dMi-2, a Hunchback-Interacting Protein That Functions in Polycomb Repression

See allHide authors and affiliations

Science  04 Dec 1998:
Vol. 282, Issue 5395, pp. 1897-1900
DOI: 10.1126/science.282.5395.1897


Early in Drosophila embryogenesis, gap gene products directly repress transcription of homeotic (HOX) genes and thereby delimit HOX expression domains. Subsequently, Polycomb-group proteins maintain this repression. Currently, there is no known molecular link between gap and Polycomb-group proteins. Here, dMi-2 is identified as a protein that binds to a domain in the gap protein Hunchback that is specifically required for the repression of HOX genes. Genetic analyses show that dMi-2 participates in both Hunchback and Polycomb repression in vivo. Hence, recruitment of dMi-2 may serve as a link between repression of HOX genes by Hunchback and Polycomb proteins.

The design of animals depends on spatially restricted expression of HOX genes (1). In the early Drosophila embryo, segmentation gene products that are locally expressed delimit the domains of HOX gene expression (2, 3). Gap proteins, such as Hunchback (Hb), bind directly to regulatory sequences of HOX genes and repress their transcription in cells outside of HOX expression domains (4,5). Although HOX genes need to be continuously repressed in these cells and in their descendants, gap proteins are only transiently available. The role of the Polycomb-group (PcG) gene products is to maintain repression of HOX genes throughout development (1,6–8). To identify proteins that may act as a molecular link between the Hb repressor and PcG proteins, we used Hb protein as a bait in a yeast two-hybrid screen.

Using LexA-Hb as bait, we isolated cDNAs representing six different genes (9). In interaction tests with various unrelated LexA baits, proteins encoded by three of the six cDNAs interacted exclusively with Hb (Fig. 1A). Among these proteins, the hip76 clone product exhibited the strongest interaction with Hb. We isolated multiple cDNA clones (10) that span a complete open reading frame (ORF) encoding a 1982–amino acid protein with high sequence similarity to the human autoantigen Mi-2 (11). We refer to the Drosophila protein as dMi-2. dMi-2 contains five conserved sequence motifs (11) that are also present in the two human Mi-2 proteins and in twoCaenorhabditis elegans ORFs (12): two chromodomains (13), a DNA-stimulated adenosine triphosphatase (ATPase) domain (14), two PHD finger motifs (15), a truncated helix-turn-helix motif resembling the DNA-binding domain of c-myb (16), and a motif with similarity to the first two helices of an HMG domain (17).

Figure 1

Identification of dMi-2 as a Hb-interacting protein. (A) Yeast two-hybrid assay. Six Hb-interacting proteins (hip) were isolated. Parentheses denote number of times isolated. Blue color intensity of yeast colonies grown on X-gal plates indicated strength of interaction [– and + signs in (A) and (B)]. Only hip57, hip66, and hip76 exclusively interacted with LexA-Hb, with hip76 showing the strongest interaction. (B) Mapping of dMi-2–interacting sequences (amino acids 1653 to 1982) in the Hb protein. LexA-Hb fusion proteins were tested for reporter gene activation in yeast without (NONE) or with a dMi-2 activation domain (AD) fusion. With the exception of LexA-Hb(2-487) and LexA-Hb(2-344), these fusions did not autoactivate transcription (NONE). Repression assays (29) demonstrated that all LexA-Hb fusion proteins bind to LexA operator sites in yeast nuclei. The D domain (black box) together with sequences directly COOH-terminal to it is sufficient to bind to dMi-2. F1, finger domain 1; D, D domain; F2, finger domain 2. (C) Lesions present in dMi-2 alleles.dMi-24 shows an insertion of 4 base pairs after codon 398 that results in a frameshift and consequently a predicted premature termination within the first PHD finger domain. IndMi-25 , a strictly conserved Gly (Gly737) that is present in all ATPase domains of the SWI2/SNF2 family is substituted by Asp; indMi-26 , the base substitution changes a Trp codon in the ATPase domain (Trp801) into a termination codon.

To map the dMi-2–interacting domain in Hb, we generated Hb fragments and tested them for dMi-2 interaction in yeast two-hybrid assays (Fig. 1B). dMi-2 interacted very strongly with sequences overlapping the D domain, a stretch of amino acids that is conserved between Hb proteins of different insect species (18). Mutations in the D box cause extensive derepression of HOX genes of the Bithorax complex (BXC) (2) (see below). Both D box alleles are premature termination codons, suggesting that the D domain and its COOH-terminal flanking sequences are critical for repression of BXC genes (19). Our interaction tests (Fig. 1B) show that this protein portion of Hb interacts with dMi-2. In vitro binding assays with bacterially expressed dMi-2 and Hb proteins confirmed that these proteins bind directly to each other (20). Thus, dMi-2 binds to a portion of Hb that appears to be critical for repression of BXC genes.

In situ hybridization to polytene chromosomes revealed thatdMi-2 maps to subdivision 76D (21). In a screen for zygotic lethal mutations in this region, we identified five complementation groups (21).

To test whether any of these five complementation groups encode dMi-2, we sequenced the dMi-2 coding regions of several alleles (22). All three sequenced alleles of one of the complementation groups showed individual base changes within conserved domains of dMi-2 (Fig. 1C). The identification of the dMi-2gene by these molecular lesions is further supported by a rescue test with a dMi-2 transgene (23).

dMi-2 homozygotes survived until the first or second larval instar. Mutant embryos and larvae showed no obvious mutant phenotypes. Specifically, expression of BXC genes such as Ultrabithorax(Ubx) and Abdominal-B (Abd-B) was completely normal in these mutant embryos (Figs. 2C and 3, A and B). This normal expression may be due to maternally deposited dMi-2 RNAs or proteins that persist through subsequent development. Consistent with this, we found that all early embryos from adMi-2 deletion stock (including those lacking the gene) showed the same high levels of dMi-2 RNA (Fig. 2A).

Figure 2

Expression and function of dMi-2 in germ cells and in embryos. (A) Embryos from a Df(3L) kto2stock (the deletion removes dMi-2) hybridized with a digoxygenin-labeled dMi-2 antisense RNA probe. Up to the blastoderm stage (left), all embryos were labeled uniformly; this RNA is therefore maternally deposited. In 12-hour-old Df(3L) kto2homozygotes, the maternal dMi-2 RNA became undetectable (right). Zygotic expression of dMi-2 in wild-type embryos (middle). (B) No eggs were obtained from dMi-2 mutant germ cells (* denotes that rare abnormal eggs were laid).dMi-22 and dMi-24 mutant germ cells carrying a dMi-2 transgene (“T”) (23) developed into normally shaped eggs, but no embryos were obtained. (C) Synergy between hb and dMi-2. Eleven- to thirteen-hour-old embryos stained with an antibody to Ubx (2). dMi-24 mutants (left) showed wild-type morphology and Ubx expression (arrowhead indicates the anterior Ubx boundary in parasegment 5);hb9K57 mutants (middle) showed derepression of Ubx anterior to the morphological gap (asterisk) (3).dMi-24 hb9K57 double mutants (right) showed much more extensive derepression (arrow) anterior to the gap. Embryos are oriented with the anterior to the left and the dorsal side up.

We thus attempted to generate embryos from mutant dMi-2 germ cells (23). However, germ cells that are mutant for any of the seven tested dMi-2 alleles failed to develop (Fig. 2B). This failure can be rescued by a dMi-2 transgene (Fig. 2B) (23), demonstrating that dMi-2 is essential for the development of germ cells. We therefore could not generate embryos that lack dMi-2 protein.

Next we tried to detect a genetic interaction between dMi-2and hb. hb9Q mutants [carrying a premature stop codon upstream of the first finger domain (19)] showed only slight anterior derepression ofUbx in embryos because of perdurance of maternalhb products (2, 18).hb9K57 mutants (carrying a D box lesion) showed more extensive anterior derepression of Ubx (Fig. 2C); this mutant protein is thought to have dominant-negative effects on the persisting maternal wild-type product (2).dMi-24 hb9K57 double mutants showed much more extensive derepression of Ubx thanhb9K57 mutants (Fig. 2C). Similarly,dMi-24 hb9Q double mutants showed more extensive derepression than hb9Q mutants alone. These results demonstrate a synergy between hb anddMi-2 that is consistent with our finding that dMi-2 binds to Hb. Furthermore, it provides strong evidence that dMi-2functions in the repression of BXC genes.

We next tested whether dMi-2 protein participates in PcG repression. As in the case of dMi-2, maternally deposited PcG product often rescues homozygous mutant PcG embryos to a considerable extent (1, 6–8). Extensive derepression of HOX genes can be observed if such homozygous embryos are also mutant for another PcG gene (7). We thus examined embryos homozygous for the PcG gene Posterior sex combs (Psc) anddMi-2 and found that Ubx and Abd-Bwere derepressed more extensively in this double mutant than inPsc homozygotes alone (24) (Fig. 3A). A similar result was found if dMi-2 was combined with other PcG mutations (24); these double mutants consistently led to much enhanced homeotic transformations compared with the single PcG mutants (Fig. 3A). Thus, there is a synergy between dMi-2and PcG genes. dMi-2 behaves like the PcG mutations Enhancer of Polycomb and Suppressor 2 of zeste, neither of which on their own cause a homeotic phenotype but do so in combination with other PcG mutations (8). This suggests that dMi-2 functions in PcG repression.

Figure 3

dMi-2 and PcG genes synergize to repress homeotic genes. (A) Thirteen- to sixteen-hour-old embryos stained for Ubx (top row, side views) or Abd-B protein (middle row, ventral views of central nervous system). No derepression of either Ubx or Abd-B is seen in dMi-2mutants. Many more cells misexpress Ubx and Abd-B in Psc dMi-2 double mutants compared with Psc mutants (arrowheads). Not all cells that misexpress Ubx are seen in these focal planes. The embryonic cuticle pattern of dMi-2 mutants is indistinguishable from wild-type cuticles (bottom row, left). Stronger homeotic transformations of the thoracic denticle belts (arrowheads) were observed in dMi-2 Pc double mutants compared withPcmutants (thePce9 mutation used here is a hypomorph).dMi-2 alleles were dMi-24 (top row),dMi-26 (middle row), anddMi-29 (bottom row). (B) Derepression of Ubx in wing discs of third instar larvae with the indicated genotype. In dMi-2 heterozygotes, like in wild-type, Ubx protein is not expressed in wing discs. Wing discs ofPc heterozygotes show derepression of Ubx(arrowhead) that is enhanced in dMi-2/Pc transheterozygotes (on average, three times as many cells express Ubx). (C) Homeotic transformations due to derepression of Scr. Each number in the second and third columns is the total number of sex comb teeth on second and third legs of 20 sibling males of the genotypes indicated. The ratio of the numbers in the second and third columns shows the enhancement caused by the mutation X. All dMi-2mutations enhance the transformation comparable to a Pclmutation. Mutations in Trithorax-like (Trl), the gene encoding the GAGA factor, do not enhance this transformation.

Next we examined imaginal discs for derepression of HOX genes as well as the phenotypes of their adult derivatives. Clonal analysis suggested that dMi-2 is required for the survival of somatic cells (23). We therefore tested whether dMi-2 mutations exhibited gene-dosage interactions with PcG mutations. Whereas larvae heterozygous for Polycomb (Pc) mutations showed slight derepression of Ubx (Fig. 3B), larvae transheterozygous for both Pc and dMi-2mutations showed more extensive derepression (Fig. 3B). Furthermore, derepression of the HOX gene Sex combs reduced(Scr) in the second and third leg discs of Pcheterozygotes results in the formation of a first leg structure, the sex comb, on the second and third legs (25). The extent of this homeotic transformation reflects the number of cells that misexpress Scr protein. We found that this homeotic transformation was far stronger in dMi-2/Pc transheterozygotes than in adults heterozygous for Pc alone (Fig. 3C), which is consistent with more extensive derepression of Scr in the double mutant. These results are further evidence that dMi-2 acts together with PcG proteins to repress HOX genes.

Previous studies led us to propose that Hb directly or indirectly recruits PcG proteins to DNA to establish PcG silencing of homeotic genes (5, 26). Our present data suggest that dMi-2 might function as a link between Hb and PcG repressors. Although dMi-2 contains two motifs with similarity to DNA-binding domains (the myb and HMG domains), dMi-2 does not seem to bind to DNA on its own. Therefore, Hb may recruit dMi-2 to DNA. Xenopus Mi-2 was recently purified as a subunit of a histone deacetylase complex (27) with nucleosome remodeling activity. In yeast and in vertebrates, several transcription factors repress transcription by recruiting histone deacetylases (28). It is possible that in Drosophila, nucleosome remodeling and deacetylase activities of a dMi-2 complex—recruited to homeotic genes by Hb—may result in local chromatin changes that allow binding of PcG proteins to the nucleosomal template. Alternatively, the proposed hb–dMi-2 complex might directly bind a PcG protein and recruit it to DNA. Finally, the involvement of dMi-2 in PcG silencing suggests that this process may involve deacetylation of histones.

  • * These two authors made equal contributions to this work.

  • Present address: School of Biology and Biochemistry, University of Bath, Bath BA2 7AY, UK.

  • Present address: Max-Planck-Institut für Entwicklungsbiologie, Spemannstrasse 35/III, 72076 Tübingen, Germany.

  • § To whom correspondence should be addressed. E-mail juerg.mueller{at}


View Abstract

Navigate This Article