Functional dominance among Hox genes: repression dominates activation in the regulation of dpp

Hox genes are used by all animals to create morphological diversity along their rostro-caudal body axis (Beeman et al., 1993; McGinnis and Krumlauf, 1992; Wang et al., 1993), and probably by evolution to create morphological diversity among animal species (Averof and Akam, 1995; Averof and Patel, 1997; Carroll, 1995). A large body of literature has been published describing the structure, expression patterns and mutant phenotypes of Hox genes in different species. The analysis of Hox mutant phenotypes has demonstrated the conserved functions of Hox genes in axial patterning; loss-of-function mutations generally transform posterior structures into copies of anterior ones.

A poorly understood aspect of Hox function is how different Hox genes compete for the control of positional identity. In many instances where non-paralogous Hox genes are co-expressed, it is the more posteriorly expressed Hox gene that is the one controlling segment identity. This is perhaps most evident in ectopic expression experiments carried out in Drosophila and in Mus. In these experiments segmental transformations are often observed anterior but not posterior to the normal expression domain of the misexpressed Hox gene (Gibson and Gehring, 1988; Gonzalez-Reyes and Morata, 1990; Lufkin et al., 1992; Mann and Hogness, 1990; Struhl, 1983; reviewed by Botas, 1993; Duboule and Morata, 1994). For example, generalized overexpression of the murine gene Hoxd-4 causes homeotic transformations of segments anterior but not posterior to its normal expression domain (Lufkin et al., 1992). Likewise, ubiquitous expression of the Drosophila Hox gene Ultrabithorax (Ubx) transforms parasegments anterior, but not posterior, to its normal anterior boundary of expression (Gonzalez-Reyes and Morata, 1990; Mann and Hogness, 1990). This functional dominance is known as phenotypic suppression (Gonzalez-Reyes and Morata, 1990; Gonzalez-Reyes et al., 1990) or posterior prevalence (Duboule, 1991). There are cases reported, however, in which the anterior Hox gene is the dominant one (Heuer et al., 1995) or in which both Hox genes cooperate (Wang et al., 1993) or show dose-dependent competition (Lamka et al., 1992) for determining segment identity. A shortcoming of these observations is that they are made at the anatomical level; the scarcity of direct targets of Hox regulation has prevented the investigation of functional dominance of Hox genes at the molecular level.

One of the better characterized targets of Hox regulation is the Drosophila gene decapentaplegic (dpp). dpp encodes a member of the TGF-β superfamily of proteins (Padgett et al., 1987). In the visceral mesoderm, dpp is expressed in parasegment 7 (PS7), where it is required for the formation of the second midgut constriction (Immerglück et al., 1990; Panganiban et al., 1990). Expression of dpp is positively regulated by Ubx in PS7, and negatively regulated by abd-A in PS8-12 (Hursh et al., 1993; Immerglück et al., 1990, Panganiban et al., 1990; Reuter et al., 1990). Regulation of dpp by Ubx and abd-A takes place through a 680-bp visceral mesoderm-specific enhancer (dpp674). This enhancer contains UBX/ABD-A protein binding sites defined by DNase I protection assays (Capovilla et al., 1994; Manak et al., 1995; Sun et al., 1995). Ubx regulation of dpp674 is direct, as shown in experiments in which expression from a mutated enhancer is reconstituted by compensatory changes in the UBX protein that alter its DNA-binding specificity (Capovilla et al., 1994; Sun et al., 1995).

Here we use dpp674 to investigate functional dominance among Hox genes molecularly. We find that posterior prevalence does not adequately describe the competition of Hox genes for dpp674 regulation. Instead we find that whenever a Hox gene functions as a repressor, it prevails over others that function as activators. In addition, we show that ABD-A represses dpp674 directly through four sites located in the 3′ half of the dpp674 enhancer. The ability of each of these sites to mediate repression does not correlate with their in vitro binding affinity to ABD-A. Finally, we found that abd-A is able to activate expression through the 5′ half of dpp674, although normally repression by ABD-A through the 3′ half prevails over activation. These observations suggest that a molecular mechanism underlying posterior prevalence is repression prevailing over activation.

All lacZ reporter constructs have been generated using the CaSpeR-hs43-AUG-lacZ vector (Chab; V. Pirrotta, unpublished data). All enhancer fragments are derived from dpp1400, which is a 1400-bp XhoI fragment of dpp mapping to 75.2-76.6 in the dpp map (Capovilla et al., 1994; St Johnston et al., 1990). dpp674 is a 680-bp BamHI-XhoI fragment in the 3′ half of dpp1400. dpp674lacZ was made by subcloning dpp674 in the BamHI-XhoI sites of Chab. dpp265 is a 267-bp BamHI-PstI fragment in the 5′ half of dpp674; dpp419 is a 419-bp PstI-XhoI fragment in the 3′ half of dpp674. dpp265lacZ and dpp419lacZ were generated by cloning the respective fragments in the BamHI-PstI (dpp265) or PstI-XhoI (dpp419) sites of pBluescript KS(+), excising them with BamHI and XhoI and subcloning them in the BamHI-XhoI sites of Chab. Mutant enhancer fragments have been cloned in Chab in the same manner as the wild-type ones.

To obtain the Ubx-Ia⁺ cDNA, a nearly full-length Ubx-Ib cDNA was excised from pfX3712 (Beachy et al., 1988) as a partial EcoRI fragment and subcloned in the EcoRI site of pBluescript KS(+) to generate BS-Ib3712. The 1.9-kb StuI-HpaI fragment of BS-Ib3712 was replaced with the same fragment from the Ubx-Ia⁺ cDNA (Lopez and Hogness, 1991). UAS:Ubx⁺ was generated by cloning this full-length Ubx-Ia⁺ cDNA in the EcoRI site of pUAST, in the 5′→3′ orientation appropriate for protein synthesis.

UAS:abd-A⁺ flies were obtained from Greig and Akam (1993). To generate the UAS:abd-A^K50 construct the abd-A cDNA was excised from the UAS:abd-A⁺ plasmid (Greig and Akam, 1993) with NotI and KpnI and subcloned in pBluescript KS(+). The 50^th codon of the homeodomain (CAG) was mutagenized in the first position to encode Lys (AAG) and the resulting NotI-KpnI abd-A^K50 cDNA was cloned in the NotI/KpnI sites of pUAST.

Site-directed mutagenesis was carried out using the Muta-Gene Phagemid in vitro mutagenesis kit (Bio-Rad, Richmond, CA). The dpp1400 fragment cloned in pBluescript KS(+) (Capovilla et al., 1994) was used as a template to generate the dpp674^1-4BCD enhancer. The mutant enhancers were sequenced and cloned in Chab as described above. Below are shown the oligonucleotides used to mutagenize each site, in comparison with the wild-type sequence (altered nucleotides indicated in bold):

site 1: CAGTTATGGTGGCCATTAAGTTTTATCGATGGCGC;

site 1^BCD: CAGTTATGGTGGGGATTAGATCTTATCGATGGCGC;

site 1^RE: CAGTTATGGTGCCCGGGGAGTTTTATCGATGGC;

site 2: GCAATTCACACCCAATTAGTAATAAATTTG;

site 2^BCD: GCAATTCACACGGGATTAGTAATAAATTTG;

site 2^RE: GCAATTCACAGCGGCCGCGTAATAAATTTGAATGC;

site 3: GATCAAAGGCCTATCAATTAGCACCCATTTCG;

site BCD: GATCAAAGGCCTAGGGATTAGCACCCATTTCG;

site 3^RE: GATCAAAGGCCTAGGGCCCCGCACCCATTTCG;

site 4a^wt: CGATTTCCCCATGCCCATTTGGCCGTGCAATGTTTG;

site 4a^RE: CGATTTCCCCATGCGGCCGCGGCCGTGCAATGTTTG;

site 4b: CGTGGAGTACTACCCATTTGGCTTCCCATTTCG-ATTTCCCCATGCC;

site 4b^BCD: CGTGGAGTACTAGGGATTAGGCTTGGGATTACG-ATTTCCCCATGCC;

site 4b^RE: CGTGGAGTACTAGGGCCCGGGCTTGGGCCCCCG-ATTTCCCCATGCC.

In all cases, eggs were laid for 6 hours and aged for 6-8 hours before fixation (all at room temperature). Mouse anti-β-galactosidase antibodies (Promega) were used at a 1:4000 dilution. Histochemical detection of β-galactosidase was carried out using biotinylated horse anti-mouse immunoglobulin (IgG) and avidin-horseradish peroxidase (Vectastain Elite kit, Vector labs). At least two different lines of each construct were utilized for the experiments reported. Particular care was taken in equalizing the conditions of staining of embryos within Figs 2 and 4. In addition, double blind tests were carried out by independent investigators to classify the embryos of Fig. 4 with respect to the level of expression in PS8-12 relative to PS7.

Hox alleles Ubx^9.22, Ubx¹⁰⁹, Scr^C1, Antp^Ns+rvC3, Ubx^MX12, abd-A^M1 and Abd-B^M8 are described in FlyBase (http://www.morgan.harvard.edu). iab-2^D24 is a null abd-A allele provided by E. B. Lewis. Males of Hox mutant stocks were crossed to females carrying the dpp674lacZ transgene on the II chromosome. Females heterozygous for dpp674lacZ and for the Hox mutant chromosome were crossed to males of the corresponding Hox mutant stock and the progeny embryos analyzed. Mutant embryos were selected for their abnormal pattern of lacZ expression.

For the GAL4/UAS experiments, stocks containing the appropriate lacZ and UAS transgenes balanced with CyOwglacZ or TM6bUbxlacZ balancer chromosomes were generated by standard crosses. This allowed us to determine unambiguously the genotype of the experimental embryos. Males of these stocks were crossed to twiGAL4; 24BGAL4 virgin females and the progeny embryos were analyzed.

The footprinting probes of Fig. 3 were obtained from the transformation plasmids dpp674lacZ, dpp674^2,3,4RElacZ, dpp674^1,3,4RElacZ, dpp674^1,2,4RElacZ, dpp674^1,2,3RElacZ. These plasmids were digested with XhoI and end-labeled by filling with Klenow. The probes were released by digesting with BamHI. ABD-A was expressed from construct pET3a/abd-A (Appel and Sakonju, 1993) in BL21 (DE3) LysS E. coli cells. The extract was prepared as described by Appel and Sakonju (1993), except that the first 2 M guanidine hydrochloride pellet was resuspended in 2 M urea buffer. After spinning at 10,000 g for 10 minutes, the supernatant was used without further purification.

DNase I footprinting and chemical cleavage sequencing reactions were carried out as described by Vachon et al. (1992). The reactions were electrophoresed on a 5% Long Ranger (J. T. Baker) gel containing 8 M urea.

Expression of dpp674lacZ was monitored in different Hox mutant backgrounds. Following the ubiquitous expression of Ubx in the visceral mesoderm, dpp674lacZ expression is activated ectopically, but only anterior to the Ubx expression domain (Fig. 1B). This is in agreement with posterior prevalence: the function of the posteriorly expressed Hox gene (abd-A) overcomes the function of the anteriorly expressed one (Ubx). In this case, repression by abd-A prevails over activation by Ubx. Overexpression of abd-A represses dpp674lacZ expression completely (Fig. 1C). Repression of dpp674lacZ expression by abd-A is not merely a consequence of abd-A repressing Ubx because (1) ectopic Ubx expression activates dpp674lacZ anteriorly but not posteriorly in the abd-A domain (Fig. 1B; see also Capovilla et al., 1994; Manak et al., 1995); (2) expression of dpp674lacZ is severely reduced but not eliminated in Ubx mutants (Capovilla et al., 1994), whereas it is eliminated after abd-A overexpression (Fig. 1C).

However, regulation of dpp674lacZ expression by other Hox genes does not conform to the idea of posterior prevalence. In embryos lacking Ubx and abd-A but not Abd-B activity, dpp674lacZ is activated in the Abd-B domain (PS11-12; Fig. 1D). As shown in Fig. 1E, Abd-B is responsible of this expression, because it is not observed in embryos lacking Abd-B function. This result was unexpected because normally neither dpp nor dpp674lacZ are expressed in PS11-12. Because Abd-B is expressed in the posterior portion of the abd-A expression domain (PS11-12) (Tremml and Bienz, 1989), this result implies that the activating function of Abd-B is normally suppressed by the repressing function of abd-A. Thus, repression prevails over activation and the most posterior Hox gene (Abd-B) does not dominate in the regulation of dpp.

A mutant dpp674 enhancer in which the homeodomain binding sites 1-4 have been substituted with restriction enzyme sequences (dpp674^1-4RE) drives lacZ expression in PS7-12 (Fig. 2A). This suggests that sites 1-4 are required for repression in the abd-A expression domain. abd-A may repress dpp directly by binding to the dpp674 enhancer in vivo, or indirectly through another homeodomain-containing protein. This latter possibility is suggested by observations that indicate that some targets of Hox regulation encode homeodomain proteins themselves, e.g. Distal-less (Dll; Vachon et al., 1992), empty spiracles (ems; Jones and McGinnis, 1993) and apterous (ap; M. Capovilla and J. Botas, unpublished results).

To investigate whether ABD-A represses dpp expression by binding to sites 1-4, we used the strategy of altering the binding sites and making compensatory mutations in the protein to restore high affinity binding (Capovilla et al., 1994; Schier and Gehring, 1992). The cores of sites 1-4 were changed to TAATCCC sequences, generating the mutant dpp674^1-4BCD enhancer. dpp674^1-4BCDlacZ expression is very similar to that directed by dpp674^1-4RE described above (compare Fig. 2A with B). TAATCCC sequences are bound with low affinity by Antennapedia-type homeodomain proteins like ABD-A, but proteins containing lysine instead of glutamine in position 50 of the homeodomain (i.e. ABD-A^K50) bind with high affinity to these sites (Hanes and Brent, 1989; Treisman et al., 1989). An abd-A^K50 transgene was expressed throughout the visceral mesoderm using a twiGal4; 24BGal4 line, and its effects on dpp674^1-4BCDlacZ expression were studied. We found that dpp674^1-4BCDlacZ expression in PS8-12 is eliminated in approximately 70% of the embryos of two independent dpp674^1-4BCDlacZ lines tested (Fig. 2C). The remaining 30% (approx.) embryos show weak posterior expression (not shown). In contrast, dpp674^1-4BCDlacZ expression in PS8-12 is unaffected in control embryos expressing a abd-A⁺ transgene from the same driver (Fig. 2D); this is consistent with the low affinity binding of ABD-A⁺ to sites 1-4BCD (not shown). Thus, we conclude that ABD-A regulates dpp directly, and not via an intermediate homeobox-containing gene.

Surprisingly, the control embryos also show ectopic anterior activation of lacZ, suggesting that ABD-A activates transcription through the 5′ portion of dpp674 (see last section of Results).

To gain insight into the mechanisms of repression by ABD-A, we determined whether DNA-binding affinity of ABD-A in vitro correlates with its strength of repression in vivo. DNase I footprinting experiments show that ABD-A binds to sites 1 and 2 with higher affinity than to sites 3 and 4 (Fig. 3A). Fig. 3 also shows that the binding affinity of ABD-A for individual binding sites is not altered by mutating adjacent sites (compare Fig. 3B-E with A). The level of protection of site 5 (wild type in all constructs) can be used as a control for the amount of ABD-A in each experiment (asterisks in Fig. 3A-E). The concentration of ABD-A required to protect each site is the same whether it is surrounded by wild-type or mutant sites. This indicates the absence of ABD-A cooperativity in binding to nearby sites on this enhancer.

To assess the roles of individual sites in mediating ABD-A repression in vivo, each of the mutated sites 1-4 were reverted to wild type in the context of dpp674^1-4RElacZ reporter gene construct. Embryos heterozygous for each of these lacZ reporter constructs and embryos heterozygous for the dpp674^1-4RElacZ transgene were stained under the same conditions and the levels of lacZ expression were compared (Fig. 4). The presence of wild-type sites 1 (high-affinity site) or 4 (low-affinity site) almost eliminate lacZ expression in the abd-A domain (Fig. 4B,E, respectively). In contrast, wild-type site 2 (high-affinity site) has a weak effect and site 3 (low-affinity site) does not mediate repression at all (Fig. 4C,D, respectively).

In summary, the ability of each binding site to mediate repression is not a function of its affinity for ABD-A measured in vitro.

The experiments described above show that abd-A represses dpp674 expression through the binding sites located in its 3′ half. These experiments also indicate that abd-A is able to activate through the 5′ half of dpp674 (Fig. 2D).

To further investigate the ability of abd-A to activate through one portion of the enhancer and to repress through another portion, we divided dpp674 in two parts. The 3′ portion of dpp674 (dpp419) mediates activation by UBX in PS7 and repression by ABD-A in PS8-12 (Fig. 5A and data not shown). However, abd-A does not repress expression from dpp265 (the 5′ portion of dpp674, see Fig. 5B). lacZ expression from dpp265 is very similar to lacZ expression from a mutant dpp674 in which sites 1-4 are mutated (compare Fig. 5B with Fig. 2A,B). In principle, lacZ expression in the abd-A domain may be explained in two simple ways. One possibility is that abd-A may be unable to repress the activity of activators present in its spatial domain. The existence of activators other than Ubx is suggested by the study of dpp674lacZ expression in Hox mutants. For example, in Ubx mutants (Capovilla et al., 1994), or in embryos mutant for all the Hox genes expressed in the visceral mesoderm (Fig. 1E), dpp674lacZ expression is reduced but not completely eliminated. An alternative possibility is that abd-A might be itself the activator responsible for expression in PS8-12. To distinguish between these possibilities, we examined lacZ expression driven by dpp265 in embryos lacking abd-A function, and in embryos after ectopic abd-A expression. To investigate the consequences of lack of abd-A function on dpp expression it is necessary to remove Ubx and abd-A functions in the same embryos, because Ubx activates dpp and is itself repressed by abd-A (Bienz and Tremml, 1988). Fig. 5C shows that in embryos mutant for Ubx and abd-A, dpp265lacZ expression is eliminated in PS8-10, the portion of the abd-A expression domain that does not overlap with the Abd-B domain. Expression is observed in the Abd-B domain (PS11-12) and is dependent on Abd-B function (not shown). Fig. 5D shows that dpp265lacZ expression is extended anteriorly after ectopic abd-A expression. In summary, these experiments confirm that abd-A is the activator responsible for dpp265lacZ expression in PS8-12. Interestingly, dpp265 does not mediate activation by abd-A when adjacent to the dpp419 sequences (Fig. 1C), an observation that is reminiscent of posterior prevalence.

We used dpp674, a direct target of Hox regulation (Capovilla et al., 1994; Sun et al., 1995; this paper), to investigate how Hox proteins compete for the regulation of common target genes. In accordance with the idea of phenotypic suppression/posterior prevalence, abd-A (functioning as a dpp674 repressor) prevails over the more anterior Ubx (that functions as an activator). However, in disagreement with posterior prevalence, the repressor abd-A prevails over the more posterior Abd-B that functions as an activator. These results suggest that a molecular mechanism underlying posterior prevalence is the dominance of repression over activation. This hypothesis is supported by the observation that abd-A functions as an activator through the 5′ portion of the enhancer and as a repressor through its 3′ portion. When these portions are fused together in the full enhancer, repression by abd-A prevails over activation. Although repression by ABD-A^K50 prevails over activation by ABD-A⁺ in PS8-12, it does not completely prevail over activation by UBX in PS7 (see Fig. 2C). One possibility is that factors other than UBX contribute to activation of dpp specifically in PS7, as suggested by dpplacZ expression in Ubx abd-A mutant embryos (see Fig. 1D).

Our findings suggest the possibility that other cases of functional dominance may be explained in terms of Hox proteins functioning as repressors prevailing over Hox proteins functioning as activators. For example, in accordance with posterior prevalence, repression of Distal-less (Dll) by Ubx prevails over Dll activation by genes of the Antennapedia (Antp) complex (O’Hara et al., 1993). Similarly apterous (ap) repression by Ubx prevails over Antp activation in the central nervous system (M. Capovilla and J. Botas, unpublished results). In contrast, and in violation of posterior prevalence, repression of centrosomin (cnn) by Antp dominates activation by Ubx in the visceral mesoderm (Heuer et al., 1995). In another case, a phosphorylation-defective Antp protein (ANTP^[1,2,3,4]A) has novel functions in addition to the wild-type Antp functions. At least some of its novel functions are a consequence of ANTP^[1,2,3,4]A ability to misregulate regulatory targets of other Hox genes. ANTP^[1,2,3,4]A violates posterior prevalence by repressing empty spiracles (ems) expression, which is normally activated by Abd-B. However it is interesting that the novel function of ANTP^[1,2,3,4]A as a dpp activator cannot overcome repression by abd-A, hence respecting posterior prevalence (Jaffe et al., 1997). Thus the attractiveness of this model is that it explains the cases of posterior prevalence in which the posterior gene is the repressor, but it also explains other cases of functional dominance in which posterior prevalence is violated.

We found that different HOX binding sites mediate different transcriptional activities. Repression by ABD-A is mediated only by certain binding sites. The ability of individual ABD-A binding sites to mediate repression does not correlate with their affinity for ABD-A measured in vitro. Specifically, we found that a low-affinity site (binding site 4) is better able to mediate repression by ABD-A than a high-affinity site (binding site 2). These results suggest the existence of cofactors involved in the regulation of dpp674 by ABD-A. One candidate for such a factor is the product of extradenticle (exd) (Chan et al., 1994; Rauskolb and Wieschaus, 1994); however, exd is not required for abd-A repression of dpp (Rauskolb and Wieschaus, 1994) or dpp674lacZ (M. Capovilla and J. Botas, unpublished). Thus Hox specificity cannot be explained solely by HOX/EXD cooperative binding, and unidentified cofactors interacting differentially with ABD-A and other Hox products probably exist. These factors may alter ABD-A binding specificity and/or may function as corepressors or coactivators, altering ABD-A activity as a transcription factor.

The above hypothesis on Hox functional dominance implies that in many cases posterior Hox genes function as repressors whereas anterior Hox genes function as activators of specific target genes. Posterior Hox genes would generally determine posterior body patterns by repressing target genes activated by more anterior Hox genes. However, it is of course unlikely that posterior Hox genes function exclusively as repressors. They probably also function as activators of some targets; as noted above the gene ems is a good candidate for direct activation by Abd-B (Jones and McGinnis, 1993). In these cases the cross-regulation between Hox genes (posterior Hox genes repress the expression of more anterior Hox genes) would ensure the dominance of posterior Hox genes. From the viewpoint of evolution, the easier way to create additional posterior patterns might be to generate ‘new’ Hox genes that repress existing targets rather than activate new targets or combinations.

We thank Ed Lewis, Gines Morata, Eric Wieschaus, Jaime Castelli-Gair and Kathy Matthews and the Bloomington Stock Center for Drosophila strains. We are grateful to Gerard Karsenti, Angela Algeri and Sharon Bickel for encouragement and suggestions, and to Giuseppa Pennetta for critical reading of the manuscript. This work was supported by a grant to J. B. from the National Science Foundation. M. C. was the recipient of an NIH postdoctoral training grant.