Dissecting the regulatory landscape of the Abd-B gene of the bithorax complex

The three homeotic genes of the Drosophila bithorax complex(BX-C), Ubx, abd-A and Abd-B, specify the identity of the 3rd thoracic segment (T3) and all abdominal segments (A1 to A8) of the fly(Lewis, 1978; Sanchez-Herrero et al., 1985). These segments derive from parasegments (PS) 5 to 14 that form during early embryogenesis. The expression of Ubx, abd-A and Abd-B is regulated by a complex cis-regulatory region covering 300 kb of DNA. Molecular genetic analysis has subdivided this large cis-regulatory region into nine segment/parasegment-specific subregions [abd/bx, bxd/pbx, iab-2 to iab-8 (Lewis, 1978; Bender et al., 1983; Karch et al., 1985; Peifer et al., 1987; Duncan, 1987; Celniker et al., 1990; Sanchez-Herrero, 1991) (for a review, see Maeda and Karch,2006)]. Remarkably, the genes and their regulatory regions are aligned on the chromosome in the order of the body segments they specify. Genetic studies using chromosomal rearrangement breaks have shown that when cis-regulatory regions are removed from the complex, specific parasegments are transformed into copies of more anterior parasegments. The clustering of mutations with similar phenotypes along the chromosome has led to a model in which each regulatory region is important in specifying homeotic gene expression in one parasegment. For example, iab-5 is thought to control the expression of Abd-B in PS10/A5, while iab-6 is thought to control expression of Abd-B in PS11/A6.

Although it is clear that each specific cis-regulatory region is important for the parasegment specific expression of a BX-C homeotic gene, it is still unclear if each regulatory region works autonomously in this function. Abd-B, for example, is expressed in PS10-13 in a stepwise gradient from anterior to posterior. Its specific expression pattern is determined by the four cis-regulatory regions iab-5 to iab-8. Because of the stepwise increase in Abd-B expression and the fact that mutations affecting the cis-regulatory regions transform parasegments into a copy of the parasegment immediately anterior to them, it has been thought that the pattern of Abd-B expression is the result of the additive effect of sequential activation of the cis-regulatory domains(Lewis, 1978; Celniker et al., 1990).

How individual cis-regulatory domains work also remains a mystery. And although many elements have been identified within the BX-C using transgenic assays [such as initiation elements, maintenance elements (ME) cell type- or tissue-specific enhancers, chromatin insulators and the Promoter Targeting Sequence (PTS)], there is still a surprising lack of appropriate mutations to determine how all of these elements interact and function within their native environment (for a review, see Maeda and Karch, 2006). Many of the mutations used to map the cis-regulatory regions are chromosomal rearrangements breaks. Such breaks often separate more than one regulatory region from its relevant target promoter, resulting in homeotic transformations of multiple segments. Rearrangement breaks also introduce foreign DNA next to the sequences of the BX-C, and so position effects are observed. Because of these limitations, it has been very difficult to confirm the function in the context of the BX-C of elements studied in transgenes.

By mobilization of three P-element enhancer trap lines, we have isolated a number of internal deficiencies that affect the iab-3 to iab-8 cis-regulatory regions. Among the deficiencies isolated are ones that remove initiator elements, ones that remove cell-type specific enhancers, ones that remove insulators and two deficiencies that cleanly removes two PTSs. Combining our genetic analysis with results from transgenic reporter constructs allowed us to define two new initiators (IAB6 and IAB7a),a new ME and various cell type-specific enhancers. Overall, our work provides compelling functional evidence that the Abd-B region of the BX-C is modular, with the expression of Abd-B in each parasegment being primarily controlled by a single cis-regulatory domain. Our work also provides strong genetic evidence for a temporal transition in cis-regulation, in which the initiator elements dictate the later activity state for an entire cis-regulatory domain.

Overlapping fragments from the walk(Karch et al., 1985) have been inserted upstream from the Ubx-lacZ reporter gene in the CasperUbx-lacZ P-element vector(Qian et al., 1991). All cloning details are available upon request. Germline transformation was performed essentially as described by Mihaly et al.(Mihaly et al., 1997). Positions of the constructs are given in nucleotide numbering along the BX-C sequence of Martin et al. (Martin et al.,1995). The coordinates of each of the fragments used for the transgenic assays is listed in Table 1.

The iab-5,6^J81 and 5,6^J82 deletions were recovered during an experiment aimed at isolating local hops of the bluetail transposon (Galloni et al., 1993; Mihaly et al.,1997). A chromosome (called J82.blt) containing the deletion present in iab-5,6^J82 and the starting bluetail transposon was first isolated. In order to assign the phenotype associated with the deletion alone, we remobilized the transposon from J82.blt Among the progeny,we recovered iab-5,6^J82 (a clean excision) and the iab-5,6^J81, which probably resulted from an aborted gene conversion event. The iab-6,7^IH, Fab-6,7¹ and Fab-6,7² deletions were recovered by imprecise excision of the bluetail transposon. Local hops and imprecise excision of the iab-7^blt transposon were screened on the basis of their failure of complementation with the iab-4,5,6^DBchromosome. The extents of the deletions were mapped by whole-genome Southern and by PCR analysis. Fab-3,5^DVwas recovered on the basis of its dominant phenotype during an excision screen of the HCJ200 P-element present in iab-4 (Bender and Hudson 2000). The iab-7²⁸⁰ deletion was recovered by mobilization of the fs(3)5649 transposon, as described by Barges et al. (Barges et al.,2000).

As meiotic recombination attempts to recover the R73 deletion separated from the Fab-7¹ deletion had failed, we generated a new R73 deletion by gene conversion using the fs(3)5649 chromosome(Gloor et al. 1991). Convertant chromosomes carrying the R73 deletion were screened by PCR. TheΔ330 deletion was also recovered during the screen, probably resulting from an abortive conversion event.

Gene conversion initiated by the excision of the P-element in the iab-7^blt line was used to direct specific mutations in the Fab-7 boundary (for details, see Hogga et al., 2001). A deletion removing the sequence covering the minor nuclease hypersensitive site was created by first introducing two loxP sites in the NcoI and BsmI sites at positions 84,827 and 85,028, respectively. The deletion was recovered after a cross to flies expressing Cre(Siegal and Hartl, 1996).

In order to dissect genetically the 3′ cis-regulatory domain of the Abd-B gene, we isolated a number of deficiencies derived from the imprecise excisions of three P-elements [iab-7^blt, fs(3)5649 and HCJ200; see, respectively, Galloni et al.(Galloni et al., 1993), Barges et al.(Barges et al., 2000) and Bender and Hudson (Bender and Hudson,2000)]. Most of these deficiencies were recovered because of their homeotic transformations of the posterior abdominal segments.

The deficiency breakpoints were first determined by whole genome Southern analysis and then confirmed by PCR analysis. Their positions are reported along the DNA sequence (SEQ89E) (Martin et al., 1995). The homeotic transformations associated with these deficiencies are described below, specifically focusing on the features of the posterior abdominal segments in adult males. The Abd-B expression pattern observed in the ventral nerve chord of mutant embryos is also described. Fig. 1A shows the extents of these deficiencies. The iab cis-regulatory domains are positioned relative to the boundaries and data presented below.

Males homozygous for iab-7²⁸⁰ have their 7th abdominal segment completely transformed into a copy of A6(Fig. 2B). In mutant embryos,the Abd-B expression pattern in PS12/A7 is replaced by the PS11/A6-specific pattern (Fig. 3). The homeotic transformations and the Abd-B expression patterns observed in iab-7²⁸⁰ are exactly the same as those seen in iab-7^Sz, a mutation that removes the entire iab-7 cis-regulatory region between Fab-7 and Fab-8(Galloni et al., 1993). In agreement with this, the iab-7²⁸⁰ mutation deletes 20 kilobases (kb) between sequence positions 64,370 and 84,012(Fig. 1A), thus removing the entire iab-7 cis-regulatory region. On the proximal side, a second smaller deletion occurred, removing DNA belonging to iab-6, between 89,392 and 86,650 (Fig. 1A) The wild-type appearance of PS11/A6 in adults and embryos indicate that no essential function of iab-6 is affected by this deletion.

The iab-6,7^IH mutation deletes 26 kb (96,559-70,265; Fig. 1A), removing about three-quarters of the iab-6 region and nearly two-thirds of the iab-7 region. In flies homozygous for this deletion, both A6 and A7 are transformed to differing degrees towards A5, consistent with the removal of important iab-6 and iab-7 enhancers(Fig. 2D). The adult phenotype of this deletion shows that although iab-6 activity is completely abolished (as seen by the complete transformation of A6 into A5), iab-7 is still partially functional [revealed by the only partial transformation of A7 into A5 (Fig 2D)]. The Abd-B expression pattern in the embryonic nerve chord shows remarkable correspondence to these adult phenotypes(Fig. 3). In PS11/A6, where iab-6 is inactivated, Abd-B expression follows the PS10/A5-specific pattern. If iab-7 were also completely inactive, we would predict that the PS10/A5-specific Abd-B expression pattern would also be reiterated in PS12/A7. Instead, we observe a weaker version of the normal PS12/A7-specific pattern that corresponds to a level of expression intermediate between that of normal PS11/A6 and PS12/A7.

Comparing iab-6,7^IH to iab-7²⁸⁰allows us to map important regulatory elements within the iab-6 and iab-7 regions. The iab-7²⁸⁰ deletion does not alter iab-6 function, while iab-6,7^IH deletion is null for iab-6. This indicates that sequences crucial for iab-6 function lie between the proximal breakpoints of these two deletions (89,392 and 96,559; see Fig. 1A). Meanwhile, regarding iab-7, the comparison of these two deletions demonstrates that there are at least two separable elements important for iab-7 activity. One element uncovered by the iab-6,7^IH deletion and one located in between the distal breakpoints of the two deletions. Consistent with this, we and others had previously isolated a weak PS12 specific initiator element mapping between the distal breakpoints of these two deletions (67,072 and 67,806)(Zhou et al., 1999; Barges et al., 2000). We postulate the existence of another PS12-specific initiator element.

In order to further narrow down the regions containing these enhancers, we cloned the regions implicated by our genetic analysis in front of a Ubx-lacZ reporter gene and studied the expression pattern of the resulting constructs in transgenic flies.

Genetic analysis had localized the new iab-6 regulatory element to a ∼7 kb region between the proximal break points of the iab-6,7^IH and the iab-7²⁸⁰ deletions. When a ∼8 kb fragment approximating this region is placed in front of the Ubx-lacZ reporter construct (fragment 7.8H in Fig. 1B), it drives expression of the lacZ reporter in a parasegment-specific manner, with strong expression in PS11, PS13 and PS14 (data not shown, but see below). The parasegment-specific expression pattern driven by this element suggests that it is capable of reading the parasegmental address and is, therefore, an initiator element. By subcloning overlapping neighboring fragments into the same vector, we were able to narrow down the iab-6 initiator to a∼3 kb region contained within fragment 7.8H. Consistent with this, a∼2.8 kb fragment (2.8XN in Fig. 1B) from this region is fully able to recapitulate the PS-specific pattern of fragment 7.8H (Fig. 1D). We have therefore, named this fragment IAB6.

The IAB6 fragment is capable of maintaining strong lacZ expression in PS11 and PS13 at later stages of embryogenesis (see Fig. S1B in the supplementary material), indicating that the fragment contains a maintenance element in addition to the initiator. The presence of a maintenance element is further supported by the observation that transgenic flies carrying the IAB6 construct show pairing sensitive repression of the mini-white reporter gene in 30% of the lines (three out of 10 lines), a hallmark characteristic of MEs(Kassis, 1994; Gindhart and Kaufman, 1995; Pirrotta and Rastelli,1994).

To identify the second putative iab-7 activating element, we once again cloned overlapping fragments from the implicated area into the Ubx-lacZ reporter construct. A ∼3 kb fragment (3H; Fig. 1B) located at the proximal edge of the iab-7 cis-regulatory region is able to direct the early expression of the Ubx-lacZ reporter gene in PS12 and PS14, indicating that this element is also an initiator element(Fig. 1E). This expression pattern is very similar to that already described for another PS12 initiator element located about 10 kb away, at the other edge of the iab-7cis-regulatory region (Zhou et al.,1999; Barges et al.,2000). We will refer to these two initiator elements as IAB7a and IAB7b, with IAB7a being the more proximal element. It should be noted that the level of PS12 and PS14-specific expression of IAB7a is much higher than that of the previously described IAB7b element. In transgenic lines carrying IAB7a,the lacZ expression pattern fades away at later stages of embryogenesis, indicating that no maintenance element is present within this fragment.

Homozygous iab-5,6^J82 males have their A5 and A6 segments transformed into A4, indicating that both iab-5 and iab-6 are affected by this mutation(Fig. 2E). These adult homeotic transformations are paralleled by the absence of Abd-B staining in PS10/A5 and PS11/A6 in homozygous iab-5,6^J82 embryos(Fig. 3). In agreement with the complete loss of iab-6 function, the entire iab-6 domain is deleted. However, despite the fact that iab-5 function is inactive,the deletion affects only one-third of the iab-5 region(Fig. 1A; 83,791-106,497). This observation indicates that sequences necessary for iab-5 activity reside within the distal one-third of the iab-5 region. Consistent with this finding, Busturia and Bienz(Busturia and Bienz, 1993) had previously identified an IAB5 initiator in this region that can drive Ubx-lacZ expression in PS10, PS12 and PS14 (see also Fig. 1C).

The iab-5,6^J81 deletion shares its proximal endpoint with iab-5,6^J82, but it is much shorter and extends only 10.5 kb distally, removing the proximal one-third of iab-6(Fig. 1A; between 95,869 and 106,497). The transformation of PS10/A5 into PS9/A4 and the loss of Abd-B expression in PS10 corroborate that sequences covering the IAB5 initiator element are indispensable for iab-5 function(Fig. 2F; Fig. 3). Interestingly, iab-6 function is partly affected in iab-5,6^J81,even though the IAB6 initiator is untouched by this deletion. In embryos, the level of Abd-B expression appears normal in PS11/A6(Fig. 3). However, in homozygous adult males, there is a partial transformation of A6 into A5(Fig. 2F). Although our previous data indicates that the IAB6 initiator is required for iab-6activity, we must conclude, based on iab-5,6^J81, that it is not sufficient for iab-6 function and that there are other elements, probably adult tissue-specific enhancers that are required for full iab-6 activity.

The iab-4,5,6^DB deletion has already been described in previous reports (Karch et al.,1985; Celniker et al.,1990; Sanchez-Herrero,1991; Crosby et al.,1993). It removes DNA between map positions 95,464 and 124,218(Fig. 1A). Based on the PS-specific expression pattern of enhancer trap P-elements that are inserted in the BX-C (the HCJ200 line); Bender and Hudson(Bender and Hudson, 2000) have redefined the position of the iab-4 cis-regulatory region at position 127,367. Thus, iab-4,5,6^DB appears to remove most of iab-4, the entire iab-5 region and the proximal half of the iab-6 cis-regulatory region (Fig. 1A). In agreement with this, homozygous iab-4,5,6^DB males have their 4th and 5th abdominal segments transformed into the 3rd (Fig. 2G) (Karch et al.,1985).

Interestingly, the 6th abdominal segment in homozygous iab-4,5,6^DB males is partly transformed towards A3 or A4(Fig. 2G), indicating that the remaining half of iab-6 is only partially functional even though the IAB6 initiator is still present. The latter observation supports the conclusions drawn from iab-5,6^J81, which suggests that the IAB6 initiator is necessary but not sufficient for iab-6 function. This conclusion is further corroborated by the similarity of the embryonic Abd-B expression pattern in PS10/A5 and PS11/A6 of embryos homozygous for either iab-4,5,6^DB or iab-5,6^J81(Fig. 3). In both cases, there is no expression in PS10/A5 (both deletions remove iab-5), and Abd-B expression in PS11/A6 is very similar to the wild-type pattern.

We have isolated two deletions in the iab-7 region that are of special interest as they remove the region implicated in Promoter Targeting Sequence (PTS) activity. Zhou and Levine(Zhou and Levine, 1999)studied an element from the BX-C in reporter constructs. They suggested that this PTS element helps a distal enhancer bypass an intervening insulator to reach its target promoter. Zhou and Levine mapped the PTS to an 820 bp region deleted from the Fab-7^R73 chromosome. Unfortunately, this chromosome was isolated as a revertant of Fab-7¹(Gyurkovics et al., 1990) and contains the Fab-7¹ deletion, making phenotypic analysis difficult. We have recovered a chromosome carrying solely the 820 bp deletion from the Fab-7^R73 chromosome implicated in PTS activity,that we named iab-7^R73(Fig. 1A; deletion of DNA between 64,229 and 65,049; see Materials and methods). Fig. 2C shows a homozygous male. A5 and A6 acquire normal identity, while A7 is weakly transformed into A6. Thus, a clean removal of the PTS sequence only weakly impairs iab-7 function, with no effect on iab-5 and iab-6function.

We also isolated a smaller deletion (Δ330^iab7)that shares the same proximal breakpoint with iab-7^R73 but extends only ∼670 bp (Fig. 1A). As homozygous Δ330^iab7 males and females are indistinguishable from wild-type flies, sequences crucial for the iab-7^R73 phenotype must lie within the 152 bp region on the distal side of the iab-7^R73 deletion.

Recently, a second PTS element originating from iab-6 has been described (PTS-6) (Chen et al.,2005). This second PTS element was mapped to the minor DNase I hypersensitive site of Fab-7(Karch et al., 1994) (from positions 84,827 to 85,028). During the course of site-directed mutagenesis studies on the Fab-7 boundary and the iab-6 cis-regulatory domain, we isolated a ∼700 bp deletion (84,661-85,367) that removes the Fab-7 minor hypersensitive site and ∼500 bp of iab-6that we have named ΔHS^*iab6. AlthoughΔ HS^*iab6 entirely removes the new PTS-6 element, it is not associated with any abnormal phenotype.

Fab-6,7¹ and Fab-6,7² are two deletions associated with gain-of-function phenotypes with very similar breakpoints and phenotypes. Both deletions share the same distal deficiency breakpoint at position 83,791 kb (the site of insertion of the blttransposon). On the proximal side, however, the deficiency breakpoint of Fab-6,7¹ maps at position 101,096, while that of Fab-6,7² maps to position 102,687(Fig. 1A). Both of these deletions remove the Fab-7 boundary element along with the entire iab-6 domain. Interestingly, instead of showing a homeotic transformation of the 6th abdominal segment into the 5th, like iab-6,7^IH, Fab-6,7¹ and Fab-6,7² exhibit a striking dominant gain-of-function phenotype, in which, the 5th and 6th male abdominal segments are largely reduced in size (Fig. 4). In extreme cases (such as shown in Fig. 4C), homozygous males harbor only rudiments of the 5th and 6th abdominal segments. This phenotype corresponds to a strong transformation of A5 and A6 towards A7 (in wild-type males, the 7th abdominal segment present in embryos and larvae does not secrete any visible tergite or sternite cuticle). In the CNS of embryos, we find the same level and pattern of Abd-Bexpression in PS10/A5 and PS12/A7. These patterns are higher than the wild-type PS10/A5 expression pattern and lower than the pattern in PS12/A11(Fig. 4F).

Based on the similarity of these phenotypes to the previously isolated Fab-7 and Fab-8 deletions, we believe that the Fab-6,7 deletions must delete a putative Fab-6 boundary element that normally separates iab-5 from iab-6. The presence of this boundary element can be best observed by comparing the Fab-6,7 mutations to the iab-6,7^IH deletion. Although in the Fab-6,7 mutants, we observe a fusion between iab-5 and iab-7, in iab-6,7^IH mutants, iab-7 is still capable of behaving in an autonomous manner. These observations imply that there must be a boundary element mapping to the region between the proximal breakpoints of these deletions (96,559 and 101,096) that keeps these domains separate.

Comparing the extent of the Fab-6,7¹ (or Fab-6,7²) mutation with that of the iab-5,6^J82 mutation sheds light on the dominant gain-of-function frontabdominal phenotype. All three deletions share the same distal endpoint, indicating that the origin of the Fab-6,7¹ and Fab-6,7² GOF phenotypes must lie within the DNA segment that remains present in Fab-6,7¹ and Fab-6,7², and that is deleted in iab-5,6^J82. Interestingly, this DNA segment contains the IAB5 initiator element (see above)(Busturia and Bienz, 1993)that, in the Fab-6,7 mutants, now becomes juxtaposed to the complete iab-7 domain (the Fab-7 boundary is removed in the Fab-6,7 alleles, and in iab-5,6^J82). We believe that the dominant gain-of-function phenotype is caused by the IAB5 initiator element ectopically activating iab-7 in PS10, an interpretation corroborated by the observation that in both mutations, the PS10/A5 and PS12/A7 Abd-B expression levels are similar and higher than the level normally found in PS10 of wild-type embryos.

Surprisingly, in both Fab-6,7 deletion mutants, PS11 expression of Abd-B differs from that seen in PS10 and PS12. As the iab-6domain is entirely deleted in these flies, we expected to see an A7/A5 fusion pattern of expression similar to that seen in PS10 and PS12. Interestingly,this is not the case. The pattern of Abd-B expression in PS11 is greatly reduced from the level seen in its neighboring parasegments. As other large boundary mutants, such as the Fab-3,5^DV mutation described below, do not behave similarly, we cannot explain this result at this time.

The third deletion we identified that has a gain-of-function phenotype is Fab-3,5^DV. Fab-3,5^DV is a deletion derivative of the HCJ200 enhancer trap line that lies at the distal edge of iab-3 at position 127,367 (Bender and Hudson, 2000). An imprecise excision created a bidirectional deletion of 25 kb between positions 112,381 and 137,391, removing the Mcp boundary along with the entire iab-4 region and about 10 kb of the distal side of iab-3. Thus, Fab-3,5^DVfuses the proximal side of the iab-3 region to iab-5. The deletion does not affect the previously identified IAB3 initiator element(Simon et al., 1990). In Fab-3,5^DV, it appears that the juxtaposition of the IAB3 initiator element to iab-5 results in the ectopic activation of iab-5 in A3/PS8. Adult males have A3 and A4 transformed into a copy of A5 (Fig. 4A). Likewise, in embryos, the Abd-B expression pattern normally seen in PS10 appears ectopically in PS8 and PS9 (Fig. 4D).

We have extended the reporter gene analysis that allowed us to identify the IAB6 and IAB7a initiator elements, to scan the whole Abd-B 3′cis-regulatory region for additional enhancers and maintenance elements. We did this by inserting overlapping DNA fragments(Fig. 1B) spanning most of the region from iab-5 to iab-8 in front of the Ubx-lacZreporter gene in transgenic flies and studying either the expression pattern of the resulting constructs, or the pairing-sensitive repression of the mini-white cassette contained on the transgene (see Fig. S1 in the supplementary material).

The proper identity of PS10/A5 through PS13/A8 is determined by the anterior-to-posterior stepwise increase in the level of Abd-B. This pattern of expression is achieved through the action of four large cis-regulatory domains. Using reporter constructs, a number of groups have identified several types of regulatory elements from within the cis-regulatory region. These elements include initiator elements, maintenance elements(polycomb response elements/PREs), cell/tissue-type specific enhancers,chromatin insulators and promoter targeting sequences (PTSs) (for a review,see Maeda and Karch, 2006). Unfortunately, with the exception of the boundary elements, genetic proof for the role of each of these elements has been confused because of the lack of discrete mutations. Thus far, most of the mutations used for the genetic analysis were rearrangement breaks (Lewis,1978; Karch et al.,1985; Sanchez-Herrero et al.,1985; Celniker et al.,1990). Because these breaks remove extremely large regions of the cis-regulatory region and place foreign DNA in its place, the phenotypes resulting from these mutants are often complex and difficult to interpret. Here, we have described a collection of internal deletions within the Abd-B region of the BX-C. Using these deletions, we have been able to infer genetically the role of many of the elements previously identified. Our results, when combined with previous data, point to four important conclusions regarding the Abd-B region of the BX-C. First, as already discussed in prior publications (Karch et al.,1985; Celniker et al.,1990; Sanchez-Herrero,1991; Galloni et al.,1993; McCall et al.,1994; Bender and Hudson,2000) (for reviews, see Peifer et al., 1987; Maeda and Karch,2006), the Abd-B region is modular by nature and can be divided into discrete functional units or domains. Second, in a wild-type situation, each domain appears to be sufficient to specify the appropriate level of Abd-B expression in a particular parasegment. Third, each domain is itself modular and made up of a similar set of definable regulatory elements. Finally, the activity of each domain is absolutely dependent on the presence of an initiator element.

In 1978, E. B. Lewis first proposed a model for BX-C patterning that stated that once segment-specific functions are activated along the anteroposterior axis, they remain active in more posterior segments and contribute to the identity of those segments (Lewis,1978). His hypothesis was based primarily on the posterior to anterior transformations observed in flies lacking distal regions of the BX-C. For example, embryos that lacked all BX-C sequences distal to the bxddomain, had all abdominal segments posterior to the first abdominal segment(A1) transformed into copies of A1. The transformations toward A1 suggested that everything necessary for A1 development was present in all posterior segments and that it was an accumulation of other segment-specific factors that caused the deviation in segmental identity. In agreement with the Lewis model, we, and others, have previously shown that enhancer trap lines inserted in the BX-C and transgenic initiator constructs are expressed in a pattern that starts from a specific parasegment and continues in more posterior parasegments (Simon et al.,1990; Qian et al.,1991; Müller and Bienz,1992; Galloni et al.,1993; McCall et al.,1994; Zhou et al.,1999; Barges et al.,2000; Bender and Hudson,2000).

Although these findings strongly suggested that the domains that become active in the parasegment they specify remain active in the more posterior parasegments, our analysis of internal deletions in the BX-C indicate that anterior domains are not required for the specification of more posterior parasegments. This is best illustrated by deletions that remove most or all of a single cis-regulatory domain. For example, the iab-7^Szand iab-7²⁸⁰ deficiencies both delete the entire iab-7 domain. If iab-7 was required for the identity of the more posterior parasegment, PS13/A8, as predicted by the Lewis model, one might expect that deleting iab-7 would affect PS13 identity. This is not what is seen. Flies homozygous for either of these mutations have PS12/A7 transformed into PS11/A6, while PS11/A6 and, more importantly, PS13/A8 are not affected in any detectable way. This phenotype indicates that although iab-7 is absolutely required in PS12/A7, it is dispensable for the identity of all other parasegments. However, PS12 is still transformed into a perfect copy of PS11, indicating that iab-6 remains capable of functioning in PS12/A7. Therefore, we believe that more anterior domains remain capable of functioning in more posterior parasements, but only in the absence of an active posterior domain. This is consistent, for example, with the observation that bxd⁺ is required in segments more posterior to PS6/A1, as bxd is the most posteriorly activated regulatory domain of the Ubx gene.

Internal deletions that affect more than one cis-regulatory region confirm these conclusions. For example, in iab-6,7^IH, where both the iab-6 and the iab-7 domains are deleted, both A6 and A7 acquire an A5 identity. However, PS13/A8 identity is again not affected by the deletion even though, in this case, two more-anterior domains are missing. Once again, the more-anterior iab-5 domain remains capable of acting posterior to PS10/A5, but only in the absence of iab-6 and iab-7. Similar results are seen with both the iab-5,6^J81 and the iab-5,6^J82deletions, where PS12/A7 identity is unaffected in the absence of iab-5 (and also iab-6 for iab-5,6^J82).

Although it is possible that deletions of anterior domains lack a visible phenotype because of the masking effect of the stronger enhancers in more posterior domains, it is clear from our results that the graded Abd-Bexpression pattern cannot be accounted for solely based on the summation of individual regulatory domains. Instead, each regulatory domain is, by itself,sufficient to generate an appropriate pattern of Abd-B expression in a specific parasegment (see also Crosby et al., 1993).

The autonomous nature of the cis-regulatory domains implies the existence of elements within each domain that can independently sense and respond to positional cues. At present, the best candidates for serving these functions are among the enhancers identified by transgenic assays. Two basic types of enhancers have been found within the cis-regulatory regions of the BX-C. Some enhancers are capable of directing the expression of reporter genes in specific types of cells or tissues, irrespective of parasegmental boundaries. These enhancers are usually turned on relatively late during embryogenesis(Simon et al., 1990; Busturia and Bienz, 1993; Pirrotta et al., 1995). An example of this kind of enhancer is seen in fragment 4.4E from the iab-6 cis-regulatory region that is capable of driving Abd-Bexpression in the visceral mesoderm (see Fig. S1 in the supplementary material). Other enhancers, however, direct expression of reporter genes from the end of cellular blastoderm stage in a segmentally restricted manner,indicating that they are capable of responding to a specific segmental address(Simon et al., 1990; Qian et al., 1991; Müller and Bienz, 1992; Shimell et al., 2000). These include the previously identified IAB5, IAB7b and IAB8 enhancers that express from PS10, PS12 and PS13, respectively(Fig. 1C-F)(Busturia and Bienz, 1993; Zhou et al., 1999; Barges et al., 2000). Here, we described two new elements of this type: IAB6 with a PS11 specificity and IAB7a with a PS12 specificity. Taken together, these findings demonstrate that each cis-regulatory region contains at least one PS specific enhancer. These specific enhancers have been previously referred to as `initiators'.

Analysis of the phenotype of internal deletions suggests that the removal of the IAB5 or IAB6 initiator completely abolishes the function of iab-5 or iab-6, respectively. For example, although the iab-5,6^J81 deletion leaves intact about two-thirds of the iab-5 domain, it deletes the IAB5 initiator element and is completely null for iab-5 function. The iab-6,7^IH is another example. This deletion removes the IAB6 initiator and, despite the fact that one quarter of iab-6 is not affected by the deletion, iab-6is completely inactivated. These results suggest that the IAB5 and IAB6 initiator elements are indispensable for iab-5 and iab-6function, respectively.

Although the removal of initiator elements completely abolishes the function of a regulatory domain, the presence of an initiator is not sufficient to recapitulate the activity of the whole domain. This is clearly seen in the iab-5,6^J81 and iab-4,5,6^DBdeletions. In both of these mutants, the IAB6 initiator is still present, yet there is a partial transformation of A6 towards A5. These results highlight the importance of other regulatory elements, such as cell-type-specific enhancers, within the iab regions that help control Abd-Bexpression.

Thus far, all available data suggest that iab-5 and iab-6each harbor only one initiator element. The situation in iab-7 and iab-8, however, appears more complex. In both cases, two initiator elements have been identified in each domain. Comparison of the iab-7²⁸⁰ and the iab-6,7^IH phenotypes suggests that within iab-7, the two initiators contribute to the final Abd-B expression pattern. Although iab-7²⁸⁰is completely null for iab-7, the iab-6,7^IHmutation leaves the IAB7b initiator intact and does not completely inactivate iab-7. This is consistent with the idea that complete loss of iab-7 function can be achieved by only the removal of both PS12 initiators in cis.

iab-8 may represent another example of a regulatory region containing more than one initiator element. An IAB8 initiator has been identified between the Fab-8 boundary and the 3′end of the Abd-B transcription unit (Fig. 1) (Zhou et al.,1999; Barges et al.,2000). However, mutations that separate the IAB8 element from the Abd-B transcription unit have no detectable effect on iab-8function, suggesting the presence of a second element within iab-8that can compensate for the loss of the IAB8 initiator. Indeed, Estrada et al.(Estrada et al., 2002)described the existence of a second iab-8-specific initiator element upstream of the Abd-Bm promoter.

Within the BX-C, Abd-B expression is strictly restricted to specific parasegments in all tissue. However, in transgenic constructs, only initiator elements seem to sense a parasegmental address. Outside the complex,the cell- and tissue-specific enhancers found within the BX-C do not. For example, although an isolated fragment, 11X, from the iab-7 region directs the expression of a Ubx-lacZ construct in the epidermis of each parasegment (see Fig. S1 in the supplementary material), normal Abd-B expression in the epidermis is restricted to the parasegments posterior to PS12. The discrepancy between the transgenic and the native expression of Abd-B implies that something within each domain must act to control the activity state of the entire domain. Because the initiator elements seem to be required for the function of each domain and can sense parasegmental position, we believe that initiator elements act as domain control regions that sense a parasegmental address and relay this information to all other elements within the domain. Although the exact mechanism of this transmission is unknown, it is clear that a part of the mechanism involves the regulation of maintenance elements.

The activity pattern of cis-regulators is set up early in embryonic development by the gap and pair rule genes that specifically instruct initiator elements (Irish et al.,1989; Müller and Bienz,1992; Qian et al.,1991; Shimell et al.,1994; Casares et al.,1995). By the time of mid-embryogenesis, the products of the gap and pair rule genes disappear. However, the activity of cis-regulatory elements is continuously required until the end of development. The activity state of each regulatory domain is maintained throughout the rest of development by maintenance elements(Müller and Bienz, 1991; Simon et al., 1993; Chan et al., 1994; Chiang et al., 1995; Fritsch et al., 1999) [for definition of ME see Brock and van Lohuizen(Brock and van Lohuizen,2001)]. We believe that maintenance elements respond to the activity state of initiator elements and modify the chromatin structure of each domain accordingly. Consistent with this, maintenance elements (MEs) have generally been found to contain binding sites for Polycomb-Group(Pc-G) and trithorax-Group proteins (Trx-G)(Strutt et al., 1997; Brown et al., 1998; Fritsch et al., 1999; Tillib et al., 1999; Horard et al., 2000; Busturia et al., 2001; Mishra et al., 2001; Hur et al., 2002; Dejardin et al., 2005). Both Pc-G and trx-G proteins are known to be involved in modifying chromatin structure, with Pc-G proteins being involved in compacting chromatin (silencing) and Trx-G proteins being involved in opening chromatin (for reviews, see Ringrose and Paro, 2004; Brock and Fisher, 2005).

Here, we have described two new maintenance elements in iab-6. The occurrence of multiple MEs/PREs in a single cis-regulatory region has been documented before and may represent the normal situation, rather than the exception (Chiang et al.,1995). The iab-5 cis-regulatory region, for example, also contains at least two MEs (Busturia and Bienz, 1993; Busturia et al.,1997; Müller et al.,1999). Although there has been only one ME identified within each of the iab-7 and iab-8 cis-regulatory regions(Hagstrom et al., 1997; Mihaly et al., 1997; Barges et al., 2000), it is likely that they too, contain more than one ME. Supporting this, a deletion of the single known iab-7 ME (PRE) has very little consequence on the phenotype of PS12/A7, suggesting a functional redundancy with another ME within iab-7 (Mihaly et al.,1997).

Like cell-type-specific enhancers, maintenance elements in transgenic constructs do not seem to be parasegmentally regulated. Indeed, one of the most common assays used to identify new maintenance elements requires them to maintain a pattern of expression initiated by the bx initiator(Müller and Bienz, 1991; Simon et al., 1993; Chan et al., 1994; Fritsch et al., 1999; Busturia et al., 2001). Therefore, like that of cell-type specific enhancers, ME function seems to be dictated by initiator elements.

Confirmation of the role of initiator elements in the control of both cell-type specific enhancers and MEs, comes from the analysis of a specific class of gain-of-function mutants. These mutations result from the simple deletions of regions located between the cis-regulatory domains. As a result of these deletions, neighboring cis-regulatory domains fuse into single units of gene regulation. Thus, these regions have been called domain boundaries to highlight their role in separating domains. An example of a boundary deletion is the Fab-7 deletion (Gyurkovics et al., 1990; Mihaly et al.,1997). In the Fab-7 deletion, less than 800 bp are deleted yet there is an almost complete transformation of PS11/A6 into PS12/A7. Surprisingly, the cells not transformed towards a PS12-type fate do not become PS11 like, but PS10 like. Normally in PS11/A6, iab-6 is active while iab-7 is silent. In Fab-7 mutants, one gets a fusion of the two domains. The result of this fusion is that there are two different parasegment specific initiator elements in the same domain. We believe that the resulting phenotypes stem from the competition between these two initiator elements on the various enhancers and maintenance elements. In most cells in PS11, the iab-6 initiator correctly reads the parasegmental address as PS11 and instructs all enhancers and maintenance elements (including the stronger iab-7/PS12 enhancers) to turn on Abd-B expression. Because of the iab-7 enhancers, there is an overexpression of Abd-B in PS11 to the level of PS12. Meanwhile,in other cells, the iab-7 initiator(s) correctly reads the parasegmental address as PS11, remains silent, along with other elements of the domain (including the elements originating from iab-6).

Although the genetic evidence highlighting a role for the boundary regions in Abd-B gene regulation is quite clear, the mechanism by which boundaries function is still mysterious. In transgenic constructs, boundaries act as insulator elements: elements capable of blocking enhancer-promoter interactions when intercalated between them(Hagstrom et al., 1996; Zhou et al., 1996; Zhou et al., 1999; Barges et al., 2000). The role of boundaries as elements capable of separating domain specific enhancers and silencers may allude to the validity of this transgenic assay. However, it is clear that within the BX-C, boundaries cannot solely act as insulator elements, and that there must be some mechanism by which these insulators can be bypassed when in the proper context.

An element deriving from the region next to the Fab-8 boundary has been proposed to aid distal enhancers in bypassing intervening boundaries. This element has been termed the promoter targeting sequence (PTS)(Zhou and Levine, 1999). In transgenic constructs, the PTS permits distal enhancers to overcome the insulating effect of Fab-8. However, the in situ function of the PTS is much less clear. Part of the reason for this is that some of the phenotypes originally attributed to a PTS deletion, such as the partial transformation of A5 towards A4, are, in fact, the consequence of the haplo-insufficiency of the Abd-B deletion used by Zhou and Levine. Another factor complicating the evaluation of the function of the PTS in the BX-C is that, until now, the only available mutation that deletes the PTS region was a revertant of Fab-7¹ (Fab-7^R73), and therefore carried a second deletion in the Fab-7 region.

Here, we described two mutations that remove putative PTS elements. The first is a mutation that deletes the R73/PTS-7 region without affecting the rest of the BX-C. And the second is a deletion that removes the newly described PTS-6 element from iab-6 (Chen et al., 2005). In both cases, the phenotypes associated with the PTS deletions are relatively small. In contrast to the phenotype of the double deletion, Fab-7^R73, flies homozygous for the single iab-7^R73 deletion exhibit only a rather mild transformation of A7. The morphology of A6, however, is wild type in these flies, indicating that the R73 region has only a weak regulatory activity that is restricted to iab-7. Meanwhile, theΔ HS^*iab6 deletion that removes the entire PTS-6 element is completely wild type in appearance. Taken together, our results indicate that if the PTS hypothesis is correct, each domain must contain multiple, redundant PTS elements.

Based on our data and the data of others(Simon et al., 1993; Crosby et al., 1993; Pirrotta et al., 1995), we propose a model to summarize our current understanding of how the BX-C functions (Fig. 5). We believe that each domain autonomously controls the expression of Abd-B within one parasegment. Accordingly, each regulatory domain in the BX-C is a modular array of all of the elements necessary for Abd-B expression in a particular parasegment. Key to the functioning of each domain is the initiator element. Acting as a centralized domain control region, the initiator element determines whether the appropriate gap and pair-rule gene products are present to justify the activation of the domain. If the proper positional address is read, the initiator then sends a signal to the rest of the domain to activate various enhancers. Because a domain lacking an initiator seems to be silenced,we imagine that this signal is made up primarily of an inhibitory signal to the ME/PREs. With the PREs placed into a non-silencing state, the enhancers become free to activate Abd-B. If the proper parasegmental address is not found, then the Pc-G proteins silence the domain. Throughout all of this,the domain boundaries keep each domain independent and free from competing influences.

Despite the fact that a rough map of the `regulatory landscape' of the Abd-B domain is now available, a number of questions regarding the function of the regulatory elements remain to be answered. These include the problem of how different regulatory elements within individual domains interact and communicate. For example, what is the nature of the signal from initiator to ME? One possibility currently being discussed is the idea that transcription across PREs might be the inhibitory signal(Bender and Fitzgerald, 2002; Hogga and Karch, 2002; Rank et al., 2002; Schmitt et al., 2005). Although it is still unclear whether transcription plays an instructive role in BX-C regulation, it is interesting to speculate that initiators may in fact generate transcription across PREs to inactivate them. Until recently,answering these types of questions within the BX-C have been unreasonable. Perhaps now, through the use of targeted mutagenesis and gene conversion, we will finally be able to tackle precise questions in the BX-C. Indeed, our results raised many questions, now ripe for investigation.

We thank Kirsten Hagstrom and Paul Schedl for the identification of the iab-6PRE. We also thank Eva Favre and Jorge Faustino for excellent technical assistance. This work was supported by the Swiss National Foundation, by the Human Frontier Science Organization and by the Sate of Geneva. J.M. is an EMBO/HHMI Scientist and a Bolyai János Research Scholar. H.G. is supported by grants from OTKA and by the NIH as a subcontractor.