ABSTRACT
The Drosophila melanogaster Polycomb group protein Posterior Sex Combs is a component of Polycomb repressive complex 1 and is central to Polycomb group-mediated silencing. A related Polycomb group gene, Suppressor 2 of zeste, is thought to be partially redundant in function. The two proteins share a small region of homology but also contain regions of unconserved sequences. Here we report a biochemical characterization of Suppressor 2 of zeste. Like Posterior Sex Combs, Suppressor 2 of zeste binds DNA, compacts chromatin, and inhibits chromatin remodeling. Interestingly, the regions of the two proteins responsible for these activities lack sequence homology. Suppressor 2 of zeste can also replace Posterior Sex Combs in a functional complex with other Polycomb group proteins, but unlike with their biochemical activities, complex formation is mediated by the region of Suppressor 2 of zeste that is homologous to that of Posterior Sex Combs. Our results establish Suppressor 2 of zeste as a functional homolog of Posterior Sex Combs and suggest that the two proteins operate via similar molecular mechanisms.
Polycomb group (PcG) genes were initially identified in Drosophila melanogaster cells for their role in maintaining Hox gene silencing during development (18, 25, 37). Subsequently, more than 15 PcG genes have been discovered (4, 40, 41). PcG genes are conserved in mammals, where they also maintain Hox gene expression and are further implicated in key processes such as X inactivation (17), genomic imprinting (12), and the self-renewal of stem cells (41). PcG gene products operate in protein complexes and are thought to maintain gene silencing through modifications of chromatin structure.
To date, a number of PcG protein complexes have been identified and characterized in Drosophila, and in most cases, analogous complexes also exist in mammalian cells (7, 11, 22, 24, 29, 31, 39, 43). One complex, Polycomb repressive complex 1 (PRC1), contains four major PcG subunits: Posterior Sex Combs (PSC), Polyhomeotic (Ph), Polycomb (Pc), and dRING1. These four proteins form a stable PRC1 core complex that can compact chromatin (15), prevent chromatin remodeling by the Swi/Snf ATPase (16), and inhibit transcription (20). In addition, mammalian homologs of PSC and dRING1 function together as an E3 ligase for the ubiquitylation of H2A, and dRING1 has been linked to the ubiquitylation of H2A in Drosophila (49).
One protein in PRC1, PSC, is sufficient for chromatin compaction and inhibition of remodeling and transcription (14, 15, 19). Psc mutant alleles encoding proteins that lack in vitro activities but retain the ability to form complexes have strong phenotypes, and truncated PSC proteins that lack the ability to form complexes are defective in gene silencing in vivo (19). These data suggest that both the formation of PRC1 and the inhibitory effects of PSC may be important for PcG-mediated gene silencing.
Genetic studies suggest that a related and adjacent gene of PSC, Suppressor 2 of zeste [Su(Z)2], could be a functional homolog. Mutations in Psc cause misexpression of Hox genes, but Psc-Su(z)2 double mutants show much more severe misexpression than Psc single mutants (3, 47). Overexpression of either Su(Z)2 or PSC also leads to similar bristle abnormalities (44). Furthermore, polytene chromosome staining revealed that the two proteins colocalize with other PRC1 components, Ph and Pc, at many sites. Both localization patterns are disrupted by a temperature-sensitive mutation in the PcG gene Enhancer of zeste (36). Finally, in clonal analyses of wing imaginal discs, deletion of both Su(Z)2 and PSC results in Hox gene derepression, which can be rescued by ectopic expression of either protein (3). These studies suggest that Su(Z)2 and PSC are at least partially redundant and that Su(Z)2 might form a PRC1-like complex, although no Su(Z)2 complexes have been identified to date.
At the protein level, Su(Z)2 and PSC share a small homology region of approximately 200 amino acids (37.4% identity) that begins at residue 19 for Su(Z)2 and at residue 246 for PSC (Fig. 1A). This homology region of PSC is sufficient and necessary to mediate interactions with other PcG proteins in PRC1 (19). Outside this region, the two proteins do not possess strict sequence homology. The nonhomologous sequences of both proteins are characterized by similar amino acid contents that are distinct from the average in the Swiss-Prot protein database (5). Interestingly, the nonhomologous region of PSC is necessary and sufficient for the described effects of PSC on chromatin (19), which raises the question of how Su(Z)2 can be a functional homolog of PSC given that there is no primary sequence conservation in the region important for these activities.
Su(Z)2 binds DNA. (A) Schematic representations of Su(Z)2 and PSC. The two proteins share a small region of homology (HR) but also contain larger, nonhomologous regions (NR). (B) FLAG-tagged Su(Z)2 and PSC were expressed in Sf9 cells. Purified Su(Z)2 and PSC (lanes 1 and 2, respectively) were resolved by 8% SDS-PAGE and stained with colloidal blue. Asterisks indicate two contaminants: a 70-kDa band that likely corresponds to HSC70 and a 55-kDa band that is likely β-tubulin. (C) Representative filter binding experiment. Su(Z)2 or PSC was titrated into binding reaction mixtures containing 20 pM 157-bp 32P-labeled DNA. The top filter (nitrocellulose) captures protein-DNA complexes, and the bottom filter (charged nylon) captures unbound DNA. (D) Calculation of apparent Kds of Su(Z)2 and PSC from DNA (1.58 ± 0.21 nM and 0.17 ± 0.04 nM, respectively). Data points are the average results of at least three experiments, and error bars are the standard errors of the means (SEM). Curve fitting is done by nonlinear regression using GraphPad Prism, with R2 values of 0.97 and 0.96 for Su(Z)2 and PSC, respectively. The inset shows the x axis on a log scale.
Because PSC seems to play a central role in PRC1 functions, we are interested in whether Su(Z)2 is indeed a functional homolog. We compared the two proteins using a number of in vitro assays and present evidence for their functional homology.
MATERIALS AND METHODS
Protein expression and purification.FLAG epitope-tagged Su(Z)2 and PSC were expressed in Sf9 cells using the Bac-to-Bac baculovirus expression system (Invitrogen) and purified as described previously (16). For the purification of the Su(Z)2 complexes, Sf9 cells were infected with FLAG-Su(Z)2 or its truncations and the Ph-proximal, Pc, and dRING1 viruses. Protein concentrations were determined by a Bradford assay (Bio-Rad), and purity was estimated with colloidal blue (Pierce)-stained protein gels.
Chromatin assembly.The 12-nucleosomal (G5E4) and 6-nucleosomal array templates used in this study have been described previously (15, 48). Chromatin templates were assembled with purified HeLa core histones by gradient salt dialysis as described previously (8, 15, 45).
Filter binding.A double filter analysis of DNA binding (52) was carried out with a 32P-labeled 157-bp TPT fragment as described previously (16, 30). DNA bound to filters was quantified on a Typhoon Trio variable mode imager (GE Healthcare). Active protein concentrations were determined using DNA in excess of protein and at concentrations at least fivefold above the Kd (dissociation constant). Binding constants were obtained from first-order fits using Prism version 5 (GraphPad Software).
Restriction enzyme accessibility (REA) assay.The experiments were performed as described previously (16), except the mononucleosome experiments contained 4 mM MgCl2. Chromatin templates were not radiolabeled, but DNA was instead visualized using SYBR gold staining (Invitrogen) and quantified using the Typhoon imager.
EMSA.The samples used in the binding reactions from REA assays were split at the end of the preincubation period, and half of each sample was used in the electrophoretic mobility shift assays (EMSAs). Samples were resolved in 0.6% 0.5× Tris-borate-EDTA agarose gels and visualized with SYBR gold staining on the Typhoon imager.
Micrococcal nuclease (MNase)/DNase I assays.The experiments were performed as described previously (16, 45), except that six-nucleosomal arrays were used and digestion products were visualized in agarose gels with SYBR gold staining on the Typhoon imager.
Electron microscopy.Sample preparation for the visualization of chromatin compaction was performed as described previously (15). Samples were applied to glow-discharged, carbon-coated grids, stained with uranyl acetate, and rinsed with water to achieve a positive stain as described previously (32, 53). Following inspection, grids were rotary shadowed with platinum and carbon, and visualization was carried out as described previously (15) using dark-field optics.
Sequence alignment and amino acid composition.Protein sequences for Su(Z)2 and PSC from 10 Drosophila species were obtained from the FlyBase (release FB2008_04) and Ensembl (release 49) databases, respectively. The FlyBase or Ensembl gene identification for each sequence is listed in Table S1 in the supplemental material. Sequences were aligned using ClustalW2 (European Bioinformatics Institute), and conservation was determined using BioEdit version 7.0.9 (Ibis Biosciences). Amino acid composition for the homology and nonhomologous regions of each protein was determined using Vector NTI version 10.3.0 (Invitrogen), and the Swiss-Prot average amino acid composition was obtained from the Swiss-Prot protein knowledgebase (release 55.3).
Nuclear extracts for coimmunoprecipitation.Nuclear extracts were prepared from wild-type yellow white (yw) Drosophila heads (42), S2 tissue culture cells, and Kc cells. Nuclei were isolated from the heads (46), S2 cells, or Kc cells, and the nuclear extract was prepared as described previously (1).
Coimmunoprecipitation.Nuclear extracts (100 μl) were incubated with protein A-agarose (Sigma) at 4°C for 1 h to preclear the extracts of proteins that may nonspecifically interact with the agarose beads. After the removal of the agarose beads, antibodies against Su(Z)2, Ph, Pc, or dRING1 (5, 2, 2, and 1 μl, respectively) were added, and binding was performed at 4°C for more than 12 h. Bound proteins were captured on protein A-agarose by incubating them at 4°C for 1 h. Agarose beads were washed with BC buffer (20 mM HEPES [pH 7.9], 0.2 mM EDTA, 20% glycerol), boiled in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer, and immunoblotted for PcG proteins.
RESULTS
To characterize the biochemical properties of Su(Z)2 and compare them with those of PSC, both proteins were expressed in Sf9 cells and purified by immunoaffinity chromatography (Fig. 1B). The activities of both proteins were analyzed side by side in multiple assays.
Su(Z)2 binds DNA.PSC binds tightly to free DNA, with no known sequence bias (16). To determine if Su(Z)2 shares this activity, we measured the Kds of Su(Z)2 and PSC for a 157-bp DNA using double filter binding (52). Su(Z)2 binds to DNA with a Kd of 1.58 ± 0.21 nM, an affinity ∼10-fold lower affinity than that of PSC, which binds to DNA with a Kd of 0.17 ± 0.04 nM (Fig. 1C and D; Table 1). Note that the Kd obtained for PSC is in excellent agreement with previous results (16).
Summary of biochemical activities of various Su(Z)2 and PSC proteins and complexes used this studya
We also used filter binding assays to quantify the fractions of the Su(Z)2 and PSC preparations that were active for DNA binding. Assuming they are binding as monomers, the average active fractions for Su(Z)2 and PSC are 27.8% ± 4.1% and 27.3% ± 2.0%, respectively. The average active fractions of different protein preparations are summarized (Table 1), and the concentrations of all proteins and complexes used in this study are active concentrations unless otherwise stated.
Su(Z)2 inhibits Swi/Snf-mediated chromatin remodeling of nucleosomal arrays.PSC has been shown to prevent chromatin remodeling by inhibiting the action of Swi/Snf (16), and we tested Su(Z)2 to see if it contains a similar activity. To track chromatin remodeling, we used REA assays. Restriction sites on nucleosomal DNA are generally inaccessible to restriction enzymes (35), but they can be exposed by ATP-dependent chromatin remodeling, such as that mediated by Swi/Snf (Fig. 2B, first two lanes) (26). Inhibition of chromatin remodeling can be quantified by the extent of the digestion of a unique restriction site on the chromatin template.
Su(Z)2 inhibits chromatin remodeling on nucleosomal arrays and mononucleosomes. (A) Schematic representation of the nucleosomal array used in REA assays. The array contains two central nucleosomes (gray) flanked by five 5S-positioned nucleosomes on each site. Two restriction sites (HhaI and XbaI) are normally occluded by nucleosomes but are exposed upon Swi/Snf-mediated chromatin remodeling. Conversely, the SacI site is normally exposed but is occluded upon chromatin remodeling. These nucleosome positions have not been mapped but are inferred from the enzyme accessibility data. (B) Su(Z)2 and PSC inhibit chromatin remodeling on nucleosomal arrays in REA assays. Representative data from REA assays with HhaI. The first two lanes are the negative and positive controls [with or without Swi/Snf but no Su(Z)2 or PSC], respectively, demonstrating that the HhaI site becomes accessible upon Swi/Snf-mediated chromatin remodeling. nuc, nucleosome. (C) Summary of REA assays on nucleosomal arrays using HhaI. Percent inhibition is calculated as {[percent uncut with Swi/Snf and Su(Z)2 or PSC] − [percent uncut with Swi/Snf]}/{[percent uncut without Swi/Snf] − [percent uncut with Swi/Snf]} × 100%. (D) Summary of the effects of Su(Z)2 on REA, demonstrating that Su(Z)2 does not prevent restriction enzyme access to the chromatin template. (E) Representative data from REA assays on nucleosomal arrays using SacI. The first two lanes demonstrate that the SacI site becomes less accessible upon Swi/Snf-mediated chromatin remodeling. (F) Summary of REA assays, demonstrating that Su(Z)2 blocks the decrease in SacI accessibility mediated by chromatin remodeling. Percent inhibition is calculated as in Fig. 2C. Data points are the average results (± SEM) of at least three experiments. (G) Summary of an REA assay using PstI on mononucleosomes. Data points are the average results (± SEM) of three experiments. (H) Representative data from EMSAs and REA assays on mononucleosomes. Mononucleosomes were incubated with increasing concentrations of Su(Z)2 or PSC, and the binding reactions were split to monitor nucleosome binding (EMSA; top panel) and inhibition of chromatin remodeling (REA; bottom panel). Arrows indicate the positions of bound mononucleosomes.
Similarly to PSC, Su(Z)2 inhibited chromatin remodeling in a concentration-dependent manner (Fig. 2B and C). For both proteins, 50% inhibition occurs at an approximate ratio of 0.5 protein per nucleosome (Fig. 2C; Table 1). To confirm that Su(Z)2 is inhibiting chromatin remodeling rather than interfering directly with restriction enzymes, we showed that Su(Z)2 did not change the level of restriction enzyme digestion by HhaI or XbaI in the absence of Swi/Snf (Fig. 2D).
The SacI site is unique among the restriction sites used, in that it is exposed on the assembled template but is occluded after chromatin remodeling (Fig. 2E, first two lanes). Inhibition of chromatin remodeling should therefore correlate to an increase, instead of a decrease, in the amount of restriction enzyme digestion. When Su(Z)2 is incubated with chromatin templates prior to chromatin remodeling and SacI digestion, the SacI site remains accessible (Fig. 2E and F). Taken together, the data from the REA assays suggest that, like PSC, Su(Z)2 inhibits ATP-dependent chromatin remodeling.
Su(Z)2 inhibits remodeling on mononucleosomes less efficiently than nucleosomal arrays.Previously, we observed that PSC inhibits the remodeling of mononucleosomes less efficiently than nucleosomal arrays (16). We compared the abilities of Su(Z)2 and PSC to utilize mononucleosomes as substrates and found that Su(Z)2 inhibits remodeling of mononucleosomes more effectively than PSC (Fig. 2G). However, inhibition on mononucleosomes is still less efficient than on 12-nucleosome arrays, with 50% inhibition at an approximate ratio of 2.5 to 0.5 Su(Z)2 protein per nucleosome, respectively (Table 1).
To determine if the low level of inhibition for remodeling the mononucleosomes reflects the weak abilities of PSC and Su(Z)2 to bind these substrates, nucleosome binding was monitored by EMSAs (Fig. 2H, top) in reactions used to test inhibition of chromatin remodeling (Fig. 2H, bottom). Both Su(Z)2 and PSC alter the mobility of mononucleosomes. At low protein concentrations that do not inhibit remodeling, a small shift in the mobility of mononucleosomes is observed. At higher concentrations of Su(Z)2 and PSC, a second slow-migrating species is observed. The appearance of the slow-migrating complex correlates with inhibition of chromatin remodeling for both Su(Z)2 and PSC, but the level of inhibition with Su(Z)2 is higher than that achieved at any concentration of PSC tested.
Su(Z)2 compacts chromatin but maintains nucleosome organization.The ability of Su(Z)2 to inhibit chromatin remodeling suggests that it might alter chromatin structure. To determine whether Su(Z)2 alters global nucleosome positioning or nuclease accessibility, we examined the effect of Su(Z)2 on MNase and DNase I digestion of six-nucleosome arrays. Nucleosomes are spaced on the arrays with repeating 5S positioning sequences, and digestion of the templates by nucleases results in a distinct ladder (Fig. 3A, lanes 1 to 4 and 9 to 12). Chromatin templates bound by Su(Z)2 showed a similar ladder after nuclease digestion (Fig. 3A, lanes 5 to 8 and 13 to 16), suggesting that Su(Z)2 does not grossly alter nucleosome organization.
(A) Su(Z)2 decreases the sensitivity of chromatin to MNase and increases its sensitivity to DNase I. Six-nucleosome templates were incubated with or without Su(Z)2 and treated with increasing concentrations of MNase or DNase I. Note the threefold difference in MNase and DNase I between samples with and without Su(Z)2. Lane M contains markers for nucleosome positions made by partial digestion of nucleosomal arrays with EcoRI, which cuts in the linker DNA between the nucleosomes. (B) Su(Z)2 compacts nucleosomal arrays in vitro. Representative electron micrographs of six-nucleosome templates alone (i to iii) or with a ratio of one Su(Z)2 protein to eight nucleosomes (nuc) (iv to vi), one Su(Z)2 protein to four nucleosomes (vii to ix), and one Su(Z)2 protein to two nucleosomes (x to xii). The scale bar in panel xii represents 100 nm and is for all images.
We compared the amounts of nuclease required for digestion in the presence or absence of Su(Z)2, and we found that three- to fivefold more MNase and three- to fivefold less DNase I were needed to digest Su(Z)2-bound chromatin templates than templates not bound by Su(Z)2 (Fig. 3A). Although we do not yet have an explanation for this observation, similar results were obtained with PSC (16). This modulation of sensitivity to nucleases is also consistent with Su(Z)2 and PSC altering the structure of chromatin.
PSC has been observed to induce the compaction of nucleosome templates in vitro (15). We examined Su(Z)2 to see if it can also compact chromatin. When visualized by electron microscopy, six-nucleosome templates showed a typical, extended, beads-on-a-string conformation (Fig. 3Bi to iii). However, increasing concentrations of Su(Z)2 cause progressive compaction of the templates (Fig. 3Biv to xii). At a ratio of one Su(Z)2 to eight nucleosomes (Fig. 3Biv to vi), some of the six nucleosomes are brought together, and at a higher ratio of one Su(Z)2 to two nucleosomes (Fig. 3Bx to xii), all six nucleosomes are compacted into a single particle. The size of some of these larger particles implies that they may contain more than one compacted array. Together, these results indicate that Su(Z)2, like PSC, can compact chromatin without globally changing the organization of nucleosomes.
The nonhomologous region of Su(Z)2 is sufficient for DNA binding and inhibition of chromatin remodeling.Su(Z)2 and PSC each have a small region of homology and a larger nonhomologous region (Fig. 1A). For PSC, these separate regions correspond to two functional regions: the homology region mediates complex formation, while the nonhomologous region has inhibitory effects on chromatin (19). We hypothesize that Su(Z)2 may be similarly divided into two functional regions. Therefore, we expressed and purified Su(Z)2 truncations corresponding to the homology region [Su(Z)2-N] and the nonhomologous region [Su(Z)2-C] (Fig. 4A) and tested their biochemical activities.
The C-terminal region of Su(Z)2 binds DNA and inhibits chromatin remodeling. (A) Schematic representation of Su(Z)2 truncations. Su(Z)2-N (amino acids 1 to 217) is the small N-terminal region that is homologous to PSC, and Su(Z)2-C (amino acids 218 to 1365) is the large C-terminal region that is not homologous but contains an amino acid composition that is similar to that of PSC. HR, homology region; NR, nonhomologous region. (B) FLAG-tagged Su(Z)2-N and Su(Z)2-C were expressed in Sf9 cells. Purified Su(Z)2-N and Su(Z)2-C (lanes 1 and 2, respectively) were resolved by 10% SDS-PAGE and stained with colloidal Coomassie blue. Asterisks indicate contaminants, as in Fig. 1B. (C) Su(Z)2-C but not Su(Z)2-N binds DNA. Representative filter binding experiment of Su(Z)2-N or Su(Z)2-C with DNA. The indicated protein concentrations are active concentrations for Su(Z)2-C but are total concentrations for Su(Z)2-N, because Su(Z)2-N is not active in DNA binding. On average, Su(Z)2-C is 22.5% ± 2.1% active. (D) Calculation of the apparent Kd of Su(Z)2-C and DNA (0.17 ± 0.05 nM), as in Fig. 1D. Data points are the average results (± SEM) of at least three experiments, with an R2 value of 0.94. (E) Su(Z)2-C but not Su(Z)2-N inhibits chromatin remodeling in REA assays (representative data). The indicated protein concentrations are total concentrations for both proteins. nuc, nucleosome. (F) Summary of REA assays. Percent inhibition is calculated as in Fig. 2C. Data points represent the average results (± SEM) of at least three experiments.
Using filter binding, we found that Su(Z)2-C but not Su(Z)2-N binds DNA, with average active fractions of 22.5% ± 2.1% and less than 1.0%, respectively. While Su(Z)2-N failed to bind significantly to DNA even at a concentration greater than 50 nM (Fig. 4C), the Kd measured for Su(Z)2-C is 0.17 ± 0.05 nM (Fig. 4D; Table 1). The affinity of Su(Z)2-C for DNA is about 10-fold higher than that of full-length Su(Z)2 and is similar to that of full-length PSC (Fig. 1D; Table 1).
We next tested if the truncations can inhibit chromatin remodeling in REA assays. Because Su(Z)2-N does not bind DNA, total concentrations were used for both proteins in order to compare their inhibitory activities. Su(Z)2-C but not Su(Z)2-N prevented Swi/Snf-mediated HhaI accessibility (Fig. 4E and F), with 50% inhibition occurring at 3 nM. For a comparison with full-length Su(Z)2, the Su(Z)2-C concentration was corrected for the fraction of active molecules. Using the active concentration, Su(Z)2-C inhibits chromatin remodeling at a ratio of 0.7 Su(Z)2-C per nucleosome (Table 1), which is very similar to the inhibition ratio of full-length Su(Z)2. Our data indicate that, like PSC, the large C-terminal nonhomologous region of Su(Z)2 is sufficient and necessary for DNA binding and inhibition of chromatin remodeling.
The nonhomologous region of Su(Z)2 is conserved among Drosophila, Anopheles, and Aedes species but is divergent from that of PSC.Our results indicate that PSC and Su(Z)2 share similar activities but, surprisingly, these activities are mediated by the nonconserved rather than the conserved regions of the proteins. Su(Z)2 and PSC likely arose from gene duplication, since they are adjacent. However, large portions of their sequences have diverged. To pursue how the two protein sequences have diverged, we compared each gene across the Drosophila, Anopheles, and Aedes species, as well as PSC and Su(Z)2 to each other across species. We obtained annotated Su(Z)2 and PSC protein sequences of 10 Drosophila, 1 Anopheles, and 1 Aedes species from the FlyBase and Ensembl databases (see Table S1 in the supplemental material), indicating that the two genes diverged at least before the order Diptera. When the Su(Z)2 protein sequences were aligned, the homology region showed a high level of conservation across the two genera, and the nonhomologous region shared segments that are also conserved, although not as much as that of the homology region (Fig. 5Ai). Our data suggest that the two protein regions are under different evolutionary constraints. PSC proteins across the Drosophila, Anopheles, and Aedes species showed similar patterns of conservation (Fig. 5Aii). However, when Su(Z)2 and PSC protein sequences were aligned together, we found that only the homology region is conserved, while the nonhomologous region is divergent except for single residues (Fig. 5Aiii). Of the 20 positions that are absolutely conserved, the majority are lysine (7/20), serine (5/20), and proline (4/20) residues.
The average amino acid compositions but not the primary protein sequences of Su(Z)2 and PSC are conserved. (A) The protein sequences of Su(Z)2 (i), PSC (ii), or both (iii) from across the Drosophila, Anopheles, and Aedes species were aligned, and the percent conservation at each amino acid position was plotted. The shaded portions correspond to the homology region of Su(Z)2 and PSC. (B) The amino acid compositions of the Su(Z)2 and PSC nonhomologous and homology regions (NR and HR, i and ii, respectively) were averaged across the Drosophila and Anopheles species and plotted against the average composition of Swiss-Prot proteins. For Su(Z)2, the Drosophila sechellias and one of three Aedes aegypti sequences (FBpp0202915 and AAEL011179-PA, respectively) contained fewer than 50 residues after the removal of the aligned HR and were therefore not included in the NR average. Two other A. aegypti sequences (AAEL011181-PA and AAEL015203-PA) contained fewer than 50 residues aligned to the HR and were accordingly not included in the HR average. Differentially shaded areas represent acidic (DE), basic (HKR), polar (CNQSTY), and nonpolar (AFGILMPVW) residues.
In addition, we found that the amino acid compositions of Su(Z)2 and PSC are conserved across the Drosophila, Anopheles, and Aedes species. On average, Su(Z)2 and PSC from the two genera deviate from the Swiss-Prot protein database's overall amino acid composition. The two proteins are enriched in lysine, proline, serine, and threonine but are deprived of aspartic acid, glutamic acid, tryptophan, and a number of other nonpolar residues. This specific amino acid composition was previously observed in Su(Z)2 and PSC from Drosophila melanogaster (5), and our analysis demonstrates that this particular composition is conserved through the Drosophila, Anopheles, and Aedes genera. Furthermore, this dramatic deviation from the protein database average is represented only in the nonhomologous region of the proteins (Fig. 5Bi), whereas the average results of the homology regions approach those of the protein database (Fig. 5Bii).
Su(Z)2 forms a functional complex with other PcG proteins.PSC is found in PRC1 with three other PcG proteins: Pc, Ph, and dRING1 (16, 43). We examined whether Su(Z)2 can form an analogous complex with these PcG components. To determine if Su(Z)2 can replace PSC in a PRC1-like complex, we coexpressed FLAG-Su(Z)2, Ph, Pc, and dRING1 in Sf9 cells and purified FLAG-Su(Z)2 and associated proteins. FLAG-Su(Z)2 was copurified with roughly stoichiometric amounts of the three other proteins (Fig. 6A), and Western blot analyses confirmed that all four proteins are present in the complex (Fig. 6B). We repeated the purification with FLAG-Ph instead of FLAG-Su(Z)2 and recovered an equivalent complex (Fig. 6A and B). The following biochemical characterization was performed using both FLAG-Su(Z)2 and FLAG-Ph complexes, and their activities were similar.
Su(Z)2 forms a complex with other PcG proteins. (A) Sf9 cells were infected with baculoviruses encoding Su(Z)2, Ph, Pc, and dRING1, with FLAG tags on either Su(Z)2 or Ph. Immunoaffinity-purified samples with FLAG-Ph and FLAG-Su(Z)2 (lanes 1 and 2, respectively) were resolved by 8% SDS-PAGE and stained with colloidal blue staining. The asterisk indicates a 70-kDa band that likely corresponds to HSC70. The intensity of each band was measured by densitometry and normalized to the molecular weight of each protein. Stoichiometry is indicated as a ratio to the FLAG-tagged component. (B) Western blots of the samples from panel A. Nuclear extract (input, I), flowthrough (F), and purified proteins (elution, E) were analyzed for the presence of each protein subunit by Western blotting. (C) The Su(Z)2 complex binds DNA. Representative filter binding experiment of the Su(Z)2 complex with DNA. (D) Calculation of the apparent Kd of the Su(Z)2 complex and DNA (1.06 ± 0.13 nM) as in Fig. 1D. Data points are the average results (± SEM) of at least three experiments, with an R2 value of 0.97. (E) The Su(Z)2 complex inhibits chromatin remodeling on nucleosomal arrays in REA assays (representative data). nuc, nucleosome. (F) Summary of REA assays on nucleosomal arrays. Percent inhibition is calculated as in Fig. 2C. Data points are the average results (± SEM) of at least three experiments. (G) The Su(Z)2 complex binds to mononucleosomes (top) but does not inhibit chromatin remodeling in REA assays (bottom) at complex-to-nucleosome ratios that completely inhibit remodeling of nucleosomal arrays. (H) Su(Z)2 coimmunoprecipitates with Ph, Pc, and dRING1 from nuclear extract from Drosophila heads. WB, Western blot; IP, immunoprecipitation.
First, we tested to see if the Su(Z)2 complex binds DNA using filter binding and found that it binds DNA with a Kd of 1.06 ± 0.13 nM (Fig. 6C and D). The average active fraction of the preparations is 22.3% ± 1.2% (Table 1). The Su(Z)2 complex thus has an affinity for DNA that is comparable to that of Su(Z)2 but is different from those of PSC and the PSC complex (Table 1). Notably, the PSC complex and PSC also have similar Kds (Table 1), suggesting that the affinity of each complex for DNA is mostly determined by either Su(Z)2 or PSC, the active subunits.
Next, we examined whether the Su(Z)2 complex can inhibit chromatin remodeling in REA assays. On nucleosomal arrays, increasing concentrations of the Su(Z)2 complex prevented Swi/Snf-mediated accessibility of HhaI (Fig. 6E), and 50% inhibition occurred at approximately 0.5 complex per nucleosome (Fig. 6F; Table 1). The complex thus has activity similar to Su(Z)2 (Fig. 2C; Table 1). The Su(Z)2 complex binds to mononucleosomes by an EMSA but does not inhibit their remodeling. We were unable to obtain Su(Z)2 complexes at high enough concentrations to compare their abilities to produce the low mobility complex and inhibition of mononucleosome remodeling observed at high concentrations of Su(Z)2 (Fig. 6G). Taken together, our results indicate that Su(Z)2 can form a functional complex with other PcG proteins and that this Su(Z)2 complex contains biochemical activities similar to those of Su(Z)2 alone.
To determine if Su(Z)2 might be found in a PRC1-like complex in vivo, we performed coimmunoprecipitation experiments on nuclear extracts from Drosophila heads and tissue culture S2 and Kc cells. We chose heads as our source of nuclear extract because Su(Z)2 is highly expressed in the brain and head, according to microarray expression data (10). Su(Z)2, Ph, Pc, and dRING1 were detected in immunoprecipitations from nuclear extracts from heads, S2 cells, and Kc cells by using antibodies with each of these proteins (Fig. 6H; see Fig. S2A in the supplemental material), suggesting that the four proteins may exist in a complex. We note that dRING1 is likely present at lower levels than the other PcG proteins in a nuclear extract from the head than in S2 or Kc cell nuclear extracts (see Fig. S2B in the supplemental material), suggesting that Su(Z)2 complexes lacking dRING1 may exist in head tissue.
The homology region of Su(Z)2 mediates complex formation.To test the hypothesis that the homology region of Su(Z)2 mediates interactions with other PcG proteins, we purified FLAG-Su(Z)2-N and FLAG-Su(Z)2-C from cells coexpressing Ph, Pc, and dRING1. Although all proteins were expressed in each case, a stable four-protein complex was obtained with Su(Z)2-N but not with Su(Z)2-C (Fig. 7A and B, lanes 1 and 2). Our results demonstrate that the small N-terminal homology region of Su(Z)2 is sufficient and necessary for complex formation.
The N-terminal homology region of Su(Z)2 is sufficient and necessary to mediate complex formation. (A) Insect Sf9 cells were infected with baculoviruses for FLAG-Su(Z)2-N, Ph, Pc, and dRING (lane 1), FLAG-Su(Z)2-C, Ph, Pc, and dRING1 (lane 2), or FLAG-Su(Z)2-N, Pc, and dRING1 (lane 3). Immunoaffinity-purified samples with FLAG-Su(Z)2-N and FLAG-Su(Z)2-C were resolved by 10% SDS-PAGE and stained with colloidal blue. The asterisk and dagger indicate protein bands that likely correspond to HSC70 and Ph degradation, respectively. (B) Western blots of samples from panel A. Nuclear extract (input, I), flowthrough (F), and purified proteins (elution, E) were analyzed for the presence of each protein subunit by Western blotting. (C) The Su(Z)2-N complex and the Su(Z)2-N ΔPh complex bind DNA with lower affinities than the Su(Z)2 full-length complex. Representative filter binding experiment of the Su(Z)2-N complex or the Su(Z)2-N ΔPh complex with DNA. The indicated protein concentrations are total concentrations, because both complexes are minimally active in DNA binding. (D) Calculation of the apparent Kd for the Su(Z)2-N complex and DNA (9.47 ± 1.46 nM in total concentration), as in Fig. 1D. Data points are the average results (± SEM) of at least three experiments, with an R2 value of 0.97. Data for the full-length Su(Z)2 complex from Fig. 6D were normalized to the total concentration and replotted for comparison. (E) The Su(Z)2-N and Su(Z)2-N ΔPh complexes inhibit chromatin remodeling on nucleosomal arrays in REA assays less efficiently than the full-length Su(Z)2 complex (representative data). The indicated protein concentrations are the total concentrations. nuc, nucleosome. (F) Summary of REA assays on nucleosomal arrays. Percent inhibition is calculated as in Fig. 2C. Data points are the average results (± SEM) of at least three experiments. Data for the full-length Su(Z)2 complex from Fig. 6G were normalized to total concentrations and replotted for comparison.
We next analyzed the Su(Z)2-N complex for its biochemical activities. Using filter binding, we found that the Su(Z)2-N complex was minimally active (2.2% ± 0.5%) in DNA binding (Table 1), and therefore, we carried out subsequent assays using the total concentration of the complex. The Su(Z)2-N complex has a measurable affinity for DNA, with a Kd of 9.47 ± 1.46 nM (Fig. 7C and D; Table 1). In REA assays on nucleosomal arrays, the Su(Z)2-N complex prevented chromatin remodeling (Fig. 7E), and 50% inhibition occurred at approximately seven complexes per nucleosome (Fig. 7F; Table 1). In both DNA binding and inhibition of chromatin remodeling, the Su(Z)2-N complex is about two- to fourfold less efficient than the full-length complex (Fig. 7D and F; Table 1). These results might seem unexpected, as Su(Z)2-N alone contains no biochemical activities (Fig. 4), but we previously obtained similar results with complexes containing truncations of PSC (19).
In the absence of the C-terminal region of Su(Z)2, other subunits such as Ph, which can inhibit chromatin remodeling, might contribute to the activities observed (16, 20). In the case of PSC complexes containing truncated PSC proteins, the removal of Ph reduced their activities (15, 19). We thus expressed and purified a three-protein complex containing Su(Z)2-N, Pc, and dRING1 (Fig. 7A and B, lane 3) and tested its activity. This Su(Z)2-N ΔPh complex was minimally active (1.5% ± 0.1%) in DNA binding (Table 1) and failed to bind significantly to DNA even at a concentration greater than 50 nM (Fig. 7C). In REA assays on nucleosomal arrays, the Su(Z)2-N ΔPh complex was less efficient than the Su(Z)2-N complex with Ph in inhibiting chromatin remodeling, suggesting that the presence of Ph in the complex contributes to its activity. The observation that the Su(Z)2-N ΔPh complex has more activity than Su(Z)2-N could mean that sequences outside the C-terminal region or in other subunits contribute to its activity. It is also possible that trace amounts of Sf9 cell PSC or Su(Z)2 copurify with the complex (Table 1). Nevertheless, our data indicate that the N-terminal region of Su(Z)2 is sufficient and necessary for interaction with other PcG proteins, while the C-terminal region is important for the inhibitory effects of the complex on chromatin.
DISCUSSION
We establish biochemically that Su(Z)2 and PSC are functional homologs. The two proteins have similar in vitro activities: DNA binding, chromatin compaction, and inhibition of chromatin remodeling (Fig. 1, 2, and 3). Furthermore, like PSC, Su(Z)2 can form a functional PRC1-like core complex with other PcG proteins (Fig. 6).
Su(Z)2 as a functional homolog of PSC.Our observation that Su(Z)2 and PSC have similar biochemical activities provides a possible explanation for why the two proteins can compensate for each other in some cases in vivo. However, the phenotypes of the Su(Z)2 and PSC mutations are complex and have differences. For example, PSC mutations often involve anterior-posterior transformations in the embryo, while Su(Z)2 mutations show only segmental defects but no signs of transformations. Furthermore, lethal embryos from PSC mutations display a distinctive abnormal head morphology, but only a small number of lethal embryos from Su(Z)2 mutations exhibit relatively minor head defects (2). Thus, the two proteins do not play identical roles in development, and we also observed small differences in their biochemistry, which might contribute to the differences in their mutant phenotypes. However, it is likely that other factors also contribute, as discussed below.
The ability of Su(Z)2 to form PRC1-like complexes might be important for the targeting of Su(Z)2. Interactions between Pc and histone H3 methylated at K27 and between Pc and Pleiohomeotic (Pho) have been implicated in targeting PRC1-class complexes (27, 28). These interactions might similarly target PRC1-like Su(Z)2 complexes. However, PSC and Su(Z)2 are likely to have some distinct targets. Polytene chromosome staining data indicate that while the two proteins colocalize to more than 50 sites, a larger number of binding sites are unique for either Su(Z)2 or PSC (36). Although it is possible that some sites of colocalization were not detected due to the limitations of polytene chromosome staining, the unique sites may contain loci whose lack of silencing in the Su(Z)2 or PSC mutant contributes to their different phenotypes. The mechanism for the differential targeting of PSC and Su(Z)2 could reflect the formation of complexes by either protein with additional proteins beyond the four core PcG components. PRC1 contains many other PcG and non-PcG subunits (38), and these factors may be important for in vivo targeting. Alternatively, differential targeting may be achieved by the varied abilities of the two proteins to interact with DNA and nucleosomes, which would predict that targets unique for Su(Z)2 or PSC have different arrangements of free DNA and nucleosomes.
Su(Z)2 and PSC may also function at different developmental times. Such differential deployment has been shown for other PcG members, namely, ESC and ESC-like, which are likely used alternatively in the PcG complex PRC2 during embryonic and postembryonic stages (23, 33, 50). Mutations in PSC are lethal during embryogenesis, whereas mutations in Su(Z)2 are lethal in the later larval and pupal stages (54), suggesting that Su(Z)2 might function at a later developmental phase. Microarray expression data also show that in embryos, PSC is highly expressed, but Su(Z)2 mRNA is virtually undetectable (34), whereas in adults, Su(Z)2 is abundantly expressed, but PSC mRNA could not be detected in multiple tissues (10). Because previous purifications of PcG complexes were performed with embryos, it is possible that a Su(Z)2 complex has been missed. Furthermore, a loss of Su(Z)2 significantly increases the likelihood of transdetermination in Drosophila imaginal discs (21), implicating the protein in restricting cellular plasticity later in development. One potential rationale for employing related but distinct proteins during different stages of development may be that they serve slightly different functions. PcG-mediated silencing can be removed by the targeting of a strong transcriptional activator during embryogenesis but not in larval stages (9), suggesting that the molecular mechanism of silencing may be altered during development. It is possible that the variations in biochemistry between Su(Z)2 and PSC and their differential deployment may contribute to an increased strength of silencing as the organism develops.
The nonhomologous region of Su(Z)2 is sufficient and necessary for inhibitory effects on chromatin.We showed that the nonhomologous region of Su(Z)2, like that of PSC, behaves almost identically to that of the full-length protein in our in vitro assays (Fig. 4). Our results are consistent with previous data, demonstrating that Su(Z)2 and its nonhomologous region behave as repressors of reporter gene constructs in mammalian cells when artificially recruited via fusion to the LexA DNA binding domain (6). Interestingly, the nonhomologous region alone is a stronger repressor in this context, and we observed that it has a 10-fold-higher affinity for DNA (Table 1).
In addition, our protein sequence alignments indicate that the nonhomologous regions of Su(Z)2 and PSC are divergent (Fig. 5). Even though we established that these regions of both proteins are sufficient for inhibitory effects on chromatin (Fig. 4), the question of how two divergent sequences can perform such a wide range of similar functions remains. One hypothesis is that the amino acid composition, rather than primary sequence, is important for interactions with chromatin. In particular, Su(Z)2 and PSC are enriched with lysine but are low in aspartic and glutamic acid contents. The basic natures of these two proteins (pI values of 9.41 and 9.77, respectively) might explain their high affinities but nonspecificities for DNA. Furthermore, key lysine residues are conserved between Su(Z)2 and PSC proteins across the Drosophila, Anopheles, and Aedes species, suggesting that charge interactions are important in Su(Z)2- or PSC-chromatin contacts.
An alternative but not mutually exclusive hypothesis is that the nonhomologous regions of Su(Z)2 and PSC are intrinsically disordered. Intrinsically disordered proteins often organize into well-folded three-dimensional structures when in contact with their substrates (13). The disordered regions of Su(Z)2 and PSC might gain similar three-dimensional shapes in the presence of chromatin, thereby providing them with similar activities. Intrinsic disorder can be predicted with high accuracy based solely on amino acid composition: proline and serine are the strongest disorder-producing residues, whereas tryptophan, phenylalanine, tyrosine, and cysteine have high order-producing tendencies (51). Average amino acid compositions of the nonhomologous regions of both Su(Z)2 and PSC are consistent with their being intrinsically disordered (Fig. 5). While this intrinsic disorder hypothesis might explain how two divergent protein sequences can have similar biochemical functions, their molecular mechanisms remain unclear.
In summary, our data establish Su(Z)2 as a functional homolog of PSC and demonstrate that both proteins have a conserved region that is important for their interactions with other PcG proteins and an unconserved region that is important for inhibitory effects on chromatin. Comparisons between the two divergent sequences with similar activities should facilitate the dissection of how these proteins exert their effects on chromatin in vitro and on gene silencing in vivo.
ACKNOWLEDGMENTS
We thank R. Jones, R. Kingston, J. Müller, and V. Pirotta for the antibodies and S. Kunes for the Drosophila strain. We also thank C. Woodcock for continued support with electron microscopy experiments, R. Kulathinal and A. Rizvi for help with sequence analysis, and T. Chu, J. Daniele, R. Islam, and S. Maniatis for guidance with experiments involving Drosophila. We are grateful to R. Emmons and T. Wu for insights and discussions on the genetics of PSC and Su(Z)2. We acknowledge N. Follmer, L. Kobrossy, B. Lengsfeld, and other members of the Francis laboratory for helpful comments on this project.
This work was supported by NIH (grant GM078456-01).
FOOTNOTES
- Received 3 July 2008.
- Returned for modification 29 July 2008.
- Accepted 22 October 2008.
- Accepted manuscript posted online 3 November 2008.
- Copyright © 2009 American Society for Microbiology