Drumstick is a zinc finger protein that antagonizes Lines to control patterning and morphogenesis of the Drosophila hindgut

Cell rearrangement drives many morphogenetic processes. During vertebrate gastrulation, non-epithelial mesodermal cells intercalate and converge toward the midline in a process designated convergent extension, resulting in dramatic elongation of the embryonic axis (Warga and Kimmel, 1990; Keller et al., 1985). Rearrangement of non-epithelial cells also occurs during invertebrate development, including the formation of Drosophila ovarian terminal filaments and the migration of ovarian border cells (Godt and Laski, 1995; Montell, 1999).

Cells can also rearrange while remaining constrained within an epithelial sheet. This type of rearrangement is responsible for elongation of both the Drosophila germ band and the C. elegans dorsal epidermis (Irvine and Wieschaus, 1994; Heid et al., 2001). Similarly, epithelial cell rearrangement causes elongation of developing tubes, including the sea urchin archenteron, C. elegans intestine, and Drosophila posterior spiracles (Ettensohn, 1985; Leung et al., 1999; Brown and Castelli-Gair Hombría, 2000).

For a field of epithelial cells to change shape in a coherent and directed manner, cell rearrangement within the epithelium must be oriented. In the case of the elongating Drosophila germ band, this orientation depends on the patterning of the anteroposterior axis of the embryo (Irvine and Wieschaus, 1994). Later, however, during development of epithelial organs, cell rearrangement is oriented to newly established axes that are specific to the organ itself. Examples include the elongation of the insect Malpighian tubule relative to signals from the tip cell, epithelial migration during tracheal system development relative to localized FGF signals in surrounding mesoderm, and eversion of the leg imaginal disc in response to segmentally localized Notch signaling (Skaer, 1993; Bradley and Andrew, 2001; Rauskolb and Irvine, 1999). A central problem in epithelial morphogenesis, therefore, is to understand how the positional cues (i.e. spatial patterning) that orient cell rearrangement are established.

The Drosophila hindgut provides a model system in which to investigate patterning and its role in orienting cell rearrangement. The embryonic hindgut is a single-layered epithelium that elongates by both cell rearrangement and cell shape change (Skaer, 1993; Lengyel and Liu, 1998; Iwaki et al., 2001). Genes controlling each step of the morphogenetic process (specification and internalization of the primordium, maintenance and elongation, and specification and patterning of subdomains) have been identified (Lengyel and Iwaki, 2002). In particular, the genes drumstick (drm) and lines (lin) are required for both patterning in the prospective small intestine and for the cell rearrangement that drives elongation of the hindgut epithelium. While drm is required to commit cells to the small intestine fate, lin represses this fate (Iwaki et al., 2001).

By analysis of single and double mutants, we show here that drm and lin interact genetically, and that lin is epistatic to drm. The lin gene encodes a novel, globally expressed protein that is thought to be a transcriptional regulator (Hatini et al., 2000). We show here that drm encodes a small, zinc finger protein expressed in a dynamic, spatially localized pattern that is consistent with the drm mutant phenotype. Ectopic expression and biochemical interaction studies indicate that Drm antagonizes Lin activity, probably through direct binding. Thus, a relief-of-repression mechanism allows expression of genes that define the small intestine. The specification of this domain is essential for the oriented cell rearrangement that elongates the hindgut.

Previously described mutant alleles used were: drm¹ (Liu et al., 1999), lin² (Hatini et al., 2000), Df(2L)tim⁰² (Myers et al., 1995), and Df(2L)ed¹ (Reuter and Szidonya, 1983). The strong UAS-lin 8 (used here) and other weaker UAS-lin stocks (Hatini et al., 2000), the byn-GAL4 hindgut-specific driver (Iwaki and Lengyel, 2002), and the 14-3fkh-GAL4 driver (Fuss and Hoch, 1998) have been previously described. The drm^P1 lin² double mutant was generated by recombination. UAS-lacZ is from the Bloomington Stock Center (Indiana University, Bloomington, IN).

Df(2L)drm^P1 and Df(2L)drm^P2 were generated by mobilization of the l(2)k10101 P element (Török et al., 1993). New point mutations in drm were generated by ethyl methanesulfonate mutagenesis (Grigliatti, 1986). Briefly, mutagenized cn bw sp males were crossed to Gla/SM6b females; 17,757 F₁ progeny chromosomes were tested for failure to complement either Df(2L)drm^P1 or Df(2L)drm^P2. To minimize the effects of second site mutations, newly isolated drm chromosomes were recombined with ast¹ dpp^d-ho ed¹ dp^ov1 cl¹ and screened for cross-over events between ed (24D3) and dp (25A), thus replacing ∼80% of the mutagenized chromosome. To identify altered nucleotides, new drm chromosomes were balanced over CyO-GFP (Casso et al., 2000) for selection and sequencing of homozygous genomic DNA. The three exons were amplified individually by PCR (performed in triplicate), subcloned into pCR2.1 (Invitrogen), sequenced in both directions, and compared to the cn bw sp parent chromosome.

Antibody staining was performed according to standard protocols (Ashburner, 1989) using the following antibodies: anti-Crumbs (Crb, 1:100; labels the apical surface of ectodermally derived tissues including the hindgut) (Tepass et al., 1990) and anti-En (1:5) were obtained from the Developmental Studies Hybridoma Bank at The University of Iowa, Department of Biological Sciences (Iowa City, IA); anti β-gal (1:1000; Promega). In situ hybridization to whole-mount embryos was performed as previously described (Tautz and Pfeifle, 1989) using digoxigenin-labeled RNA probes (Roche Molecular Biochemicals). To avoid cross-reactivity with other members of the odd family, a ‘zinc-fingerless’ drm-specific probe was synthesized from a PCR-amplified fragment of the drm cDNA 3′UTR. The upstream and downstream primer sequences were 5′-CAAAACTCAGCAAACAACTA-3′ and 5′-GAGTGTGTGTGTATGTGTGT-3′, respectively. The specificity of this probe was confirmed by its inability to hybridize to embryos homozygous for the drm^P1 deficiency (which deletes drm, but not sob, odd, or bowl). Other probes were made from cDNA templates of lin (Hatini et al., 2000), upd (Harrison et al., 1998) (also known as outstretched, os – FlyBase) and hh (Lee et al., 1992). For transverse sections, embryos were labeled with anti-Crb prior to mounting in Epon. Two μm serial sections were cut and stained with 0.01% Toluidine Blue, 0.025% Methylene Blue, 0.025% sodium tetraborate. Whole-mount embryos were analyzed with a Zeiss (Thornwood, NY) Axiophot microscope using differential interference contrast optics. Sections were analyzed by phase contrast microscopy. Images were acquired with a Sony DKC-5000 digital camera and embryos were staged based on morphology (Campos-Ortega and Hartenstein, 1997). Larval feeding assays were performed as described previously (Pankratz and Hoch, 1995).

Standard techniques were used for molecularly characterizing the drm gene (Sambrook et al., 1989). The breakpoints of the drm^P1 and drm^P2 deletions were determined by plasmid rescue (3′ end) and inverse PCR (5′ end) from the residual P element. The drm¹ I element insertion was localized by Southern blot and molecularly mapped by inverse PCR. Total RNA for northern blot was isolated from 0- to 17-hour embryos using TRIzol (Invitrogen), and poly(A) RNA isolated by PolyATtract (Promega). Probes were constructed by radiolabeling BamHI fragments of DS01379 P1 phage clone (Kimmerly et al., 1996) for Southern blot, or LD26791 EST clone (Research Genetics) for northern blot. The drm-specific primers used for 5′ and 3′ RACE were 5′-GATATTGCGTGCTGTTGTTGGTGCT-3′ and 5′-CTTAAACGCCGGGCACCAGTACACA-3′, respectively. Amplification was performed on RACE-adapted cDNA from 0 to 4-hour embryos using the Marathon cDNA Amplification Kit (Clontech).

To make constructs for GAL4:UAS expression, the following oligonucleotides were used:

(a) 5′-CAGTAGATCTCAGTACACATACCGCCTGTCAA-3′,

(b) 5′-GTCAGAATTCAAGATGCGTCCGAAGTGCGAGTT-3′,

(c) 5′- GTCAGAATTCATGTTTGCTGTAATGCGAAT-3′,

(d) 5′-CTGAGGTACCACAACGGGGTTTTACGCTTGATA-3′,

(e) 5′-CTGACTCGAGATATGTTTATGTGGTATTTAAG-3′,

(f) 5′-GACTCTCGAGTTACTCGCACTTCGGACGCATCTTG-3′,

(g) 5′-CTGAAGATCTGCGAAGGAAATGTTACTGA-3′,

(h) 5′-CCTGATGATCCACGAGTGCACCCACAAGTCC-3′,

(i) 5′-CCTGAGATCACCTATTCGGGCGAGGTGTGCGGC-3′.

The various drm domains (see Fig. 6) were amplified from either cn bw sp cDNA or the LD26791 EST clone using the following restriction site-modified oligo pairs (cut sites underlined above): Drm (a and e); C-Drm (b and e); N-Drm (c and f). Amplified products were digested and subcloned into either BglII/XhoI-digested (for Drm) or EcoRI/XhoI-digested pUAST (Brand and Perrimon, 1993). Point mutations were created by site-directed mutagenesis with the QuikChange XL Kit (Stratagene) using the following oligonucleotides and their complements (altered nucleotides underlined above): R46C (h) and C57G (i). HA-tagged derivatives of each construct were made by incorporating the HA epitope sequence, in frame, into the PCR primers. A UAS-DrmEST construct expressing the full-length drm cDNA was also generated by subcloning the 2.1 kb XhoI fragment from LD26791 into pUAST. The gain-of-function phenotypes of UAS-Drm and UAS-DrmEST were indistinguishable (data not shown). The drm genomic DNA rescue construct was generated by subcloning the 18.1 kb SgrA1 (blunt-ended)/NotI fragment from DS01379 into NotI/HpaI-digested pCaSpeR-4 (Thummel and Pirrotta, 1992). The drm-GAL4 construct was generated by amplifying 7.1 kb of the drm promoter (primers d and g, cut sites underlined above), digesting with KpnI and BglII, and subcloning upstream of GAL4 in KpnI/BamHI-digested pGaTB (Brand and Perrimon, 1993). The 10.3 kb KpnI/NotI drm-GAL4 fragment was then subcloned into pCaSpeR-4.

Transformation was performed using standard techniques (Ashburner, 1989). Briefly, constructs were coprecipitated at a 2:1 ratio with the pP{Delta2-3} transposase plasmid (Laski et al., 1986), resuspended overnight at 4°C in injection buffer (5 mM KCl, 0.1 mM sodium phosphate, pH 7.8), and filtered through a 0.2 μm cellulose acetate Spin-X column (CoStar) before loading into needles pulled from 1 mm Kwik-Fil borosilicate glass capillaries (World Precision Instruments). DNA was injected using a Transjector microinjector (Eppendorf) into w¹¹¹⁸ embryos under 80% ethanol.

Schneider S2 cells were plated in 6 cm culture dishes 24 hours prior to calcium phosphate transfection with 2.5 μg UAS-Myc-Lin, 2.5 μg UAS-HA-Drm (or Drm derivatives), and 3 μg ubiquitin-GAL4 in 600 μl volume. Transfectant was removed after 12 hours, and cells were harvested 48 hours later. Individual plates of cells were lysed in 400 μl NET buffer (50 mM Tris pH 7.5, 400 mM NaCl, 5 mM EDTA, 1% NP-40, 50 μg/ml phenylmethylsulfonyl fluoride, 1 μg/ml leupeptin, 1.4 μg/ml pepstatin) for 30 minutes and incubated with mouse anti-HA (1:40; monoclonal HA.11 clone 16B12, Covance). Antibody-antigen complexes were collected with Protein G Sepharose (Invitrogen), washed, and resuspended in 100 μl 2× Laemmli buffer. Ten microliters of each sample was separated by SDS-PAGE and transferred to a PVDF membrane. Lin protein was detected using rabbit anti-Myc primary antibody (1:1000; polyclonal c-Myc A-14; Santa Cruz Biotechnology) and HRP-conjugated anti-rabbit secondary antibody (1:3000; Vector Laboratory). Proteins were visualized by ECL Plus detection (Amersham Pharmacia Biotech).

In both drm and lin embryos, the hindgut is wider and shorter than that of wild-type embryos (Fig. 1A-C) (Iwaki et al., 2001). Beyond this superficial similarity, however, drm and lin hindguts are quite distinct. The drm hindgut is smaller in diameter, and its epithelium consists of an undulating layer of columnar cells resembling those of the immature wild-type hindgut primordium (Fig. 1F). In contrast, the lin hindgut appears distended, consisting of a uniform layer of cuboidal cells similar in appearance to those of the wild-type small intestine (Fig. 1G). The strongly Crb-stained boundary cells, which form two parallel rows running the length of the large intestine, are duplicated in drm but are absent in lin hindguts (Fig. 1E-G) (Iwaki et al., 2001).

As revealed by gene expression studies, drm and lin hindguts are improperly patterned, with opposite effects on specification of the large and small intestine. Expression in the dorsal large intestine of engrailed (en) is retained in drm but absent from lin hindguts (Fig. 1I-K). Expression in the small intestine of unpaired (upd), encoding a ligand for the JAK-STAT pathway) is missing from drm but greatly expanded in lin hindguts (Fig. 1M-O). Similarly, expression in the small intestine of hedgehog (hh) is reduced in drm but greatly expanded in lin hindguts (Fig. 1Q-S). These data indicate that, in drm embryos, the large intestine is present and the small intestine is greatly reduced or absent; in lin embryos, in contrast, the small intestine is greatly expanded and the large intestine is missing.

Because their phenotypes are opposite and easily distinguishable, we used epistasis analysis to ask whether drm and lin interact genetically. By all criteria applied, the hindgut of the drm lin double mutant is remarkably like that of the lin single mutant. As in lin embryos, the drm lin hindgut is short and distended (Fig. 1D), consists of cuboidal epithelial cells, and lacks boundary cell rows (Fig. 1H). Similarly, en expression is absent, and both upd and hh expression are expanded posteriorly (Fig. 1L,P,T). All of these observations taken together show that lin is epistatic to drm in the hindgut.

Since similarities exist in the genetic regulation of foregut and hindgut development (Pankratz and Hoch, 1995), we asked whether lin is also epistatic to drm in the foregut. In drm embryos, the proventriculus (a multi-layered valve-like structure at the foregut-midgut junction that forms by epithelial folding) does not form, and the foregut is long and narrow (Fig. 1V) (Liu et al., 1999). In both lin and drm lin embryos, the proventriculus also fails to form, but the foregut is short and wide (Fig. 1W,X). We conclude that lin is epistatic to drm in the foregut as well.

The gene expression and phenotypic data presented here and previously (Iwaki et al., 2001) demonstrate that lin represses and drm promotes the small intestine fate. Taken together with the epistasis of lin to drm, this indicates that drm specifies small intestine by antagonizing the repressive effect of lin. To understand the molecular basis for this relief of repression, we have molecularly characterized the drm gene and analyzed the interaction, both in vivo and in vitro, between Drm and Lin.

The drm¹ allele was originally identified as a spontaneously occurring lethal mutation mapping between 23F6 and 24A2 (Liu et al., 1999). drm was mapped more precisely by using P element mobilization to generate two overlapping deletions, drm^P1 and drm^P2 (Fig. 2A and Table 1). The left and right breakpoints of these deletions were defined molecularly (see Materials and Methods), thereby localizing drm to a ∼60 kb interval between tim and sob. Southern blot analysis of drm¹ genomic DNA revealed an insertion within the 11.9 kb BamHI fragment, characterized by inverse PCR as an I element upstream of the predicted gene CG10016 (Rubin et al., 2000b). Three expressed sequence tag (EST) clones from this gene have been identified (Rubin et al., 2000a); the longest, LD26791, is 2.3 kb and contains three exons (Fig. 2B).

Surprisingly, the LD26791 EST contains only one small open reading frame (ORF) with two zinc finger motifs (Fig. 2D). To address the possibility that larger proteins might arise from this gene by alternative splicing, we characterized CG10016 transcripts by several approaches. A northern blot of embryonic RNA with probes from LD26791 identifies a single 2.5 kb transcript, similar in length to the EST clones (Fig. 2C). RT-PCR with intron-spanning primers failed to identify any splice variants (data not shown). An alternative first exon was identified in a small fraction (∼10%) of 5′ RACE products, and several alternative poly(A) addition sites were identified by 3′ RACE; none of these variants affects the ORF length. Sequence of all EST clones, RT-PCR products, 5′ and 3′ RACE products, and genomic DNA (Adams et al., 2000) is consistent with the conclusion that CG10016 is an 8.5 kb transcription unit. Three exons are spliced to form a 2.5 kb mRNA encoding an 81 amino acid, 10,120 Da predicted protein (Wilkins et al., 1998).

The position of the I element in drm¹, and the endpoints of the deletions in drm^P1 and drm^P2, strongly suggest that CG10016 is the drm gene. To obtain further evidence supporting this relationship, we performed rescue and mutagenesis experiments. We generated germline transformants of an 18.1 kb fragment of genomic DNA (Fig. 2A) that includes CG10016 plus 6.9 kb of upstream and 2.7 kb of downstream DNA (but no other predicted genes). Although this construct does not rescue drm lethality, it does rescue the drm hindgut and foregut phenotypes (data not shown). We also generated and molecularly characterized five new EMS-induced alleles (see Materials and Methods). Four of these are point mutations (in three different residues) within the first zinc finger, and one is a roo element insertion in the second exon splice donor site (Fig. 2D, Table 1); no mutations were found outside the ORF. In all of these mutants, the proventriculus is both morphologically and functionally defective (assessed by larval feeding assays; data not shown), while the hindgut is defective only in flies carrying the stronger alleles. These phenotypic characteristics support the following allelic series: drm⁵<drm²<drm⁴<drm¹<drm³=drm⁶<drm^P1=drm^P2. Based on phenotypic rescue experiments and molecular mapping of drm alleles, we conclude that CG10016 is the drm gene.

Drm is a member of the Drosophila odd-skipped (odd) family of zinc finger encoding genes that includes odd, sister of odd and bowl (sob), and bowel (bowl) (Nüsslein-Volhard and Wieschaus, 1980; Coulter et al., 1990; Hart et al., 1996; Wang and Coulter, 1996). These genes map close to each other (Fig. 3A), suggesting that the family has arisen by relatively recent duplication. Like bowl and sob (but not odd), drm contains a splice donor site within the R74 codon of the second zinc finger. Interestingly, this splice site has been conserved evolutionarily, as it is also present in both the mouse and human odd-skipped related (Osr) genes Osr1 and Osr2 (So and Danielian, 1999; Lan et al., 2001).

The Drm protein contains two zinc finger motifs (compared to four in Odd and five in both Sob and Bowl; Fig. 3B). The zinc fingers in Odd, Sob, and Bowl conform to the canonical C₂H₂ structure (C-X₂-C-X₁₂-H-X₃-H) that is most commonly associated with a DNA-binding function, but in some cases, can have protein-binding capability (Rosenfeld and Margalit, 1993; Mackay and Crossley, 1998). In Drm, the first zinc finger conforms to the canonical C₂H₂ sequence and has a high degree of similarity (∼95%) to the first finger of the other Odd family members. The second zinc finger of Drm is divergent; the primary sequence conforms to the canonical C₂H₂ sequence up to the H73 residue, but the second His residue is replaced by a Cys, with H-X₄-C spacing between the latter two zinc-coordinating residues. This residue spacing is found in other C₂HC fingers with demonstrated protein-binding activity. Computer modeling (SWISS-MODEL) (Guex and Peitsch, 1997) with respect to the known structure of the Drosophila U-shaped (Ush) C₂HC zinc finger shows that the Drm C₂HC finger is theoretically capable of folding around a zinc ligand (Liew et al., 2000). Another distinguishing feature of Drm is the divergent linker region between its zinc fingers. The most common linker, found in over 50% of known C₂H₂ fingers, consists of five residues with the consensus sequence TG(E/Q)(K/R)P (Wolfe et al., 2000). The Odd, Sob and Bowl linkers all have the conserved sequence TDERP, whereas the Drm linker (KSPEIT) is different both in sequence and length (Fig. 3C). Since its C₂H₂ and C₂HC zinc fingers are, in principle, capable of either DNA or protein binding, Drm may function by either or both of these mechanisms.

Consistent with the gut defects observed in drm embryos, drm is expressed dynamically in cells that will give rise to the foregut and hindgut (Fig. 4). In the posterior gut primordium, drm mRNA is first detected at stage 5 in a ventral crescent at 10% embryo length (EL) (Fig. 4A). Cells in this region are fated to give rise to the epithelia of the posterior midgut, Malpighian tubules and hindgut (Campos-Ortega and Hartenstein, 1997). At stage 6, the posterior crescent expands dorsally to encircle the amnioproctodeal plate (Fig. 4B). By stage 7, drm is expressed in a ring within the proctodeal invagination (Fig. 4C). During germ band extension (stages 8 and 9), drm expression is seen in a region overlapping the junction of the posterior midgut and hindgut primordia (Fig. 4D,E). At stage 10, drm is expressed transiently in the evaginating buds of the Malpighian tubules (Fig. 4F). Between stages 11 and 13, drm mRNA is detected in the posterior midgut, the ureters of the Malpighian tubules, and the most anterior cells of the small intestine (Fig. 4G,H); expression in these domains persists throughout the remainder of embryogenesis.

drm is also expressed in the anterior gut primordium starting at stage 5 (Fig. 4A). Expression increases in this domain until stage 10 when it is internalized as the invaginating stomodeum (Fig. 4F). drm expression is then refined to a narrow ring of cells at the junction of the foregut and anterior midgut (Fig. 4H). Like odd and sob (Hart et al., 1996), drm is also expressed in a seven-stripe segmental pattern at stage 5 (Fig. 4A); this pattern evolves into fourteen stripes that mark the anterior margin of each segment (Fig. 4H).

The domains in which drm is expressed during embryogenesis include a region predicted to become small intestine (Fig. 4C-G). By stage 13, however, drm mRNA is not seen in the small intestine (Fig. 4H). To determine if drm is expressed early in cells fated to become small intestine, we generated a drm-GAL4 driver and used it to drive UAS-lacZ. Owing to the perdurance of both GAL4 and β-gal, this provides an historical summation of the drm expression pattern. By stage 13, β-gal is detected in a much larger domain of the posterior gut in drm-GAL4:UAS-lacZ embryos than is drm mRNA in wild-type embryos (Fig. 4I, compare with 4H). Most importantly, the presence of β-gal in the small intestine of stage 16 embryos (Fig. 4K) demonstrates that some of the drm-expressing cells in the early embryo do indeed give rise to the small intestine.

If, as suggested above, spatially localized expression of drm in the anterior hindgut allows specification of the small intestine by antagonizing lin, then expression of drm throughout the hindgut should inhibit endogenous lin, thereby producing a lin-like hindgut phenotype. We tested this by driving UAS-drm using the hindgut-specific byn-GAL4 driver. The result of this gain-of-function expression of drm is a hindgut phenotype that resembles lin loss-of-function mutants: the hindgut is short and distended, consisting of a uniform layer of cuboidal cells similar to those of the wild-type small intestine, and the boundary cell rows are absent (Fig. 5A,C). en is not expressed in the hindgut of these embryos, but upd and hh expression is greatly expanded posteriorly (Fig. 5E,G,I). Like lin mutations, the byn-GAL4:UAS-drm combination is lethal. Overall, it appears that in both lin mutants and embryos expressing drm throughout the hindgut the small intestine is expanded at the expense of the large intestine. We conclude that drm functions in the hindgut by antagonizing lin.

If drm acts primarily by inhibiting lin activity in the hindgut, then overexpression of lin might overcome the repressive effect of endogenous drm, thereby producing a drm-like phenotype. Consistent with this notion, overexpression of lin produces a hindgut phenotype that resembles drm loss-of-function mutants: the hindgut is short and wide, consisting of an undulating layer of columnar cells with a duplication of the boundary cell rows (Fig. 5B,D). As seen in drm embryos, en is expressed throughout the dorsal hindgut, upd expression is absent, and hh expression in the small intestine is reduced (Fig. 5F,H,J). Like drm mutations, the byn-GAL4:UAS-lin combination is lethal. Since weaker hindgut drivers (e.g. 14-3fkh-GAL4) and weaker UAS-lin transformants do not produce a phenotype as severe as with the strong byn-GAL4 driver or the strongest UAS-lin stocks (R. Green, unpublished observations), we conclude that the repressive effect of drm can be titrated only by relatively high levels of lin.

The epistasis and in vivo overexpression studies indicate that drm inhibits lin activity, either directly or indirectly. As we observe no reduction of lin expression in the drm expression domains of wild-type embryos (Hatini et al., 2000; Iwaki et al., 2001), or any changes in drm or lin expression in lin or drm mutant embryos, respectively (data not shown), we conclude that drm does not affect lin at the transcriptional level. We therefore used a biochemical approach to ask whether Drm might interact physically with Lin. When expressed together in cultured cells, full-length Drm and full-length Lin interact with each other, as demonstrated by coimmunoprecipitation assays (Fig. 6B, lane 1), indicating that the two proteins are in a complex. Since Drm and Lin interact with each other in a yeast two-hybrid assay (data not shown), the Drm-Lin interaction is likely mediated by direct binding.

To map the protein interaction domain, we generated deletion and point mutation constructs of Drm (Fig. 6A) and tested their ability to coimmunoprecipitate with Lin (Fig. 6B). The N-terminal portion of Drm (N-Drm) is unable to bind Lin, while the C-terminal portion of Drm, containing the two zinc finger motifs (C-Drm), retains full Lin-binding activity. Because these results map the protein-protein interaction domain to the zinc fingers, we tested the requirement of each individual zinc finger for Lin-binding activity. A mutation in the C₂H₂ first finger (R46C, identical to the mutation in the drm⁶ null allele) abolishes Lin-binding activity, while disruption of the C₂HC second finger (C57G, a substitution in one of the conserved zinc-binding cysteine residues) reduces Lin-binding activity, but does not abolish it completely. To confirm these results, we made similar point mutations in the truncated C-Drm construct and tested their effect on Lin-binding activity. Again, a mutation in the C₂H₂ first finger in C-Drm(R46C) abolishes Lin-binding activity while disruption of the C₂HC second finger in C-Drm(C57G) reduces Lin-binding activity (Fig. 6B). We conclude that the C₂H₂ zinc finger is essential for binding to Lin, while the C₂HC finger contributes to binding, perhaps by stabilizing the interaction.

We next asked which portions of the Drm protein are required to block Lin function in vivo by expressing, throughout the hindgut, the same Drm deletion and point mutation constructs used in the coimmunoprecipitation studies (Fig. 6A). We assayed the activity of these mutated proteins by their ability to induce a lin-like hindgut phenotype when ectopically expressed with byn-GAL4 (similar to the gain-of-function phenotype induced by ectopic expression of full-length Drm). Expression of N-Drm, which lacks the zinc fingers, results in a morphologically wild-type hindgut (Fig. 6C), while expression of C-Drm, which contains only the zinc finger motifs, produces a morphologically lin-like hindgut (i.e. short and distended and lacking boundary cell rows; Fig. 6C, compare with Fig. 5A). Because these results map the in vivo Lin-inhibiting activity to the zinc fingers, we tested the requirement of each individual zinc finger for Lin-inhibiting activity. A mutation in the C₂H₂ first finger (R46C) abolishes the Drm gain-of-function phenotype, while disruption of the second C₂HC finger (C57G) has no effect on Drm activity (Fig. 6C, compare with Fig. 5A). These effects are observed whether the point mutations are in full-length Drm (R46C and C57G) or in the truncated C-terminal portion of Drm (C-Drm(R46C) and C-Drm(C57G); data not shown). Taken together, the results demonstrate that the first zinc finger, but not the second, is required for Lin-inhibiting activity in vivo.

In summary, Drm constructs that coimmunoprecipitate with Lin are able to repress lin activity in vivo, while Drm constructs lacking Lin-binding activity in vitro are not able to repress lin activity in vivo (see Fig. 6A). We conclude that the C₂H₂ first finger is essential for both the Lin-binding and Lin-antagonizing functions of Drm. The C₂HC second finger, while contributing to Lin binding, is not absolutely required for Lin-inhibiting activity. Thus the drm and lin antagonism observed by genetic approaches is mediated by a physical interaction between the Drm and Lin proteins.

We showed previously that drm is required for specification of the small intestine, while lin represses this fate (Iwaki et al., 2001). Here we demonstrate that lin is epistatic to drm, indicating that lin acts downstream of drm in a genetic pathway. In our molecular characterization, we determine that, in contrast to the uniform expression of lin (Hatini et al., 2000), drm is expressed in a highly localized pattern at the anterior of the hindgut primordium. We further show that uniform expression of drm throughout the hindgut results in a lin-like phenotype. When co-expressed in cultured cells, Drm and Lin are found in a complex. Taking into account these genetic, molecular and biochemical data, we propose a model (Fig. 7) in which Lin (or a complex containing Lin) functions as a transcriptional regulator that represses the genetic program promoting small intestine fate. In the anterior of the hindgut, Drm blocks the activity of Lin by direct physical association, thereby allowing specification of small intestine fate. The Drm-Lin interaction in the hindgut thus defines a pathway in which the spatially localized expression of a small regulatory protein inhibits the repressive effect of a uniformly expressed transcriptional regulator.

Putative transcriptional regulators have been shown to be required for a number of processes of epithelial cell rearrangement. die-1 encodes a zinc finger protein required for intercalation of dorsal epidermal cells in C. elegans (Heid et al., 2001), grain a GATAc factor required for stigmatophore elongation (Brown and Castelli-Gair Hombría, 2000), and ribbon (rib) a BTB domain protein required for tracheal branching morphogenesis (Bradley and Andrew, 2001). All of these genes are expressed throughout and are required within the cells undergoing cell rearrangement. In contrast, drm is expressed only at one end of the prospective hindgut, but is required for cell rearrangement throughout much of the epithelium.

As the drm, bowl and lin phenotypes suggest that juxtaposition of properly specified small and large intestine is required for hindgut cell rearrangement (Iwaki et al., 2001), we speculate that expression of signaling molecules specifically in the small intestine orients the rearrangement of cells within the large intestine. Hedgehog (Hh), Wingless (Wg) and Serrate (Ser) are expressed in the termini of the hindgut, but they appear to play either a minor or no role in controlling morphogenesis (Iwaki et al., 2001; Iwaki and Lengyel, 2002). One signaling molecule that is required for hindgut cell rearrangement is Upd, the Drosophila ligand activating JAK/STAT signaling, which is expressed specifically in the small intestine (K. A. J. and J. A. L., unpublished). Further characterization of the upd pathway and its downstream effectors may provide insight into how patterning of the small intestine affects oriented cell rearrangement.

Both epistasis and ectopic expression experiments indicate that, in the hindgut, Drm acts primarily through Lin. The association between Drm and Lin, demonstrated both in cultured cells and in yeast, suggests that, rather than binding to DNA, Drm functions by binding to protein. Drm contains two zinc fingers that differ structurally, and perhaps functionally, from those of its family members. The first, canonical C₂H₂ finger lacks the conserved TG(E/Q)(K/R)P linker present in other Odd family zinc fingers, which is required in other canonical C₂H₂ fingers for stabilization of high-affinity DNA-protein binding (Laity et al., 2000). Furthermore, as shown here, this C₂H₂ finger is both necessary and sufficient for Lin-binding in vitro and Lin-inhibiting activity in vivo. A protein-binding function for the Drm C₂H₂ finger is consistent with the demonstrated role of other C₂H₂ fingers, including those of the Ikaros family, in homo- and heterodimerization (Sun et al., 1996; Morgan et al., 1997; Kelley et al., 1998). Although we cannot rule out a DNA-binding function for Drm, our data are consistent with a model in which the primary role of the C₂H₂ first finger is in binding to Lin.

C₂HC fingers have also been shown to act as protein binding motifs: both mammalian FOG and Drosophila Ush bind to zinc fingers in their partner proteins via specific C₂HC fingers (Tsang et al., 1997; Fox et al., 1999; Haenlin et al., 1997). When co-expressed in cultured cells and assessed by coimmunoprecipitation, the Drm C₂HC second finger does indeed contribute to the binding of Drm to Lin (although it is not absolutely necessary). In overexpression studies in vivo, however, the Drm C₂HC finger was not necessary for Lin-inhibiting activity. Consistent with this in vivo observation, no mutations in the Drm C₂HC finger were identified in our screen of 18,000 mutagenized chromosomes. Thus, while a protein-binding function is suggested by the in vitro data, we have not detected a required in vivo function for the Drm C₂HC finger.

In conclusion, our characterization of Drm adds to the growing number of C₂H₂ zinc fingers that act as protein-protein interaction motifs (Mackay and Crossley, 1998). The Drm protein is unique in that it is extremely small (10 kDa), is expressed in a highly spatially localized pattern, and appears to function entirely (at least in the hindgut) by associating with, and thereby antagonizing, the activity of a globally expressed protein, Lin.

lin is expressed relatively uniformly throughout the embryo, yet has tissue- and stage-specific effects, and both activating and repressing activities. A number of these functions have been shown to depend on spatially localized cofactors or signals. In the posterior spiracles and eighth abdominal belt, lin functions together with Abdominal B (which is expressed specifically in the posterior) to activate empty spiracles, cut and spalt (Castelli-Gair, 1998). In the dorsal epidermis, as a result of localized wg signaling, lin activates wg and represses veinlet expression (Hatini et al., 2000); lin is required to promote quaternary (4°) and represses tertiary (3°) cell fates (Bokor and DiNardo, 1996; Hatini et al., 2000). In the hindgut, lin represses expression of upd and hh (characteristic of small intestine fate); establishment of the small intestine requires the locally expressed cofactor Drm, which relieves repression by Lin.

Antagonism of repressors is an important mechanism by which transcriptional activity is spatially regulated (Courey and Jia, 2001). In Drosophila, localized cell signaling by receptor tyrosine kinases (e.g. Torso and Sevenless) or the canonical wg pathway is required to relieve global repression (in a number of cases by Groucho), thereby allowing proper patterning and/or differentiation of portions of the embryonic termini, segments, midgut and eye (Paroush et al., 1997; Jimenez et al., 2000; Waltzer and Bienz, 1998; Cavallo et al., 1998; Rebay and Rubin, 1995). While these relief-of-repression pathways are dependent on localized cell signaling, the Drm-Lin interaction is, to our knowledge, the first described example in which relief of repression is initiated by spatially restricted production of a small transcriptional regulator.

Many questions remain about the Drm-Lin interaction and its mechanism of action. First, it is not clear how Lin acts as a transcriptional regulator. In the hindgut, and perhaps in other tissues, Lin might act as part of a repressosome complex (Courey and Jia, 2001), binding DNA directly or associating with chromatin on the basis of an interaction with other DNA-binding proteins. Binding of Drm to a Lin-containing complex might then inhibit the repressive activity of Lin. Second, activation of gene expression requires not only the absence (or inactivation) of repressors, but also the presence of transcriptional activators (Struhl, 1999). Thus, there must be as-yet-unidentified transcriptional activators that promote expression of small intestine-specific targets once Drm has lifted the repression by Lin.

We have shown that drm is expressed dynamically not only in the hindgut, but also in the foregut and in the developing epidermis. This raises the question of whether Drm-mediated Lin inhibition might occur in tissues other than the hindgut. Both the spatially restricted expression of drm and the epistasis of lin to drm in the foregut epithelium (Fig. 1U-X) are consistent with the idea that Drm also blocks Lin activity in the proventriculus primordium, thus allowing proper folding morphogenesis. In the dorsal epidermis, stripes of drm expression roughly correspond to defects in patterning seen in drm mutants (V. H. and S. DiNardo, unpublished observations); these defects are opposite to those seen in lin embryos (Hatini et al., 2000), suggesting that Drm-mediated relief of Lin repression may occur in dorsal epidermis patterning. We have also observed lin expression in leg imaginal discs and a required role for lin in leg development (V. Hatini and S. DiNardo, unpublished observations); taken together with the potential role of drm in the developing leg (Cohen et al., 1991), these results suggest that Drm and Lin might also control aspects of leg development. We conclude that, by antagonizing Lin, Drm controls multiple events of epithelial patterning and thereby morphogenesis.

Another gene in the odd family, bowl, has a hindgut phenotype very similar to that of drm (Wang and Coulter, 1996; Iwaki et al., 2001), suggesting that bowl might also function in the same pathway as drm and lin. However, in contrast to the spatially localized expression of drm, but similar to the spatial and temporal expression of lin, bowl expression appears uniform throughout the hindgut primordium until stage 12 (Wang and Coulter, 1996). Thus, despite its similarity in both mutant phenotype and sequence to drm, bowl function in the hindgut is likely different from that of drm. In mouse and humans, the genes Osr1 and Osr2 encode Odd-related proteins with three and five zinc fingers, respectively, that are most similar to those of Sob and Bowl (Fig. 3C) (So and Danielian, 1999; Lan et al., 2001). Mouse Osr1 and Osr2 are expressed during embryogenesis in dynamic, spatially localized patterns correlated with sites of active morphogenesis (So and Danielian, 1999; Lan et al., 2001). Given that Drm (and perhaps Bowl) is involved in a pathway with Lin, it may be significant that a mammalian protein, predicted from an EST clone and sharing homology with Lin over a 200 amino acid region, was recently identified (Venter et al., 2001). It will be interesting to determine if any parallels exist between the expression and function of these mammalian homologs and the Drosophila Lin and Odd family of proteins.

We thank Silvia Wenjuan Yu and Monica Martinez for excellent technical assistance, Joan Hooper for assistance in making the UAS-DrmEST germline transformant, the Bloomington Stock Center for fly stocks, Albert J. Courey, Utpal Banerjee, Cheryl Kerfeld and members of the Lengyel laboratory for helpful discussions, and Stephen DiNardo for his encouragement, active participation and support to V. H. This work was supported by NIH grants GM08042 to the UCLA Medical Scientist Training Program (to support R. B. G.), GM45747 to S. DiNardo (to support V. H.), and HD09948 to J. A. L.