Small Ubiquitin-like Modifier (SUMO)-mediated Repression of the Xenopus Oocyte 5 S rRNA Genes*

Background: Xenopus oocyte 5 S rRNA genes become permanently repressed after the gastrula-neurula transition. Results: TFIIIA interacts with the SUMO ligase PIAS2b and the corepressor CtBP, which occupy the oocyte but not somatic 5 S rRNA genes. Conclusion: Activation of CtBP by SUMOylation leads to repression of the oocyte genes. Significance: TFIIIA mediates repression as well as activation of 5 S rRNA transcription. The 5 S rRNA gene-specific transcription factor IIIA (TFIIIA) interacts with the small ubiquitin-like modifier (SUMO) E3 ligase PIAS2b and with one of its targets, the transcriptional corepressor, XCtBP. PIAS2b is restricted to the cytoplasm of Xenopus oocytes but relocates to the nucleus immediately after fertilization. Following the midblastula transition, PIAS2b and XCtBP are present on oocyte-type, but not somatic-type, 5 S rRNA genes up through the neurula stage, as is a limiting amount of TFIIIA. Histone H3 methylation, coincident with the binding of XCtBP, also occurs exclusively on the oocyte-type genes. Immunohistochemical staining of embryos confirms the occupancy of a subset of the oocyte-type genes by TFIIIA that become positioned at the nuclear periphery shortly after the midblastula transition. Inhibition of SUMOylation activity relieves repression of oocyte-type 5 S rRNA genes and is correlated with a decrease in methylation of H3K9 and H3K27 and disruption of subnuclear localization. These results reveal a novel function for TFIIIA as a negative regulator that recruits histone modification activity through the CtBP repressor complex exclusively to the oocyte-type 5 S rRNA genes, leading to their terminal repression.

The Xenopus 5 S rRNA genes have long served as a model for the developmental regulation of transcription, providing some of the first evidence that chromatin actively participates in the control of gene expression (1). There are two major multigene 5 S rRNA families as well as an assortment of trace and pseudogenes in the Xenopus laevis genome. The oocyte-type genes number more than 20,000 copies per haploid genome and are expressed during oogenesis and for a brief period when zygotic transcription begins at the midblastula transition. The genes are found in the subtelomeric region of most chromosomes in clusters exceeding 1000 repeats (2,3). The somatic-type genes are expressed at all stages of development and are sufficient to support ribosome biogenesis in adult cells. A single cluster of 400 copies is located on chromosome 9 (3).
Transcription of 5 S rRNA genes by RNA polymerase III requires the formation of an initiation complex comprised of TFIIIA, 3 TFIIIB, and TFIIIC. TFIIIA is necessarily the first factor to bind to the internal promoter and is solely dedicated to transcription of 5 S rRNA genes (4). TFIIIA is extraordinarily abundant in Xenopus oocytes, with ϳ10 12 molecules/cell, which reflects a second role for the protein (5,6). In early-stage oocytes, TFIIIA also forms a storage ribonucleoprotein particle for accumulated 5 S rRNA until it is used later for ribosome assembly. Although the levels of TFIIIA decline at the end of oogenesis, there are still ϳ10 7 molecules/cell when zygotic transcription begins at the midblastula transition, representing more than a 100-fold excess of factor over the total number of 5 S rRNA genes (6,7). The affinity of TFIIIA for the promoters of the oocyte and somatic genes is essentially identical (8), so the declining levels of TFIIIA alone cannot account for the silencing of the oocyte-type genes that is complete at the gastrula-neurula transition (9).
The final repressed state of the oocyte-type genes occurs during the period in which the maternally inherited histone H1 variant, H1M, is replaced by the adult H1A subtype (10 -12). There is substantial evidence that H1A-dependent positioning of nucleosomes in somatic cells ultimately determines the differential expression of the two types of 5 S rRNA genes (13)(14)(15)(16)(17)(18)(19)(20)(21). In the case of the somatic 5 S rRNA genes, H1A binds to the 5Ј side of the nucleosome, leaving much of the internal promoter sequence exposed, whereas H1A binds to the 3Ј side of the nucleosome positioned on the oocyte-type genes, occluding access to TFIIIA (14,15,18). This distinct arrangement of nucleosomes relies on the sequences that flank the two types of genes, which are A:T-rich in the case of the oocyte genes and G:C-rich in the case of the somatic genes (19,20). Although there is a detectable difference in the acetylation of histone H4 associated with somatic (hyperacetylated) compared with oocyte (hypoacetylated) 5 S genes in adult (kidney) cells (22), this modification does not appear to influence the binding of TFIIIA (14).
We have shown that TFIIIA is phosphorylated primarily on serine 16 by protein kinase CK2 (23). Phosphorylation begins during oogenesis and appears to be stimulated by progesterone-induced maturation of the oocyte to an egg (24). TFIIIA, in which serine 16 is replaced with glutamic acid, cannot support the transcription of oocyte-type 5 S rRNA genes in vitro or in vivo, whereas transcription of the somatic genes is unaffected. Template exclusion assays demonstrated that transcription complexes containing the phosphomimetic variant can form on the oocyte-type genes but are inactive. Therefore, this posttranslational modification of TFIIIA can account for the decreased expression of the oocyte-type genes relative to the somatic-type genes when zygotic transcription begins at the midblastula transition and provides a mechanism for initial repression of the former when levels of TFIIIA are still high (25).
Serine 16, which is located in the ␤ strand of the first zinc finger, is well exposed on the surface of DNA-bound TFIIIA (26,27). Consistent with this position of the amino acid, we found no evidence that phosphorylation of TFIIIA affects its binding affinity for the promoters of either the oocyte-or somatic-type genes nor for 5 S rRNA, suggesting that the functional effect of modification likely alters an interaction with another protein (24). To address this question, we carried out a yeast two-hybrid screen of a Xenopus expression library using the first four fingers of TFIIIA as bait and identified PIAS2b, a SUMO ligase, and XCtBP, a transcriptional corepressor regulated by SUMOylation, as candidates. Using ChIP to track the occupancy of the 5 S rRNA genes during early embryogenesis, we detected the presence of these proteins as well as SUMO1 and SUMO2/3 solely on the oocyte-type genes. Repressive histone modifications (H3K9me2 and H3K27me3) appear on the oocyte genes concomitant with the binding of XCtBP. The ChIP assays also revealed that some amount of TFIIIA remains associated with the oocyte 5 S rRNA genes throughout this period of development, consistent with a role in establishing the repressed state of these genes. Immunohistochemical staining of embryos with anti-TFIIIA antibody reveal a pattern around the nuclear periphery that corresponds to the subnuclear arrangement of telomeres in somatic Xenopus cells (2), providing further evidence that this transcription factor, contrary to previous speculation, is not entirely displaced from the oocyte-type genes.
These results account for the stoichiometric excess of TFIIIA over the total number of 5 S rRNA genes during early embryogenesis and demonstrate that, rather than being lost from the oocyte-type genes, a limiting amount of the factor remains associated with them and, seemingly, acts as a negative regulator. The chromatin structure established by the H1A-dependent positioning of nucleosomes appears to be only one aspect of a more complex process in which TFIIIA itself recruits chromatin modifying activity that contributes to repression and to the subnuclear localization of the oocyte 5 S rRNA genes.

EXPERIMENTAL PROCEDURES
Several experimental procedures have been described in detail earlier, including preparation of oocytes and embryos, in vivo transcription assays, RNase protection assays, pulldown assays, 5Ј Rapid Amplification of cDNA Ends (RACE), in vitro RNA synthesis, and screening of a yeast two-hybrid library (24, 28 -30).
Plasmids and Nucleic Acids-The sequence encoding the first four zinc fingers (amino acids 1-136) or the first three (amino acids 1-103) of TFIIIA were amplified by PCR and cloned into the BamHI and NdeI sites of plasmid pGBKT7 (Clontech), which contains the GAL4 DNA-binding domain, to generate the bait plasmids pGBKT7-4F and pGBKT7-3F, respectively. The plasmid pACT2-C54, containing X. laevis PIAS2b, was isolated during the screen of the Xenopus oocyte cDNA library using the first four fingers of TFIIIA as bait. A 2200-bp NcoI/XhoI restriction fragment from pACT2-C54 was cloned into the NcoI/SalI sites of pGBKT7 to generate pGBKT7-C54. Plasmid pC54RACE was generated by cloning a 530-bp EcoRI and XhoI product from a 5Ј RACE assay of PIAS2b mRNA into corresponding sites in pBSKSϩ. Plasmid pACT2-B36, containing X. laevis XCtBP, was also isolated during the screen of the Xenopus oocyte cDNA library, and a 2000-bp BamHI/XhoI restriction fragment from it was subcloned into pGADT7 to generate pGADT7-XCtBP.
Plasmid pETC54Ab contains the sequence encoding the last 107 amino acids of PIAS2b inserted into pET23b (Novagen). The truncated protein is expressed with a His 6 tag at the carboxyl terminus. Plasmid pETC54FH6 encodes full-length PIAS2b also with a His 6 tag at the carboxyl terminus.
The plasmid pSG9M-Gam1 was a gift from Dr. Susanna Chiocca (European Institute of Oncology, Milan, Italy) and contains the complete coding sequence of Gam1 carrying an N-terminal myc-tag (31,32). The plasmid was linearized with XhoI for in vitro synthesis of Gam1 mRNA. The plasmid pXE90 was a gift from Dr. Randall Moon (University of Washington School of Medicine, Seattle, WA) and contains the complete coding sequence for X. laevis XCtBP (33). It was used as the template for PCR amplification and for subcloning the XCtBP open reading frame into pET23b to create pETXCtBP.
Yeast Two-hybrid Screen-A X. laevis oocyte cDNA library (Clontech) that contains 3.5 million independent clones was amplified according to the instructions of the manufacturer, and Saccharomyces cerevisiae strain AH109 was used as the host for the screen. The bait plasmid, pGBKT7-4F, encodes the GAL4 DNA binding domain fused to amino acids 1-136 of TFIIIA (DNA-BD/4F). Library prey plasmids (pACT2) contain a leucine selective marker that allows for selection of the plasmid when transformed cells are grown in medium lacking leucine (S.D./-leu). Similarly, the bait vector with a tryptophan gene is grown in medium lacking tryptophan (S.D./-trp). ADE2 (adenine) and HIS3 (histidine) are growth reporter genes that are only expressed when a two-hybrid interaction takes place between the bait and the prey. After transformation with bait and prey plasmids, cells were resuspended in 10 ml of TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) (0.5 ml for the controls) and spread on the appropriate nutrient selection plates. For the library transformation, 0.2 ml were spread on 150-mm S.D./ -leu/-trp/-ade/-his (quadruple dropout) plates, whereas 0.1 ml of dilutions (1:100, 1:10, and 1:1) was spread on smaller, 100-mm plates to facilitate calculation of the transformation efficiency. Control transformations (pGBKT7-53 ϩ pGADT7-T and pGBKT7-laminC ϩ pGADT7-T) in 0.1 ml were spread on 100-mm plates. Plates were incubated for 3-7 days at 30°C before colonies were inspected for size and robustness of growth. Colonies were streaked on quadruple dropout plates, and colony lift ␤-galactosidase assays were performed for those that grew back. Each colony was spotted in duplicate in a gridlike pattern, and colonies were allowed to grow for 2 days at 30°C. Colonies were lifted onto a dry piece of filter paper that was submerged gently in liquid nitrogen for 20 s to lyse cells. Frozen filters were thawed at room temperature, colony side up, and placed on a second filter paper presoaked in Z buffer (113 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, and 2 mM MgSO 4 (pH 7.0)) containing 0.334 mg/ml X-gal (Promega) and 40 mM ␤-mercaptoethanol. Assays were developed for less than 8 h in a 30°C incubator, and colonies were scored. ␤-galactosidase assays were repeated at least three times for each clone. Colonies positive for the colorimetric assay were picked from the replica plate, restreaked on quadruple dropout plates, and used for prey plasmid isolation.
Expression and Purification of Recombinant Proteins-MBP-TFIIIA fusion protein was expressed from the plasmid pMTF according to the published procedure (34). The coding sequences for full-length PIAS2b, a truncated version containing the C-terminal 107 amino acids (used for antibody production), and full-length XCtBP were inserted into pET23b (Novagen) and expressed in Escherichia coli Rosetta(DE3) (Novagen). Proteins were purified by affinity chromatography using nickelnitrilotriacetic acid-agarose resin (Qiagen). 35 S-labeled proteins were synthesized in TNT rabbit reticulocyte lysate (Promega).
Antibody Production, Purification, and Labeling-Antibodies against full-length human CtBP1 (catalog no. sc-17759), human SUMO1 (catalog no. sc-5308), and SUMO2/3 (catalog no. sc-32873) were purchased from Santa Cruz Biotechnology. Those for H3K9me2 (catalog no. 9753) and H3K27me3 (catalog no. 9756) were from Cell Signaling Technology. Antibodies to the carboxyl terminus of PIAS2b were raised in rabbits by injecting the animals with a histidine-tagged polypeptide carrying the C-terminal 107 amino acids bound to nickel-nitrilotriacetic acid-agarose resin. For affinity purification of antibodies, samples (5 mg) of PIAS2b or TFIIIA were coupled to an N-hydroxysuccinimide-activated HiTrap column (1 ml, Amersham Biosciences Pharmacia) following the protocol of the manufacturer. Rabbit serum (12 ml) was diluted to 120 ml with cold 10 mM Tris-HCl (pH 7.5), cleared by centrifugation, and loaded onto the column. After washes with 10 mM Tris-HCl (pH 7.5) and 0.5 M NaCl, the immunoglobulin was eluted with 0.1 M glycine (pH 2.5) and collected in 900-l fractions into tubes containing 100 l of 1 M Tris-HCl (pH 8.0). Purified TFIIIA antibody was labeled with Alexa Fluor 568 carboxylic acid succinimidyl ester (Invitrogen) following the protocol of the manufacturer. The concentration of labeled protein was determined from A 280 using 203,000 cm Ϫ1 M Ϫ1 as the molar extinction coefficient for IgG after accounting for the absorption of the dye at 280 nm with a correction factor of 0.46. Alexa Fluor 568-labeled goat anti-rabbit antibody was purchased from Invitrogen.
Pulldown and Coimmunoprecipitation Assays-Plasmids pGBKT7-C54 and pGADT7-XCtBP were used to prepare 35 Slabeled PIAS2b or XCtBP, respectively, using the TNT T7-coupled rabbit reticulocyte lysate system (Promega). For binding assays to PIAS2b, a 15-g sample of MBP-TFIIIA was incubated with 20 l of the TNT reaction in the presence and absence of the oocyte 5 S gene (5 g) for 2.5 h at 4°C. A similar amount of MBP-L5 was used as a negative control. The total volume of the binding reactions was maintained at 200 l using pulldown buffer (10 mM HEPES (pH 7.4), 100 mM NaCl, 2 mM EDTA, 1% Triton, 1 mM dithiothreitol, and protease inhibitor mixture (Roche)). Amylose resin (15 l) was added to each tube, and the reactions were incubated for 3 h at 4°C on a rotating wheel. A negative control included incubation of amylose resin with 20 l of the PIAS2b TNT reaction alone. The resin was washed four times with 300 l of pulldown buffer, and the reaction was suspended in SDS loading buffer. Samples were resolved on a 10% SDS-polyacrylamide gel, followed by autoradiography to visualize the 35 S-labeled protein. For binding assays to XCtBP, protein A-Sepharose beads with 10 g of affinity-purified antibody to either TFIIIA or PIAS2b were incubated with 2 g each of the indicated proteins for 1 h at 4°C, followed by two washes with 1 ml of NET buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 0.01% Nonidet P-40). IgG-bound proteins were subsequently incubated with 20 l of XCtBP TNT reaction in a total volume of 300 l of NET buffer for 1 h at 4°C. Negative control reactions included protein A beads incubated only with XCtBP TNT reaction or an assay that included all three proteins with beads that were not adsorbed with antibody. Bound complexes were washed twice with 1 ml of NET buffer, resuspended in loading buffer, and resolved by SDS-polyacrylamide gel electrophoresis followed by autoradiography.
Embryo Preparation and Microinjection-Preparation of oocytes and embryos followed standard protocols (35). Embryos were maintained in 1/3 Marc's Modified Ringer and injected (5-30 nl) in the animal hemisphere using a Narishige programmable microinjector (IM-300) or a Drummond Nanoject II and allowed to develop at room temperature.
Immunohistochemistry and Confocal Microscopy-Embryos and oocytes were collected at the appropriate stage of development and dispensed ϳ30/well in a 24-well culture dish. Cells were fixed with freshly prepared MEMFA (0.1 M MOPS (pH 7.4), 2 mM EDTA, 1 mM MgSO 4 , and 3.7% formaldehyde). Fixed oocytes were incubated with 5 g of PIAS2b IgG (or TFIIIA IgG) in 0.5 ml of TBSN (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.2% Nonidet P-40) containing 4% (w/v) BSA for 24 h. Samples were washed with five changes of TBS buffer (10 mM Tris-HCl, pH 7.4, 155 mM NaCl) over a 24-hour period at room temperature. This was followed by incubation with 5 l of Alexa Fluor 568 labeled goat anti-rabbit antibody (Invitrogen) in 0.5 ml of TBSN containing 4% BSA, for 24 h. Samples were protected from exposure to light from this point onward. Nonspecifically bound IgG was removed by five 1-ml washes of TBS over a period of 24 h. Oocytes were transferred to 1.5-ml tubes and dehydrated with four 1-ml washes with methanol lasting 5 min each. Samples were cleared in 1 ml of Murray's solution (benzyl alcohol:benzyl benzoate, 2:1) and stored at 4°C, protected from light, until images were collected. Embryos were stained following this procedure, except antibodies (primary and secondary) were incubated for 48 h to allow for better penetration, and washes were extended over a period of 36 h. Samples were washed with TBS, dehydrated, and cleared as outlined above. Nuclei (DNA) were counterstained with a 5 M solution of TOPRO-3 in TBS for 30 min or a 0.3 M solution of DAPI for 10 min at room temperature. Excess stain was removed by washing with TBS. A Bio-Rad MRC 1024 scanning confocal system attached to a Nikon Diaphot 200 inverted microscope or a Nikon A1R-MP confocal microscope was used to collect images of labeled specimens.
Chromatin Immunoprecipitation Assays-Chromatin samples were prepared essentially according to the method used in Ref. 36 as modified in Ref. 37. Isolated nuclei were fixed in 1% (final concentration) formaldehyde, and chromatin was solubilized by sonication. Samples were centrifuged at high speed for 10 min (4°C) to remove insoluble material, and aliquots were flash-frozen in liquid nitrogen and stored at Ϫ80°C. The amount of DNA in each chromatin sample was determined using the PicoGreen dsDNA fluorometric quantitation assay (Invitrogen) following the protocol of the manufacturer. Samples were precleared by incubation with 100 l of swollen protein A-Sepharose beads containing 1 g/l carrier tRNA and 1 g/l herring sperm DNA. Each immunoprecipitation assay contained 0.1 g (200 l) of precleared chromatin, 5 g of affinity-purified antibody (PIAS2b and TFIIIA), commercially purchased antibody (CtBP, SUMO1, SUMO2/3, H3K9me2, and H3K27me3), or 5 l of preimmune serum and protein A resin. After a series of washes, immune complexes were eluted by incubation with 185 l of freshly prepared NS buffer (0.1 M NaHCO 3 and 1% SDS) for 30 min at room temperature, followed by a second elution with 100 l. Eluents (285 l) were mixed with 15 l of 5 M NaCl, and the cross-linking was reversed by overnight incubation at 65°C. For PCR reactions, the MgCl 2 concentration for the oocyte gene amplification was 4 mM, and that for the somatic gene was 2.5 mM. The oocyte gene was amplified with primers ChIPO1 (5Ј CCA CAG TGC CGC TGA CAA G 3Ј) and ChIPO2 (5Ј CAG CAG CAC CTT TTG GCT CC 3Ј). The somatic gene was amplified with primers ChIPS3 (5Ј GGC CCC AAC AAC GCA GCA C 3Ј) and ChIPS2 (5Ј GCA GCT AGC TGT CTG GCT GTT G 3Ј).
Gam1 Embryos-For assays of SUMOylation activity, onecell embryos were injected with Gam1 mRNA (0.5 ng) or H 2 O and allowed to develop to the early neurula stage (ϳ16 h postfertilization). Ten embryos were collected and homogenized in 11 l of SUMO reaction buffer (Boston Biochem) and cleared of cell debris by centrifugation at 16,000 ϫ g for 5 min in a microcentrifuge. An aliquot of extract was added to the assay mix containing 60 M SUMO1, 5 M Ubc9 (E2 enzyme), 5 M E2-25K (substrate peptide), 25 mM magnesium-ATP for a total volume of 25 l. A positive control reaction contained 500 nM SAE1/SAE2 (E1 enzyme) in place of the cell extract. Reaction mixtures were incubated at 37°C for 4 h, terminated by the addition of SDS loading solution, and analyzed by Western blotting using antibody specific to the E2-25K peptide (Boston Biochem). For transcription assays, one-cell embryos were injected with 0.08 Ci of [␣-32 P]UTP alone or with 7 ng of capped TFIIIA mRNA or 5 ng of capped Gam1 mRNA. Embryos were cultured in 1/3ϫ Marc's Modified Ringer until early gastrula (ϳ10 h post-fertilization). Embryos from each sample set were collected and homogenized in 500 l of SETS (150 mM NaCl, 10 mM EDTA, 50 mM Tris-HCl (pH 7.5), and 0.5% SDS). RNA was isolated as described before (24), and three embryo equivalents of RNA were loaded onto a 10% polyacrylamide denaturing gel, separated by electrophoresis, and visualized using autoradiography. Films were scanned and analyzed using National Institutes of Health ImageJ software to determine the ratio of 5 S rRNA to tRNA in each sample. For confocal analysis, one-cell embryos were injected with 2.5 ng of Gam1 mRNA. Embryos at the indicated stage were fixed and processed for immunohistochemical analysis as described above.

PIAS2 and XCtBP Bind to the N Terminus of TFIIIA-Serine
16 is the major site of phosphorylation in TFIIIA (23). Substitution with glutamic acid at this position serves as a putative phosphomimetic that cannot support transcription of the oocyte-type genes in vitro or in vivo despite binding and recruiting at least one other polymerase III transcription factor (24). This observation suggests that this region of TFIIIA is involved in an interaction that regulates the differential transcription of 5 S rRNA genes. A yeast two-hybrid screen of a X. laevis cDNA library was undertaken to identify proteins that associate with this segment of TFIIIA. TFIIIA is comprised of nine C2-H2 zinc finger domains followed by an 8-kDa transcriptional activation domain (34). To focus on possible interactions within the site of phosphorylation, the bait used for the screen encompassed the first 136 amino acids of TFIIIA, which spans zinc fingers 1-4.
The initial screen of the library was made on the basis of a double (adenine/histidine) nutritional selection. Individual colonies were taken through a second round of nutritional selection before proceeding to colony lift ␤-galactosidase assays. The colorimetric assay was repeated at least three times for each clone prior to isolation of prey plasmids. Of 219 positive colonies from the nutritional selection, 33 exhibited moderate to robust ␤-galactosidase activity and were taken for further characterization. There were multiple isolates of the BЈ⑀ subunit of protein phosphatase 2A (accession number AAG22076) and a member of the FAX-ZFP family of zinc finger proteins (38). In addition, clones encoding the Xenopus ortholog of the SUMO ligase, PIAS2 isoform ␤ (accession number NM 001091503.1) and XCtBP (accession number Q9W758) were unique isolates that gave particularly robust responses in ␤-galactosidase assays. A shorter prey construct encompassing only the first three zinc fingers of TFIIIA also gave a strong positive response with these two proteins. The magnitude of the ␤-galactosidase activity for the interaction of PIAS2b with the TFIIIA fragments is comparable with that measured with p53, a known target of this SUMO ligase (Fig. 1A) XCtBP is a Xenopus homolog of the corepressor protein CtBP. There are two paralogs of the protein in mouse and human, but, as noted by Brannon et al. (33), the relationship of the frog protein to these two isoforms is not clear. Although XCtBP has slightly higher identity to human CtBP1 (80%) than to CtBP2 (76%), it is currently considered a homolog of the latter. In many cases, CtBP family members use a domain near their N terminus to target binding partners that contain a fiveamino acid motif, PXDLS. TFIIIA does not contain this sequence, indicating the interaction occurs through one of the other regions of CtBP that are known to engage in proteinprotein interactions (40). For example, the carboxyl-terminal end of XCtBP binds to the transcription factor XTcf-3 (33). Furthermore, the XCtBP clone isolated from the screen is missing the first 38 amino acids. This also indicates that the PXDLSbinding domain is not involved in its association with TFIIIA.
The interactions of PIAS2b and XCtBP with TFIIIA were confirmed using pulldown assays (Fig. 1, B and C). The association of PIAS2b with TFIIIA was tested in the presence and absence of a stoichiometric amount of oocyte-type 5 S rRNA gene, which had no detectable effect (Fig. 1B), demonstrating that the interaction of these two proteins does not require a DNA-induced conformation in TFIIIA. We also detected an interaction between XCtBP and PIAS2b in the pulldown assays (Fig. 1C). Although PIAS2b has been implicated in the SUMOylation of CtBP (41), this is the first evidence for a direct interaction between the ligase and this target. Of note, the binding of PIAS2b and XCtBP to TFIIIA are not mutually exclusive in these assays, indicating the possibility that the proteins can form a ternary complex.
Expression of PIAS2b and CtBP during Early Xenopus Development-To understand how PIAS2b and XCtBP might play a role in the transcription of the 5 S rRNA gene families, we examined the expression of these proteins during development from oocyte to late neurula embryo. Northern blot analysis showed that the PIAS2b transcript is highest in stage I oocytes and then declines rapidly to undetectable levels by stage IV of oogenesis. This measurement was confirmed using a more sensitive RNase protection assay that revealed a similar decrease but with a perceptible amount of transcript still present at the end of oogenesis ( Fig. 2A). Burn et al. (42) have shown that PIAS2 mRNA continues to decrease slowly after fertilization until the neurula stage, when there is an apparent resumption of transcription. A Western blot analysis of developmentally staged oocytes detected the appearance of PIAS2b protein at stage II that increases over the remainder of oogenesis (Fig. 2B). The total amount of PIAS2b protein remains constant from the mature oocyte up though the late neurula stage (on a per-embryo basis). Therefore, early transcription of the PIAS2b gene in the oocyte is sufficient to reach and maintain a constant amount of the protein that persists though the early stages of embryo development.
The expression of XCtBP is different from PIAS2b. Brannon et al. (33) have shown that the level of maternal transcripts remains constant from the unfertilized egg through gastrulation. Zygotic transcription of XCtBP is then activated at the gastrula-neurula transition and peaks at the tailbud stage. Western blot assays first detect XCtBP at the end of oogenesis, with the amount of protein remaining relatively uniform up to the late neurula stage (Fig. 2C). The expression patterns of these two proteins indicate that there is a possible role for PIAS2b in the oocyte, whereas the earliest availability of XCtBP occurs in the mature egg, if not later in development.
Subcellular Relocalization of PIAS2b during Early Embryogenesis-The SUMOylation machinery and SUMO target proteins are mostly, but not exclusively, nuclear (43). Given that TFIIIA has two distinct compartmentalized activities, we examined the distribution of PIAS2b to determine whether its interaction with the transcription factor is possibly limited to a specific subcellular location and, consequently, one of these functions. Whole-mount immunohistochemical staining of staged oocytes and embryos with affinity-purified antibody directed at Xenopus PIAS2b was used to track the subcellular location of this SUMO ligase over time (Fig. 3). Consistent with the Western blot assay, PIAS2b protein is present in early-stage oocytes and increases throughout oogenesis, maintaining a constant concentration as the volume of the cell increases (Fig.  3A).  35 S-labeled PIAS2b. The fusion protein was recovered by binding to amylose resin. Precipitated samples were analyzed by SDS-PAGE followed by autoradiography. C, reticulocyte lysate containing 35 S-labeled XCtBP was incubated with either TFIIIA or PIAS2b alone or with both proteins together. Complexes were collected by immunoprecipitation using either anti-TFIIIA (lanes 1 and 2) or anti-PIAS2b (lanes 3 and 4) antibody adsorbed to protein A-Sepharose. Control assays used unmodified protein A-Sepharose. Precipitated samples were analyzed by SDS-PAGE followed by autoradiography.
Unexpectedly, PIAS2b protein is located solely in the cytoplasm at all stages of oogenesis. The apparent exclusion of the protein from the nucleus was confirmed by manual dissection to prepare enucleated oocytes and germinal vesicles that were analyzed by Western blotting (Fig. 3B). Extract prepared from the former gave a robust signal (Fig. 3B, top panel), whereas the weak reaction of nuclear extract (Fig. 3B, bottom panel) is likely due to contaminating PIAS2b adsorbed nonspecifically to the nuclear membrane. To our knowledge, this is the first observation of a strict cytoplasmic compartmentalization of a member of the SUMOylation machinery. Following fertilization, PIAS2b is disbursed throughout the cells of the early embryo. However, nuclear enrichment of the ligase can be detected beginning around the midblastula stage (Fig. 3, C and D) and continues until the majority of the protein is located in the nucleus by the early gastrula stage (Fig. 3, E and F). Because most known targets of PIAS2b are nuclear proteins, its exclusive localization in the cytoplasm of oocytes appears to be a means to regulate the activity of the ligase globally during early development.
SUMOylation Activity Is Required for Repression of the Oocyte 5 S rRNA Genes-The interaction of TFIIIA with a SUMO ligase and with a transcriptional corepressor that is activated by SUMOylation raises the question whether this posttranslational modification plays any role in the regulation of the 5 S rRNA genes, especially given the observed nuclear localization of PIAS2b during gastrulation that coincides temporally with the terminal repression of the oocyte-type genes. To address this point, we inactivated SUMOylation activity by the expression of the avian adenovirus protein Gam1, which binds to the SAE1 subunit of the E1-activating enzyme and triggers its proteolytic destruction (31,44). The advantage of this strategy, as opposed to an antisense methodology, is that the effect is immediate and eliminates both existing and de novo accumulation of the E1 enzyme.
To establish that SUMOylation activity can be effectively reduced by expression of Gam1, one-cell embryos were injected with mRNA (0.5 ng) encoding the protein (or water as a negative control). Whole cell extract was prepared from early neurula embryos (stage 14, 16 h post-injection) and used as a source of E1 activity for in vitro assays. Extract from water-injected embryos supports SUMOylation of a substrate peptide (Fig. 4A, lane 2), whereas there was no detectable activity in the extract prepared from Gam1-injected embryos (Fig. 4A, lane 3). This effect is long-lived. Detectable SUMOylation activity in embryos injected at the one-cell stage does not appear until midneurula (22 h post-fertilization).
One-cell embryos injected with mRNA encoding TFIIIA express elevated levels of 5 S rRNA by the gastrula stage because of activation of the normally repressed oocyte-type genes (7). We compared this effect with embryos injected with Gam1 mRNA (Fig. 4, B and C). The SUMOylation-deficient embryos exhibit a comparable level of enhanced 5 S rRNA synthesis, attributable to the oocyte-type genes, relative to control embryos injected with water. This result implicates SUMOylation activity in the establishment of the repressed state of the oocyte 5 S rRNA genes. Western blot analysis of Gam1-injected and control embryos shows that there is no increase in the levels of TFIIIA in the former to account for the increased synthesis of 5 S rRNA (Fig. 4D), supporting the view that the increased transcriptional activity in the SUMOylation-deficient embryos is due to a loss of repression.
Occupancy of the 5 S rRNA Genes during Early Embryogenesis-The effect of Gam1 on the transcription of the oocyte 5 S rRNA genes implicates SUMOylation in the terminal repression of these genes. We turned to ChIP assays to test for the presence of SUMO, PIAS2b, and XCtBP on 5 S rRNA genes that would further support a role for this posttranslational modification in their regulation. The ChIP assays were used to monitor the oocyte (Fig. 5A) and somatic (Fig. 5B) genes at time points following the resumption of transcription at the midblastula transition. Both SUMO1 and SUMO2/3 were present on the oocyte 5 S rRNA genes at late blastula stage and remained at a constant level through all time points. There are  . Depletion of SUMOylation activity prevents repression of oocyte 5 S rRNA genes. A, SUMO activation (E1) activity in extract prepared from Gam1-injected embryos was measured by Western blot analysis using a 25-kDa SUMO substrate peptide. Lane 1, peptide alone; lane 2, peptide incubated with extract from water-injected (control) embryos; lane 3, peptide incubated with extract from Gam1-injected embryos. The reactions in lanes 2 and 3 also contained E2 (Ubc9) enzyme, SUMO1, and ATP. Positions of molecular weight markers (in kilodaltons) are indicated. B, one-cell embryos were injected with [ 32 P]UTP and TFIIIA mRNA, which activates transcription of the oocyte-type genes after the midblastula transition, or Gam1 mRNA. RNA was isolated at early gastrula stage, and three embryo equivalents were analyzed by electrophoresis/autoradiography. Lane 1, control embryos injected with [ 32 P]UTP only; lane 2, injection with TFIIIA mRNA (7 ng); lane 3, injection with Gam1 mRNA (5 ng). C, autoradiographs for four experiments were scanned, and 5 S rRNA was quantitated relative to tRNA using ImageJ software. The ratio is given in arbitrary units. Error bars indicate mean Ϯ S.D. D, one-cell embryos were injected with Gam1 mRNA (lanes 1, 3, and 5) or water (lanes 2, 4, and 6). Whole cell extract was prepared, and four embryo equivalents were analyzed by Western blot developed with TFIIIA antibody. several potential targets of SUMOylation that can account for this result, including PIAS2b (45), CtBP (41,46), histone proteins (47,48), and corepressors/chromatin modifiers (reviewed in Ref. 49). The detection of SUMO1 and SUMO2/3 may indicate that there are multiple targets associated with the oocyte genes. However, some proteins can be modified by more than one paralog (50), and poly-SUMO2/3 chains can be capped by SUMO1 (51).
Contrary to current models for the regulation of the Xenopus 5 S rRNA genes, the ChIP assays unexpectedly revealed an approximately constant amount of TFIIIA associated with the oocyte-type genes at all time points. The ChIP assays here can only be used qualitatively because the DNA content of early Xenopus embryos changes substantially because of the loss of amplified rDNA that dissipates during the first 15 h after fertilization (52,53). Initially, the amplified, extrachromosomal rDNA is ϳ30 pg/cell compared with 12 pg of total chromosomal DNA (54). Notwithstanding this complication, we estimate that 1% or less of the oocyte-type 5 S rRNA genes are occupied by TFIIIA. However, this would correspond to ϳ10 copies of TFIIIA per tandem repeat of the oocyte-type genes. TFIIIA and PIAS2b are present concurrently on the oocyte genes beginning at the earliest time point, indicating that the transcription factor immediately recruits the ligase or that the two bind as a preformed complex (Fig. 5A). TFIIIA and PIAS2b remain bound to the oocyte 5 S rRNA genes through the late neural stage.
The CtBP proteins are transcriptional corepressors that are found as a component of large multisubunit complexes that contain histone-modifying enzymes, chromatin-associated proteins, and other corepressors (55,56). Notably, CtBP corepressor activity is controlled by SUMOylation, with PIAS2b serving as a specific E3 ligase for this protein (41). The ChIP assays detect some initial binding of XCtBP at the late blastula stage that reaches a constant level during the gastrula stage and remains constant or, in some replicates of this experiment, declines at the last time point, late neurula stage. This behavior is consistent with the CtBP complex providing chromatinmodifying activity exactly at the developmental stage during which the oocyte-type 5 S rRNA genes become permanently repressed (9).
There is some evidence for differences in chromatin structure between the two types of X. laevis 5 S rRNA genes in somatic cells. Histone H4 on the oocyte-type genes is hypoacetylated in contrast to the somatic-type, which is hyperacetylated (22). The CtBP complex carries several chromatinmodifying activities, including histone deacetylases, methylases, and a demethylase that ostensibly contribute to its activity as a transcriptional corepressor. The two CtBP-associated methylases, EHMT1 and EHMT2, target H3K9 and, secondarily, H3K27. Both modifications are a signature of transcriptionally repressed chromatin. Dimethylation of H3K9 is first detected at late blastula stage and plateaus at late gastrula stage. The less pronounced trimethylation of H3K27 is first detected at early gastrula stage and slowly increases during progression to early neurula stage (Fig. 5A). Therefore, within the sensitivity of the ChIP assays, the association of XCtBP with the  DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51 oocyte-type genes is coincident with the detection of histone H3 methylation.

Repression of Xenopus Oocyte 5 S rRNA Genes
The somatic 5 S rRNA genes showed no evidence of occupation by any SUMO-modified proteins, PIAS2b, or CtBP. We detected only the presence of TFIIIA (Fig. 5B). As expected, the repressive chromatin marks, methylated H3K9 and H3K27, are absent from the somatic-type genes.
Because reduction of SUMOylation activity increases transcription of the oocyte-type genes (Fig. 4B), we asked whether there is a corresponding change in the occupancy of the 5 S rRNA genes in Gam1-injected embryos. With the repeated caveat that these ChIP assays are semiquantitative at best, the amount of SUMO protein associated with the oocyte genes is decreased, albeit not eliminated (Fig. 5C). Replicates of the experiment determined that the binding of TFIIIA is delayed and reduced moderately. Likewise, there is a small but detectable decrease in the amount of PIAS2b and XCtBP. There is a much more discernible decrease in H3K9me2 and, especially, H3K27me3, consistent with SUMOylation controlling the activity of the CtBP complex (41,55). Therefore, the deficiency in SUMOylation has affected the chromatin structure of the oocyte-type genes, accounting for their increased transcription. Interestingly, the ChIP assay shows a considerable loss, to nearly undetectable levels, of TFIIIA from the somatic 5 S rRNA genes in the Gam1-injected embryos (Fig. 5D). It is possible that the reduction of SUMOylation activity allows the oocyte genes, which outnumber the somatic genes by 50:1, to compete more efficiently for the declining amounts of TFIIIA. This phenomenon would also explain why the increased transcription of 5 S rRNA genes in Gam1-injected embryos never reaches the same level as embryos injected with mRNA encoding TFIIIA, which elevates the amount of the transcription factor to levels that greatly exceed the total number of 5 S rRNA genes (7).

TFIIIA at the Nuclear Periphery of Somatic Cells-
The current model for the terminal repression of the oocyte 5 S rRNA genes has a positioned nucleosome excluding the binding of TFIIIA to the internal promoter. However, the ChIP assays demonstrate that some amount of TFIIIA is normally associated with the oocyte genes past the time when they become fully repressed. In X. laevis, the tandem repeats of the oocyte-type genes are found at the end of the long arm of most chromosomes that are distributed around the nuclear periphery of interphase somatic cells (2). We undertook immunohistochemical staining of staged embryos with antibody directed against TFIIIA to obtain supporting evidence for the presence of the factor on the oocyte-type genes and to determine whether this subnuclear localization pattern was established early in development (Fig. 6). In a two-cell embryo (10 9 molecules of TFIIIA/cell), there is still an appreciable amount of the transcription factor located in the cytoplasm, although enrichment of the factor in the nucleus is apparent (Fig. 6A). At midblastula stage, much of the transcription factor is nuclear, and the first indication of placement around the nuclear periphery is detectable (Fig. 6B). This arrangement is well established at late blastula stage (Fig. 6C) and is retained though midneurula stage (Fig. 6D). The single repeat of the somatic-type genes cannot account for these staining patterns. Rather, they are consistent with the ChIP experiments that detected the presence of TFIIIA on the oocyte 5 S rRNA genes during this same period. In addition, the methylation of histone H3K9 on the oocyte-type genes is in accord with two recent reports showing that this modification is required for positioning transcriptionally repressed chromatin at the nuclear periphery (57, 58).
Because inhibition of SUMOylation activity by Gam1 leads to considerably reduced methylation of histone H3 on the oocyte-type genes, we asked whether loss of SUMOylation activity also affects their subnuclear localization. Immunohis- tochemical staining of Gam1-injected embryos (Fig. 6E) did show changes in TFIIIA staining to widely different degrees relative to water-injected controls (Fig. 6F). Staining of the nuclear periphery was decreased in some cases, and the pattern of TFIIIA that did remain in this region appeared to be aggregated into fewer, but, in some cases, larger, structures. Notwithstanding the severity of the phenotype, we conclude that the loss of SUMOylation can perturb the normal peripheral arrangement of the oocyte-type genes. However, it is important to caution that the 5 S rRNA repeats are adjacent to the telomeres, whose positioning at the nuclear periphery also appears to require SUMOylation activity (59 -61). Therefore, at this time, it is not possible to judge whether the disrupted staining pattern of TFIIIA in Gam1 embryos arises from changes in the chromatin structure of the 5 S rRNA genes, of telomeres, or both. The degree to which the heterochromatic structure of the oocyte 5 S rRNA genes may contribute to the subnuclear positioning of telomeres will be important to investigate.
SUMOylation of TFIIIA-An important question is whether TFIIIA is also subject to SUMOylation. Although there is no consensus sequence (YKXE, where Y is a hydrophobic residue) in TFIIIA, a possible site (CKEE 46 ) is close to the major site of phosphorylation at serine 16. We attempted to detect SUMOylated TFIIIA using immunoprecipitation of endogenous protein from whole cell extract followed by analysis using either Western blots or mass spectrometry. Neither approach was conclusive. Weak bands that potentially correspond to SUMOylated TFIIIA (at ϳ50 kDa) were detected on Western blots with antibodies to SUMO1, and peptides corresponding to SUMO1 and SUMO2 were detected in the mass spectra of immunoprecipitated TFIIIA. It is well established that the steady-state levels of SUMOylated protein are generally low, typically less than 5% (62)(63)(64). Nonetheless, the results from both analyses were not sufficient to establish that there is a pool of SUMOylated TFIIIA in embryos. We determined that TFIIIA can be SUMOylated in vitro (Fig. 7). However, it is not possible at this time to say whether the transcription factor is an actual target of this modification in vivo.

DISCUSSION
There is considerable evidence that histone H1 positions a nucleosome over much of the internal promoter of the X. laevis oocyte 5 S rRNA genes to the exclusion of TFIIIA, whereas the nucleosome that occupies the somatic 5 S rRNA genes permits binding of TFIIIA and formation of an active transcription ini-tiation complex (13)(14)(15)(16)(17)(18)(19)(20)(21). These observations, when considered in light of the increasing amounts of somatic histone H1 and the declining amounts of TFIIIA that occur contemporaneously during the early stages of embryogenesis, have led to a model based on a simple competition between factor binding and nucleosome organization. This model is supported by the observation that reduction of H1 levels leads to a delay in repression of the oocyte-type genes (13,16). In this study, we present evidence that the differential repression of the oocyte 5 S rRNA genes does not occur though a passive mechanism of simple competition, but, rather, that TFIIIA itself actively orchestrates this process, which requires SUMOylation activity, the corepressor XCtBP, and histone modifications that can account for the ultimate subnuclear localization of the oocytetype genes to the nuclear periphery.
The ChIP assays reported here have revealed that TFIIIA is associated with some fraction of the oocyte-type genes well into neurula stage. The confocal images of embryos immunohistochemically stained with TFIIIA antibody provide a direct visual confirmation. Complete repression of the oocyte-type genes occurs at the gastrula-neurula transition (9), but there is still an excess of TFIIIA over the total number of 5 S rRNA genes at this developmental stage (6,7). Additional evidence that TFIIIA is not limiting during this period comes from two observations. Ribozyme depletion of histone H1 is sufficient to activate transcription of oocyte 5 S rRNA genes at least through the gastrula-neurula transition (13,16), and Gam1 depression of SUMOylation activity has the same effect. In combination, these observations indicate that the level of TFIIIA does not account for the inactivity of the oocyte-type 5 S rRNA genes. Rather, it is formation of a repressive chromatin structure that depends on the recruitment of the corepressor complex XCtBP and its activation by PIAS2b-directed SUMOylation.
The question that arises is how TFIIIA is able to function simultaneously as an activator (somatic-type genes) and apparent repressor (oocyte-type genes) of transcription. It is firmly established that the A:T-rich sequences that flank the oocyte-genes are necessary for repression of their transcription (19,20,65). Sera and Wolffe (18) determined that the A:T-rich sequence immediately downstream of the oocyte genes between nucleotide positions 123 and 144, which is the predicted binding site for histone H1, is the major determinant for the repressive positioning of nucleosomes on these genes. Therefore, it is noteworthy that TFIIIA binds to the 5 S rRNA genes in an extended conformation, oriented with its N-terminal end at approximately nucleotide 96. Our two-hybrid experiments delimit PIAS2b and XCtBP binding to this end of TFIIIA, meaning that these two proteins are positioned near or at this A:T tract. PIAS2b, like all members of the PIAS family, possesses a SAP domain that has high affinity for A:T-rich sequences (66). By virtue of this DNA-binding domain, PIAS2b potentially stabilizes TFIIIA association with a subset of oocyte genes during a period when much of the region is being organized into a repressive chromatin structure. We speculate that TFIIIA, occupying the binding site for the core octamer, along with PIAS2b, occupying the A:T-rich site of H1, can be integrated with no disruption into the ordered array of nucleosomes that occupy the oocyte 5 S rRNA gene repeats in somatic  DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 35477
cells (67)(68)(69). When histone H1 was removed from fractionated chromatin containing oocyte-type genes, simple addition of purified RNA polymerase III resulted in some transcription of 5 S rRNA (67). This is further evidence that transcription complexes are not fully displaced from the oocyte-type genes in compacted chromatin. This model also accounts for the absence of PIAS2b from the somatic-type genes, which are flanked by G:C-rich sequences. We propose that the combined interactions of PIAS2b with TFIIIA and with the A:T-rich sequence flanking the oocyte-type genes enables stable binding of this complex, with subsequent recruitment and activation of the XCtBP corepressor complex (Fig. 8).
We microinjected mRNA encoding PIAS2b to which we added a nuclear localization element (SV40 large T antigen). This added element did not appear sufficient to mislocalize the protein to the nuclei of oocytes and indicates that the protein is actively held in the cytoplasm. A mechanism for this retention of PIAS2b may be its association with TFIIIA, most of which resides in the cytoplasm of oocytes bound to 5 S rRNA. Mutagenic analysis of PIAS3L has demonstrated that three domains (SAP, PINIT, and SP-RING), shared among most members of the PIAS family, are required for proper nuclear localization and retention (70), so an interaction between any one of these three regions with TFIIIA could potentially result in entrapment of the ligase in the cytoplasm. The expression of PIAS2b during oogenesis and its cytoplasmic retention through an interaction with TFIIIA could provide a kinetic advantage to the formation of TFIIIA-PIAS2b complexes on the oocyte genes following germinal vesicle breakdown because the expression of the adult isoform of histone H1, required for repression of the oocyte-type genes, does not occur until the midblastula transition (71).
The strict cytoplasmic localization that we observe for PIAS2b in oocytes can also account for previously unexplained results from early in vitro transcription studies. Peck et al. (72) reported that transcription of oocyte-type relative to somatictype 5 S rRNA genes was suppressed by more than an order of magnitude in whole cell extract prepared from oocytes compared with transcription in nuclear extract (or microinjected nuclei). These authors noted that the lower oocyte:somatic transcription ratio in whole oocyte extract was similar to that measured in developing embryos. Significantly, the inactive oocyte-type genes in this extract were complexed with TFIIIA. We propose that the subcellular localization of PIAS2b in oocytes is sufficient to account for the differential transcription activity of these two extracts, supporting our contention that the accessibility of the oocyte-type genes to PIAS2b following germinal vesicle breakdown is the event that initiates repression.
To our knowledge, these experiments provide the first evidence that a member of the CtBP corepressor family is involved in the regulation of genes transcribed by RNA polymerase III, although there are examples of cross-talk between the eukaryotic RNA polymerases through other shared coregulators (see, for example, Refs. [73][74][75]. XCtBP potentially accounts for much if not all of the chromatin modifications found on the oocyte-type genes by virtue of its presence in a large complex (ϳ1.5 MDa) that contains histone demethylase, histone methyltransferase, histone deacetylase, and SUMO ligase activities as well as an assortment of other transcriptional coregulatory proteins (55). SUMOylation of CtBP1 has been reported to control transcriptional repression, with PIAS2b serving as a specific SUMO ligase in HeLa cells (41). At least eight components of the CtBP complex can be SUMOylated, which potentially explains our detection of both SUMO1 and SUMO2/3 on the oocyte-type genes (49).
The histone modifying activities in the CtBP complex can account for the presently known differences between histones associated with the oocyte versus somatic 5 S rRNA genes. The deacetylases HDAC1 and HDAC2 can not only establish the hypoacetylated state of histone H4 on the oocyte 5 S rRNA genes (22), but they are likely needed for the eventual methylation of H3K9 and H3K27 that can be accomplished with the methyltransferases EHMT1 and EHMT2, which are also con- FIGURE 8. Model for the repression of the oocyte-type 5 S rRNA genes following the midblastula transition. A, repression of the oocyte-type genes is largely due to histone H1-dependent positioning of a nucleosome over the binding site for TFIIIA (nucleotides 45-96). The A:T-rich sequence flanking the 3Ј end of the gene, which is the binding site for the linker histone, is required for repression. In addition, ϳ1% of the oocyte-type genes are bound by TFIIIA, which can interact with PIAS2b and the CtBP corepressor complex. This complex may be stabilized by an interaction between the SAP domain of PIAS2b and the A:T-rich sequence that immediately flanks the 3Ј end of the oocyte-type genes. B, the somatic-type genes are flanked by G:C-rich sequences, providing no contact site for PIAS2b. Histone octamers are weakly positioned on the somatic-type gene, and the binding site for TFIIIA is exposed. Binding of the transcription factor is sufficient to reposition the octamer that possibly leads to dissociation of the H1 linker histone (18).
stituents of the complex. Consistent with the differential modification of histones on the two types of 5 S rRNA genes observed here, a genome-wide analysis of human polymerase III genes revealed that histones flanking inactive tRNA genes are, likewise, minimally acetylated but marked by H3K9me3 and H3K27me2/3. Therefore, just as for polymerase II genes, these histone modifications act as a signature of chromatin associated with repressed polymerase III genes (76).
There have been several observations linking transcriptional repression to the positioning of genes at the nuclear lamina (77), and there is now evidence that methylation of H3K9 plays an essential role in this process (57, 58). Transcriptionally silent regions found in lamin-associated domains are highly enriched for H3K9me2 (reviewed in Ref. 78), consistent with the strong ChIP signal we measured for this modification exclusively on the oocyte-type genes. Experiments in Caenorhabditis elegans indicate that mono-or dimethylation of H3K9 is specifically required for perinuclear localization with trimethylation needed for transcriptional repression that may be reinforced by methylation at H3K27, which we also find exclusively on the Xenopus oocyte-type genes (58). In combination, our results indicate that maintenance of the transcriptional repression of the oocyte 5 S rRNA gene arrays relies on H3K9me-dependent localization to the nuclear periphery.
The involvement of SUMOylation in the global repression of the Xenopus oocyte 5 S rRNA genes has striking parallels with other complexes that promote the formation of extended heterochromatic domains (49,79). The zinc finger protein Sp3 is converted from an activator to a potent repressor of transcription by conjugation to SUMO1 (or SUMO2 in vitro), with PIAS1 serving as a specific ligase (80,81). The modification results in the localization of Sp3 to the nuclear periphery in a pattern remarkably similar to that of TFIIIA (80). The SUMOylated form of Sp3 seeds the formation of a repressive complex that establishes heterochromatic marks, including H3K9me3 (82). The Polycomb group (PcG) and PML proteins, similarly, form multimeric complexes, regulated by SUMOylation, that establish and maintain longrange heterochromatic structures (46,83,84). With these examples in mind, it is not surprising that the repression of the tandem repeats of the oocyte-type genes seems to use a similar strategy for the formation of heterochromatin that becomes spatially segregated at the nuclear periphery.
We have shown that a small amount of TFIIIA remains associated with the Xenopus oocyte 5 S rRNA genes during the period when they become permanently repressed and that the repression depends on SUMOylation activity. The resulting repressive chromatin modifications also stabilize the localization of the oocyte-type genes to the nuclear periphery in somatic cells. The steady-state level of SUMO modification of most proteins is generally quite low, the "SUMO enigma" (64), and it has not been possible to determine with certainty whether TFIIIA itself is SUMOylated in vivo. Notwithstanding this point, it is clear that TFIIIA can act as a scaffold to bring together PIAS2b and XCtBP specifically on the oocyte-type genes, with the E3 ligase mediating the SUMOylation and, therefore, the activity of the chromatin-modifying complex.