Analysis of a Polycomb Group Protein Defines Regions That Link Repressive Activity on Nucleosomal Templates to In Vivo Function

Abstract

Polycomb group (PcG) genes propagate patterns of transcriptional repression throughout development. The products of several such genes are part of Polycomb repressive complex 1 (PRC1), which inhibits chromatin remodeling and transcription in vitro. Genetic and biochemical studies suggest the product of the Posterior sex combs (Psc) gene plays a central role in both PcG-mediated gene repression in vivo and PRC1 activity in vitro. To dissect the relationship between the in vivo and in vitro activities of Psc, we identified the lesions associated with 11 genetically characterized Psc mutations and asked how the corresponding mutant proteins affect Psc activity on nucleosomal templates in vitro. Analysis of both single-mutant Psc proteins and recombinant complexes containing mutant protein revealed that Psc encodes at least two functions, complex formation and the inhibition of remodeling and transcription, which require different regions of the protein. There is an excellent correlation between the in vivo phenotypes of mutant Psc alleles and the structure and in vitro activities of the corresponding proteins, suggesting that the in vitro activities of PRC1 reflect essential functions of Psc in vivo.

The genes of the Polycomb group (PcG) and trithorax group (trxG) maintain the spatially restricted patterns of homeotic gene expression in the embryo. For the most part, PcG proteins propagate patterns of transcriptional repression, while trxG proteins maintain patterns of transcriptional activation (6, 20, 43, 56). Accordingly, mutations in PcG genes cause derepression of homeotic genes, and those in trxG genes fail to maintain transcriptional activation, both of which result in homeotic transformations. We are interested in understanding the mechanism by which PcG genes maintain silencing.

Genetic and biochemical studies have shown that PcG proteins function as part of multiprotein complexes that act to repress gene expression. Repression has been proposed to occur by creating a chromatin structure refractory to remodeling and transcription, by directly inhibiting the general transcription machinery, or by a combination of mechanisms (9, 15, 19, 22, 30, 36, 38, 40, 44, 47). Two functionally distinct PcG complexes have been purified from Drosophila embryos, ESC/E(Z) and Polycomb Repressive Complex 1 (PRC1). The ESC/E(Z) complex can methylate histone H3 at lysine 27 through the E(Z) SET domain (10, 13, 31, 41), and this methylation has been proposed to act as a molecular mark that targets other PcG complexes to the correct location (10, 11, 13, 31, 41). PRC1 contains stoichiometric amounts of the PcG proteins Polycomb (Pc), Polyhomeotic (Ph), dRing1 (dRing1/Sce) (24, 26), and Posterior sex combs (Psc) as well as a number of other proteins (51, 52). Unlike ESC/E(Z), PRC1 has no known enzymatic activities. However, in vitro studies show that it can inhibit both chromatin remodeling by the hSWI/SNF complex (22, 52) and transcription of a chromatin template by RNA Pol II (30). A PRC1 core complex (PCC) consisting of Pc, Ph, dRing1, and Psc can recapitulate the in vitro inhibition activities of PRC1, suggesting these four proteins confer the basic activities of PRC1. Remarkably, the inhibitory activities of this core complex can be recapitulated by high concentrations of Psc alone, suggesting that Psc plays a central role in mediating PCC function. Although Ph can also inhibit chromatin remodeling and transcription in our in vitro assays, it does so less efficiently than does Psc, suggesting that its contribution is minor compared to that of Psc.

Psc consists of 1,603 amino acids and contains an N-terminal region of 246 amino acids that is homologous to the products of the vertebrate proto-oncogene bmi-1 and the tumor suppressor gene mel-18, as well as the Drosophila genes Suppressor 2 of zeste [Su(z)2] and lethal (3)73Ah (7, 28, 59). This region of homology contains a C3HC4 Ring finger motif and a helix-turn-helix-turn-helix motif (HTHTH) and, in the case of Su(z)2, has been shown to be involved in locus-specific binding of polytene chromosomes (48, 53). Studies of bmi-1 and mel-18 have implicated the Ring finger in subnuclear localization and cellular transformation (12), and Ring finger motifs in other proteins have been demonstrated to function as E3 ubiquitin or SUMO ligases (18). The HTHTH motif has been implicated in skeletal transformation due to overexpression of Bmi-1 (3) and transcriptional repression by tethered Bmi-1 (12). Two-hybrid studies in yeast suggest that both of these protein motifs may be involved in mediating protein-protein interactions between Psc, Pc, and Ph (32); however, these interactions have not been addressed within the context of an intact complex. Outside of the N-terminal homology region, Psc has no recognizable structural motifs and no obvious homology to other proteins. However, the C-terminal regions of Psc and Su(z)2 have similar amino acid compositions, suggesting they may contain a conserved function that cannot be easily discerned from the amino acid sequence (7, 59). This function may involve transcriptional repression or protein-protein interactions, as has been suggested by transgene analyses of Su(z)2 (9). Interestingly, Psc appears to share functional similarities with Su(z)2 and lies adjacent to it in the Su(z)2 Complex [Su(z)2-C] (for example, see references 5, 45, 54, 57, and 62).

Psc has been genetically characterized, and 11 ethylmethane sulfonate (EMS)-induced alleles have been grouped into two phenotypic classes (62). The first is made up of six strong loss-of-function alleles (Psc¹, Psc^h27, Psc¹⁴⁴⁵, Psc^e25, Psc^e24, and Psc^e23) that are lethal when heterozygous with one another as well as with deficiencies for the locus. The second class consists of moderate loss-of-function alleles (Psc^h28, Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb) that show various degrees of viability when heterozygous with each other, strong loss-of-function alleles, or deficiencies. Based on several genetic assays, the two classes of alleles can be ordered into a phenotypic series progressing from most to least severe: Psc¹ > [Psc^h27, Psc¹⁴⁴⁵, Psc^e25] > Psc^e24 > Psc^e23 > Psc^h28 ≫ Psc^h30 ≫ [Psc¹⁴³³, Psc^e22, Psc^epb] (alleles within brackets cannot be ordered relative to each other). Three aspects of this series merit further discussion.

First, Psc¹ displays complex genetic behavior. Although considered loss-of-function with regard to viability, it also displays dominant and gain-of-function attributes (for example, see references 1, 29, 61, and 62). Second, Psc^h28 represents a transition point between the strong and moderate alleles (62). Like the moderate alleles, it shows variable viability when heterozygous with other moderate alleles and at least one strong allele, Psc^e23. On the other hand, it resembles the strong alleles by its lethality when heterozygous with deficiencies. Third, moderate alleles display intragenic complementation when heterozygous with certain members of the strong class (62). This complementation suggests that Psc contains at least two genetically separable functional domains.

To characterize the structure of Psc, we sequenced the EMS-induced alleles. Most contain nonsense mutations and collectively form a deletion series that reveals an excellent correlation with the phenotypic series. To determine the relationship between the in vitro inhibitory activities of Psc and PRC1 and their in vivo function, we characterized the mutant proteins with regard to their in vitro impact on chromatin remodeling and transcription, both as individual subunits and in the context of the PCC. Our studies delineate distinct functional regions within Psc that are important for its in vitro inhibitory activities as well as its in vivo functions, suggesting that the in vitro activities are a good indication of the role of Psc in vivo.

MATERIALS AND METHODS

Drosophila stocks.

Psc alleles and the marked chromosomes and stocks that carry them have been described previously (33, 42, 62). The CyO-19 green fluorescent protein (GFP)-marked balancer chromosome (CyO, P{w[+mC] =GAL4-Kr.C}DC3, P{w[+mC] = UAS-GFP.S65T}DC7) was obtained from the Bloomington Stock Center.

Molecular analysis of Psc alleles.

Psc alleles were kept in stock heterozygous with the CyO-19 balancer chromosome. To obtain genomic DNA for sequencing, 2-h egg lay collections were made from each heterozygous stock, aged 16 to 20 h, and then homozygous mutant embryos (non-GFP) were collected in groups of 25, placed in 1.5-ml Eppendorf tubes, and frozen at −80°C. Genomic preparations were made from each batch of embryos using the Berkeley Drosophila Genome Project protocol, and then PCR was performed using primer sets designed to amplify each exon as well as at least 50 bp of flanking intronic DNA. PCRs were gel purified and cleaned with the QIAGEN gel extraction kit, and then double strand sequence was collected across all exons via direct sequencing from the gel-purified fragments using the Dana Farber/Harvard Cancer Center sequencing center. Because of their different genetic backgrounds, the sequences of Psc¹ (GenBank DQ022536), Psc¹⁴³³ (DQ022537), and Psc¹⁴⁴⁵ (DQ022538) were compared to the Psc cDNA (GenBank X59275) (7), while the sequences of the remaining alleles (DQ022539 to DQ022546) were compared to a Psc cDNA compiled from genomic sequence (AE003820.3). These reference sequences contain polymorphisms, resulting in ∼20 amino acid differences between the encoded proteins. Furthermore, because the polymorphisms include in-frame deletions and insertions, the respective amino acid sequences do not align precisely. We describe all mutant proteins using the amino acid numbering associated with X59275, because this sequence was used in our bacculovirus vectors (see below). Psc¹, Psc¹⁴³³, and Psc¹⁴⁴⁵ share three amino acid polymorphisms (K519N, A545T, and S577A) not present in either reference sequence, while Psc¹⁴³³ and Psc¹⁴⁴⁵ also have a Q1167L polymorphism. Psc^h28 has two missense mutations (I762V and G1041E) located C-terminal to a stop codon, and several alleles have silent nucleotide changes (R.B.E., data not shown).

Western analyses.

Approximately 50 embryos collected from flies heterozygous or homozygous for Psc mutations were suspended in 6× sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (6% SDS, 400 mM Tris [pH 6.8], 30% glycerol, 30 mM EDTA, 0.0012% bromphenol blue) and boiled for 5 min before loading on an 8% SDS-PAGE gel. Blots of proteins from Psc¹⁴⁴⁵ homozygous embryos were stained with Amido Black (Sigma) after blotting. Psc was visualized using a mixture of two monoclonal anti-Psc antibodies, 6E8 and 7E10, which were a gift from P. N. Adler (37, 54).

Protein expression and purification.

Mutagenesis was carried out using the QuickChange in vitro mutagenesis kit (Stratagene). The cDNA sequence for Psc was inserted into the pFastBac vector for bacculovirus and mutagenized directly (Psc¹, Psc^h27, Psc^e24) or replaced with segments of the cDNA that had been subcloned into pBluescript and mutagenized (all other alleles). All mutations were verified by sequencing, and the Psc^e23 mutation was verified further by directly sequencing bacculoviral DNA. A FLAG tag was placed at the N termini of Psc proteins as indicated below. Psc and PCC proteins were expressed in Sf9 cells and purified as described previously (22). All purifications used buffers with 10 μM ZnCl₂. Washes contained 0.05% NP-40. Proteins were eluted in BC300 with 0.4 mg/ml FLAG peptide and 10 μM ZnCl₂.

Functional assays.

Wild-type and mutant Psc and PCC were assayed for inhibition of chromatin remodeling as previously described (22). Reactions contained between 0.01 ng and 1 ng end-labeled 5S array and 9 to 10 ng unlabeled 5S array templates. Psc and PCC were diluted in BC300 with 1 mg/ml bovine serum albumin and incubated with the template for 20 min before addition of hSwi/Snf. The assays for inhibition of transcription were carried out as previously described (30), using 10 ng of 5S array templates.

Analyses of Psc proteins in wing imaginal disks.

cDNA fragments encoding PSC^1-572, PSC^1-872, or PSC^456-1603 were cloned into the CaSpeR-hs vector to obtain hs-Psc^1-572, hs-Psc^1-872, and hs-Psc^456-1603 (plasmid maps are available from J.M. on request), and several independent transformant lines were isolated for each construct. Transgenic lines carrying a hs-Psc construct have been described previously (5). To determine the ability of the transgenes to repress homeotic gene activity in clones of cells homozygous for Su(z)2^1.b8, a deficiency that removes Psc and Su(z)2 (5, 8), the following strains were made: FRT42 Su(z)2^1.b8/CyO; hs-Psc, FRT42 Su(z)2^1.b8/CyO; hs-Psc^1-572, FRT42 Su(z)2^1.b8/CyO; hs-Psc^1-872, and FRT42 Su(z)2^1.b8/CyO; hs-Psc^456-1603. Marked clones of Su(z)2^1.b8 mutant cells were then generated by crossing flies from these strains to hs-flp; FRT42 hs-nGFP and inducing clones during the first instar stage. Clones were analyzed 96 h after induction by staining with antibodies against Abd-B and GFP as previously described (5). Rescue function of the hs-Psc transgenes was analyzed by applying 1 h of heat shocks every 12 h over a 96-h period, starting at the time of clone induction. In animals carrying the hs-Psc or the hs-Psc^1-572 transgene, Abd-B was tightly silenced in all Su(z)2^1.b8 clones in every animal analyzed. In animals carrying the hs-Psc^1-872 or hs-Psc^456-1603 transgene, misexpression of Abd-B in Su(z)2^1.b8 mutant clones was indistinguishable from the misexpression observed in Su(z)2^1.b8 mutant clones in animals lacking a hs-Psc transgene. At least two independent transgene insertions were analyzed with each construct and found to give the same result.

Expression of wild-type and mutant forms of Psc in embryos was induced as follows. For Fig. 7B, transgenic embryos were heat shocked five times (20-min heat shocks at 37°C every 2 h; first heat shock applied between 2 and 4 h of development) prior to fixation and stained with antibody against Abd-B. For Fig. 7C, a single 30′ heat shock at 37°C was applied, and total embryo extracts were prepared for analysis 30 min after completion of the heat shock.

FIG. 7.

Silencing function of wild-type and mutant Psc proteins in vivo. (A) Wing imaginal disks immunostained for Abd-B (red) and GFP (green), which serves as a marker for wild-type cells. Abd-B is not expressed in wild-type cells (top left) but is strongly misexpressed in clones lacking Psc and Su(z)2 (marked by lack of GFP signal) in the absence of a Psc transgene (no TG); the mutant clones also display a tumor-like phenotype (5). Transgenes expressing wild-type Psc or PSC^1-872 restore silencing of Abd-B and prevent the tumor-like phenotype in the mutant clones. In comparison, PSC^1-572 and PSC^456-1603 are ineffective. Psc⁻ Su(z)2⁻ mutant clones sort out from the surrounding wild-type tissue and form vesicles; in the “no TG,” “hs-Psc^1-572,” and “hs-PSC^456-1603” images, all cells in each of the clones express Abd-B, but not all cells are visible in this single confocal section. (B) Ectopic expression of Abd-B in some cells in the nervous system (arrows) (not all cells showing ectopic expression are visible in this focal plane) when PSC^1-572, but not wild-type Psc, PSC^1-872, or PSC^456-1603, is overexpressed in embryos. Note that under the conditions used to express Psc proteins in imaginal disks in the experiment shown in Fig. 7A (see Materials and Methods), expression of PSC^1-572 does not result in ectopic expression of Abd-B protein in wild-type (GFP-positive) imaginal disk cells. (C) Expression of mutant forms of Psc in transgenic embryos. Western blot analysis of embryonic extracts (EE) from transgenic embryos expressing PSC^1-572, PSC^1-872, and PSC^456-1603, respectively. Embryos were not heat shocked (−hs) or were heat shocked (+hs) for 30 min and then were incubated for another 30 min prior to extract preparation. In each case, a Psc protein signal of the expected size is present only in extracts from heat-shocked embryos; the corresponding recombinant Psc proteins (RP), purified from baculovirus, were loaded as size references. The blot was stripped and reprobed with an antibody against alpha-tubulin to demonstrate comparable loading of extracts. We could not detect the endogenous wild-type Psc protein, suggesting that transgene-encoded Psc proteins are expressed at significantly higher levels.

RESULTS

Structures of the Psc alleles correlate with their phenotypic classifications.

We identified the molecular lesions associated with 11 EMS-induced alleles of Psc: Psc¹, Psc^h27, Psc¹⁴⁴⁵, Psc^e25, Psc^e24, Psc^e23, Psc^h28, Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb (33, 42, 62). Genomic DNA was isolated from homozygous mutant embryos, amplified by PCR, sequenced directly, and analyzed as described in Materials and Methods. Most alleles contained nonsense mutations predicted to truncate the Psc protein. Eight encode simple truncations: Psc^h27 (K111*), Psc¹ (E521*), Psc^h28 (K760*), Psc^h30 (Q910*), Psc¹⁴³³ (Q1075*), Psc^e22 (Q1098*), Psc^epb (Q1098*), and Psc¹⁴⁴⁵ (Q1453*). Although Psc^e22 and Psc^epb encode the same nonsense mutation, they likely represent independent mutational events, since they were derived from separate mutageneses (62). Psc¹, Psc^h28, Psc¹⁴³³, and Psc¹⁴⁴⁵ have additional amino acid changes that represent either polymorphisms or changes that are not translated (Materials and Methods). One allele, Psc^e24 (M373K and E378*), contains both missense and nonsense mutations, and another, Psc^e23 (C268Y), carries a missense mutation in the C3HC4 ring finger domain. Finally, Psc^e25 had no changes in the coding region. These results are summarized in Fig. 1A.

FIG. 1.

Molecular analysis of mutant Psc proteins. (A) Sequence analysis. Diagrams depict wild-type Psc (black), proteins of the moderate alleles (red), strong alleles (green), and engineered deletion constructs (white). Products of strong alleles that are likely associated with expression defects are shown in light green, and asterisks indicate locations of missense amino acid changes. The homology region (gray shading), Ring finger (dark stripes), and HTHTH motifs (light stripes) are indicated. (B) Western analysis of animals heterozygous for mutant Psc alleles. (C) Western analysis of embryos homozygous for Psc¹⁴⁴⁵ or Psc^e23. Equal protein loading was confirmed by Amido Black staining (for Psc¹⁴⁴⁵) or by Western blotting using an antibody to TATA binding protein (TBP) (for Psc^e23; Santa Cruz). (D and E) Recombinant Psc proteins corresponding to mutant alleles (∼400 ng) (D) or engineered deletions (∼500 ng) (E) were separated on 4 to 15%-gradient SDS-PAGE gels and stained with colloidal Coomassie. An asterisk (D and E) indicates contaminating heat shock protein 70 (Hsp70).

To verify the sequencing results, Psc^mutant/+ embryos were collected for each mutant allele, and total embryonic protein was examined by Western analysis using anti-Psc monoclonal antibodies (37). The Western blots showed that embryos heterozygous for seven of the alleles (Psc^e24, Psc¹, Psc^h28, Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb) expressed altered proteins consistent with the size predicted by the sequence data (Fig. 1B). Altered bands were not seen for the remaining four alleles, Psc^h27, Psc¹⁴⁴⁵, Psc^e23, and Psc^e25 (Fig. 1B). This observation was not surprising for Psc^h27, even though Psc^h27 is transcriptionally competent (2) and is expected to produce a truncated protein, since its corresponding protein is predicted to lack the homology region used to generate the anti-Psc monoclonal antibodies (37). Altered bands were also not expected for Psc¹⁴⁴⁵, Psc^e23, and Psc^e25, since they were predicted to make proteins of wild-type, or nearly wild-type, size. Our findings are in agreement with those of previous studies of protein sizes for Psc¹, Psc¹⁴⁴³, and Psc¹⁴⁴⁵ (37).

To assess expression of Psc¹⁴⁴⁵, Psc^e23, and Psc^e25, homozygous mutant embryos were collected for each allele and analyzed by Western blotting. Homozygous Psc^e23 embryos expressed approximately normal levels of a protein of the size expected for full-length Psc (Fig. 1C), suggesting that the phenotype of Psc^e23 results from the alteration in its Ring finger motif. In contrast, embryos homozygous for either Psc¹⁴⁴⁵ or Psc^e25 did not express detectable levels of Psc (Fig. 1C), although recovery of protein from Psc^e25 embryos was too low to determine conclusively whether this allele produces a protein (data not shown). We suggest that the phenotypes of Psc¹⁴⁴⁵ and Psc^e25 might be due to lack of synthesis or instability of the protein. Association of Psc^e25 with a posttranslational defect is especially likely, since Psc^e25 is transcriptionally competent (2).

The deletion series formed by the following truncation alleles, Psc^h27, Psc^e24, Psc¹, Psc^h28, Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb, shows an excellent correlation to the phenotypic series in which the strong loss-of-function alleles encode the shortest proteins, while the moderate alleles encode the longest. Notably, the transition between the moderate and strong truncation alleles occurs at amino acid 760, which is the truncation point of the transition allele Psc^h28. All of the remaining moderate alleles have nonsense mutations located after this amino acid. With the exception of Psc¹⁴⁴⁵ and Psc^e25, which do not express detectable levels of protein in vivo, all of the strong alleles have mutations located in the N-terminal 521 amino acids of the protein. Thus, the different classes of Psc alleles are associated with distinct mutant protein structures.

A central region of Psc is important for in vitro inhibition of both chromatin remodeling and transcription.

To determine how mutations in Psc affect its in vitro activities as a single subunit, we expressed mutant Psc proteins using a baculovirus expression system, purified them by immunoaffinity chromatography via the FLAG epitope that was engineered at the N terminus of each protein (Fig. 1D), and then determined their ability to inhibit both chromatin remodeling and transcription. We tested mutant proteins corresponding to the following eight alleles: Psc¹, Psc^h27, Psc^e24, Psc^e23, Psc^h28, Psc^h30, Psc¹⁴³³, and Psc^epb. With the exception of Psc^h27, whose expression pattern is not known, each of these alleles expressed mutant protein in embryos (Fig. 1B and C). Psc^e22 was not tested because the mutant protein it produces is predicted to be identical to that encoded by Psc^epb, and neither Psc¹⁴⁴⁵ nor Psc^e25 was tested because they do not express detectable levels of protein in vivo.

First, we tested whether the mutant proteins could inhibit chromatin remodeling by the human SWI/SNF complex by using a restriction enzyme accessibility assay (22). hSWI/SNF makes a HhaI site within a nucleosomal template accessible to digestion in an ATP-dependent manner, thereby providing a quantitative assay for chromatin remodeling (35), and PRC1, PCC, and Psc are all able to inhibit this remodeling reaction (22). End-labeled nucleosomal templates were preincubated with purified mutant Psc proteins for 20 min, and then HhaI, hSWI/SNF, and ATP were added. Reactions were incubated for 1 h, at which point the extent of HhaI digestion of the nucleosomal template was measured by resolving end-labeled products on an agarose gel and quantifying the proportion of undigested template (Fig. 2A).

FIG. 2.

In vitro activities of mutant Psc proteins. (A) Inhibition of chromatin remodeling by mutant Psc. End-labeled nucleosomal arrays were incubated with 0 nM, 4 nM, 8 nM, 16 nM, or 32 nM Psc, and the extent of chromatin remodeling by hSwi/Snf was determined by quantifying the increase in HhaI accessibility. (The first and third lanes of each panel demonstrate the extent of chromatin remodeling in the absence of Psc.) Lanes with no hSwi/Snf contain 0 nM and 32 nM Psc. (B) Summary of inhibition of chromatin remodeling by mutant Psc. Percent inhibition was calculated as described previously (22):

Bars represent the mean activity at 16 nM Psc; values were compared by Student's t test, and asterisks indicate values that are statistically different from wild-type values (P < 0.01). Note that titrations were carried out for all proteins in this figure and Fig. 5; the bar graph shows a concentration in the linear range for activity of wild-type Psc. (C) Comparison of inhibition of chromatin remodeling by Psc proteins representing the different phenotypic classes. (D) Inhibition of transcription by mutant Psc. Nucleosomal array templates were incubated with 0 nM, 6 nM, 12 nM, 24 nM, or 48 nM Psc before addition of Gal4-VP16, HeLa nuclear extract, and nucleoside triphosphates. (E) Summary of inhibition of transcription by mutant Psc. Results are presented as percentages of transcription activity in the absence of Psc. Bars represent the mean activity at 12 nM Psc; values were compared by Student's t test, and asterisks indicate values that are statistically different from wild-type values (P < 0.01). (F) Comparison of inhibition of transcription by Psc proteins representing the different phenotypic classes. All error bars represent standard errors (n ≥ 2).

Mutant Psc proteins representing the simple truncations can be divided into two groups, those that inhibit chromatin remodeling and those that do not. Mutant proteins corresponding to the moderate alleles Psc^h30, Psc¹⁴³³, and Psc^epb had inhibition activities indistinguishable from that of the wild type (Fig. 2A to C). In contrast, those corresponding to the strong alleles Psc^h27 and Psc^e24 had significantly lesser abilities to inhibit chromatin remodeling (Fig. 2A to C). Interestingly, mutant protein corresponding to the Psc^h28 transition allele was unable to inhibit chromatin remodeling, highlighting the similarity between Psc^h28 and the strong alleles. Mutant Psc¹ protein also failed to inhibit remodeling, suggesting that it behaves like proteins representing the strong loss-of-function alleles. Finally, mutant Psc^e23 protein displayed wild-type inhibition activity, indicating that an intact Ring finger motif is not essential for in vitro inhibition of chromatin remodeling.

Second, we tested the ability of mutant Psc proteins to inhibit transcription. Mutant proteins were incubated with a chromatin template, and then the Gal4-VP16 transcriptional activator, HeLa nuclear extract, and nucleoside triphosphates were added. Transcription levels in each reaction were analyzed by performing primer extension on the reaction product and quantifying the resulting products as described previously (30).

The activities of the mutant proteins paralleled those observed in our assays of chromatin remodeling. Mutant proteins corresponding to the moderate alleles Psc^h30, Psc¹⁴³³, and Psc^epb again had inhibition activities that were not statistically distinguishable from that of the wild type (Fig. 2D to F), while mutant proteins corresponding to the strong alleles Psc^h27 and Psc^e24, as well as the Psc^h28 transition allele and Psc¹, were significantly impaired in their ability to inhibit transcription of a chromatin template (Fig. 2D to F). Once again, the Psc^e23 protein had an activity similar to that of wild-type Psc (Fig. 2D to F). These observations suggest that with the exception of Psc^e23, there is a straightforward correlation between the in vitro inhibition activities of the truncation alleles and their phenotypic class (Fig. 2C and F).

As part of a parallel study of Psc activity, we made three deletion constructs of the Psc cDNA (Fig. 1B), each of which was engineered to have a FLAG epitope at the N terminus. Two encode C-terminal deletions, called PSC^1-872 and PSC^1-572, which consist of the first 872 and 572 amino acids of Psc, respectively. The third encodes an N-terminal deletion, PSC^456-1603, which contains amino acids 456 to 1603. The construct breakpoints were chosen prior to our sequence analysis of the Psc alleles, so it was fortuitous that the breakpoints of the first two constructs happened to flank the nonsense mutation of the transition allele Psc^h28.

We expressed and purified these proteins (Fig. 1E) and then tested their abilities to inhibit chromatin remodeling (Fig. 3A and B) and transcription of a chromatin template (Fig. 3C). PSC^1-872 had activities that were mildly reduced relative to those of wild-type Psc in both assays, while the activities of PSC^1-572 were severely reduced (Fig. 3A to C). PSC^456-1603 was also able to inhibit chromatin remodeling and transcription with activities similar to those of wild-type Psc (Fig. 3A to C), indicating that the sequences C-terminal to 456 are sufficient for in vitro inhibitory activities. These data suggest that a portion of Psc that is important for full inhibition lies within or overlaps the region between amino acids 456 and 872. Since PSC^1-572 and the protein corresponding to Psc^h28 (K760*) are also unable to support significant inhibition, the region between amino acids 760 and 872 might be of particular importance.

FIG. 3.

In vitro activity of mutant Psc proteins of engineered deletion constructs. (A) Inhibition of chromatin remodeling by mutant Psc (concentrations as indicated in panel B). All lanes are from the same experiment and are arranged so that identical concentrations of Psc are vertically aligned. The titration of PSC^1-572 was extended to concentrations higher than that for the other proteins because of the lower activity of PSC^1-572. (B) Summary of inhibition of chromatin remodeling by mutant Psc. Standard error is shown. (C) Inhibition of transcription by mutant Psc (6, 12, 24, 48 nM).

Mutant Psc proteins can alter in vitro activity of the PCC.

Since Psc functions as part of a multiprotein complex (23, 49, 51, 52), we expected that mutant Psc proteins would affect the structure and/or activity of the complex. Accordingly, we used a recombinant PCC consisting of Pc, Ph, dRing1, and mutant Psc to test individual mutant Psc proteins for their ability to allow the core complex to form and inhibit chromatin remodeling. Insect cells were coinfected with four baculovirus constructs designed to express the mutant Psc protein (without FLAG epitope tags), wild-type Pc and dRing1, and Ph carrying an N-terminal FLAG epitope tag. FLAG-Ph and associated proteins were purified by immunoaffinity chromatography from nuclear extracts prepared from the coinfected cells. Since the Psc^h27 protein cannot be detected with available antibodies, we used a FLAG-tagged Psc^h27 protein (and Ph without the FLAG tag) in experiments with this protein.

We found that mutant Psc proteins corresponding to the moderate alleles Psc^h30, Psc¹⁴³³, and Psc^epb, the transition allele Psc^h28, and the strong alleles Psc¹, Psc^e24, and Psc^e23 all allowed complex formation (Fig. 4A). To confirm that these mutant core complexes had stoichiometric amounts of each of the four subunits, we examined their composition by colloidal Coomassie staining and Western analysis (Fig. 4A and B). PCC subunits were found at approximately equal levels in each of the mutant complexes (Fig. 4A and B). Mutant protein corresponding to Psc^h27 was impaired for complex formation: Pc, dRing1, and Ph copurified with the FLAG-tagged Psc^h27 protein in substoichiometric amounts. This defect in complex formation might have been more severe than observed, because apparent read-through of the nonsense codon at position 111 produced small amounts of full-length FLAG-tagged Psc (data not shown). While it is possible that the FLAG tag on the Psc^h27 protein contributed to poor complex formation, we have not previously seen adverse effects of this tag on formation of PCC recombinant complexes. We also found that while PSC^1-872 and PSC^1-572 allowed complex formation, PSC^456-1603 did not (Fig. 4C, D, and E). PSC^456-1603 was tested for complex formation by coinfecting FLAG-tagged PSC^456-1603 with Ph, Pc, and dRing1 and then identifying the proteins that copurified with FLAG-PSC^456-1603. Decreased amounts of Ph, dRing1, and Pc were seen in the eluates (Fig. 4E, lane E). Taken together, these data indicate that efficient complex formation can be achieved with the N-terminal 572 amino acids of Psc. Assuming that complex formation with the Psc^e24 protein does not require the missense mutation at amino acid position 373, our data suggest that complex formation may be possible with as little as the N-terminal 378 amino acids. This would be consistent with the inability of PSC^456-1603 to support complex formation.

FIG. 4.

Formation of PCC with mutant Psc proteins. (A) Approximately 800 ng PCC containing mutant Psc proteins was separated on a 4 to 15%-gradient SDS-PAGE gel and stained with colloidal Coomassie. The identity of the Psc subunit of the complex is indicated [e.g., PCC (e23) for PCC containing Psc^e23 protein]. Positions of mutant Psc (circles) and other subunits are indicated. *, degradation product of Ph; **, Hsp70. (B) Western blots to assess relative stoichiometry of PCC subunits in complexes formed with mutant Psc (1 pmol in blots with anti-Pc or anti-dRing1; 100 fmol in blots with anti-Psc or anti-FLAG). PCC containing the Psc^e24 or Psc¹ protein was probed with the 6E8 antibody, while all others were probed with the 7E10 antibody. Because equal amounts of protein were loaded in each lane, intensities of bands for each antibody should be similar in all lanes if PCC subunit stoichiometries are similar in all complexes. The lower region of the gel probed with 7E10 is not shown because it was probed with another antibody. (C) PCC (1 μg) carrying mutant Psc proteins was separated on a 4 to 15%-gradient SDS-PAGE gel and stained with colloidal Coomassie. (D) Two hundred nanograms PCC carrying mutant Psc proteins was separated on 8% SDS-PAGE gels, and the Western blots were probed with antibodies against PCC subunits or with an anti-FLAG antibody to detect Psc. (E) Insect cells were infected with viruses encoding FLAG-tagged PSC^456-1603 and other PCC subunits, and immunoaffinity purification was performed on nuclear extracts from infected cells. Input (I), flow-through (FT), and elution (E) samples were analyzed for the presence of PCC subunits by Western blotting.

Core complexes containing mutant Psc proteins were then tested for their ability to inhibit chromatin remodeling and transcription in vitro. These assays are complicated, because PCC subunits other than Psc, such as Ph, might contribute to the activity of a complex containing a mutant Psc protein (22, 30). Nevertheless, as detailed below, we found that some mutant Psc proteins compromised PCC activity.

In comparison to the ability of wild-type PCC to inhibit chromatin remodeling, we found that the activities of core complex preparations containing proteins of the moderate alleles Psc^h30, Psc¹⁴³³, and Psc^epb were indistinguishable from that of the wild type, and those of preparations containing proteins of the transition allele Psc^h28 or the strong alleles Psc¹ and Psc^e24 were significantly impaired (Fig. 5A to C). We also found that complexes prepared with PSC^1-872 were similar to the wild type, while those containing PSC^1-572 were relatively ineffective at inhibiting chromatin remodeling (Fig. 6A and B). Interestingly, although complexes containing the Psc¹ or PSC^1-572 proteins had reduced activity, they still inhibited chromatin remodeling better than did either of these proteins alone (compare Fig. 2B to 5B and 3B to 6B). This suggests either that the context of a complex enhanced the activity of mutant Psc subunits or that other PCC subunits, such as Ph, contributed activity in our assays (22, 30). Consistent with the latter possibility, removing Ph from PCC carrying PSC^1-572 or Psc^e24 protein further reduces the activity of the complex (N.J.F., unpublished result). Finally, whereas the Psc^e23 protein was similar to the wild type when assayed as a single subunit (Fig. 2A to C), core complex preparations containing this mutant protein were less active than preparations of wild-type PCC (Fig. 5A to C). This reduced activity of complexes formed with the Psc^e23 protein might partially explain the strong loss-of-function phenotype of Psc^e23 in vivo.

FIG. 5.

In vitro activities of PCC carrying mutant Psc proteins. (A) Inhibition of chromatin remodeling by mutant PCC. PCC (0, 5, 10, 20, or 40 nM) carrying wild-type or mutant Psc was incubated with end-labeled nucleosomal array and then tested for chromatin remodeling as described in the legend to Fig. 2A. (B) Summary of inhibition of chromatin remodeling by wild-type and mutant PCC at 12 nM; activity and statistical comparisons were calculated as in Fig. 2B. (C) Comparison of inhibition of chromatin remodeling by PCC carrying Psc proteins representing the different phenotypic classes. (D) Inhibition of transcription by mutant PCC (0, 1.7, 3.5, 7, and 14 nM) was assayed as described in Fig. 2D. (E) Summary of inhibition of transcription by wild-type and mutant PCC at 7 nM. Results are presented as a percentage of transcription activity in the absence of PCC in each reaction, and statistical comparisons were calculated as described for Fig. 2E. (F) Comparison of inhibition of transcription by PCC carrying Psc proteins representing the different phenotypic classes. Standard errors are shown (n = 2 to 8).

FIG. 6.

In vitro activities of PCC carrying mutant Psc proteins of engineered deletion constructs. (A) Inhibition of chromatin remodeling by mutant PCC (0.5 to 32 nM; see panel B). All lanes are from the same experiment and are arranged so that identical concentrations of PCC are vertically aligned. The titration of PCC carrying PSC^1-572 was extended to concentrations higher than those for the other PCC preparations because of the lower activity of PCC containing PSC^1-572. (B) Summary of inhibition of chromatin remodeling assays by mutant PCC. Standard errors are shown. (C) Inhibition of transcription by mutant PCC (6, 16, 48 nM).

We next measured the ability of PCC containing mutant Psc proteins to inhibit transcription and found a pattern of activities similar to those found in our assays of chromatin remodeling, although there was more overlap among the allele classes in this assay. Here, complexes containing protein corresponding to the moderate allele Psc^epb but not Psc^h30 or Psc¹⁴³³ were less active than those containing wild-type Psc. Because the protein encoded by this allele was active in all of the other assays, the importance of this reduced activity is unclear. Complexes containing proteins corresponding to Psc^h28, Psc¹, or Psc^e24 were less active than those containing wild-type Psc, although the difference was significant only for Psc^e24 (Fig. 5D to F). Our data further indicated that preparations containing PSC^1-572 were impaired for inhibition of transcription, while complexes containing PSC^1-872 were able to inhibit transcription (Fig. 6C). Finally, preparations containing mutant Psc^e23 proteins inhibited transcription to a level comparable to that of preparations containing proteins corresponding to the moderate alleles (Fig. 5D to F). Thus, while these results are generally consistent with the normal function of moderate alleles and impaired function of strong truncation alleles observed in other assays, they highlight the possibility that there are differences among the members within each class (see Discussion).

N-terminal 872-amino-acid fragment of Psc can maintain homeotic gene repression.

Maintenance of homeotic gene repression is a key function of the PcG genes, and several of the Psc alleles described above fail to maintain repression (5, 55, 57). Here, we use transgenic technology to determine the capacity of the construct-derived PSC^1-872, PSC^1-572, and PSC^456-1603 proteins to maintain homeotic gene repression in vivo. Our strategy rested on previous studies showing that absence of Psc and Su(z)2 in clones of cells in the wing imaginal disk results in the derepression of homeotic genes, including Abd-B. Importantly, this derepression can be prevented by expression of wild-type Psc and Su(z)2 (5) or by expression of wild-type Psc alone (Fig. 7A). Using Abd-B expression levels as our assay for repression, we therefore induced clones lacking Psc and Su(z)2 in wing imaginal disks and determined whether transgenes expressing PSC^1-872, PSC^1-572, or PSC^456-1603 would be able to maintain homeotic gene repression (see Materials and Methods).

We found that while PSC^1-872 restored repression to wing disks lacking Psc and Su(z)2, PSC^1-572 and PSC^456-1603 could not (Fig. 7A). All three proteins were expressed at comparable levels as measured in embryos (Fig. 7C). These data suggest that a function necessary for the repression of Abd-B in wing imaginal cells in these assays is encoded within the N-terminal 872 amino acids of Psc and that a region of Psc that is essential for this function is located C-terminal to amino acid 572. The failure of PSC^456-1603 to maintain repression might have resulted from the absence of a region important for repression, from an inability to support complex formation, or from an inability to appropriately target. Interestingly, in embryos, overexpression of PSC^1-572, but not of wild-type Psc, PSC^1-872, or PSC^456-1603, led to ectopic activation of Abd-B expression in some cells in the nervous system, even in the presence of wild-type Psc (Fig. 7B). This suggests that PSC^1-572 may act in a dominant-negative manner to interfere with the repression of homeotic genes.

DISCUSSION

Our molecular analysis of the genetically characterized alleles of Psc identifies at least two regions important for the function of this key PcG protein. Eight of the alleles, Psc^h27, Psc^e24, Psc¹, Psc^h28, Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb, form a deletion series spanning most of the protein (Fig. 1). We find a direct correlation between the lengths of the proteins encoded by these mutations and their abilities to inhibit chromatin remodeling and transcription in vitro as single subunits: mutant proteins 910 amino acids in length or longer display activity similar to that of the wild type in both assays, while those that are shorter than, or equal to, 760 amino acids in length support minimal activity (Fig. 2 and 5). Additional analysis of the mutant proteins derived from three deletion constructs of Psc, PSC^1-872, PSC^1-572, and PSC^456-1603, is consistent with these data and highlights the importance of a region lying between amino acids 456 and 872 (Fig. 3 and 6). Thus, a critical aspect of the in vitro inhibitory activities of Psc may lie, at least in part, between amino acids 760 and 872. We also identified a region of Psc that is important for formation of the core complex, PCC. This region is located N-terminal to amino acid 572 and may be contained in as little as the N-terminal 378 amino acids (Fig. 4).

Our in vivo clonal analyses are consistent with these in vitro observations (Fig. 7). The ability of PSC^1-872, but not PSC^1-572 or PSC^456-1603, to repress Abd-B in wing imaginal disks points to two regions of Psc, one lying in part between amino acids 572 and 872 and the other N-terminal to amino acid 456, that are important for in vivo maintenance of Abd-B repression. The location of these domains is consistent with the in vitro roles of the region between amino acids 760 and 872 and the region N-terminal to amino acid 572 for in vitro inhibition activities and complex formation, respectively.

The in vitro behavior of Psc mutant proteins as single subunits is generally indicative of their affect on PCC activity. Complexes containing mutant Psc proteins that are active as single subunits in vitro are generally more active than complexes containing mutant proteins that are only minimally active as single subunits. Nevertheless, the capacity of the latter class of complexes to support activity at all is significant and suggests either that the residual activity of minimally active Psc proteins can be enhanced in the context of a complex, perhaps through interactions with other subunits, and/or that subunits other than Psc can contribute to PCC activity. With regard to the latter possibility, we have found that Ph inhibits chromatin remodeling and transcription as a single subunit, although less efficiently than does Psc (22, 30), and that removal of Ph from complexes that contain PSC^1-572 or the Psc^e24 protein reduces activity (N.J.F., unpublished).

The ability of the Psc^e23 protein to inhibit both chromatin remodeling and transcription in vitro demonstrates that an intact Ring finger is not essential for these activities (Fig. 2). This is consistent with the ability of PSC^456-1603, which lacks the Ring finger altogether, to support in vitro inhibition activities at wild-type levels (Fig. 3 and 6). In contrast, incorporation of the Psc^e23 protein into PCC decreases in vitro inhibition activities, suggesting that in the context of the complex the Ring finger becomes important in vitro (Fig. 5). This diminished activity of PCC containing the Psc^e23 protein is intriguing in light of our observation that as a single subunit, the Psc^e23 protein approximates wild-type Psc activity in in vitro assays even though in vivo Psc^e23 is a loss-of-function, recessive lethal allele; the lethality of Psc^e23 may rest in part on disrupting normal intermolecular interactions, such as those within PRC1. Alternatively, it is possible that the Psc^e23 mutation, which disrupts a conserved cysteine in the Ring finger, disrupts a E3 ubiquitin or SUMO ligase activity, although such an activity has yet to be demonstrated for Psc or its homologues.

Recently, a separate study showed that PCC has the ability to compact chromatin, probably through the bridging of nucleosomes (21). Interestingly, this activity can also be carried out by Psc alone. Some of the mutant proteins studied here were also characterized for their ability to inhibit compaction, and those that decreased compaction also decreased inhibition of chromatin remodeling and transcription in vitro. Thus, PSC^1-572 is impaired for chromatin compacting activity, while PSC^1-872 and PSC^456-1603 behave as does full-length Psc. One interpretation of these observations suggests that the internucleosomal interactions leading to chromatin compaction contribute to the inhibition of chromatin remodeling and transcription. Given that the mutations corresponding to the mutant proteins we have characterized also impair Psc function in vivo, it is possible that similar interactions with chromatin are important for the gene silencing function of Psc in vivo.

Our assays using purified and recombinant components also suggest that several functions of Psc do not require histone modification or variants. In contrast, recent data suggest that histone modifications, histone-modifying enzymes, and histone variants might be involved in silencing by PRC1 (for example, see references 16, 17, 58, and 60). For example, complexes containing PRC1 components, particularly dRing1 and its mammalian homologues, can ubiquitylate H2A (60), although this has not yet been demonstrated with either purified PRC1 or PCC. It will be interesting to determine how these modifications might alter Psc targeting or efficiency in vivo.

In vivo genetic behavior of mutant Psc alleles suggests additional functional domains.

The parallel between the in vivo behavior of mutant Psc alleles and the in vitro activities of the corresponding mutant proteins is striking. Mutant proteins displaying significant in vitro activity correspond to four alleles of the moderate class (Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb), and those that are essentially inactive correspond to alleles of the strong class (Psc¹, Psc^h27, and Psc^e24). Furthermore, protein corresponding to the transitional Psc^h28 allele behaves in vitro as would protein corresponding to the strong class of alleles and falls, by length, at the interface between proteins of the moderate class and those of the strong class (Fig. 1). These observations suggest that the in vitro assays serve as a reasonable proxy for in vivo Psc function. Below, we extend our analysis by considering our data in light of prior genetic analyses.

Our data indicate that the region of Psc lying C-terminal to amino acid 872 is essential neither for in vitro activity nor for the maintenance of HOX gene silencing in our imaginal disk assays. Nevertheless, the four moderate alleles whose protein products are expected to be truncated at or beyond amino acid 910 (Psc^h30, Psc¹⁴³³, Psc^e22, and Psc^epb) show a loss of viability when hemizygous in vivo (62). In addition, the reduction in viability may correlate with the length of the proteins they encode. The hemizygous viabilities of Psc^e22, Psc^epb, Psc¹⁴³³, Psc^h30, and Psc^h28 are 30 to 55%, 15 to 33%, 6 to 9%, 1 to 2%, and 0% of the wild-type level, respectively. While some of these differences in viability may be attributed to variations in genetic background, the differences between Psc¹⁴³³, Psc^h30, and Psc^h28 likely reflect real changes in Psc function, since these three alleles differ by other genetic criteria as well (see below) (62).

The progressive loss of viability might be due to an increasing instability of the Psc protein as its length decreases. This interpretation, however, is not easily compatible with the observation that moderate alleles have gain-of-function characteristics and support intragenic complementation (see below). Another possibility is that the region of Psc C-terminal to amino acid 760 encodes multiple functions, which are differentially deleted by the moderate alleles. Alternatively, there may be a single key function distributed throughout the C-terminal half of Psc such that mere quantity of the region can influence function. Nonlocalized activity, which is consistent with the amino acid composition of the region, has also been proposed as a mode of action for the analogous region of Su(z)2 (53). In light of these possibilities, the reduced in vitro activities of proteins representing the moderate alleles may be a true reflection of the reduced viability associated with these alleles.

A second feature of the moderate alleles that substantiates the importance of the C-terminal end of Psc is the homeotic transformation of wing to halter tissue in adults that are hemizygous for a moderate allele or heterozygous for two such alleles (62). This phenotype suggests that the region of Psc lying C-terminal to amino acid 1098 provides an activity that is essential for normal development. Since the homeotic phenotype is gain of function, possibly antimorphic, it may be that moderate allele proteins give rise to an activity that is normally suppressed, directly or indirectly, by the C-terminal end of Psc. The possibility that one domain of Psc influences another is consistent with previous genetic analyses of Psc as well as Su(z)2 (48, 53, 62).

A third observation concerns the zeste gene, whose product is implicated in maintenance of gene activation and repression and is present in PRC1 (for example, see references 14, 27, 39, and 51). Here, we focus on z¹, a mutation of zeste that represses expression of the white gene (46), and the ability of Psc alleles to directly or indirectly alter this repression (61, 62). All six strong alleles suppress the z¹ phenotype, indicating that wild-type Psc promotes repression of the white gene by z¹. Since Psc^h28 does not modify the z¹ phenotype, it seems that the region of Psc lying between amino acids 521 and 760 is important for repression by z¹. In contrast, all moderate alleles except the transition allele Psc^h28 enhance the z¹ phenotype and indicate that loss of material beyond amino acid 1098 directly or indirectly augments repression of the white gene by z¹. Furthermore, since Psc^h30 enhances silencing while Psc^h28 does not, enhancement may require the region between amino acids 760 and 910. It may be that this region confers gain-of-function silencing activities that are normally kept in check by the region lying C-terminal to amino acid 1098. The potential gain-of-function aspect of enhancement is consistent with the gain-of-function homeotic activity of the moderate alleles.

Intriguingly, Psc¹ is a strong dominant suppressor of the z¹ phenotype and appears to antagonize wild-type Psc in a Psc¹/+ animal (61, 62). This antimorphic nature may derive from abnormal activity of the homology region, which is expected to be intact in the product of Psc¹ but disrupted in or lacking from that of Psc^e24, Psc^e23, and Psc^h27, none of which displays a similar dominant phenotype. If so, our data suggest that amino acids 521 through 760 keep the homology region in check. This interpretation is consistent with the in vivo dominant-negative nature of PSC^1-572 and studies suggesting that the homology region of Su(z)2 may be modulated by a region lying C-terminal to it (53). Our data may also pertain to one of the most intriguing aspects of z¹, which is that its ability to silence white is enhanced when white is physically paired with a homologue (43). As such, it is possible that the regions identified here as important for z¹-mediated silencing might include those that contribute to homologue pairing in vivo. (Also see reference 34.)

Finally, our understanding of Psc may be further clarified by observations of intragenic complementation (62). The most dramatic example involves Psc^e23 and Psc^h30. Heterozygosity of Psc^e23 or Psc^h30 with a deficiency for the Su(z)2-C results in 0% or 1 to 2% viability, respectively. In contrast, Psc^e23/Psc^h30 animals show 100% viability. Psc^e23 also supports complementation with Psc^h28, although at a much-reduced rate; while Psc^h28 shows 0% viability when heterozygous with a deficiency for Su(z)2-C, Psc^e23/Psc^h28 animals show 3 to 15% viability. These different levels of complementation highlight the significance of the region lying between amino acids 760 and 910.

What is the mechanism of this complementation? It may be that the presence of two Psc mutant proteins results in two correspondingly different PRC1 complexes, which together provide all essential PRC1 function; the mutant Psc^e23 protein could provide C-terminal functions, while the mutant Psc^h30 and Psc^h28 proteins could provide N-terminal activity, especially with regard to Ring finger function. Alternatively, the presence of two mutant Psc proteins within a single PRC1 complex, possibly as a heterodimer, might allow each to compensate for the other's deficiencies. It is not known whether a single PRC1 complex contains one, two, or even more copies of Psc, since only relative stoichiometries, not absolute numbers, have been determined for the proteins in complexes isolated from Drosophila (51). Evidence that Psc has the capacity to dimerize comes from two-hybrid studies with Psc, Pc, and Ph (32), as well as studies of Mel-18 and Bmi-1 (4, 25, 50). These studies suggest that the Ring finger may play a role in dimerization. However, since Psc^e23 carries a mutation in the Ring finger, if complementation involves heterodimerization, then either the Psc^e23 mutation does not preclude heterodimerization or heterodimerization occurs through another means, perhaps through the HTHTH motif or via another protein, such as dRing1 (34).

In short, our findings emphasize the importance of the C-terminal two-thirds of Psc in defining Psc function in vivo and in vitro and the N-terminal region in complex formation in vitro. These studies show that the C-terminal region plays a key role in inhibition of chromatin remodeling and transcription on nucleosomal templates and suggest that these inhibitory functions in vitro might relate to in vivo function of the protein. It is intriguing that this region of Psc is not conserved in the mammalian homologs Bmi-1 and Mel-18 and does not show obvious similarity to any protein other than that of the related Su(z)2 gene. This suggests that the presumed function of the C-terminal portion of Psc in maintaining silenced states during development is novel or, perhaps, is supplied in other organisms by proteins or domains that have yet to be identified.