Polycomb group protein complexes exchange rapidly in living Drosophila

The activity of genes in eukaryotic organisms is modulated by non-histone chromatin-binding proteins. Both global and local chromatin states are transmitted from one cell generation to another(Grewal and Moazed, 2003; Phair et al., 2004; Vermaak et al., 2003). However, data are lacking on the lifetime of chromatin-bound protein complexes, such that the actual mechanisms by which they exert their activation or repression are not understood. Early in Drosophiladevelopment, a particular expression pattern of homeobox-containing (HOX)genes is established along the anteroposterior axis of the organism(Lewis, 1978) by transcriptional activators and repressors whose presence is only temporary and encoded by gap and pair-rule genes (Gaul and Jäckle, 1989; Tautz,1988), whereas maintenance of the expression pattern is obligatory throughout development. Disturbance of the pattern leads to a switch of the determined state and homeotic transformations of the organism(Bienz and Muller, 1995; Busturia and Morata, 1988; Garcia-Bellido et al., 1976; Lewis, 1963).

The Polycomb group (PcG) and Trithorax group (trxG) of proteins are chromatin-binding proteins responsible for conserving the transcriptional state of the HOX genes and thereby cell identity. PcG proteins are responsible for the persistence of silencing whereas the trxG proteins are required for transcription in the active domains(Francis and Kingston, 2001; Levine et al., 2004; Orlando, 2003). PcG proteins are targeted to particular regions of the genome called Polycomb response elements (PREs) (Chan et al.,1994; Orlando et al.,1998; Strutt et al.,1997) where they act in multicomponent complexes to repress transcription of their target genes. The continued presence of PcG proteins on the PREs throughout development is required for silencing since deletion of the PRE (Busturia et al.,1997) or individual PcG genes(Beuchle et al., 2001) anytime during development of the organism results in gene derepression. Interestingly, although PcG complexes maintain the repression pattern for up to 10 cell generations most of the PcG protein complement dissociates at every mitosis (Buchenau et al.,1998).

There exist experimental data for the association of the PcG proteins with specific chromatin sequences, including the first observations by immunofluorescence on polytene chromosomes(Chiang et al., 1995; Franke et al., 1992; Rastelli et al., 1993). In vivo crosslinking and chromatin immunoprecipitation (ChIP analysis) of PcG proteins have preferentially detected high levels of proteins of the PCC(Polycomb core complex), and recently, also of Pleiohomeotic (Pho) and Enhancer of zeste [E(Z)], on PREs and promoters of known homeobox genes(Breiling et al., 2004; Ringrose et al., 2003; Strutt and Paro, 1997; Wang et al., 2004). Several models have been proposed for the mechanism of PcG-mediated repression, such as (1) heterochromatinization or formation of a closed chromatin conformation that does not allow access to promoters; (2) inhibition of the assembly of the preinitiation transcription complex; and (3) interference with transcription initiation and/or elongation (Min et al.,2003; Paro and Hogness,1991; Simon and Tamkun,2002). Experimental evidence can be found to support each of the models. For example, PcG complexes reduced accessibility for large RNA polymerases over large stretches of DNA in the bithorax homeobox gene cluster(BX-C) (Fitzgerald and Bender,2001), thereby inhibiting transcription of reporter genes,although restriction enzymes retained DNA access. However, the presence of PcG proteins at the Ubx promoter in wing imaginal discs(Wang et al., 2004) lends support to a direct inhibition of transcription, although perhaps only at the elongation rather than at the initiation step as has been suggested for the heat shock protein 26 (hsp26) promoter(Dellino et al., 2004).

Two different multiprotein polycomb repression complexes (PRCs) have been isolated and characterized biochemically. PRC2(Ng et al., 2000) is composed of the PcG proteins, Extra sex combs (Esc), Suppressor (12) of zeste [Su(z)12]and histone-binding Nurf-55 and Enhancer of zeste [E(Z)], the latter of which methylates histone H3 at lysine 27 both in vivo and in vitro(Cao et al., 2002; Czermin et al., 2002; Kuzmichev et al., 2002; Müller et al., 2002; Yamamoto et al., 2004), thus marking nucleosomes for assembly of repression complexes. PRC1, which contains equimolar quantities of Polycomb (Pc), Polyhomeotic (Ph), Posterior sex combs(Psc) and Sex combs extra (Sce/dRing1), all of which have been shown to be essential for PcG silencing. Other PcG and non-PcG proteins such as Sex combs on midleg (Scm), heat-shock protein cognate 4 (Hsc4) and Zeste (Z), and some transcription factors have been isolated with PRC1 in non-stoichiometric amounts, implying the presence of more than one type of polycomb repression complex (Levine et al., 2002; Mulholland et al., 2003; Saurin et al., 2001).

Whether the in vitro isolated or assembled complexes represent truly competent repression machineries is a matter of debate, as will be discussed later. In vivo data imply that functional complexes are assembled sequentially, directly and with a particular hierarchy, on the chromatin itself (Buchenau et al., 1998; Wang et al., 2004) and single PcG gene deficiencies such as E(Z) result in loss of complex formation on PREs, although all of the proteins involved in the PCC or PRC1 are still present (Rastelli et al.,1993; Wang et al.,2004). For a complete understanding of the repression mechanism,we need to know the stability and lifetime of functional repression complexes in the living organism. Recently, it was reported that Polycomb can be competed away from genomic sites by methylated histone tail peptides in permeabilized salivary gland nuclei(Ringrose et al., 2004). However, no data have been available about binding equilibria and dissociation rate constants of any multiprotein PcG chromatin complex in vivo. In this study, we addressed this problem by performing photobleaching experiments(fluorescence recovery after photobleaching, FRAP) on GFP fusion proteins of Polycomb (Pc) and Polyhomeotic (Ph), two essential members of the PCC in whole living Drosophila embryos and larval tissues to determine their diffusion, binding equilibria and residence times. We measured these values in living organisms at different stages of development to determine whether there are changes in the stability of the complexes. By taking advantage of the polytene nature of the salivary gland chromosomes, we assessed the uniformity of the complexes between individual bands. The actual forward and reverse rate constants for complex formation were determined. Most of the complexes exchange within a period of 1 minute and all of the complexes in less than 10 minutes. We discuss the compatibility of these data with present models for repression and draw inferences about the homogeneity of the repression complex.

The Ph (proximal) gene was fused to green fluorescent protein (PhGFP) using the strategy described in Netter et al.(Netter et al., 2001). It was cloned with both a UAS and a Pc promoter to provide controlled expression at different stages of fly development. The predicted Pc promoter (Neural Network Promoter prediction program of the Berkeley Drosophila Genome Project) was isolated from a PstI fragment of the Pc genomic clone (kind gift of Jürg Müller)(Paro and Hogness, 1991) by amplification of a 559 bp fragment using the primers(5′-TTTAGATCTCAATTTGTGATACAATAAGTG-3′ and 5′-CCCGAGCTCATCTTAGCAAGTAGCCGTGTC-3′) and inserted as an EcoRI fragment upstream of the Ph protein-coding sequence(kind gift from Jürg Müller). The resulted fusion was inserted as a BglII-NotI fragment into the pUAST vector(Brand and Perrimon, 1993). Transgenic lines containing the construct P[UAS,Pc:PhGFP] were generated with standard transformation protocols using the w¹¹¹⁸ host line(Spradling and Rubin, 1982)and the site of chromosome integration was determined genetically. The Pc promoter alone drives expression of PhGFP in the salivary gland nuclei. Expression of PhGFP in embryos and larval wing imaginal discs was induced with the Gal4 drivers listed below.

The following strains were used in this study: w¹¹¹⁸;P{pPc-PcGFP,w⁺}; w¹¹¹⁸; P{UAS,Pc-PhGFP,w⁺}; yw; P{en2.4-GAL4}e22c/SM5 (to drive expression of phGFP in embryos);

P{Gal4;w⁺}Bx^MS1096 (to drive expression of phGFP in wing imaginal discs) where it drives the expression of Gal4 in the whole wing blade (Capdevila and Guerrero, 1994).

All strains were maintained on standard corn-agar medium at 18°C and experiments were carried out at 25°C. The PcGFP stock was kindly provided by R. Paro (Dietzel et al.,1999) and the en:Gal4 and Bx^MS1096:Gal4 (Milan et al., 1998) drivers were provided by H. Jäckle.

For live imaging dechorionated embryos were transferred to a chamber with a coverslip bottom (LabTek) in oxygenated Tyrode's buffer (135 mM NaCl, 10 mM KCl, 0.4 mM MgCl₂, 1 mM CaCl₂, 5.6 mM glucose, 10 mM HEPES, pH 7.2). In order to prevent movement and buffer evaporation, they were covered with a polycarbonate membrane with 8 μm pores that allowed oxygen exchange (Nucleopore). Larval imaginal and salivary gland tissues were dissected in PBS and immediately transferred to similar chambers and covered with a Whatman 3M filter paper soaked in Tyrode's buffer. Imaging was performed at 21°C for a maximum of 2 hours after mounting with a 63×N.A. 1.2 water immersion objective using an inverted Zeiss LSM 510META microscope. GFP was excited with the 488 nm line of an Ar ion laser and emission collected between 505 and 545 nm with a pinhole equivalent to 2 Airy discs.

FRAP images in somatic cell nuclei were performed with an XY sampling of 0.10 μm/pixel and in polyploid salivary gland nuclei, at 0.14 μm/pixel. Photobleaching was carried out for ∼200 ms (FRAP in salivary gland nuclei)at ∼200 μW laser power (measured through the objective). Pre-bleach and post-bleach images were acquired at high scanning speed with minimal laser intensity (AOTF 2%, ∼5 μW). At later measurement times (after frame 20), the interval between scans was increased in order to reduce bleaching during monitoring.

Three dimensional inverse FRAP (3D-iFRAP) experiments in embryos and imaginal discs were carried out by bleaching the whole nucleus except for a small region surrounding a fluorescent locus of interest for ∼4 seconds. A time series of seven confocal z sections (with 5 second intervals between stacks for PhGFP and 10 seconds for PcGFP) were recorded for ∼120 seconds after bleaching. The time stacks were aligned using a 3D tracking algorithm (View5D information can be found at http://wwwuser.gwdg.de/∼rheintz/View5D/)developed in this laboratory. After alignment, the spot intensity was calculated using a weighted region of interest.

In FRAP experiments on bands of PcG proteins in salivary gland nuclei, the free signal was estimated by averaging the intensity in each frame near the spot in a region as defined by the lowest 30% voxels of the sum intensity projection over all aligned pre-bleach frames (see Fig. 7). This nucleoplasmic signal of the free protein estimated frame by frame was subtracted from each pixel and the total bound protein was determined as the sum of all pixels in the mask region of the 30% brightest pixels in the projection over all aligned pre-bleach frames.

Crude extracts were prepared from embryos of different developmental stages and larval tissues using lysis buffer [20 mM HEPES-KOH pH 7.5, 100 mM KCl, 2 mM EDTA, 0.5% Triton X-100, 0.3 U/ml aprotinin, 10 μg/ml leupeptin, 100μg/ml soy bean trypsin inhibitor, protease inhibitor cocktail tablets(Roche Diagnostics), 5 mM DTT, 1 mM MgAc₂]. Proteins were separated on NuPage 4-12% Bis-Tris or NuPage 3-8% Tris-acetate polyacrylamide gels, and western blots were probed with primary polyclonal anti-Pc (kind gift of Renato Paro), anti-Ph or anti-GFP and HRP-conjugated secondary antibodies by chemiluminescence (Amersham Pharmacia Biotech). PABP(Roy et al., 2004), S6 (Santa Cruz Biotechnology) and eIF4A(Hernández et al.,2004) antibodies were used as loading controls. The intensities of the signals on the x-ray films were quantified on a scanning densitometer(G-710, BioRad). Optical density values were extracted and normalized to the loading controls indicated in Table 1.

As the original construct of Netter et al.(Netter et al., 2001)containing only the UAS promoter caused overexpression in the salivary gland nuclei, we chose to construct a phGFP expression vector with both a Pc and a UAS promoter(Brand and Perrimon, 1993) to better control the expression of PhGFP in embryos and larvae. The Pcpromoter drives PhGFP expression in salivary gland nuclei at physiological levels but not in embryos and larval diploid tissues. Therefore, we used several Gal4 drivers to induce PhGFP expression in these tissues. The level of Ph is controlled by feedback inhibition of the Ph promoter(Fauvarque et al., 1995). The PhGFP protein was able to rescue the ph⁵⁰⁴-null homozygous mutant and, thus, is totally competent in repression. PcGFP has been functionally characterized (Dietzel et al., 1999). Expression of PcGFP rescued alleles with a mutation in the Pc chromodomain (Pc^XL5), the domain of the protein that is essential for targeting the PcG complex by binding trimethylated lysine 27 of histone H3. In addition, the PcGFP protein binds to the same polytene chromosome loci as wild-type Pc (data not shown). However, mutations in the C-terminal region of the protein or null mutations (Pc²,Pc³) were not rescued. It is possible that the repression complex cannot accommodate all Pc proteins containing the GFP fusion moiety(see below) or that some other hitherto unknown aspect of Pc protein function is impaired (Dietzel et al.,1999).

Quantitation of western blots of imaginal discs or salivary glands from wild-type and transgenic flies revealed slightly lower levels of GFP-labeled proteins than the endogenous ones, such that the ratio of total Pc protein in transgenics (including PcGFP) was only 1.6 times that in wild type and the ratio of total Ph (including PhGFP) in transgenics was 1.7 times that of wild type (Fig. 1A,B; Table 1). PhGFP expression in wing imaginal discs was induced by Bx^MS1096:Gal4 driver that induces expression in the whole wing blade. The wing blade represents 60-70% of the wing disc. Therefore, the amount of PhGFP expressed per nucleus is comparable to the amount of total untagged Ph per nucleus. There was a change in the relative expression of the proximal to the distal Phgenes in the transgenic line as seen in Fig. 1A. The relative expression levels of PcGFP and PhGFP in transgenic salivary glands determined by western blotting using an anti-GFP antibody was 1 to 2.2(Fig. 1C). In the salivary gland nuclei both PcGFP and PhGFP expressions are induced by the Pcpromoter (no Gal4 driver used in this case).

Before cellularization in Drosophila, the PcG proteins are all of maternal origin and their binding to chromatin is restricted to a few PRE sites (Orlando et al., 1998). It is not clear if the repression complexes formed are functional as zygotic transcription has not yet begun. Thus, there exists a window in development(early division cycles) in which one can measure the diffusion of the fusion proteins by classical FRAP techniques. In the preblastoderm embryos the distribution of PcGFP is rather homogeneous throughout the nucleus(Dietzel et al., 1999) and the nuclear size is large relative to somatic diploid nuclei later in development(Fig. 2A). At cycle 10, we observed a few faint aggregates of PcGFP in a uniform fluorescent nucleoplasm. We measured the diffusion constant in regions without aggregations by conventional FRAP. The number and intensity of the PcGFP aggregated loci increases as embryonic development proceeds and as nuclei decrease in size. Another development stage providing access to the free protein is in the larval salivary gland nuclei, where the chromatin is condensed in polytene chromosomes leaving regions of free nucleoplasm. From FRAP experiments in both early embryos (Fig. 2) and salivary gland nucleoplasm (not shown), we obtained similar diffusion constants for PcGFP of 0.74 μm² second^–1 and 0.41 μm² second^–1, respectively(Table 2). The amount of PhGFP induced in early pre-blastoderm embryos was insufficient for obtaining reproducible FRAP measurements. Thus, the free diffusion constant was derived exclusively from salivary gland nuclei. The value, 0.22 μm²second^–1, is only twice as slow as that of PcGFP(Table 2) as would be expected for a protein three times larger than Pc. The values for both proteins are slower than expected for free, monomeric, diffusing proteins(Verkman, 2002), indicating that the proteins may interact non-specifically in the nucleus with histones or other chromatin-bound proteins, although no specific binding to PREs occurs at this stage.

Our primary objective was to determine rate constants for PcG protein complexes in living embryos and tissues of Drosophila during stages of development when the complexes are biologically competent. Whole-mount embryos are viable and can proceed through the entire embryonic development process, i.e. through larval stages and pupation. Explanted imaginal discs cultured in vitro can develop and undergo metamorphosis by addition of hormones (Sengel and Mandaron,1969). Thus, both embryos and imaginal discs meet our criteria for living tissue. Genetic loci are locally represented in up to 2000 copies in salivary glands, a differentiated tissue exclusive to larvae, and provide us with the possibility of observing and measuring the stability of complexes that bind to single or few PRE loci. Pc and Ph bind to overlapping target genes on these polytene chromosomes(Franke et al., 1992). After the mid-blastula transition aggregates of PcG proteins appear in all 2N nuclei of the embryo. The number of such aggregates of Ph and Pc increases during embryogenesis such that, by stage 14, over 100 such loci can be distinguished in fixed whole-mount embryos (a number similar to the number of bands observed on polytene chromosomes) (Buchenau et al.,1998). The number of distinct loci in imaginal discs is an order of magnitude smaller than in embryos, suggesting that these loci are composed of higher-order aggregates (D.J.A.-J., unpublished). The distributions of the en:GAL4 induced PhGFP and Pc promoter induced PcGFP expressions in post-cellularization stages of embryogenesis and in wing imaginal discs are shown in Fig. 1D-K.

As shown in the following sections, the recovery times of the PcGFP and PhGFP complexes were at least an order of magnitude longer than the free diffusion of the macromolecules. Thus, it was necessary to measure for more than 50 seconds in order to reach equilibrium between the redistribution of bleached and unbleached proteins. A preferred method for determining dissociation rate constants under such conditions is to use inverse FRAP(iFRAP), whereby the entire nucleus (except for a small region surrounding the fluorescent complex of interest) is bleached and the depletion of fluorescence from this region is monitored over time. The rate of disappearance of the fluorescent locus will be a direct measure of the first order dissociation rate constant of the protein from the complex(Dundr et al., 2002). We found that this type of photobleaching technique fitted our system best due to the reasons described below. Nuclei and chromatin itself are not stationary in live Drosophila tissues, as shown in Fig. 3. Core histone-GFP that does not dissociate from chromatin in interphase showed similar dynamics,indicating that the movement we observe in our cells is not due to dissociation of whole complexes from the chromatin(Post et al., 2005) (see Movie 1 in the supplementary material). In addition, photobleaching of the nuclear lamin fused to RFP in embryonic and 2N larval disc Drosophila nuclei revealed no rotation of the nuclei over a period of more than 3 minutes (C. Fritsch, personal communication). In order to overcome the problem of chromatin mobility, the dissociation and residence times of the PcG chromatin-bound proteins were analyzed in three dimensions by adapting the iFRAP procedure to a version denoted 3D-iFRAP that tracks the fluorescent locus over time (see Materials and methods). The fluorescence decay of the unbleached locus and the increase in fluorescence in the bleached nucleus were monitored in the movement corrected data(Fig. 4) and the average intensities of the locus of interest were plotted over time to derive the rate constants (Fig. 5).

3D-iFRAP was used to analyze the residence times and dissociation rate constants of PcG fusion proteins in embryo and wing imaginal disc nuclei of PcGFP- and PhGFP-expressing flies (Fig. 5). Individual recovery times for PhGFP loci in stage 13 to 16 embryos showed a Gaussian distribution around 20 seconds(Fig. 5A). Thus, we fit a single exponential decay function to the average data from 30 measurements after normalization to the initial intensities (see Fig. 5B). Such a procedure resulted in a single dissociation rate constant of k_off=0.051±0.004 second^–1 in embryos. In wing imaginal disc nuclei of PhGFP-expressing flies, we found a similar distribution of dissociation rate constants, although with a slightly positive skewness (Fig. 5A). Averaging 32 individual curves, we fit a single dissociation rate constant of k_off=0.032±0.002 second^–1(Fig. 5C).

Although both PcGFP and PhGFP are localized similarly to 30-40 loci in imaginal disc nuclei, the fraction of free fusion protein differs for the two fusion proteins, as seen in Fig. 1. The iFRAP method requires the complete photo-destruction of all free protein, which results in some unintended collateral bleaching of the fluorescent locus of interest by the high NA objective used in the case of PcGFP nuclei that contain a large amount of free protein, as seen by comparing the pre-bleach and first post bleach intensity measurements in Fig. 5B-D. The reduced fluorescence intensity in the complex decreased the signal to noise ratio for the measurement and reduced the accuracy of the determination of the dissociation rate constant. We compensated for this problem by performing a global fit on all of the dissociation curves for all experiments from the wing discs for the PcGFP. From the dissociation curve depicted in Fig. 5D, a dissociation rate constant of k_off=0.034±0.003 second^–1 was determined, very similar to the dissociation rate constant of PhGFP in larval disc nuclei.

PcGFP and PhGFP bind to distinct (Fig. 6) overlapping loci in salivary gland nuclei. Classical FRAP experiments (intense photobleaching of the band and monitoring of the fluorescence recovery over time) (Fig. 6E) were conducted on individual bands to determine the dissociation rate constants for PcGFP and PhGFP complexes. Individual bands showed consistent recovery times after multiple bleachings(Fig. 6F), although not all bands in the same nucleus exhibited the same recovery times(Fig. 6B,D). The equilibrium association constant of the complex is given by the ratio of a pseudo first-order forward rate constant (which includes the concentration of the free protein) and a first-order dissociation rate constant. In the case of PcGFP, a large amount of the labeled protein is free, i.e. leading to a lower fraction of bound protein than for PhGFP (bound ratio, B_ratio, see Materials and methods, Fig. 8B). Therefore, the shape of the FRAP recovery curve will be dominated by the fast diffusion of the free protein, as is evident from the comparison of the curves in Fig. 6B,D. We accounted for the free protein component by segmenting out the bound fraction of proteins from the bleach box (Fig. 7) (as described in the Materials and methods, and further explained in the Discussion). A comparison of the recovery curves for the bound fraction (blue curve) and the unbound free component (green curve) of PhGFP and PcGFP(Fig. 7D,F) shows that the segmentation separated the fast recovery process of the free protein (which occurs in the first seconds) from the actual binding reaction. By fitting the recovery curve for the bound fraction, we computed recovery times(t₀, the time required for the fluorescence intensity to reach ∼63% of the final height of the recovery curve for the bound molecules) for both PhGFP and PcGFP from such curves(Fig. 8A). The distribution of values was very similar for both proteins with t₀ recovery times ranging from 50 to 350 seconds.

In the case of salivary gland polytene chromosomes, the individual bands represent complexes binding to one or a few genes. Thus, we can ask if the range of measured recovery times represents different exchange rates for different genes (PREs) or the same exchange rate influenced by the density of binding sites. Simulations revealed that for interpreting these FRAP experiments the complex interplay between unbleached free protein, the total amount of bound protein at a locus and the free diffusion constant must be taken into account. We simulated the expected FRAP behavior for loci with different amounts of bound protein at a locus (thus different intensities), as briefly described in the Materials and methods, using a diffusion constant of 0.5 μm² second^–1 of the free protein determined experimentally and by systematically varying the dissociation rate constants around the experimentally determined values (a complete description of the simulations will be presented elsewhere). Fig. 8B demonstrates that with a single dissociation rate constant we would expect the recovery times to vary approximately linearly with the ratio of bound to free protein, B_ratio, as defined in the Materials and methods, Eqn 5,and results in the data shown by the connected points. In Fig. 8B, the experimentally measured recovery times are also plotted against the B_ratio for each locus. In the experimental data, no strong correlation was found between the recovery time and the intensity of the locus(number of binding sites), indicating that the complexes on different genes have different stabilities, and implying that they differ in composition. Including the dependence of the dissociation time on the density of binding sites, B_ratio, and the free diffusion component(determined experimentally to be ∼0.5 μm²second^–1) as predicted from the simulations, we calculated dissociation rate constants for each of the analyzed loci using Eqn 6. These values are independent of the local concentration of sites and the effect of diffusion. Bound protein ratios were normalized to the highest B_ratio value for each protein separately and are plotted in Fig. 9A. The dissociation rate constants are similar for both proteins but about one-fifth the values found for the complexes in 2N wing disc nuclei. The forward reaction is 2nd order and the rate is dependent on the concentration of binding sites and on the local concentration of the free protein. The amount of unbound protein in the nucleus is sufficient for the binding process to occur undisturbed for both PhGFP and PcGFP cases. From the fluorescence intensity ratios, we calculated that ∼10% of the total PhGFP and ∼2% of the PcGFP are in a bound state in a salivary gland nucleus at equilibrium. As the absolute number of binding sites (C_s) is unknown, the association rate constant could not be determined independently. Instead, a pseudo-association rate constant was calculated (which is related to the actual k_on by the equality k_on*=k_on·C_s, see Materials and methods, Eqn 7) (Sprague et al., 2004) for each locus and these values for PhGFP are plotted in Fig. 9B. The pseudo-association rate constants are an order of magnitude larger than the dissociation rate constants, confirming that dissociation is rate limiting.

In this paper, we have addressed fundamental questions about the stability and lifetime of PcG repression complexes in living organisms on their target sites. Applying photobleaching microscopy and computer modeling to transgenic Drosophila fly lines expressing GFP fused PcG proteins, we were able to measure diffusion constants, dissociation rate constants and residence times of PcG complexes on chromatin. Although complexes of PcG proteins have been isolated in vitro from disrupted cells(Saurin et al., 2001) or assembled from protein components expressed in heterologous systems (Francis,2001), none of these in vitro systems recapitulates the actual repression mechanism occurring in host organisms. That is, specific binding to PREs and gene-specific repression have not been demonstrated in vitro. The in vitro experiments showing chromatin binding and interference with remodeling machines or transcription complexes demonstrated that individual members of the PRC1 could produce similar interference as the complete complex in some cases (Francis et al., 2004). Repression occurred on chromatin consisting of nucleosomes without histone tails (Shao et al., 1999),results that do not recapitulate the in vivo situation. Removal of cis-acting Polycomb repressor elements (PRE) in vivo even at later developmental stages leads to derepression of homeotic genes(Busturia et al., 1997),indicating that propagation of the silenced state requires the activity of proteins that bind directly to the PREs and not to chromatin in general. In fact, in the organism, deficiency in single PcG genes causes loss of function(Beuchle et al., 2001) and targeting of PcG complexes requires trimethylation of lysine 27 of histone H3(Cao et al., 2002; Müller et al., 2002). The discrepancy between in vitro and in vivo repression may be the result of non-physiological conditions for the binding and inhibition experiments. A further complication is the fact that there is evidence of in vivo hierarchical assembly on chromatin (Wang et al., 2004), whereas in vitro complexes are preformed and bound to chromatin. To avoid these problems, we chose to address the question of the stability of PcG complexes in live animals and tissues. Our results indicate that the repression system is flexible and based on a continuous exchange of the PcG members at specific loci in the genome.

The application of fluorescence recovery after photobleaching (FRAP) to fusion proteins of GFP (and its analogs) by confocal microscopy has allowed the study of the dynamics of the steady-state distribution of nuclear proteins in living cells (Houtsmuller and Vermeulen, 2001; Phair and Misteli, 2001; White and Stelzer, 1999). Although most analyses have been qualitative(Cheutin et al., 2003; Dou et al., 2002; Festenstein et al., 2003; Houtsmuller et al., 1999; McNally et al., 2000; Misteli et al., 2000), a quantitative analysis can be used to obtain diffusion constants and dissociation rate constants by the application of combined techniques and appropriate models (Carrero et al.,2003; Dundr et al.,2004; Dundr et al.,2002; Phair et al.,2004; Rabut, 2004). In most cases, however, a reaction dominant model has been adopted. That is, because diffusion is faster than most dissociation rates, it is often, albeit incorrectly, neglected in the analyses. Recently, Sprague et al. presented a full analytical treatment for uniformly dispersed binding sites and showed simulations for various boundary conditions and rate constants (Sprague et al., 2004). Because our complexes are not uniformly distributed we have extended this treatment to a model with discrete binding loci and present here simulations using diffusion and binding rate constants derived from our experimental data. (A full description of the model will be presented elsewhere.)

Our FRAP experiments are somewhat different from those presented previously in the literature. We used live whole animals and tissues, rather than adherent tissue culture cells. This fact required adaptation of the acquisition method to include 3D tracking of loci after inverse FRAP bleaching(3D-iFRAP) as Drosophila diploid cells show an extensive reorganization of nuclear content during measurement times in the minute range(see Fig. 3 and see Movie 1 in the supplementary material) (Post et al.,2005). After registration of the fluorescent locus over time, we were able to calculate the dissociation rate constants for PcG complexes in embryos and imaginal discs. We found very consistent and similar behavior for both proteins in the complexes in the wing discs, corresponding to dissociation rate constants of 0.032±0.002 second^–1for PhGFP and 0.034±0.003 second^–1 for PcGFP (Figs 4 and 5), slightly smaller than in embryos for which the dissociation rate constant for PhGFP was 0.051±0.004 second^–1. All of these complexes were completely exchanged in less than 3 minutes. That is, both PcGFP and PhGFP in PcG complexes in vivo exchanged in an order of magnitude shorter time than the cell cycle. These data rule out a model for repression by which PcG complexes would mask the chromatin and render it inaccessible to transcription factors or other proteins as we discuss below.

The level of Pc is crucial for the maintenance of a competent complex as can be deduced from the fact that Pc^+/–heterozygotes show homeotic transformations(Lewis, 1978). Western blotting revealed that the fusion proteins do not reach levels greater than the wild type in non-transgenic animals(Table 1). That is, the total PcGFP protein content in the mutants was 0.53 that of wild type and PhGFP was comparable with the wild-type level.

The diffusion constants for PcGFP in early embryos before complex formation(0.74 μm² second^–1) and in the nucleoplasm of salivary gland nuclei (0.41 μm² second^–1) are smaller than one might expect for a protein of ∼62 kDa, indicating that the protein may exhibit non-specific binding to chromatin. Breiling et al.(Breiling et al., 1999)demonstrated that Pc has an affinity for nucleosomes without histone tails and that the C terminus was crucial for this interaction. However, the Pc chromodomain, essential for complex assembly, has a strong preference for trimethylated Lys 27 over other methylated sites or unmodified H3 showing a K_D of 5 μM and >1000 μM, respectively, in in vitro binding studies (Fischle, 2003). As seen in Fig. 6, the diffusion of the free protein obscures the recovery kinetics of the binding process measured on individual bands in salivary glands and the curve must be decomposed to fit the recovery kinetics (Fig. 7). The fitting assumes an excess of free protein, which is true for both of our transgenic proteins, despite the lower nucleoplasmic fluorescence in the case of PhGFP (see below). In an ideal situation with infinitely fast diffusion,loci with different concentrations of binding sites of identical affinity would recover within the same time. At the diffusion constants measured experimentally in our nuclei (∼0.5 μm²second^–1), we found by simulation that two loci with the same size but with different concentrations of binding sites will recover with different times: i.e. higher concentration, longer recovery time(Fig. 8). We created masks in the images to separate out pixels that contained predominantly non-bound protein from that involved in complexes. By first fitting and removing the diffusion component we were able to fit the resulting recovery curves to a single exponential, as in a kinetic process in a standard chemical equilibrium(Fig. 7). The dissociation rate constants were in the same range as those measured in the 2N cells of embryos and imaginal discs but the means were shifted towards a value of around one-third of that for PhGFP to one-quarter of that for PcGFP. These differences could reflect some unintended bias in the selection of the bleach loci in either the 2N nuclei or the polytene bands. However, each polytene band represents thousands of complexes at one or a few PREs, rather than an average of many different complexes; thus, these data may be more robust. In either case, even when PREs are in close proximity, such as is the case of a thousand chromatids closely aligned in the polytene chromosomes, the PcG proteins are in a chemical equilibrium with unbound protein. The reduced rate constants may reflect the large local binding sites, whereby a dissociated protein does not immediately join the `free pool' but has a higher probability to rebind in the close vicinity. However, the proteins are not `trapped' in the complex but rather are able to completely exchange in under 6 minutes. The reproducibility of the recovery times of individual bands subjected to successive FRAP measurements is shown in Fig. 6F, indicating that the differences of two- to threefold in recovery times(Fig. 8) between different bands can be considered reproducible and significant.

The t₀ values calculated from these recovery curves are, however, not a direct measure of residence time because of their dependence on the effect of diffusion transport in combination with ongoing depletion from the free pool. As could be demonstrated using simulations, if the dissociation rate constant and the concentration of free protein were the same for all complexes then one would expect the recovery rate to depend linearly on the amount of bound protein (lines in Fig. 8B), which is essentially a measure of the ability of a spot to deplete the free protein pool during recovery. As seen in the same figure, the experimental data do not show such a correlation, implying that there are differences in the composition of the complexes on different genetic loci and that the dissociation rate constants,though similar (within a factor of 5, Fig. 9A), reflect the specific mixture of PcG and non-PcG auxiliary proteins on the polytene bands. Such an interpretation is compatible with the data of Rastelli et al., who showed varying occupancy of PcG proteins and Zeste on more than 100 bands by immunohistochemistry on polytene chromosomes(Rastelli et al., 1993). To rule out the possibility of a very slow component that would appear as an immobile fraction in a single exponential fit, we also fitted the data with a sum of two exponentials, but did not find a consistent second time in this case and less precision of the first time. Thus, we conclude that both PcGFP and PhGPF in repression complexes exchange within a few minutes in live Drosophila cells.

In Fig. 9B, the pseudo-association rate constants as described in the Materials and methods are plotted for PhGFP. The values are an order of magnitude larger than the dissociation rate constants and thus, dissociation is rate-limiting. We have not attempted to present pseudo-association rate constants for PcGFP for the following reasons. We assume that the number of binding sites for Pc and Ph are approximately equal as the proteins bind to overlapping sites on polytene chromosomes (Rastelli et al.,1993), isolated complexes of the proteins contain equimolar quantities of both proteins (Saurin et al., 2001) and they are targeted to the same PREs by ChIP analysis(Breiling et al., 2001). As discussed above, the off-rates are similar for the two proteins. However, we can see a larger pool of unbound PcGFP compared with PhGFP in both embryos and larval tissue (Figs 1 and 6). As determined from western blots in salivary glands, PcGFP is not present in amounts higher than PhGFP(Fig. 1C). Wang et al.(Wang et al., 2004) have shown that there is sequential recruitment of PcG complexes to the PREs, whereby Pc targets the complex to chromatin by binding to trimethylated H3K27(Cao et al., 2002; Czermin et al., 2002; Fischle et al., 2003; Wang et al., 2004). From these considerations, we postulate that the PcGFP fusion protein, although competent to target PREs with modified histones and engage in a competent complex,cannot substitute for all Pc molecules in the complexes (perhaps owing to steric hindrance of adjacent GFP moieties). Our data suggest that the unmodified Pc is preferred in the complex by a factor of about 4 or 5; thus,association rate constants calculated for PcGFP will not properly reflect the true on rate of the unmodified protein, whereas off rates should not be adversely affected.

We calculated the ratio of the bound/free fusion proteins from the segmentation of the salivary gland prebleach FRAP images to be 1:10 and from the western blots (Fig. 1) the ratio of the fusion protein to wild-type protein of 1:1. We estimated the absolute concentration of GFP protein in the salivary gland nuclei to be∼2-4 μM by comparison to the intensity of droplets of purified GFP protein in an immiscible solution in our microscope system. If the total concentration of binding sites is equivalent to the concentration of the bound Ph, we can estimate the K_D for the protein in vivo in salivary gland nuclei to be ∼5 μM.

Using photobleaching in confocal microscopy and computer simulations, we determined the stability of PcG protein complexes during development. These are the first FRAP experiments performed in whole live organisms and tissues. We found that all complexes were exchangeable throughout all developmental stages. The relatively short residence times of 2-6 minutes for Ph and Pc in the repression complex rule out models for repression that invoke blocking chromatin access. They also suggest that competition could exist between antagonistic factors at PREs and promoters, allowing modulation of the state of repression during development by changes in their balance. These data complement those found in other systems, such as the direct competition for chromatin binding sites between histone H1 and microinjected high-mobility group (HMG) proteins, as demonstrated by Catez et al.(Catez et al., 2004).

Most FRAP studies of nuclear proteins have involved components in transcription complexes or transcriptional activators that exchange in less than 2 minutes (Phair et al.,2004). The only repressor protein that has previously been investigated is heterochromatin protein 1 (HP1), a protein targeted to heterochromatin in higher eukaryotes(Cheutin et al., 2003; Festenstein et al., 2003). Although HP1 is loaded directly onto the chromatin during replication, it was found by FRAP to bind only transiently to chromatin with a maximum residence time of ∼60 seconds. Thus, both HP1 and PcG repression complexes appear to function by dynamic competition with other chromatin-binding proteins rather than by formation of a static, higher-order chromatin structure with immobilized bound repressors. Our FRAP measurements on polytene chromosomes revealed differences in the dissociation rate constants between individual bands that imply a flexible repression system of complexes with various compositions that influence the binding affinity of other members and whose turnover is in the order of a few minutes.

We conclude that: (1) activation and repression can be dynamically controlled by simple chemical equilibria; (2) reduction in PcG levels will facilitate epigenetic change and may explain why non-cycling cells can be reprogrammed more easily than cycling cells(Baxter et al., 2004); and (3)PcG complexes are exchangeable protein assemblies that maintain repression over many cell cycles by simple chemical equilibria.

We thank C. Fritsch for extensive constructive discussions and editorial assistance with the manuscript; G. Hernández for antibodies, help and advice; J. Müller for plasmids; and. Jäckle, R. Paro and F. Netter for fly strains and plasmid constructs. We are also grateful to T. Jovin for suggestions in the experimental design, and for discussions and assistance with FRAP analysis and equilibrium kinetics. The work was supported by a Max Planck Predoctoral Fellowship to G.F. This work is part of the scientific research conducted in the International PhD Program Molecular Biology-International Max Planck Research School at the Georg August University (Göttingen, Germany).