suggest that the proper silencing status of rDNAs might be critical for repression of the abaxial genes in as1 and as2 backgrounds. In addition, NUC2, another plant nucleolin (see Table 1), might be related to leaf phenotypes of as2 rh10-1, because NUC2, a negative regulator of CG methylation in rDNAs (Durut et al., 2014), is up-regulated in nuc1.
Arabidopsis nucleolar protein APUM23 with its pumilio-like RNA-binding repeats is involved in the processing of 35S pre-rRNA (Abbasi et al., 2010). The double mutants apum23 as1 and apum23 as2 produce filamentous leaves (Huang et al., 2014), suggesting involvement of APUM23 in leaf development.
It is worth noting that Arabidopsis double mutants of the ribosomal protein gene RPS6 with as2 exhibit the strongest adaxial defects of leaves among as2 mutants combined with other mutations of ribosomal protein genes (Horiguchi et al., 2011). Rps6 of budding yeast is one of five ribosomal proteins (Rps4, Rps6, Rps7, Rps9, and Rps14) that were identified as bona fide components of the processome (Bernstein et al., 2004). The remaining four out of these five genes have not yet been tested for a role in leaf development. In addition, five genes for ribosomal proteins in the small subunit have been examined and three of these five exhibit adaxial defects in the as2 background (Horiguchi et al., 2011). Seven out of ten genes for ribosomal proteins in the large subunit also exhibit the adaxial defect in the as1 and the as2 background (Pinon et al., 2008;Yao et al., 2008;Horiguchi et al., 2011;Szakonyi and Byrne, 2011). It would also be intriguing to investigate relationships between the adaxial defects and ribosomal protein genes for nucleolar complexes such as a large subunit (LSU) processome (McCann et al., 2015) in the nucleolus of Arabidopsis. It is also interesting to study relationships between AS1-AS2 and recently identified nucleolar protein AtNUFIP (Rodor et al., 2011).
How do AS1-AS2 and RH10 cooperatively repress the downstream genes?
Although our results predict that AS1-AS2 and nucleolar proteins in the putative SSU processome-like complex might cooperatively repress the abaxial genes and class 1 KNOX genes, molecular mechanisms behind the cooperative action remain unelucidated. Presently, there are two classes of hypotheses to explain this cooperative repression.

Class 1
In general, sufficient amounts of ribosomes might provide a favourable situation to control expression of a certain group of genes involved in organ development and cell proliferation. In the wildtype plant, the cooperative repression as reported here might be explained by assuming a combinatorial effect of global enhancement of translational capacity and spatio-temporal repressive regulation of the target genes such as ETT/ARF3, ARF4 and class 1 KNOX genes by AS1-AS2. In the rh10 mutant, the ratio of 18SrRNA/25SrRNA is reduced (Fig. 3D) implying a decrease in ribosome amount which could cause limit the general translational ability, and which might lead to up-regulation of these target genes in the absence of AS1-AS2; however according to this model, a question still remains concerning how the maintenance of Components of the SSU processome of yeast in A. thaliana were identified by bootstrapping phylogenetic tree analyses from 1157 amino acid sequences, as shown in Table S7. NE indicates 'Not extracted' by BLASTp (E<10 −10 ).
translational capacity might be involved in repression of transcript levels of these developmentally critical genes. Specific ribosomes are involved more or less efficiently in translations of specific transcripts (Nishimura et al., 2005;Wang et al., 2013;Xue and Barna, 2012). It is, however, less likely that RH10 might affect biogenesis of specific ribosomes.

Class 2
Recently it has been reported that the nucleolus and its peripheral regions offer a molecular architecture such as nucleolar-associated domains (NADs) (van Koningsbruggen et al., 2010;Németh et al., 2010;Padeken and Heun, 2014) that function in the regulation of gene transcription. Most rDNA loci are condensed as heterochromatin in the periphery of the nucleolus, and in the case of yeast and animal chromosomes specific sequences might also locate around the nucleolar surface, which might form distinct nuclear subcompartments (Thiry and Lafontaine, 2005;Neumüller et al., 2013). In Arabidopsis, DNA methyltransferase (MET1), nucleolar histone deacetylase (HDA6) and chromatin assembly (CAF1) genes are involved in formation of such a subcompartment White arrowheads indicate filamentous leaves. Magnified views of the shoot apices in as2-1 nuc1-1, as1-1 nuc1-1, and as1-1 parl1-2 plants, which correspond to the white-dotted boxed regions of each respective panel, are shown. (B) Quantitative analysis of leaf polarity in the wild-type and mutant plants. Rosette leaves classified into filamentous and expanded leaves were counted on each plant at 22 days. Plants (n=38-106) were used for each count. Frequencies of filamentous leaf were determined as described in Fig. 1D. (C) Magnified view of the shoot apex of an as1-1 nuc1-1 plant observed by SEM at 28 days. (D,E) Expression patterns of FILp:GFP in leaf primordia of Col-0 and respective mutant plants. Fluorescence signals are shown in a transverse section of leaf primordia of indicated plants, and of a filamentous leaf of as1-1 nuc1-1 grown on the soil at 22°C for 25 days. The view of filamentous leaves in a 28-day-old as1-1 nuc1-1 plant is shown (the bottom row, right). Arrows indicate GFP signals in abaxial epidermis; the arrowhead indicates GFP signals over entire epidermis. Green, GFP signals; red, autofluorescence. Scale bars=5 mm in (A); 500 μm in (C); 200 mm in (D); 100 mm in (E). (Pontvianne et al., 2013). AS2 and AS1, which is associated with HDA6 (Luo et al., 2012b), are colocalized to AS2 bodies adjacent to the nucleolus Luo et al., 2012a). Mutant proteins of AS2 that do not form AS2 bodies were not functional [L. Luo (Nagoya University, Nagoya, Japan), C.M. and Y.Mac., unpublished data]. If a plant has nuclear and nucleolar structural organizations similar to those of yeast and animals, the cooperative repression might be explained by assuming a certain physical interaction between these nucleus-and nucleolus-associated subcompartments, AS2 bodies, and the SSU processome-like complexes. Mutations in the RH10, NUC1, and RID2 might affect integrity of nuclear subcompartments, which might alter transcriptional patterns of genes locating in and around the subcompartments. If target genes of AS1-AS2 are associated with the subcompartment, proper structural networks between the subcompartments, the nucleolar complex and the AS1-AS2 repressor might be required for cooperative repression of the target genes.
The nucleolus can also play a role in sequestration of a certain protein in the 'detention centre'. This mechanism might affect action of proteins ( positively or negatively) that are implicated in transcriptional and post transcriptional controls (Jacob et al., 2013). The sequestration is driven by long non-coding RNAs in the nucleolus, where some of these RNAs are transcribed and/or processed. It is intriguing to investigate a molecular link between the cooperative epigenetic repression by AS1-AS2 and these nucleolar events.

Fluorescence microscopy
We crossed as2-1 rh10-1 and several fluorescence lines, FILp:GEP (Watanabe and Okada, 2003), YFP:HDT1 , and 35S: Nucleolin:GFP . We introduced FILp:GFP into rh10-1, as2-1 rh10-1, nuc1-1, as2-1 nuc1-1, rid2-1, and as2-1 rid2-1 mutants by genetic cross. Fluorescence in shoot apices was observed as described by Ishibashi et al. (2012). Rosette leaves classified into filamentous, trumpet-like, and expanded leaves on each plant grown at 22°C and 26°C were counted at 22 days. Plants (n=38-106) were used for each count. Frequencies of trumpet-like or filamentous leaf were determined as described in Fig. 1D. Fig. 7. Roles of AS1-AS2, RH10, NUC1, and RID2 in leaf development in A. thaliana. The AS1-AS2 complex and RH10 (or NUC1, RID2) act cooperatively to repress the levels of transcripts of leaf abaxial-determinant genes, which include ETT/ARF3 and ARF4, and of 'meristem' genes, namely class 1 KNOX genes. Repression of ARFs is crucial for the establishment of adaxial-abaxial polarity and then formation of flat and symmetric leaves. Class 1 KNOX genes are similarly repressed by AS1-AS2 together with genes for the nucleolar proteins, which is critical for proximal-distal axis growth of leaves and stem cell maintenance. See details in Discussion.

Histological analysis and observation of plant morphology
Plant tissues were prepared for thin sectioning as described by Iwakawa et al. (2007). Plant morphology was observed with a Scanning Electron Microscope XL30 (Philips Electron optics, Hillsboro, USA), and a Zeiss Stemi SV11 Apo Microscope Stereo (Carl Zeiss Inc., Oberkochen, Germany).

Construction of RH10-GFP
A cDNA fragment of the RH10 gene coding region was cloned into the binary vector pGWB6 containing the coding sequence for sGFP by using Gateway™ cloning technology (Invitrogen, Inc.) to produce binary vectorcarrying genes encoding the RH10-GFP fusion protein under control of the CaMV35S promoter (RH10-GFP). RH10-GFP was transformed into plants by Agrobacterium-mediated transformation (Bechtold and Pelletier, 1998). Primers for this construction are listed in Table S6.

Real-time RT-PCR
Leaves and shoot apices of mutant and wild-type plants were harvested at 15-days after vernalization, immediately frozen in liquid nitrogen, and stored at −80°C. Manipulations of RNA isolation and RT-PCR were performed as described by Iwakawa et al. (2007). Primer sets are listed in Table S6. The mean and standard deviation were calculated with values of more than three experimental replicates. Results were normalized by reference to the results for ELONGATION FACTOR1a, TUBULIN6 and ACTIN2 transcripts.

Analysis of rRNA and the precursors
Preparation of total RNAs from 15-day-old plants and hybridization with the digoxigenin-labeled probe specific for ITS1 were performed as described by Ohbayashi et al. (2011). The probe was synthesized with primers ITS1 F and ITS1 R (Table S6). Gels were stained with GelRed™ (Biotium Inc., Hayward, CA, USA).

Analysis of relative values of the 18S/25S rRNA ratio
Total RNAs were prepared from shoots excluding cotyledons of 15-day-old Col-0 and rh10-1 plants. Levels of 18S and 25S rRNAs were measured by quantitative RT-PCR ( primer listed in Table S6), and the ratio of the 18S level to the 25S level in each plant was calculated. Relative values of the rh10-1 ratio to the Col-0 ratio were calculated.

Data processing
Raw data processing was performed by using Affymetrix Gene Chip Operating Software (GCOS) (Version 1.4.0.036). We used two sets of gene expression data, Dataset-RH10-26 and Dataset-RH10-22. Each set of array data comprised 22,746 plant genes ( probe sets). Initially, we calculated the expression signals for all strains and the log2(ratio) for three strains (as2-1, rh10-1 and as2-1 rh10-1 grown at 26°C) and three strains (as2-1, rh10-1 and as2-1 rh10-1 grown at 22°C) against Col-0 by GCOS. In this experiment, we excluded 64 controls and 2,177 genes subject to cross-hybridization, according to NetAffx Annotation (www.affymetrix.com). Furthermore, for each group set, we excluded those genes for which all strain data sets showed an absent call (i.e., a detection call determined by GCOS based on the Pvalue of the one-sided Wilcoxon 5 signed-rank test; an absent call means P≥0.065, which is the default threshold in GCOS), because it indicates that the expression signal was undetectable. We also excluded those genes for which all strain data sets showed a no change call [i.e., no change call means (1-0.006)≥P≥0.006 for one-sided Wilcoxon signed-rank test], because no change indicates that the expression signal is almost equal to that of Col-0. Thus, 4759 and 5623 genes were selected for Dataset-RH10-26 and Dataset-RH10-22, respectively. Among these data, the log2(ratio) values that were >two-fold or <0.5-fold were rounded to two-fold and 0.5-fold, respectively; the log2(ratio) values with no change were rounded to 0, to avoid category proliferation in clustering.

Gene-list-4 for Arabidopsis thaliana
In the present study, we used KB-FuzzyART. This is a powerful algorithm for clustering of biological data (Takahashi et al., 2008; however, KB-FuzzyART is a knowledge-based system that requires some knowledge data. Therefore, we constructed Gene-list-1 for A. thaliana genes in a previous study (Takahashi et al., 2008). We added 61 genes (e.g., chromatinrelated genes, ACS family, LSH family and PIN genes) to update Gene-list-2 , and further added 73 genes and renamed it Gene-list-4, which comprises 486 genes. Further detailed information on Gene-list-4 is shown in Table S1.

Clustering by KB-FuzzyART
To know the effect of the rh10-1 mutation, we applied the knowledge-based fuzzy adaptive resonance theory (KB-FuzzyART) (Takahashi et al., 2008 to the clustering of the gene expression profiles that were obtained by microarray analysis (with the Affymetrix GeneChip) of mRNAs from shoot apices of Col-0, as2-1, rh10-1 and as2-1 rh10-1 grown at 26°C for 15 days (Dataset RH10-26). The 486 genes in Gene-list-4 (Table S1) were filtered and 332 genes were excluded that have undetectable calls (Abs.), no-change calls (N.C.), or no probes on the ATH1 chip against the wild type in as2, rh10-1 or as2-1 rh10-1 plants. The expression data sets of the 154 filtered genes were processed by KB-FuzzyART and classified into 25 clusters. Then, the 22,746 probes on the GeneChip were filtered and 17,052 genes were excluded. The remaining 5694 genes were assigned either as members of each cluster or as six outliers. The patterns of gene expression levels in the outliers did not match those in any clusters. The outliers comprised 229 of the 5694 genes. Profiles of the constructed clusters and gene assignments are shown in Tables S2 and S3. We sought for genes whose levels of expression were increased in the as2-1 rh10-1 double mutant when compared with those in the wild type grown at 26°C, as shown in Table S2. Expression levels of genes in Clusters h1 were upregulated in the as2 rh10-1 double mutant. Clusters h1 included KNAT2, KNAT6, STM, KAN2, KRP2, KRP5 and IPT3, which have also been identified downstream of modifier genes, such as EAL1 and ELO3 . Clusters h14 included BP/ KNAT1 and Clusters h17 included ETT/ARF3 and YAB5 (Table S3, Fig. S2A).
We also applied KB-FuzzyART to the clustering of gene expression profiles that were obtained by microarray analysis (with the Affymetrix GeneChip) of mRNAs from Col-0, as2-1, rh10-1 and as2-1 rh10-1 grown at 22°C for 15 days (Dataset-RH10-22). The expression data sets of the 129 filtered genes were processed by KB-FuzzyART and classified into 19 clusters. Then, the 22,746 probes on the GeneChip were filtered and 17,709 genes were excluded. The remaining 5037 genes were assigned either as members of each cluster or as outliers (Table S4). The outliers comprised 184 out of 5037 genes. The clusters that have patterns of increased transcript levels in the as2-1 rh10-1 double mutant did not include abaxial determinant genes and class 1 KNOX genes, except for YAB5, as shown in Table S5.
In silico homology analyses to identify Arabidopsis orthologs of proteins in the SSU processome We obtained amino acid sequences of 72 yeast proteins and 66 human proteins in the SSU processomes from NCBI RefSeq (www.ncbi.nlm.nih. gov/refseq), using gene names and accession numbers of the proteins listed by Phipps et al. (2011). Arabidopsis proteins that were homologous in terms of amino acid sequences to these yeast, and human proteins were obtained from Ensembl Release 83 (www.ensembl.org). Homology searches were conducted by BLAST Ver. 2.2.25 separately for these yeast and human sequences. Total 1157 proteins representing 482 genes of A. thaliana were extracted by BLASTp (E<e −10 ) from the Arabidopsis database commonly on the basis of these yeast and human sequences (Table S7). The bootstrapping phylogenetic tree analyses were conducted by ClustalW Ver. 2.1 with default parameters by using the website of the DNA Data Bank of Japan (DDBJ). For our phylogenetic tree analyses, we selected one by one the branch of the tree that was the closest to each protein in yeast and/or human SSU processomes to identify Arabidopsis orthologs.