RNA-binding proteins SOP-2 and SOR-1 form a novel PcG-like complex in C. elegans

Epigenetic gene silencing involves the formation of heritable repressive chromatin structures that are established in a stepwise manner, through initiation, histone modification and propagation along the chromatin fiber. Establishment of such chromatin structure appears to involve the concerted action of RNA and interacting proteins(Heard, 2004; Lippman and Martienssen,2004). The RNA components can be a non-coding structural RNA, such as Xist RNA, which functions in mammalian X chromosome inactivation(Pannings and Jaenisch, 1998),or a small interfering RNA (siRNA) derived from the RNAi machinery, which is employed in the silencing of genes within the heterochromatic domains in S. pombe and Drosophila(Volpe et al., 2002; Pal-Bhadra et al., 2004). These RNA components could be responsible for targeting the interacting protein complex to specific loci and/or in later maintaining the repressed state of the genes (Akhtar,2003; Lippman and Martienssen,2004).

The epigenetic silencing of Hox genes mediated by Polycomb group (PcG)proteins is also thought to involve the formation of localized repressive chromatin structures (Levine et al.,2004; Lund and van Lohuizen,2004). Hox genes encode homeodomain-containing transcription factors that specify the positional identities of cells along the anteroposterior axis during multicellular organism development(Gellon and McGinnis, 1998). In PcG mutants, Hox genes are ectopically expressed in regions outside their normal expression domains, resulting in homeotic transformations, in which the body structures are duplicated or lost. Two conserved PcG complexes have been identified, the ESC/E(Z) complex, containing Extra sex combs and Enhancer of zeste, and the PRC1 complex, the core components of which include Polycomb (PC), Posterior sex combs (PSC),Polyhomeotic (PH) and RING1 (Levine et al., 2004). The ESC/E(Z) complex has been shown to associate with histone deacetylases and also contains an intrinsic histone methyltransferase activity that specifically methylates H3 at lysine 9 (K9) and lysine 27 (K27)(van der Vlag and Otte, 1999; Muller et al., 2002). The PRC1 complex functions as an E3 ubiquitin ligase that is specific for H2A(de Napoles et al., 2004; Wang et al., 2004). Recent observations suggest that, like other forms of epigenetic gene silencing,PcG-mediated Hox gene repression might also involve RNA(Sun and Zhang, 2004). For example, the ESC/E(Z) complex and the PRC1 complex are required for X chromosome inactivation in mammals and they are recruited to the X chromosome in a Xist structural RNA-dependent manner, although their direct interaction has not been established (Silva et al.,2003; Plath et al.,2003; Plath et al.,2004).

PcG-mediated Hox gene repression is an ancient mechanism, conserved from flies to mammals. However, the nematode C. elegans apparently lacks PRC1 complex genes. Furthermore, although the ESC/E(Z) complex is found (e.g. mes-2 and mes-6), mutations in these genes result in only mild homeotic defects (Korf et al.,1998; Holdeman et al.,1998; Ross and Zarkower,2003). Nevertheless, C. elegans Hox genes are subject to global repression. This repression is mediated by a novel gene, sop-2. In sop-2 mutants, the onset of Hox gene expression is normal but is subsequently expressed ectopically in diverse body regions. Like the components of the PRC1 complex, SOP-2 forms nuclear bodies, called SOP-2 bodies (Zhang et al., 2003). The formation of SOP-2 bodies is tightly correlated with its function and its formation appears to require the RNA-binding activity of SOP-2(Zhang et al., 2003; Zhang et al., 2004a; Zhang et al., 2004b), although the nature of the RNA under physiological conditions remains to be elucidated(Zhang et al., 2004b).

Identification of additional components in the SOP-2 nuclear bodies will provide insight into the role of SOP-2 in Hox gene repression. Although SOP-2 bears little sequence similarity to PRC1 complex proteins, components of the PRC1 complex are also localized into distinct nuclear bodies, called PcG bodies, and also have RNA-binding activity(Netter et al., 2001; Zhang et al., 2004b). The study of the SOP-2/RNA complex is likely to shed light on the salient features of PcG-mediated gene repression. In this study, we describe the identification and characterization of a PcG-like gene in C. elegans,sor-1, that encodes a novel RNA-binding protein with limited regions of similarity to the mouse PcG protein Rae28. SOR-1 and SOP-2 colocalize in SOP-2 nuclear bodies and direct interact with each other. Our studies reveal that SOR-1 and SOP-2 define a putative PcG-like complex in global repression of Hox genes in C. elegans. Remarkably, neither SOR-1 nor SOP-2 is conserved in other organisms, not even in the congeneric species C. briggsae,suggesting a surprising lack of evolutionary constraint on an ancient regulatory system.

Most strains carrying the him-5(e1490) mutation, which produces a high frequency of males in self progenies. The following strains were used in this work.

• LGII: sop-2(bx91), muIs16(mab-5::gfp, dpy-20)

• LGIII: sor-1(bp1); sor-1(bp2); sor-1(bp3); mab-5(e1239) egl-5(n945)

• LGV: bxIs14(pkd-2::gfp, pha-1(+))

• LGX: bxIs13(egl-5::gfp, lin-15(+))

The bx91 mutation is temperature sensitive. At 15°C, animals have no obvious defects. At 25°C, animals arrest at L1 and L2 stages. Adult sop-2(bx91) animals were shifted from 15°C to 25°C and the arrested early larvae were examined for the expression pattern of Hox genes.

sor-1 mutations were isolated in screens for mutants with ectopic expression of mab-5::gfp and egl-5::gfp reporters. Approximately 17,000 haploid genomes were screened. bp1, bp2 and bp3 are located at the same genetic locus by genetic mapping and non-complementation experiments. Three factor mapping placed sor-1between unc-32(0.00) and sma-3(–0.93), about–0.23, on LGIII. Six out of eight Sma nonUnc and seven out of 28 Unc nonSma recombinants carried sor-1 mutation. PCR products from ZK1236.3 were injected into + sor-1(bp2) +/unc-32 + sma-3 together with transformation marker pRF4(rol-6(su1006)), and its ability of rescuing the early larval lethality of sor-1(bp2) was assessed. The heterozygotes transformants gave rise to adult sor-1(bp2) males and hermaphrodites, while 100% of sor-1(bp2) animals without the transgene arrested at L1 and L2 stages.

cDNA clones of sor-1, yk526e9 and yk336g5, were kindly provided by Dr Kohara (NIG, Japan) and were sequenced to confirm the predicted exon/intron junctions. The 5′ end of the sor-1 cDNA was obtained by RT-PCR using SL1 sense primer and an antisense primer that is specific to sor-1. SL1 is a trans-spliced leader that is present at the 5′end of many C. elegans cDNAs. The sor-1 mutations identified in our genetic screens were determined by sequencing the corresponding sequence of the sor-1 locus.

Single-stranded RNA (ssRNA) was transcribed from the T7 and T3-flanked PCR templates (ZK1236 nucleotides 6709-7498) with MEGAscript T3 and T7 kits(Ambion). The ssRNAs were then annealed, and injected into muIs16, bxIs13,bxIs14, mab-5 egl-5; bxIs14 and sop-2(bx91); bxIs13 animals. F1 progenies generated 4 hours after injection were scored for larval lethality,ectopic expression of Hox genes, or generation of anterior rays, as shown by the expression of pkd-2::gfp reporter.

The sor-1::gfp reporter was constructed by PCR fusion based approach (Hobert, 2002). The fused PCR products were derived from two overlapping PCR DNA fragments. One contained the DNA derived from ZK1236 (nucleotides 3272 to 9530), which includes a 3 kb promoter region and the entire ORF of sor-1. Another one contained the gfp and the unc-54 3′UTR from pPD95.67. The PCR products were co-injected with pBX-1(pha-1+) into pha-1 mutant worms and the transformants were analyzed.

The sor-1 cDNA corresponding to the N terminus of SOR-1 (amino acid 230 to 390) was cloned into the pET28 expression vector. This his-tagged fusion protein produced by E. coli BL21 was purified to be used as an immunogen in rabbits. The antisera were first absorbed with bacterial acetone powder, followed by NAB protein A spin purification kit (Pierce).

Embryos were obtained from well-fed adult hermaphrodites. The permeabilization of embryos and young larvae was performed by Freeze-Cracking methods (Albertson, 1984). The freeze-cracked slides were fixed, blocked and incubated with anti-SOR-1 antibody at a final dilution of 1:400 at room temperature for 2-4 hours. The worms were then washed three times and incubated with Rhodamine Red-X-conjugated goat anti-rabbit IgG. The specificity of SOR-1 antibody was demonstrated by lack of staining in control experiments with pre-immune sera,with antisera that were preincubated with 0.25 mg of purified SOR-1 fusion protein, or in sor-1(RNAi) embryos. sop-2(bx91) L2 larvae were shifted from 20°C to 25°C, and the embryos derived form these animals were used for immunostaining.

Constructs encoding GST-SOR-1 and GST-SOP-2 were made by subcloning portions of the cDNAs of SOR-1 and SOP-2 into pGEX-4T-1. Fusion proteins were expressed in BL21 cells and purified with glutathione sepharose resin according to the manufacturer's recommendation (Pharmacia). For in vitro systems of ³⁵S-labelled proteins, the corresponding sor-1and sop-2 were cloned into pcDNA3 as templates for transcription and translation (TNT Coupled Reticulocyte Lysate System, Promega). The GST fusion proteins (200 ng) were incubated with ³⁵S-labelled protein and 10μl glutathione sepharose beads in binding buffer [25 mM Tris.Cl (pH 7.6),150 mM NaCl, 1 mM DTT, 0.5% Triton X-100, 10% Glycerol, 1 mM PMSF] for 2 hours at 4°C. The reactions were then washed for four times with 1 ml binding buffer. Bound proteins were analyzed by SDS-PAGE and analyzed by autoradiograph.

For in vivo GST pull-down assay, SOP-2 (residues 58-140) and SOR-1 (48-200)were cloned into the vectors of pGEX-6P-1 and PET30a, respectively, and co-expressed in E. coli strain BL21(DE3). GST-4B resin was first used for purification of the complex. GST was removed by thrombin and then the digested complex was subjected to ion exchange column (Amersham-Pharmacia) for further purification.

The complex was purified to more than 95% homogeneity in assay buffer (10 mM Tris, pH 8.0, 3 mM DTT, 100 mM NaCl). The complex (1.0 ml, about 2.0 mg complex) was subjected to gel filtration analysis (Superdex200,Amersham-Pharmacia). Samples from the peak corresponding to the complex were visualized by SDS-PADE and stained with Coomassie Blue.

RNA-binding reactions contained 20 mM HEPES (pH 7.6), 100 mM KCl, 2 mM EDTA, 0.01% NP40, 1 mM DTT, labeled RNA fragment (20,000 cpm). Ten times cold irrelevant ssRNA, dsRNA and 100 times yeast tRNA (mass excess) were used for competition experiments. Reactions (25 μl) were incubated on ice for 20 minutes and electrophoresed on 4% native TBE PAGE gel and analyzed by autoradiograph.

To identify components involved in maintaining the integrity of expression domains of Hox genes in C. elegans, strains expressing the Hox gene reporter, egl-5::gfp or mab-5::gfp, were used to screen for mutants with expanded expression domains of egl-5 and mab-5,respectively. Three alleles of a new genetic locus, termed sor-1(sop-2 related-1) (bp1, bp2, bp3), were isolated from the total 17,000 mutagenized genomes. In early wild-type larvae, egl-5 expression is restricted to cells in the tail region(Fig. 1A)(Ferreira et al., 1999). In sor-1 mutants, egl-5 is ectopically expressed in head region, and sometimes in the anterior seam cells(Fig. 1B). Similarly, the expression domain of mab-5 is greatly expanded. During early wild-type larval development, expression of mab-5 is limited to the cells of posterior body region, including P9-P12, V5, and V6(Fig. 1C)(Kenyon et al., 1997). In sor-1 mutants, mab-5::gfp is ectopically expressed in the head, and in the syncytial hypodermal cell 7 (hpy7)(Fig. 1D). Taken together,these observations suggest that wild-type sor-1 is required for global repression of Hox genes.

The ectopic expression of Hox genes in sor-1 mutants could result either from ectopic Hox gene initiation or from later failure in maintaining the repressed state of Hox genes. To distinguish between these possibilities,we determined the onset of ectopic expression of egl-5::gfp and mab-5::gfp in sor-1 mutant embryos. As in the developing larvae, egl-5::gfp and mab-5::gfp reporters are expressed in a limited number of cells. Inappropriate expression domains of egl-5and mab-5 are first detected when embryogenesis reached the threefold stage, when most cell lineages are completed. At this time, egl-5::gfp begins to be expressed in the head region and mab-5::gfp is ectopically expressed in hyp7. Hence, as in the PcG mutants of other organisms, the expression of Hox genes in sor-1 mutants is correctly initiated but later becomes derepressed.

To determine whether the endogenous Hox genes are ectopically expressed in sor-1 mutants (like their reporter genes), we examined whether homeotic transformations, phenotypic readouts of mis-expression of Hox genes,occur in sor-1 mutants. We studied the development of a row of epidermal seam cells in males, in which the role of Hox genes mab-5and egl-5 in cell fate specification is well characterized(Emmons, 1999). During wild-type male development, the three most-posterior seam cells, V5, V6 and T,give rise to nine sensory rays. mab-5 is expressed in the lineages of V5 and V6, and is required for the generation of rays from these cells(Salser and Kenyon, 1996). egl-5 is expressed in the posterior branches of the V6 lineage where it specifies the identities of the V6 rays(Ferreira et al., 1999). The anterior seam cells, from V1 to V4, which do not express mab-5 and egl-5, generate three parallel longitudinal cuticular ridges, known as alae. Ectopic expression or loss of function of Hox genes mab-5and egl-5 results in homeotic transformation of the fates adopted by seam cells V1 to V6. The early larval lethality of sor-1 mutants identified in our screens prevents us from analyzing the role of sor-1 in male ray development, an event that occurs later in larval development. We therefore took advantage of sor-1(RNAi), which at the appropriate concentration causes weaker loss-of-function defects in the animals developed from the eggs laid in the first few hours after injection. sor-1 dsRNAs (gene identification see below) were injected into worms carrying a pkd-2::gfp reporter, which is specifically expressed in one of the two ray neurons and serves as a ray marker(Fig. 1E) (Barr and Sternberg,2000). We found that in sor-1(RNAi) males, pkd-2::gfp is ectopically expressed in the anterior body region instead of being confined to the posterior (89%, n=18) (Fig. 1F), indicating that ectopic rays are generated from anterior seam cells. By contrast, ectopic anterior rays are completely absent in sor-1(RNAi) mab-5 egl-5 males (n=13). Thus, loss of function of sor-1 causes anterior to posterior cell fate transformation in a Hox gene-dependant manner.

In addition to ectopic expression of Hox genes, sor-1 mutant animals display many other defects. All sor-1 mutants identified in our genetic screens arrest as homozygotes at L1 and L2 larval stages. In sor-1(RNAi) hermaphrodites, vulva defects are apparent, such as bursting vulva and protruding vulva (Fig. 1G). In addition, 20% of sor-1(RNAi) mutants(n=41) show partial hermaphrodite-to-male sexual transformation,revealed by inappropriate expression of pkd-2::gfp, a marker of male-specific neurons fates (Fig. 1H). sor-1(RNAi) hermaphrodites are also sterile because of defects in gonads and germline development. These defects cannot be readily attributed to misregulated expression of Hox genes and suggest that sor-1 may have other targets in addition to Hox genes.

The defects of sor-1 mutants are similar to those previously described for sop-2 mutants(Zhang et al., 2003). Both sor-1 and sop-2 are involved in global repression of Hox genes and in regulating the expression of non-homeotic genes. To assess the interactions between sop-2 and sor-1, we generated sop-2; sor-1 double mutants by performing sor-1(RNAi) in sop-2(bx91) animals. sor-1(RNAi) is likely to recapitulate a null or strong loss-of-function phenotype in the animals developed from the eggs laid in 4-48 hours after injection, as the same extent of ectopic expression of Hox genes was observed in sor-1 (RNAi) and sor-1(bp1) (in which about two-thirds of SOR-1 is deleted) mutant animals. Shifitng sop-2(bx91ts) to non-permissive temperature (25°C) at L4 stage appears to cause null defects in the next generation, as the same extent of defects were observed as those in sop-2(bp7) animals, a putative null allele of sop-2, in which a stop codon mutation occurs at position 80, leading to a deletion of the 655 C-terminal residues of SOP-2. We chose RNAi because both sop-2(bp7) and sor-1(bp1) homozygous animals arrest at L1 to L2 larval stages and can be maintained only as heterozygous. We examined the genetic interactions between sor-1 and sop-2 by comparing the detailed expression pattern of egl-5in sor-1(RNAi); sop-2(bx91) double mutants with that observed in sor-1(RNAi) and sop-2(bx91) single mutant animals. First, we determined the onset of ectopic expression of Hox genes. In sor-1 and sop-2 single mutants, Hox genes are not ectopically expressed until the developing embryos reach the threefold stage. By contrast, ectopic expression of egl-5 in the head region was observed earlier, at the`pretzel' stage in sop-2(bx91); sor-1(RNAi) embryos (n=15)(Fig. 2E,F), during which neither sor-1(RNAi) (n=6) nor sop-2(bx91)(n=3) shows ectopic expression of egl-5(Fig. 2A-D). Thus, sor-1 and sop-2 function synergistically in maintaining the repressed state of Hox genes in early embryogenesis.

Next, we examined the spatial limits of ectopic expression domains of Hox genes in sop-2; sor-1 animals. In sor-1 and sop-2single mutants, egl-5 is ectopically expressed in head neurons and in tail cells. The number of cells expressing the egl-5 reporter,however, is dramatically increased in sop-2(bx91); sor-1(RNAi) double mutants. In sop-2(bx91); sor-1(RNAi) mutants, the average number of cells expressing egl-5 is 47 in the head and 28 in the tail, compared with 13 in the head and eight in the tail in sor-1(RNAi) mutants, and with 32 in the head and 12 in the tail in sop-2 mutants. The increased number of egl-5-positive cells in sop-2;sor-1(RNAi) mutants is not an additive effect of sor-1 and sop-2, as the expression domains in the head and tail are dramatically expanded (Table 1; Fig. 2G-I). Moreover, in 89%(n=19) of the sop-2(bx91); sor-1(RNAi) animals, egl-5 is seen in the mid-body region, including the ventral cord and seam cells (Fig. 2J), whereas it is observed in only 13% (n=39) of sor-1 mutants and 24%(n=37) of sop-2 mutants. Therefore, sor-1 and sop-2 act synergistically in some body regions to repress Hox gene expression.

Mutations in sor-1 and sop-2 also cause synthetic larval lethality. Both sor-1(RNAi) and sop-2 mutants are arrested at early larval stages, mostly at the L1 and L2 stages. However, 98% of the sop-2(bx91); sor-1(RNAi) animals are arrested at the L1 stage(Table 1). Larval arrest is not a phenotype typically associated with misregulation of Hox genes in C. elegans, indicating that sor-1 and sop-2 also function together in regulating the expression of non-homeotic genes. In these experiments, sop-2(bx91) appears to be a null allele and sor-1(RNAi) is a strong loss-of-function, if not a null, allele. The similar defects caused by mutations in sop-2 and sor-1suggest that they may function redundantly in regulating some targets in common. Synergistic interactions between sop-2 and sor-1suggest that they also have independent functions in Hox gene repression. They could have distinct biochemical functions within a single complex. Alternatively, they could also function in separate complexes that are involved in Hox gene regulation.

The sor-1 locus was mapped by three-factor mapping between unc-32 and sma-3, a small genetic interval on chromosome III(Fig. 3A). It was cloned by transformation rescue experiments (see Materials and methods). The sor-1 rescuing DNA fragment contains a single predicted gene ZK1236.3. Most of the intron/exon boundaries were confirmed by sequencing the cDNAs of this region (Fig. 3A). We found that a 22 nucleotide leader sequence, the spliced leader 1 (SL1), is trans-spliced onto the 5′ end of sor-1 transcripts. SL1 trans-splicing occurs in about 57% of the genes in C. elegans(Blumenthal and Steward, 1997). SOR-1 encodes a protein with 810 amino acids(Fig. 3B). In sor-1(bp1) mutants, the CAG glutamine codon at position 304 is mutated to TAG amber stop codon, resulting in the truncation of 506 amino acids in the C terminus. In sor-1(bp2) and sor-1(bp3), an identical nonsense mutation occurs at position 634, leading to a deletion of the 176 C-terminal residues (Fig. 3B). Existence of these mutations in sor-1 mutant alleles further confirms the identification of sor-1. Treatment of wild-type animals with ZK1236.3 dsRNA phenocopies sor-1 mutants, causing the ectopic expression of mab-5 and egl-5 as described above. Therefore, the ectopic expression of Hox genes is due to the loss of function of sor-1.

RNA has been postulated to play an important role in PcG-mediated Hox gene repression (Sun and Zhang,2004). We therefore tested whether SOR-1 directly binds to RNA. GST-SOR-1 was purified and incubated with a radiolabeled single- or double-stranded RNA derived from the 5′UTR of the Hox gene egl-5, followed by electrophoretic mobility shift analysis (EMSA). The results showed that a fragment of SOR-1 binds efficiently to the RNA probes (Fig. 3C). Binding of SOR-1 to RNA is effectively competed by adding cold unrelated single- or double-stranded RNA, but not by adding tRNA or DNA(Fig. 3D). This rules out the possibility that SOR-1 is a general nucleic acid-binding protein or sticks to the charged phosphate backbone of nucleic acids non-specifically. RNA binding by SOR-1 in vitro does not appear to be sequence specific, as several unrelated RNA templates bind to SOR-1 with comparable efficiency (data not shown). However, this apparent lack of sequence specificity in vitro may not reflect the differential affinity for physiologically relevant RNA targets.

Using a series of nine overlapping SOR-1 fragments, the specific RNA-binding region of SOR-1 was mapped to an 87 amino acid region (amino acid 443 to 530). This region does not contain a recognizable RNA-binding motif,but is rich in proline and glutamine. A transgene deleted for the RNA binding region of SOR-1 cannot rescue the sor-1 mutant phenotypes, indicating the importance of this region for sor-1 function.

To gain further insight into the functional properties of SOR-1, we searched for putative domains in SOR-1. SOR-1 contains a region that shows weak similarity to the mouse PcG protein Rae28 (35% similarity over 381 amino acids from amino acids 115 to 496 of SOR-1) and a second region that has weak similarity to extensin 2 domain (amino acids 422 to 659 of SOR-1,6.1e^–01), which, notably, is also present in SOP-2 (amino acids 300 to 648, 5.9e^–01). The extensin motif exhibits a high degree of post-translational modification, including hydroxylation,glycosylation and crosslinking, suggesting that the function of SOR-1 and SOP-2 may be regulated by post-translational modification.

To examine the expression pattern of sor-1, we generated transgenic lines expressing sor-1::gfp. The reporter gene contains the entire coding sequence and the promoter region of sor-1, with gfp inserted in the C terminus of sor-1. Fluorescence can be observed in all nuclei in developing embryos. At larval stages, sor-1::gfp becomes weaker in all of the somatic cells. Interestingly,SOR-1::GFP forms distinct nuclear bodies besides its homogenous distribution in the nucleoplasm (Fig. 4A).

To further determine the expression pattern and level of endogenous SOR-1 protein, we raised rabbit polyclonal antibodies against the N-terminal 160 amino acids of SOR-1 (see Materials and methods). The SOR-1 antibody staining pattern was consistent with those shown by the gfp reporter. SOR-1 is nuclear localized and expressed in all cells from the one-cell stage onwards. As development proceeds, the expression level of SOR-1 appears to decline. From the two cell-stage onwards, SOR-1 is found to be inhomogenously expressed in the nuclei with obvious accumulations in distinct nuclear speckles(Fig. 4B). During larval development, SOR-1 is present in all somatic cell nuclei. Although at lower levels of expression, the nuclear bodies in which it localizes can still be formed.

The localization of SOR-1 in nuclear bodies lead us to investigate whether it is colocalized with SOP-2, a C. elegans PcG protein that is also localized in nuclear bodies, termed SOP-2 bodies(Zhang et al., 2003). Transgenic animals expressing a functional sop-2::gfp reporter were immunostained with anti-SOR-1 antibodies. The superimposable confocal images of SOR-1 (Fig. 4C) and SOP-2::GFP (Fig. 4D) in distinct nuclear bodies provide evidence that they are colocalized in SOP-2 nuclear bodies (Fig. 4E).

To determine the role of SOP-2 and SOR-1 in formation of the SOP-2 nuclear bodies, we examined the localization of these proteins in sop-2 and sor-1 mutants using anti-SOP-2 and anti-SOR-1 antibodies. Absence of SOR-1 does not affect the localization of SOP-2(Fig. 4F,G). No obvious changes in SOP-2 localization were observed in sor-1(RNAi) embryos(Fig. 4F). In sop-2(bx91) mutant embryos, both SOP-2(bx91) (46/46 cells) and SOR-1(45/45 cells) are localized in the cytoplasm(Fig. 4H-M), while the cytoplasmic localization of SOP-2 and SOR-1 is not evident in wild-type animals (>50 cells examined). In sop-2(bx91) mutant embryos, the nuclear localization of SOP-2(bx91) and SOR-1 is also seen in some cells(25/46 cells for SOP-2, and 30/45 cells for SOR-1). Thus, SOP-2 function is required for the proper localization of both SOP-2 and SOR-1.

Colocalization of SOR-1 and SOP-2 in SOP-2 bodies and dependence of the localization of SOR-1 on the function of SOP-2 prompted us to test whether these proteins directly interact. Glutathione-S-transferase (GST)-SOP-2 carried on glutathione sepharose beads was incubated with S³⁵labeled SOR-1. The retained proteins after several washes were examined by gel electrophoresis. We found that SOP-2 specifically binds to the N-terminal 503 amino acids of SOR-1 (Fig. 5A). And, vice versa, GST-SOR-1 interacts with the ³⁵S-labeled SOP-2(Fig. 5B). As SOP-2 contains an RNA-binding activity and reactions of in vitro translated proteins contain many RNAs, we further tested whether the interaction between SOR-1 and SOP-2 was RNA dependent. We found the interaction was not significantly reduced after the reaction was treated with RNase (data not shown), supporting the notion that SOP-2 and SOR-1 interact directly.

To map the minimal interacting domain in each protein, a series of GST fusion fragments of SOR-1 and SOP-2 were used in GST pull-down assay in vitro. The SOR-1 interacting domain in SOP-2 is mapped to its N terminus, rather than to its C-terminal protein-protein interacting SAM domain, which mediates the self-association of SOP-2 (Fig. 5A,C). However, the SOP-2 interacting domain in SOR-1 is located to its N terminus, which is predicted to be rich in helical structures(Fig. 5B,C).

Interaction between the N terminus of SOP-2 and SOR-1 was further demonstrated by the co-purification of SOP-2 and SOR-1 expressed in bacterial. GST-SOP-2 (amino acids 58-140) and non-tagged SOR-1 (amino acids 48-200) were co-expressed in E. coli strain BL21. SOP-2 proteins were purified by the GST affinity column followed by ion exchange and Superdex columns. The elute was subjected to SDS-PAGE and visualized by Coomassie staining. We found that SOR-1 was specifically co-purified with SOP-2(Fig. 5D).

In addition to a role in germline development(Fong et al., 2002), the MES-2/MES-6 complex is also involved in transcriptional repression of Hox genes (Ross and Zarkower,2003). To determine whether the MES-2/MES-6 complex and the putative SOP-2/SOR-1 complex function together in repression of Hox genes, we compared the expression pattern of Hox reporter genes in sop-2(bx91);mes-2/mes-6(RNAi) mutants with those in single mutants. As previously reported, egl-5::gfp is weakly ectopically expressed in mes-2/mes-6(RNAi) mutants (Table 1). However, no synergistic interactions are observed between sop-2(bx91) and mes-2/mes-6 (RNAi)(Table 1). The onset of ectopic expression and the expression domains of egl-5::gfp in sop-2(bx91); mes-2/mes-6(RNAi) is similar to those in sop-2single mutants. Lack of synergy between mutations in these two complexes and the weaker effects of mes-2 and mes-6 on Hox genes suggest that the MES-2/MES-6 complex is unlikely to be a prerequisite to recruit the SOP-2/SOR-1 complex to Hox genes.

However, sop-2(bx91) and a mes-6 mutation cause synthetic lethality and a bursting vulva phenotype. Shifting sop-2; mes-6 L1/L2 larvae from 15°C to the semi-permissive temperature 20°C, 69.3% of animals (n=322) died of bursting vulva, whereas none of the single mutants (n>100) has a bursting vulva phenotype. Shifting sop-2; mes-6 mutant eggs laid in the first 8 hours from 15°C to non-permissive temperature 23°C, 100% animals arrested at L1 stage, while sop-2(bx91) mutants alone arrested at L2 to L4 stages and all mes-6 mutant animals developed into adults. No such synergistic interactions have been observed between sop-2 and a mes-4mutation, consistent with the fact that MES-4 is not a component of the MES-2/MES-6 complex. These synergistic interactions suggest that SOP-2/SOR-1 and MES-2/MES-6 could act in a single pathway or their effects in separate pathways may be cumulative in regulating non-Hox genes.

SOR-1 and SOP-2 do not appear to be homologs of known PcG proteins in other organisms, although several properties of the SOR-1/SOP-2 complex, including localization into nuclear bodies and binding to RNA, are reminiscent of the ones of the PRC1 complex in other organisms. To determine whether SOR-1/SOP-2 defines a new global Hox gene repression system, we examined whether other organisms contain homologs of SOR-1 and SOP-2. Surprisingly, no recognizable SOR-1 or SOP-2 homologs were identified in fly, mouse and human by searching available databases.

To determine whether SOR-1 and SOP-2 are conserved among nematodes, we searched the sequence of C. briggsae, which is the closest known species to C. elegans and has almost identical morphology. The C. briggsae sequence is currently more than 98% complete and a genome-wide comparison of C. elegans and C. briggsae reveals that 89% of the C. briggsae genes have either orthologs (62%) or multiple matches(27%) in the C. elegans genome(Stein et al., 2003). As in C. elegans, the ESC/E(Z) complex is present in C. briggsae. However, C. briggsae lacks PRC1 complex genes. Surprisingly,orthologs of SOR-1 or SOP-2 could not be identified in C. briggsaegenome. The most closely related protein to SOR-1 in C. briggsae is CBP23695 (encoded by CBG18142) (2.6e^–10, 45.3%) and to SOP-2 is CBP06294 (encoded by CBG02561) (1.1e^–19, 52.4%). However,these two proteins do not appear to be the orthologs of SOR-1 and SOP-2. Regions of homology are restricted to the low complexity parts of SOR-1 or SOP-2. More importantly, functionally significant domains in SOR-1 (e.g. the RNA binding domain) and SOP-2 (e.g. the SAM domain and the RNA-binding domains) are not present in these proteins. In addition, CBP06294 has a clear ortholog in C. elegans, the gene ZK84.1 (7.4e^–175,82.1%), and CBP23695 is closer to the C. elegans gene H20J18.1a(3.7e^–16, 54.6%). Interestingly, orthologs of the genes adjacent to both sor-1 and sop-2 do exist in C. briggsae and the organization of these neighboring genes in C. briggsae is comparably co-linear with the ones in C. elegans(Fig. 6A,B). For example, as in C. elegans, there are two genes between CBG18138 and CBG18143, and one gene between CBG20942 and CBG20938 in C. briggsae. This further supports the assertion that SOR-1 and SOP-2 orthologs are absent from C. briggsae. In conclusion, the putative SOR-1/SOP-2 complex, the function of which is similar to that of the PcG complex in other organisms in maintaining the repressed state of Hox genes, may have undergone specific evolution in C. elegans, even though the underlying mechanisms appear to be conserved.

Despite a lineage-driven mode of development in C. elegans, we provide evidence that sor-1 and sop-2 function as PcG-like genes in global transcriptional repression of Hox gene. As in PcGmutants in other organisms, the expression of Hox genes is correctly initiated in sor-1 and sop-2 mutants. However, the repressed state of Hox genes fails to be maintained outside their normal expression domains during subsequent development. SOR-1 and SOP-2 may form a complex in Hox gene repression. Both GFP reporters and antibody staining indicate that SOR-1 and SOP-2 are colocalized in nuclear bodies and the localization of SOR-1 depends on SOP-2. Furthermore, SOR-1 and SOP-2 directly interact in vitro. The synergistic phenotypes of sop-2; sor-1 double mutants further support the conclusion that SOP-2 and SOR-1 function together in gene regulation. In other organisms, combination of mutations in different PcG genes typically produces more than additive effects(Jurgens, 1985). These synergistic effects could be due to the disruption of the function or integrity of the complex by a cumulative effect of two simultaneously mutations.

Although sor-1 and sop-2 work in concert in Hox gene repression, they do not appear to make an equal contribution to gene regulation. Mutations in sop-2 cause more dramatic effects than those in sor-1 mutants, demonstrated by a greater extent and degree of ectopic expression of an egl-5 reporter. Furthermore, the expression patterns of sor-1 and sop-2 are not identical. The expression level of sor-1 is weaker than that of sop-2,especially in the post-embryonic somatic cells. At later embryonic and larval stages, the majority, if not all, of SOP-2 is localized in nuclear bodies. Although SOR-1 is colocalized with SOP-2 in bodies, it also diffuses in the nuclei at all stages. Thus, SOR-1 is likely to be needed only in specific cell types and developmental stages. Differences in the timing and tissue specificity of homeotic misexpression have also been observed in PcGmutants in other organisms (Soto et al.,1995). Homeotic defects are more severe in the loss of function of Pc, Ph and Scm than those in loss of function of other PcG components. This could be due to their distinct roles in establishing repressive chromatin structures. For example, PSC is essential for compacting the nucleosomal arrays, while RING1 mediates the ubiquitination of H2A (Francis et al., 2004; de Napoles et al., 2004; Wang et al., 2004).

Other possible integral components of the SOP-2/SOR-1 complex are RNAs. Both SOR-1 and SOP-2 contain RNA-binding activity. Having two RNA-binding proteins involved in the same process supports the role of RNA in epigenetic silencing of Hox genes, although the nature of the RNA components in the SOP-2/SOR-1 complex has yet to be identified. siRNAs and structural roX1 and roX2 RNAs are important for the formation and targeting of the RITS complex and the Drosophila dose compensation complex (DCC), respectively (Verdel et al., 2004; Motamedi et al.,2004; Akhtar,2003). Analogous to other RNA-mediated epigenetic gene regulatory phenomena, the RNA components could be involved in maintaining the integrity of the PcG complex and/or targeting the PcG complex to specific loci(Fig. 6C).

In other organisms, the ESC/E(Z) complex and the PRC1 complex function together in the maintenance of the Hox gene expression. The ESC/E(Z) complex may be involved in recruiting the PRC1 complex to the target loci. Unlike in sop-2 and sor-1 mutants, which have widespread ectopic expression of Hox genes, the effect of mes-2 and mes-6mutations on Hox gene repression is subtle and is only apparent in sensitized genetic backgrounds (Ross and Zarkower,2003). Lack of synergy between sop-2(bx91) and mes-2/mes-6(RNAi) further supports the conclusion that the MES-2/MES-6 complex may not be involved in recruiting the SOP-2/SOR-1 complex in Hox gene repression.

In spite of a lack of obvious sequence similarity, several conserved properties of the PRC1 and putative SOP-2/SOR-1 complexes suggest that they appear to employ a conserved mechanism in Hox gene repression. First,components of the PRC1 complex and the SOP-2/SOR-1 complex are localized into distinct nuclear bodies, although the mechanistic role of the bodies in Hox gene repression remains unknown. Second, the protein-protein interaction SAM domain is present in both complexes. Domain-swapping experiments indicate that only the SAM domains of PcG proteins, including PH and SCM, but not the ones of non-PcG proteins, including TEL and L(3)MBT, can functionally substitute for the SAM domain of SOP-2 in terms of repression of Hox genes and proper formation of SOP-2 nuclear bodies (Zhang et al., 2004a). Overexpression of the SAM domain behaves as a dominant-negative regulator that compromises the function of the SOP-2/SOR-1 complex and the PRC1 complex in Hox gene repression (Y.S. and H.Z.,unpublished) (Peterson et al.,2004). These observations indicate an important role of the SAM domain in PcG-mediated Hox gene repression. Third, both of the SOP-2/SOR-1 complex and the PRC1 complex contain RNA-binding activity(Zhang et al., 2004b). Although a role for RNA in PcG-mediated gene silencing of Hox genes in other organisms has not been demonstrated, PcG proteins are involved in some RNA-dependent silencing processes, such as X chromosome inactivation and silencing of transgenes in fly (Plath et al., 2003; Plath et al.,2004; Pal-Bhadra et al.,2002). The evolutionary constraint on the PRC1 complex is, thus,likely to be conferred at the mechanistic level but not at the protein level. The reason for the maintenance of the PRC1 complex over very large evolutionary distance encompassing insects and vertebrates, while in nematodes a similar complex appears to be evolutionarily volatile, remains to be explained.

We thank Drs Xiaodong Wang, Scott Emmons, Robyn Lints, Rachel Palmer and members in our laboratory for their helpful comments on the manuscript. We also thank Drs Gang Zhi and Zhifu Han for the fusion proteins and the generation of SOR-1 antibodies. This work was supported by the National High Technology Projects 863.