Transcription through the iab-7 cis-regulatory domain of the bithorax complex interferes with maintenance of Polycomb-mediated silencing

The bithorax complex (BX-C) contains three homeotic genes, Ubx,abd-A and Abd-B, that are responsible for determining the identity of parasegments 5 to 14 (PS5-14). These parasegments will form the posterior thorax (T2 and T3) and all eight abdominal segments of the adult fly(A1 to A8) (Lewis, 1978;Sanchez-Herrero et al., 1985). The PS-specific expression pattern of Ubx, abd-A and Abd-Bis controlled by a large cis-regulatory region that covers 300 kb of DNA and that is subdivided into nine functionally autonomouscis-regulatory domains (abx/bx, bxd/pbx, iab-2 toiab-8) (for reviews, see Duncan,1987; Peifer et al.,1987). For example, the iab-5 cis-regulatory domain regulates Abd-B expression in a pattern that confers PS10/A5 identity to the cells of that PS. Similarly, the iab-6, iab-7 and iab-8 cis-regulatory domains activate Abd-B expression in patterns appropriate for PS11/A6, PS12/A7 and PS13/A8 identity, respectively(Celniker et al., 1990;Sanchez-Herrero, 1991). When one of the cis-regulatory domains is inactivated, the parasegment specified by the affected regulatory domain is transformed into a copy of the PS immediately anterior. Thus, in a deletion of iab-7, PS12/A7 is transformed into PS11/A6. In this case, Abd-B expression in both PS11 and PS12 is driven by the iab-6 cis-regulatory domain alone(Galloni et al., 1993).

The regulation of the BX-C homeotic genes during embryogenesis is subdivided into two phases: initiation and maintenance. In the initiation phase, the products of the gap and pair-rule segmentation genes are responsible for initiating the parasegment specific expression of the BX-C homeotic genes. These proteins interact with target sequences (called initiation elements) in the nine cis-regulatory domains(Simon et al., 1990;Qian et al., 1991;Muller and Bienz, 1992;Shimell et al., 1994). However, the products of the segmentation genes are present only transiently in the early embryo. Maintenance of the initial pattern requires thetrithorax-Group (trx-G) and Polycomb-Group(Pc-G) genes. The trx-G genes function to keep the homeotic genes on, while the Pc-G genes function to maintain the inactive state of the homeotic genes (reviewed byParo, 1990;Simon and Tamkun, 2002). Experiments with reporter constructs have identified elements, called Polycomb Response elements or PREs, in several of the BX-C cis-regulatory domains, that appear to be targets for Pc-G action. When these PREs are combined with a parasegment-specific initiation element, they maintain the parasegmentally restricted expression pattern conferred on the reporter by the initiation element (Muller and Bienz,1991; Busturia and Bienz,1993; Simon et al.,1993; Chan et al.,1994; Chiang et al.,1995; Poux et al.,1996). In addition to this maintenance activity, PREs are also able to repress the activity of the mini-white reporter gene used to establish transgenic lines. Usually, transgenic lines carrying the mini-white gene harbor darker eye color when the inserts are homozygous. When they are included in a mini-white transgene, the PREs repress or even eliminate mini-white expression when the animals are homozygous (a phenomenon referred to as the pairing-sensitive repression assay)(Kassis et al., 1991;Chan et al., 1994;Gindhart and Kaufman, 1995;Hagstrom et al., 1997;Muller et al., 1999) (for a review, see Pirrotta and Rastelli,1994).

Genetic and molecular analysis has identified chromatin domain boundaries that demarcate the cis-regulatory domains, insuring the functional autonomy of each regulatory domain(Gyurkovics et al., 1990;Galloni et al., 1993;Mihaly et al., 1997;Zhou et al., 1999;Barges et al., 2000) (for a review, see Mihaly et al.,1998). For example, in PS11, the Fab-7 boundary protects the active iab-6 cis-regulatory domain from the inactiveiab-7 domain, preventing inappropriate regulatory interactions between the two domains. Immediately adjacent to the Fab-7 boundary,lies the iab-7PRE, which is involved in maintaining inactivity ofiab-7 in parasegments anterior to PS12(Hagstrom et al., 1997;Mihaly et al., 1997).

Two classes of mutations affect the Fab-7 region. Class II mutations, such as Fab-7², delete the boundary alone and leave the nearby iab-7 PRE intact. They lead to a mixed gain- and loss-of-function phenotype in PS11/A6; there are groups of cells acquiring PS10/A5 identity, because in these cells both iab-6 andiab-7 are inactive. The remaining cells of PS11 adopt a PS12/A7 fate,because both iab-6 and iab-7 are active in these cells. This mixed gain- and loss-of-function phenotype arises because there is a competition in the fused cis-regulatory domain between positive elements iniab-6 that ectopically activate iab-7 and negative elements in iab-7 that ectopically silence iab-6. Class I mutations,such as the original Fab-7¹ allele, are larger deletions that remove not only the boundary but also the nearby iab-7 PRE. In this class of mutation, the balance between gain- and loss-of-function phenotype is shifted towards gain-of-function, and A6 is completely transformed into A7 (see Fig. 2B) (Mihaly et al.,1997).

Most chromatin domain boundaries in higher eukaryotes have been identified by their ability to block enhancer-promoter interactions when intercalated between them (enhancer-blocking assay) (for reviews, seeGerasimova and Corces, 1996;Geyer, 1997;Sun and Elgin, 1999;Bell et al., 2001). In our terminology, we call elements defined in the enhancer-blocking assay chromatin insulator. In Drosophila two insulators, scs/scs'(Kellum and Schedl, 1991;Kellum and Schedl, 1992) andgypsy (Geyer and Corces,1992; Roseman et al.,1993) have been extensively studied in the enhancer-blocking assay. We have previously described that gypsy or a minimal scs fragment (scs^min) cannot substitute for Fab-7; their enhancer-blocking activity prevents the iab-5 and iab-6 cis-regulatory domains from interacting with the Abd-B target promoter (Hogga et al., 2001). We describe the results of experiments in which we replace Fab-7 by a slightly larger scs fragment that was used in enhancer-blocker experiments by different laboratories (Kellum and Schedl,1991; Kellum and Schedl,1992; Vazquez and Schedl,1994; Dunaway et al.,1997; Krebs and Dunaway,1998; Parnell and Geyer,2000). Surprisingly, this scs fragment behaves differently than scs^min and leads to opposite gain- and loss-of-function phenotype depending on its orientation within the context of the Fab-7 region. The orientation-dependent effect is due to the presence of a promoter immediately adjacent to the scs insulator. Our results suggest that transcription through the iab-7 PRE interferes with the maintenance of a Polycomb repression complex on the iab-7 domain.

Fab-7 was replaced by SCS_prom (or _promscs)using the gene conversion strategy described by Hogga et al.(Hogga et al., 2001).

DNA techniques, fly work, antibody staining and in situ hybridization have been described previously (Mihaly et al.,1997; Hogga et al.,2001; Zhou et al.,1999). The antibody against ABD-B was kindly provided by Sue Celniker (Celniker, 1990).

Adult abdominal cuticles were mounted as described elsewhere(Mihaly et al., 1997),examined and photographed on an Axioplan microscope with a 5× lens. Only half cuticles are shown in Fig. 2. The dorsal surface of each abdominal segment has a rectangular plate of hard cuticle called the tergite (only half of the tergites are visible on the left of each panel, as well as the genitalia at the bottom). The ventral surface of abdominal segments is composed of soft cuticle called the pleura. On the ventral midline of the second (A2) and more posterior segments, there are small plates of harder cuticle called sternites. In males only six abdominal segments are visible. The tergites on A5 and A6 are pigmented and can be therefore distinguished from more anterior tergites. On the ventral side, the sixth sternites can be distinguished from the more anterior sternites by its different shape and by the absence of bristles. Homeotic transformations associated with Fab-7 are best visible in males where most (Fab-7²) or all of A6(Fab-7¹) is missing (seeFig. 2B). As A7 and A8 do not contribute to any visible cuticle after metamorphosis in males, homeotic transformations associated with Fab-8 are detected in females where A7 develops as a smaller segments than the anterior segments (seeFig. 2C).

The slightly larger scs fragment used to replace Fab-7 in the present studies is depicted in Fig. 1. The extra-DNA (relative to scs^min) consists in the 282bp MluI-PstI fragment at one edge of scs^min. In the _promscs convertant line, the 282 bp extra DNA fragment facesiab-6. In scs_prom, this extra DNA is juxtaposed toiab-7.

Fig. 2A shows the phenotype observed in homozygous males in which the _promscs construct replaces Fab-7. In these flies, A3 to A6 were transformed into a mixture of A2-A3 identity, indicating that iab-3 throughiab-6 are affected by the _promscs element. In prior experiments, we have shown that replacement of Fab-7 by the minimal scs insulator (scs^min) in both orientations results in a consistent phenotype in which iab-5 and iab-6 are prevented from interacting with Abd-B by the intervening insulator(Hogga et al., 2001). Thus,the extra 282 bp DNA element appeared to interfere at a distance withiab-3 and iab-4. Interference with iab-3 andiab-4 functions in _promscs is surprising, because thesecis-regulatory domains regulate abd-A (seeFig. 5A) and are distant from_promscs. This result implies that, _promscs exerts a negative polar effect that can spread 40 kb away into iab-3. In addition, we have previously provided evidence that, upon insulation fromAbd-B by the intervening scs^min insulator, iab-5is targeted instead to the abd-A gene, which it activates in a pattern appropriate for specifying a A5-like identity (seeHogga et al., 2001). In_promscs, A5 is transformed into A2-A3, indicating that the negative polar effect exerted by _promscs also affects iab-5/abd-Ainteraction. This phenotype was seen in four independent conversion lines and a whole genome Southern analysis has verified that there are no large rearrangements affecting the iab-3 through iab-5 regions of the _promscs chromosome. Finally, as heterozygous flies are wild type, the negative polar effect of _promscs on iab-3through iab-5 is only acting in cis.

Fig. 2B shows the phenotype observed in homozygous males in which the same scs fragment replacesFab-7 in the opposite orientation (scs_prom). Instead of observing the loss-of-function phenotype described above, where A3, A4, A5 and A6 are transformed to a more anterior segment, we found a gain-of-function phenotype in which A6 took the identity of a more posterior segment, A7. This is similar to removal of Fab-7 entirely. Because in these conversions we removed Fab-7 by introducing the Fab-7²deletion, we confirmed by sequencing that scs_prom is intact. Moreover, the same phenotype is observed in the five other independent conversion events that we recovered. The dominant gain-of-function phenotype associated with scs_prom was confirmed by the observation that heterozygotes (scs_prom/+) displayed the same phenotype, although not as severe. Because the dominant gain of function associated with scs_prom was absent from the flies in which Fab-7 was replaced by scs^min, we conclude that the same extra 282 bp fragment(MluI-PstI) is responsible for the loss-of-function phenotype in _promscs and the dominant gain-of-function phenotype in scs_prom.

Closer examination of the males shown inFig. 2B revealed that the phenotype of scs_prom flies was slightly different from the phenotype generated by the Fab-7² deletion alone. As mentioned in the Introduction, class 2 mutations such asFab-7², which remove only the boundary and leave an intactiab-7PRE, caused a mixed gain- and loss-of-function phenotype: most cells of A6 adopted A7 identity while the remaining adopted A5 identity (seeFig. 2B). If the scs_prom construct had no effect on the region, we should have observed a Fab-7² phenotype simply because of the removal of the Fab-7 element. Fig. 2B shows that this was not the case: A6 was completely transformed into A7 in homozygous scs_prom flies. This phenotype is identical to the phenotype of the class I Fab-7¹² allele where the boundary and nearby iab-7PRE are deleted(Fig. 2B). Thus, introduction of the scs_prom element converts a class II allele into a class I allele, as if the extra 282 bp fragment (MluI-PstI)interfered with the activity of the nearby iab-7 PRE. To test this hypothesis, we decided to verify how scs_prom affected PRE-mediated pairing-sensitive repression of a miniwhite reporter construct (see Introduction). Although transformants with the scs element in the_promscs orientation are pairing sensitive in 63% of the lines, when the scs element is in the opposite orientation (scs_prom) the pairing-sensitive frequency decreases to 11%(Fig. 1). Thus, our hypothesis that scs_prom within the BX-C contains an anti-PRE activity is supported by these ectopic constructs. To localize this anti-PRE activity, we analyzed the pairing-sensitivity of scs_prom derivatives in which the scs_prom fragment was progressively shaved from one end (seeFig. 1). A deletion removing 282 bp from the end (MluI-PstI deletion; scs^min)restored pairing-sensitivity to a frequency of 47%. Because a deletion that extended further towards the HpaI site did not significantly increase the proportion of pairing-sensitive lines (50%), we conclude that most of the anti-PRE element of scs is located within the MluI-PstI fragment.

Examination of homozygous scs_prom females indicated thatiab-8 was also partially activated in the sixth and seventh abdominal segments, both of which show an identity intermediate between A7 and A8, a phenotype that is reminiscent of Fab-8 boundary deletions(Fig. 2C) (Barges et al.,2001). These observations suggest that the anti-PRE activity contained in scs_prom is not only interfering with the functioning of the nearbyiab-7 PRE, but spreads across the whole iab-7 domain and reaches iab-8.

We also examined the Abd-B expression pattern in embryos(Fig. 3). In scs^min,the partial block between iab-5/iab-6 and the Abd-B promoter was visualized by the great reduction of Abd-B expression in PS10 and PS11 (Hogga et al., 2001). In scs_prom, because of the posterior transformation of A6 into A7, we were expecting to monitor a reiteration of the PS12/A7 Abd-Bexpression pattern in PS11/A6 (Galloni et al., 1993). This is not what we found. In PS10 and PS11, the expression patterns were similar to the patterns detected in scs^min. In PS12, the expression pattern was much lower than in wild type and was similar to the pattern detected in PS11. In fact, this expression pattern resembled that of an iab-7 deletion mutant(Galloni et al., 1993). This should lead to a transformation of PS12 into PS11, exactly the opposite of the phenotype that we observe in adults. Unfortunately, there are no distinguishing morphological landmarks in embryos and larvae that definitively identify these two parasegments earlier in development. It is not clear why,in the adult, the readout of scs_prom results in a gain-of-function phenotype, whereas in the embryo, Abd-B expression resembles loss of function. We believe these findings indicate that iab-7 misexpression is more pronounced in the pupa, when the adult structures are forming,strongly supporting the idea that scs_prom affects primarily the maintenance phase. Bender and Fitzgerald have described similar mutations that affect the maintenance phase of iab-2 silencing in PS6/A1(Bender and Fitzgerald, 2002). In their case, the identity of the affected abdominal segment PS6/A1 could be easily recognized in embryos and larvae from PS7/A2. Intriguingly, the dominant gain-of-function phenotype associated with their mutations was also only detectable in adult.

We have previously shown that scs contains a chromatin insulator element(scs^min) that is able to interfere with long-range enhancer-promoter interactions (Hogga et al., 2001). In addition to the insulator, the 1.2 kb scs fragment contains at one extremity an element that, in conjunction with scs^min, is able to induce a gain-of or loss-of-function phenotype when replacing Fab-7. When facing iab-7, this element destabilizes iab-7 and iab-8 silencing in A6. However, when facing iab-6, this element exerts a negative polar effect oniab-5, iab-4 and iab-3. These two gain- and loss-of-function effects are difficult to reconcile. Previously, Avramova and Tikhonov(Avramova and Tikhonov, 1999)discovered that scs contained a promoter and challenged the idea that scs was a neutral chromatin domain boundary. Their finding supported the promoter decoy hypothesis of Geyer (Geyer,1997), in which insulation is achieved by a promoter-trapping mechanism. Although the promoter described by Avramova and Thikonov (Avramova and Thikonov, 1999) maps to the other side of scs (seeFig. 1), promoter trapping could account for the polar effect on iab-5, iab-4 andiab-3. In this scheme, insertion of a promoter at the iab-6edge of Fab-7 would attract the nearby cis-regulatory elements and divert them from their normal abd-A and/orAbd-B promoter (see Fig. 5A). In this case, transcription of sequences near the promoter should be detected in parasegments affected by _promscs (and scs_prom). In order to test this hypothesis, we performed whole-mount in situ hybridization on embryos of the _promscs and scs_prom lines. We synthesized strand-specific probes from theiab-6 (probe A; Fig. 4) or iab-7 DNA (probe B;Fig. 4) flanking the scs element. Using both probes on _promscs, scs_prom and scs^min lines, we found that the promoter described by Avramova and Thikonov (Avramova and Thikonov, 1999) remained silent in the context of the BX-C. There was, however, a promoter that became active in the context of the BX-C at the other side of the scs fragment.Fig. 4 shows the results obtained with a probe from the iab-6 side of _promscs(probe A). In wild type (or Fab-7²), transcription was detected from a very early stage throughout embryogenesis in PS13 and PS14. The expression profile of this transcript was reminiscent of a transcript that originates from a promoter localized just downstream from the Abd-Btranscription unit (Zhou et al.,1999). In _promscs embryos, however, we detected additional transcripts in PS10/A5, PS11/A6 and PS12/A7. Interestingly, we failed to detect these transcripts when the truncated version of_promscs (scs^min) replaced Fab-7, indicating that the MluI-PstI fragment contains the promoter (or sequences necessary for its activity). Appearance of PS10-specific transcripts may explain the polar effect of _promscs on iab-5 if we assume that iab-5 activity is trapped by the scs promoter and thus diverted from the abd-A promoter [in the _promscs context,iab-5 is insulated from Abd-B by the intervening scs^min insulator, see Hogga et al.(Hogga et al., 2001)]. A similar mechanism would explain the polar effect on iab-4 andiab-3 (Fig. 5A). However, we failed to detect transcription in PS9 and PS8, where both regulatory domains are active (Fig. 4). The discrepancy between the observed pattern of embryonic transcription and the adult phenotype may reflect a higher affinity of the promoter to iab-3,4 in adults.

It is not entirely clear which regulatory domain activates transcription in PS12. In the _promscs context, the scs insulator is located between the promoter and the iab-7 cis-regulatory domain. As reported by Hogga et al. (Hogga et al.,2001), the insulator alone does not completely impair interactions between the distal iab-5/iab-6 cis-regulatory domains and theirAbd-B target promoter (see alsoFig. 3). Thus, in_promSCS, iab-7 may not be completely insulated from the promoter and activates transcription in PS12. As an alternative explanation we believe that transcription in PS12 in _promscs embryos results from activation by the more anterior cis-regulatory domains(iab-3, iab-4, iab-5 or iab-6), which once activated in a given parasegment, remain active in the more posterior parasegments, as first proposed in Ed Lewis' model(Lewis, 1978). It should be noted that although promoter-trapping is an attractive explanation accounting for the polar effect of _promscs on the iab-3, iab-4,iab-5 and iab-6, we cannot exclude the possibility that transcription through these cis-regulatory domains is the cause of their inactivation.

In scs_prom, the edge of scs containing the promoter is facing the iab-7 cis-regulatory domain. Not surprisingly, we detected intense transcription in PS12 with a probe from the iab-7 edge (probe B, Fig. 4). The same probe failed to detect any transcript in wild type or in embryos in whichFab-7 is replaced by scs^min. These results indicate that the PS12-specific transcript very likely originates from the same promoter that is firing in PS10-12 in _promscs. We also detected equally intense transcription in PS11. Thus, the fragment harboring anti-PRE activity when associated with scs_prom, contains a promoter that fires transcription across the iab-7PRE with which it interferes. As mentioned above, the anti-PRE activity contained in scs_prom not only interferes with the functioning of the nearby iab-7PRE, but spreads across the whole iab-7 domain and reaches iab-8. Notably, the same transcription pattern in PS11 and PS12 is observed with a strand-specific probe originating from iab-8 (data not shown, seeFig. 4).

Fig. 5B suggests how the transcription from scs_prom might cause the posterior transformation in the sixth and seventh abdominal segments (PS11 and PS12). The promoter in scs_prom is activated in PS11 and PS12(Fig. 4), sending transcripts across the iab-7 and iab-8 regulatory regions. In PS11, both of these regulatory regions are normally silent, but the act of transcription apparently reverses the silencing, causing the cells in PS11 to differentiate in the same way as those of PS12 or PS13. Likewise, in PS12, transcription across the iab-8 region activates it, transforming PS12 cells towards PS13 character. We do not observe transcription from scs_prom in PS13. In this parasegement, however, the iab-8 promoter is activated,giving rise to leftwards transcription(Zhou et al., 1999). It is possible that this leftwards transcription interferes with rightwards transcription from scs_prom. It is perhaps surprising that the transcription from scs_prom begins in PS11, as the PS12-specific regulatory region (iab-6) is separated from the promoter of scs_prom by the scs insulator. However, the insulator in an scs^min conversion at the same site does not completely insulateiab-6 from the Abd-B target promoter(Hogga et al., 2001) (see alsoFig. 3), making it likely thatiab-6 can activate the scs_prom across the insulator. It is also possible that the function of an insulator depends on neighboring sequences. If, for example, insulator function is enhanced by a nearby PRE,then partial loss of the iab-7PRE function might weaken the scs_prom insulator.

There are precedents where transcription has been suggested to play a role in chromatin remodeling. For example, the human β-globin locus is subdivided into three chromatin domains, each of which become more accessible to nuclease digestions upon gene activation(Ashe et al., 1997;Gribnau et al., 2000;Plant et al., 2001). Interestingly, large intergenic transcripts delineate each of these domains and chromatin remodeling of each domain is preceded by its transcription. Another example has been reported by Rank et al.(Rank et al., 2002). Using a transgenic context, they found that transcription across a PRE could interfere with silencing. Finally, in the accompanying paper, Fitzgerald and Bender provide evidence that transcription across the iab-2 cis-regulatory domains in PS6/A1 interferes with iab-2 silencing, resulting in the posterior transformation of PS6/A1 into PS7/A2(Bender and Fitzgerald, 2002). In this case, the identity of the affected abdominal segment can easily be recognized in embryos and larvae. Despite the existence of intense transcription of iab-2 in embryos, the dominant gain-of-function phenotype associated with iab-2 misexpression is only detectable in the adult. Thus, as in our case, transcription across the iabregulatory regions appear to interfere with silencing during the late maintenance phase, when the adult structures are forming.

If transcription can interfere with Pc-G silencing, what are the mechanisms responsible for this activity? Factors that affect RNA polymerase II (RNAPII)transcript elongation have been shown to have an effect on chromatin. For example, it has been suggested that histone acetyl transferases (HAT) such as PCAF (Cho, 1998) or ELP3 (Wittschieben, 1999) assist RNAPII in relieving inhibition caused by nucleosome arrays. Although active chromatin requires acetylation of specific lysine residues in the H3 and/or H4 histone tails (Moazed, 2001), the recent purification of Pc complexes suggests that histone deacetylation is required for establishing a stable long-term Pc-G silencing complex(Saurin et al., 2001;Tie et al., 2001). In the case of scs_prom, perhaps the frequent passage of RNAPII and its associated histone acetylation activities though the PREs interferes with the assembly of the Pc-G silencing. Involvement of acetylated histones in antagonizing PcG-dependent silencing is supported by the findings of Cavalli and Paro (Cavalli and Paro,1999) showing that high levels of acetylated histone H4 are associated with non-repressive PRE sequences. Alternatively, it has been recently found that variant histone H3.3 was deposited on active chromatin during transcription, providing a mechanism for the immediate activation of genes that are silenced by histone modification(Ahmad and Henikoff, 2002). It may be possible that transcription across iab-7 (and alsoiab-8) results in deposition of new nucleosome marked by H3.3,interfering thereby with the maintenance of silencing by the Pc-G complex.

It has been known for a long time that the large cis-regulatory regions of the bithorax complex are transcribed(Lipshitz et al., 1987;Sanchez-Herrero and Akam,1989; Cumberledge et al.,1990). In blastoderm stage embryos, the iab-2 thoughiab-8 regions can be divided into three domains, each transcribed in a region that extends from a specific anterior limit to the posterior limit of the segmented part of the embryo. These domains are only broadly defined but their order on the chromosome reflects the anterior limit of expression for each of them. In the light of our data, it is tempting to speculate that transcription of the iab domains convey a regulatory signal,preventing assembly of the Polycomb-repressing complex on theiab domains that need to remain active. If this is true, transcripts should appear in the anteriormost parasegments/segments where eachcis-regulatory domain in activated. However, so far, we have not seen transcripts in every regulatory region, which would account for the sequential activation of each regulatory domain. Moreover, this model predicts that theiab-7 PRE and iab-7 domains should be transcribed from PS12,where iab-7 is first active. So far transcripts across theiab-7 domain have only been detected in PS13 and 14 (this study)(Zhou et al., 1999). Thus, it remains unclear whether intergenic transcription plays a role in wild-type animals to create and/or maintain open chromatin, or whether the existence of intergenic transcripts is the consequence of an open structure. However, our experiments strongly suggest that forced transcription through an inactivecis-regulatory domain interferes with the maintenance of silencing,highlighting an incompatibility between transcription and Pc-G mediated silencing. This activity probably reflects a fundamental mechanism to protect an actively transcribed gene from being inactivated by the Pc-G proteins that are present in all cells.

We are indebted to H. Gyurkovics for numerous discussions and suggestions during the course of this work, and to W. Bender and R. Paro for sharing unpublished data. We also acknowledge R. Meada and Y. Moshkin for critical reading of the manuscript. We thank J. Weiss for his help with transcripts detection on whole embryos, A. Spierer for injections and J. Vasquez for providing the initial SCS clone. Finally, we thank E. Favre and G. Faustino for excellent technical assistance. This work was supported by the Swiss national Found, the State of Geneva and the Human Frontier Science Organization.