A Mutational Analysis of dishevelled in Drosophila Defines Novel Domains in the Dishevelled Protein as Well as Novel Suppressing Alleles of axin
Andrea Penton, Andreas Wodarz, Roel Nusse

Abstract

Drosophila dishevelled (dsh) functions in two pathways: it is necessary to transduce Wingless (Wg) signaling and it is required in planar cell polarity. To learn more about how Dsh can discriminate between these functions, we performed genetic screens to isolate additional dsh alleles and we examined the potential role of protein phosphorylation by site-directed mutagenesis. We identified two alleles with point mutations in the Dsh DEP domain that specifically disrupt planar polarity signaling. When positioned in the structure of the DEP domain, these mutations are located close to each other and to a previously identified planar polarity mutation. In addition to the requirement for the DEP domain, we found that a cluster of potential phosphorylation sites in a binding domain for the protein kinase PAR-1 is also essential for planar polarity signaling. To identify regions of dsh that are necessary for Wg signaling, we screened for mutations that modified a GMR-GAL4;UAS-dsh overexpression phenotype in the eye. We recovered many alleles of the transgene containing missense mutations, including mutations in the DIX domain and in the DEP domain, the latter group mapping separately from the planar polarity mutations. In addition, several transgenes had mutations within a domain containing a consensus sequence for an SH3-binding protein. We also recovered second-site-suppressing mutations in axin, mapping at a region that may specifically interact with overexpressed Dsh.

WNT signaling molecules are crucial for cell-cell communication, cell fate specification, embryonic axis formation, and growth control during development of vertebrates and invertebrates (Wodarz and Nusse 1998; Yamaguchi 2001). Genes that transduce signals from Wnt proteins are conserved across species (Cadigan and Nusse 1997; Peifer and Polakis 2000). In Drosophila, wg is required during embryogenesis for segmental patterning, muscle development, and midgut formation and to specify appendage primordia. Frizzled (Fz) and Frizzled-2 (DFz2), two seven-pass transmembrane proteins with cysteine-rich extracellular domains, act as Wg receptors (Bhanotet al. 1996). Wg binds to these receptors and to the Arrow LDL-related coreceptor (Bhanotet al. 1996; Tamaiet al. 2000; Wehrliet al. 2000), resulting in the activation of Dishevelled (Dsh). dsh encodes a cytoplasmic phosphoprotein (Klingensmithet al. 1994; Theisenet al. 1994; Yanagawaet al. 1995) and its activity is required for the transduction of the Wg signal (Noordermeeret al. 1994; Siegfriedet al. 1994). It is thought that Dsh acts by binding to and inhibiting the Axin protein that negatively regulates Wg signaling (Kishidaet al. 1999; Liet al. 1999a; Smalleyet al. 1999; Salicet al. 2000). Axin is a scaffolding protein that binds to APC, another negative regulator of Wg signaling, and to Armadillo (Arm), a homolog of β-catenin (Behrenset al. 1998; Hartet al. 1998; Ikedaet al. 1998; Sakanakaet al. 1998). Axin negatively regulates Arm by facilitating the action of Zeste white-3 (Zw-3; Siegfriedet al. 1992), a homolog of the vertebrate serine/threonine kinase GSK3-β (Hartet al. 1998; Ikedaet al. 1998). Zw-3 phosphorylates Arm, leading to its destruction by the ubiquitin pathway (Peifer and Polakis 2000). Inhibition of Zw-3 results in the stabilization of Arm. Stabilized Arm forms a complex with the HMG-box DNA-binding protein dTCF/Pangolin (Brunneret al. 1997; van de Weteringet al. 1997) and activates transcription of Wg target genes.

In addition to mediating Wg signaling events, dsh is required to generate planar polarity, also known as tissue polarity, by regulating both the correct orientation of ommatidia within the Drosophila eye and the alignment of bristles on the adult epidermis (Theisenet al. 1994; Krasnowet al. 1995; Struttet al. 1997). This second function of dsh appears to be distinct from its role in Wg signaling and may result in the activation of a Jun-N-terminal kinase pathway (Boutroset al. 1998; Liet al. 1999b). dsh acts downstream of frizzled (fz) in planar polarity signaling (Krasnowet al. 1995; Struttet al. 1997). Although fz acts redundantly with Dfz2 during Wg signaling, it is required for planar polarity (Adleret al. 1990; Bhanotet al. 1999; Chen and Struhl 1999). In contrast, there is no evidence for an involvement of Dfz2 in planar polarity signaling. Recent studies suggest that Fz and Dfz2 have different affinities for Wg and also differ in signaling specificity in regions other than the ligand-binding domain (Boutroset al. 2000; Rulifsonet al. 2000; Strapps and Tomlinson 2001). Thus, Dsh may interact with Dfz2 to regulate events that are associated with Wg signaling and with Fz to regulate planar polarity.

Three conserved domains have been identified in the Dsh protein, including an amino-terminal DIX (Dishevelled, Axin) domain, a central PDZ (Postsynaptic density 95, Discs Large, Zonula occludens-1) domain, and a C-terminal DEP (Dishevelled, Egl-10, Pleckstrin) domain (Boutros and Mlodzik 1999; Figure 4A). The DEP domain is required for planar polarity signaling. A viable dsh allele, dsh1, possesses a lesion within this domain (Axelrodet al. 1998; Boutroset al. 1998) and disrupts planar polarity but not Wg signaling (Axelrodet al. 1998; Boutroset al. 1998; Figure 3, C and D). While the structure of the DEP domain has been elucidated (Wonget al. 2000), it is not known how this module works. Interestingly, the DEP domain is necessary and sufficient to translocate the Dsh protein to the membrane when Frizzled proteins are expressed (Axelrodet al. 1998).

Although the functions of these domains have been examined by deletion analysis, rigorous mutagenesis studies have not been undertaken in vivo (Boutros and Mlodzik 1999). We designed genetic screens to identify additional dsh alleles that are deficient for planar polarity signaling and to find dsh alleles that are deficient in Wg signaling. These studies led to the identification of a putative src homology 3 (SH3)-binding motif in Dsh that is essential for Wg signaling, suggesting that SH3 proteins may be important regulators of Wg signaling. We also investigated the role of potential phosphorylation sites for the function of Dsh in Wg and planar polarity signaling by site-directed mutagenesis. Besides mutations in dsh itself, we isolated mutations that act as second-site modifiers of the phenotype caused by overexpression of Dsh in the eye. Some of these modifiers encode novel mutations in axin.

MATERIALS AND METHODS

Planar polarity screens: A total of 23 rounds of mutagenesis were performed for both the planar polarity screens and the dsh misexpression screen. w1118 males were isogenized on the X chromosome. In rounds 1–9 of the mutagenesis these males were mutagenized overnight with a 15–30 mm solution of EMS (Sigma, St. Louis) in 1% sucrose using established procedures (Ashburner 1989). Mutagenized males were crossed to y w dshv26/FM7, TW9 virgin females (Wieschauset al. 1984; Perrimon and Mahowald 1987) and 20,000 F1 B+ females were screened for planar polarity defects (Figure 1A). During subsequent rounds of the mutagenesis 0.5–3.3 mm of the mutagen 1-ethyl-1-nitrosourea (ENU) in 1% sucrose was used instead of EMS, because ENU rarely induces chromosomal rearrangements and produces a wider spectrum of mutations than does EMS (Ashburner 1989; Pastinket al. 1990). A total of 37,840 additional females were screened using this protocol. v males, isogenized on the X chromosome, were mutagenized as above with ENU and crossed to dshA3 virgin females (Figure 1B). A total of 21,000 flies were screened and an additional dsh allele, dshA21, was identified. The mutation frequency of EMS was estimated to be between 23 and 30% and between 16 and 34% for ENU. The mutation frequency was determined by the frequency of sex-linked lethal mutations. To this end, individual B females of the genotype w1118/FM7, TW9 were crossed to wild-type males. Progeny from this cross were scored for the presence or absence of males. Since FM7, TW9 is lethal in males, only males that have no lethality on the mutagenized w1118 chromosome were obtained.

dishevelled misexpression: A BamHI/EcoRI fragment of the dshmyc construct (Yanagawaet al. 1995) was cloned into the pUAST (Brand and Perrimon 1993) vector. Injection of UAS-dshmyc into y w flies was performed (Spradling 1986), and 46 lines were obtained. These lines were crossed to GMR-Gal4 (Moses and Rubin 1991) and the phenotype of adult eyes was examined. Lines 3-8 and 1-16 produced a moderate phenotype when crossed to GMR-Gal4 as compared to GMR-Gal4; UAS-wg flies or to other lines of UAS-dsh (Figure 2C; data not shown). The addition of the myc epitope did not interfere with UAS-dsh function in this assay when compared to a line containing UAS-dsh without the myc tag. The cytological insertion positions of these lines were determined. The insertion positions mapped to 64C, line 3-8, and 85B/C, line 1-16. Since lines 3-8 and 1-16 were homozygous viable, mapped to the third chromosome, and appeared to have single insertion sites, they were utilized for the misexpression screen.

A UAS-dsh1 construct was generated from DNA that was PCR amplified from dsh1 genomic DNA preparations (see below). A PFLM1/Blp1 fragment was isolated from this PCR product and cloned into PFLM1/Blp1-digested pBS-dshmyc. This construct was cloned into the pUAST vector and UAS-dsh1myc flies were obtained as above. Twenty-four lines were isolated. Their phenotypes were similar to wild-type UAS-dsh when crossed to GMR-Gal4.

Mutant lines of UAS-dsh that showed a modification of the original eye phenotype were crossed to da-Gal4, which drives ubiquitous expression in the embryo, and to 69B-Gal4, which drives expression in the wing (Brand and Perrimon 1993), to determine the effect of misexpression in these tissues.

dishevelled misexpression screen: w/Y; Sp/Cyo; UAS-dsh males were mutagenized overnight with a 25 mm solution of EMS in rounds 6–9 of the mutagenesis. In subsequent rounds, 0.5–2 mm solution of ENU (Sigma) in 1% sucrose was used. These males were isogenized on the third chromosome. Mutagenized males were crossed to w; GMR-Gal4 virgin females and the eyes from ∼90,000 F1 flies were screened (Figure 1C).

Of these 90,000 flies, 104 mutant lines were generated (Table 1). Twenty-five lines were isolated with apparently wild-type eyes (Figure 2F), 68 lines had eyes larger than those of parental lines but smaller than those of wild-type flies (Figure 2D), and 11 lines had eyes that were almost wild type in size but rough, suggesting that they retained less function in Wg signaling than did the previous class (Figure 2E; Brunneret al. 1997). A total of 101 of these lines map to the third chromosome as would be expected if they contain lesions within the UAS-dsh transgene. Three lines map to the fourth chromosome and could not be mapped further due to the absence of recombination on the fourth chromosome.

These screens were designed to identify specific amino acid changes rather than large deletions or truncations of the Dsh protein and we therefore examined the expression of the Dsh protein encoded by the transgene in 49 lines using anti-myc antibodies. Twenty-eight lines expressed Dsh protein at robust levels and 26 of these lines were sequenced. A total of 36 lines have disruptions either in Dsh protein expression or in sequence of the UAS-dsh transgene. Included in this group are all 25 lines that possess wild-type eyes and 11 lines that contain partial loss-of-function mutations. Twelve lines do not contain disruptions in Dsh protein expression or in the sequence of the UAS-dsh transgene, suggesting that they possess second-site modifiers of the GMR-GAL4; UAS-dsh eye misexpression phenotype. Five of these lines were mapped (see below). The remaining 55 lines have not been analyzed further but the data are available on request.

Generation of dsh mutants lacking potential phosphorylation sites or protein domains: To generate mutants of dsh in which conserved serine or threonine residues were substituted by alanines, we used the oligonucleotide-mediated site-directed mutagenesis method of Kunkel (Sambrooket al. 1989). The mutagenesis reactions were carried out using single-stranded DNA prepared from the myc-tagged, full-length dsh cDNA cloned in pBluescript (Yanagawaet al. 1995) and the following mutagenesis primers: Dsh-ST1 CTCGCATATCAAGCG GCAGCCGTGCTCGCCGCCGATCTCGAGTCG (T178A, S181A, S183A, S186A, S187A); Dsh-ST2 GATCTCGAGGCGGCCGC TCTCTTTGGCGCCGAAGCCGAGCTCACG (S191A, T192A, S193A, T197A, S199A); Dsh-ST4 TCGCGCGCCGCGGCGTAC GCCGCCATAGCCGACTCG (T235A, S236A, S237A, S239A, S240A, T242A); Dsh-ST5 CCGACGCGGCCATGGCCCTAAA TATTATTGCCGTCGCCATCAAC (S244A, T245A, S247A, T252A, S254A); Dsh-CK2 CTCTTTGGCGCCGAATCC (T197A); Dsh-CK3 TCGTACAGCGCTATAACCGAC (S240A); Dsh-CK4 GAGAACATGGCCAACGACGAG (T314A). The resulting mutations are given in parentheses. The mutated Dsh cDNAs were cut out of the Bluescript vector with SalI and XbaI and were ligated into pCasPeR-hs cut with XbaI and StuI after filling in the SalI site of the Dsh fragment. The same strategy was used to subclone the deletion mutants DshΔbasic, DshΔPDZ, and DshΔDEPD-2 (Yanagawaet al. 1995) into pCasPeR-hs.

Rescue of dsh loss-of-function alleles: Hemizygous dshv26 animals die at late third instar or early pupal stages (Klingensmithet al. 1994). To test for rescuing capacity of the transgenes, we crossed heterozygous y dshv26/FM7 females with males carrying a heat-shock-inducible dsh construct on one of the autosomes. Developing larvae and pupae were exposed to a 30-min heat shock every 24 hr until hatching of adult flies. Rescued dshv26 hemizygous males were easily identified by yellow body color and eyes with wild-type shape (B+).

To score the rescuing capacity of our transgenes with respect to the tissue polarity phenotype of dsh1, we analyzed the orientation of wing hairs and thoracic bristles of hemizygous dsh1 males carrying a dsh transgene on one of the autosomes.

Mapping of UAS-dsh modifiers: UAS-dsh males that suppressed the Dsh eye and wing misexpression phenotype but that did not encode mutations within UAS-dsh, UAS-dsh M, were crossed to ru h th st cu sr e ca/TM3 Sb females. UAS-dsh M/ru h th st cu sr e ca females were crossed to ru h th st cu sr e Pr ca/TM6B males and recombinant males were scored for these recessive markers. Three individual males of each recombinant class (for example ru or ru h or ru h th st) and the reciprocal class were crossed to w; GMR-Gal4 females and Pr+ flies were scored for the presence of UAS-dsh and for the modifier. If UAS-dsh was not present then w; GMR-Gal4/+; recombinant/+ males were crossed to w; GMR-Gal4/Cyo; UAS-dshT/TM6B females and w; GMR-Gal4/+; UAS-dsh/(+ or recombinant) flies were scored for the presence of the modifier. Of these flies, 50% are expected to contain the modifier. Five lines, UAS-dsh M8-3, UAS-dsh M8-4, UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dsh M16-21, were strong suppressors and these mapped to ca at position 3-100.7. We more finely mapped three of them, UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dsh M16-21, by crossing w; ra M/TM6B males to w; P{w(+mC)}dcoj3B9/TM3, Sb females. w; ra M/P{w(+mC)}dcoj3B9 females were crossed to Pr ra ca/TM6B males and males that recombined in the interval between ra and dco of the genotype ra P{w(+mC)}dcoj3B9/Pr ra ca or ++/Pr ra ca were crossed to w; GMR/Cyo; UAS-dsh/TM6B and the presence or absence of the modifier was noted. ra maps at position 3-97.3 and dco maps to the right of ca at cytological position 100B2-4. Recombinants of the genotype ra M P{w(+mC)}dcoj3B9 were obtained. Interestingly, these flies suppressed the GMR-Gal4; UAS-dsh eye phenotype more strongly than did ra M flies, suggesting that dco interacts in this assay. dco encodes Drosophila casein kinase I and its mammalian homolog is implicated in Wnt signaling (Klosset al. 1998; Peterset al. 1999; Sakanakaet al. 1999). w; ra M/TM6B males were crossed to P{w(+mC)}axin/TM6B females and recombinants were mapped in the same manner as the above cross. From these two crosses we concluded that M lies near axin and proximal of dco in the interval 99D4-5 to 100B2-4. We were unable to obtain recombinant ra M P{w(+mC)}axin flies. This indicated that this modifier might be axin and the axin gene in three of these lines, UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dsh M16-21 was sequenced (see below). These lines did indeed contain novel axin mutations.

Sequence analysis: dsh is a single exon gene, facilitating sequencing of alleles. Genomic DNA of adult flies of the genotype dsh1, dshA3, or dshA21 was isolated and the entire coding region was sequenced. PCR primers from positions 870, 5′ of the coding region, and 2914, 3′ of the coding region, (Berkeley Drosophila Genome Project) were used to amplify genomic DNA and these PCR products were directly sequenced using the ABI system.

Isolates that affected the phenotype of UAS-dsh generated in the misexpression screen were crossed to GMR-Gal4 and third instar eye discs were dissected from these flies. These discs were stained with the 9E10 anti-myc antibody obtained from the hybridoma facility at The University of Wisconsin using established protocols (Blair 1992). Since the myc tag in UAS-dsh is at the carboxy end of the protein, UAS-dsh alleles that encode truncations within the Dsh protein would not be labeled with the anti-myc antibody. The entire coding region was sequenced from lines that produced intact protein. Genomic DNA from these lines was isolated and DNA was amplified using PCR. Primers were designed to sequences within the UAS-dsh transgene to ensure that the UAS-dsh gene but not endogenous dsh gene would be amplified. PCR-amplified UAS-dsh products were reamplified using the original 5′ or 3′ primer along with an internal primer to obtain enough DNA for sequencing. Sequencing was done as above.

Genomic DNA was isolated from lines UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dsh M16-21. These lines are now called axin8-13, axin8-66, and axin16-21. DNA fragments were PCR amplified using primers to axin genomic sequence (Berkeley Drosophila Genome Project) at positions 740 and 1660, 1310 and 3455, 2640 and 4046, and 3432 and 4805. These fragments represent the entire coding region of axin and were sequenced as above.

Adult wing mounting: Wings were mounted in Euparal and incubated overnight at 65°.

RESULTS

Screen for planar polarity alleles of dsh: The viable dsh1 allele causes planar polarity defects, resulting in aberrant orientation of bristles and hairs if present in hemizygous male adults or in combination with a dsh null allele (dshv26; Figure 3, C and D). Since dsh1 survives over dshv26, planar polarity mutants will produce surviving adults and can be identified in an F1 screen. We used dshv26 heterozygotes to screen for new dsh alleles (Figure 1A). Using the mutagen EMS, we screened through 20,000 flies and obtained 1 dsh allele, dshA3 (Table 1). We then used the mutagen ENU to screen an additional 37,840 flies. Although a total of 15 potential positive dsh alleles were observed in the F1 generation, only 1 transmitted to the F2. We reasoned that alleles of dsh may have a higher probability of being recovered if we screened over a dsh allele that retained wg signaling function. dshA3 flies are viable and fertile in contrast to dsh1 flies that are only semiviable. Thus an additional screen was performed in combination with dshA3 using the mutagen ENU. A total of 21,000 flies were screened, and 20 potential new dsh alleles were observed, but only 1 allele, dshA21, was recovered in the F2. dshA3 and dshA21 have phenotypes similar to dsh1 (Figure 3, C and D) and display planar polarity abnormalities in the wing and thorax. The defects of dshA21 flies are variable, indicating that this allele has variable expressivity. dshA3 and dshA21 encode mutations within the DEP domain like dsh1 (Table 2 and Figure 4).

View this table:
TABLE 1

Statistics of the mutagenesis screens

Figure 1.

—Screens for new dsh alleles that affect planar polarity and Wg signaling. (A) dsh is located on the X chromosome at cytological position 10B6. w1118 males were mutagenized and mated to virgin females that are heterozygous for a dsh null allele (dshv26) and for the balancer chromosome (FM7, TW9). F1 males were not recovered because they were null for dsh or possessed the FM7, TW9 balancer chromosome that is lethal in hemizygous flies (see materials and methods). The thoraces of F1 females were examined for polarity defects that occur in combination with dshv26. They were distinguished from flies carrying the FM7, TW9 chromosome because they were Bar+. Females that had planar polarity phenotypes were mated to FM7/Y males. Surviving males in the F2 generation were screened for planar polarity defects, visible in hemizygous males. Stocks were established from these males. (B) Male flies marked with vermillion (v) were mutagenized and mated to female flies homozygous for the dshA3 mutation. Thoraces of female flies were screened for polarity defects. Females that possessed defects were mated to FM7/Y males and v F2 males were screened for polarity defects. v maps at position 1-33 and dsh maps at 1-34.5, making recombination between v and dshA3 unlikely, and permitting us to distinguish the dshA3 chromosome from new dsh mutations. (C) Males of the genotype w; Sp/Cyo; UAS-dsh were mutagenized and crossed to w; GMR-Gal4 females. The eyes of F1 flies were screened to determine if they were larger or smaller than eyes from the parental lines (Figure 2). Candidate F1 males were crossed to w; TM3, Sb/TM6B females and w; GMR-Gal4/+; UAS-dsh*/TM3, Sb or w; GMR-Gal4/+; UAS-dsh*/TM6B flies in the F2 generation were analyzed to determine whether the modified phenotype was still present. If so, individual sibling males were crossed to w; GMR-Gal4 females. The UAS-dsh transgene was marked with the mini-(w)+ gene, which allowed its detection in the F2 and F4 generations.

Screen for dsh alleles that disrupt Wg signaling: dsh alleles interfering with Wg signaling are hemizygous and homozygous lethal, complicating the isolation of large numbers of these alleles. We used the GAL4:UAS (Brand and Perrimon 1993) system to produce mutations in a UAS-dsh transgene that are compromised in their ability to activate Wg signaling when overexpressed. dsh was tagged with the myc epitope in the carboxy terminus to distinguish it from the endogenous gene, and UAS-dsh transgenic flies were generated. The GMR enhancer drives expression of the GAL4 protein in cells in developing eyes (Moses and Rubin 1991). If expression of UAS-wg or UAS-dsh is driven by GMR-GAL4, then adult eyes are dramatically reduced in size (Figure 2, B and C). This effect occurs during pupation since GMR-Gal4; UAS-dsh third instar eye discs have no abnormalities (data not shown). Flies that carry a UAS-dsh1 construct exhibit the same reduced eye phenotype as flies that carry a wild-type UAS-dsh construct (data not shown), indicating that this phenotype is not generated by planar polarity signaling.

To obtain mutations in the transgene, males carrying the UAS-dsh transgene were mutagenized with ENU and crossed to GMR-GAL4 females (Figure 1C). We examined F1 progeny for either wild-type eyes (indicating an inactivating mutation in the transgene) or eyes that were larger than those of the parental lines (indicating a partial loss of function in the transgene). Tables 1 and 2 present the results.

Characterization of UAS-dsh alleles: UAS-dsh mutant lines were analyzed further to determine whether these mutations attenuate Wg signaling in other developmental processes. These lines were crossed to 69B-Gal4, which is expressed in the ectoderm during embryonic and larval development (Brand and Perrimon 1993). When parental UAS-dsh lines were crossed to 69B-Gal4, most of the offspring animals died during pupal stages. Severely abnormal wings were observed in adult survivors (Figure 3E). These wings had ectopic margin bristles in the interior of the wing, a defect that is caused by excessive Wg signaling (Diaz-Benjumea and Cohen 1995). We observed other defects that included venation abnormalities, blisters, and disruptions in planar polarity, although the latter was difficult to assay because the wing was severely deformed. The process of planar polarity is sensitive to levels of signaling since both a lack of signaling and a gain of signaling cause pattern disruptions. Since dsh transduces both Wg and planar polarity signaling events, these defects were expected. Many UAS-dsh lines displayed wings that either were wild type or carried planar polarity disruptions when crossed to 69B-Gal4 (Figure 3, F and H; Table 2). Pupal lethality was also suppressed.

UAS-dsh lines were also crossed to da-Gal4, which is expressed ubiquitously in the embryo (Brand and Perrimon 1993; Wodarzet al. 1995). Ectopic Wg signaling is known to cause lethality and cuticular abnormalities during embryogenesis (Noordermeeret al. 1992) and parental lines displayed embryonic or larval lethality when crossed to da-Gal4. In contrast, most UAS-dsh lines either were viable or displayed pupal lethality when they were crossed to da-Gal4 (Table 2). Thus mutations within UAS-dsh that attenuated Wg signaling during eye development also attenuated Wg signaling in other developmental processes.

View this table:
TABLE 2

Missense mutations that disrupt dsh or UAS-dsh function

UAS-dsh mutations map to several domains. Seven UAS-dsh alleles encode missense mutations within the DIX domain and these alleles all strongly disrupt Wg signaling in our assays (Table 2). We identified only one UAS-dsh mutation in the PDZ domain, UAS-dsh8-1, (Table 2), which only mildly disrupts ectopic Wg signaling during eye development. It is lethal when crossed to da-Gal4 and semilethal when crossed to 69B-Gal4, and surviving adults have wings resembling those of the parental lines (Figure 3G).

Three UAS-dsh alleles possess mutations within a novel domain that is located between the PDZ and the DEP domains (Figure 4A). Interestingly, this domain contains a motif that encodes a class I consensus core sequence for an SH3 protein-binding site (Figure 4B; Kayet al. 2000). The consensus site consists of the sequence RTEPVRP at position 352–358 and is surrounded by proline-rich sequence. These UAS-dsh alleles strongly disrupt Wg signaling events indicating that this domain may be important for Wg signaling (Table 2; Figures 2E and 3F).

Three alleles of UAS-dsh map to a region within the DEP domain, which extends from position 440 to 459 (Table 2; Figure 4). These alleles abrogate Wg signaling (Table 2) implying that a subregion of the DEP domain might be utilized in Wg signaling events. This is in contrast to studies that show that the DEP domain is dispensable for Arm accumulation in Drosophila cl-8 cells (Yanagawaet al. 1995).

In vitro mutagenesis of potential phosphorylation sites of Dsh: While a forward mutational screen for novel alleles is powerful, it is not feasible to uncover mutations in multiple sites. In the case of phosphorylation, proteins are often modified at several sites and such sites could be redundant. Several protein kinases are known to interact with Dsh. In particular the area surrounding a basic region in Dsh (amino acids 178–254) is phosphorylated by CK1, CK2, and PAR-1 (Willertet al. 1997; Peterset al. 1999; Sunet al. 2001). This region contains 27 serine and threonine residues and many of these are conserved (Figure 4C). To uncover potential phosphorylation sites, we replaced most of the conserved serine and threonine residues in the region from amino acids (aa) 178 to 254 by alanine using in vitro mutagenesis. This was done in clusters of 4–6 serine and threonine residues at a time (Figure 4C). Subsequently, several of these mutated clusters were combined to generate larger regions devoid of serine and threonine residues. In addition, constructs were generated that lacked the basic domain, the PDZ domain, or the complete C terminus including the DEP domain (Yanagawaet al. 1995).

Figure 2.

—Eye phenotypes of UAS-dsh isolates when crossed to GMR-Gal4. Drosophila wild-type eyes (A) are characterized by 750 ommatidia that develop during the third larval instar as the morphogenetic furrow progresses from posterior to anterior in the eye disc (Bate and Martinez Arias 1993). The GMR promoter is expressed posteriorly to the morphogenetic furrow in developing photoreceptor cells (Moses and Rubin 1991). When GMR-Gal4 is crossed to UAS-wg (B) or UAS-dsh (C), adult eyes that are severely reduced in size result. This defect is caused by degeneration that occurs during pupal development. One class of UAS-dsh isolates (D) possesses eyes that are larger than those of parental lines (C) but smaller than those of wild type (A), indicating a partial loss of function in UAS-dsh. Another class of UAS-dsh isolates possesses eyes that are almost wild type in size (E) but rough, suggesting that very little signaling function of UAS-dsh remains. The remaining UAS-dsh isolates appear to lack signaling since their eyes are wild type in appearance (F).

The mutagenized Dsh constructs were introduced into flies and subjected to a series of functional assays. We first asked whether the different transgenes were able to rescue the loss-of-function phenotype of the amorphic allele dshv26 (Klingensmithet al. 1994). Developing larvae and pupae were exposed to a 30-min heat shock every 24 hr until hatching of adult flies. With this regimen, rescue of dshv26 to viability was observed for all Dsh transgenes except for DshΔPDZ and DshΔDEPD-2 (Table 3). Rescue was also observed in dshv26 hemizygous males derived from germline clones (data not shown). From these results we conclude that those transgenes that rescue lethality are fully functional in the absence of endogenous Dsh.

In a second assay, we tested the transgenes for rescue of the tissue polarity phenotype of dsh1 (Perrimon and Mahowald 1987; Axelrodet al. 1998; Boutroset al. 1998). Our results are summarized in Table 3. Six of the 11 transgenes fully rescued the tissue polarity phenotype of dsh1, while 5 did not show any rescue. The five constructs that failed to rescue the dsh1 phenotype are DshΔPDZ, DshΔDEPD-2, DshST124, DshST4, and DshST45 (see Figure 4C). The latter three constructs all have in common the substitution of serine/threonine residues by alanines in the highly conserved cluster ST4 (Figure 4C). Intriguingly, cluster ST4 is located within the region of Dsh required for binding of and phosphorylation by the protein kinase PAR-1 (Sunet al. 2001). As PAR-1 is essential for polarity signaling during oogenesis in Drosophila (Shulmanet al. 2000; Tomancaket al. 2000; Coxet al. 2001; Huynhet al. 2001), it may also affect planar polarity by phosphorylation of sites located in cluster ST4. This cluster also contains a consensus phosphorylation site for CK2 (Figure 4C), but this serine residue (S240) is probably not required for tissue polarity signaling. A mutant Dsh protein carrying a substitution of serine 240 by alanine fully rescues tissue polarity signaling in a dsh1 mutant background (Table 3). The same is true for the point mutants changing the other three potential CK2 consensus phosphorylation sites in the region between aa 178 and 320 to alanines (Table 3; see Figure 4C).

All the constructs that rescued the tissue polarity defects of dsh1 also showed a wild-type tissue polarity pattern in rescued dshv26 hemizygous males (data not shown). From this result we conclude that these transgenes are sufficient to rescue all aspects of Wingless signaling and tissue polarity signaling in the complete absence of endogenous Dsh. Interestingly, with respect to tissue polarity signaling, full rescuing activity of the heat-inducible transgenes was obtained at basal expression levels without heat shock. Transgenes that did not rescue the tissue polarity defects without heat shock also failed to rescue when raised under the heat-shock regimen described for rescue of dshv26 (data not shown). Thus, the amount of Dsh required for its function in tissue polarity signaling is apparently lower than the amount required for function in Wg signaling.

Figure 3.

—The planar polarity phenotype and wing phenotypes of UAS-dsh isolates when crossed to 69B-Gal4. The hairs and bristles of wild-type flies are oriented in a regular manner such that all hairs or bristles in a particular region are pointing in the same direction. The bristles of wild-type thoraces point from the anterior direction to the posterior direction (A) and the hairs of wild-type wings point in a proximal-to-distal direction (B). In flies that are mutant for dsh1 (C and D) the orientation of bristles and hairs is disorganized and irregular whorls are seen on the thorax (C, arrows) and wings (D, arrows). B and D (left) show a ×5 magnification of wings from a wild-type and a dsh1 mutant fly, respectively. The five regions of the wing are indicated, region I between the anterior wing margin and L2, region II between L2 and L3, region III between L3 and L4, region IV between L4 and L5, and region V between L5 and the posterior margin. B and D (right) show a ×40 magnification of wings from a wild-type and a dsh1 mutant fly. In this and all subsequent figures, region IV is characterized for planar polarity defects. Anterior is up, posterior is down (A and C). Proximal is left and distal is right (B–H). L is longitudinal vein. The phenotype of parental line UAS-dsh (E) is quite severe when crossed to 69B-Gal4. Survivors from this cross have ectopic mechanosensory bristles near the wing margin (top, arrow), severe vein and planar polarity defects (bottom, arrows) and some individuals have blisters (data not shown). UAS-dsh8-12 and UAS-dsh16-1 flies have planar polarity defects (F and H, arrows), but no ectopic bristles and only mild venation defects when crossed to 69B-Gal4. UAS-dsh8-1 flies have a partial loss-of-function phenotype when crossed to GMR-Gal4, but their phenotype resembles parental lines when crossed to 69B-Gal4 (G).

Mutations in axin act as suppressors of Dsh overexpression phenotypes: Twelve UAS-dsh stocks with attenuated misexpression phenotypes in the eye did not disrupt Dsh protein expression and had no mutations within the UAS-dsh transgene. These stocks are referred to as UAS-dsh M (for modifier of UAS-dsh). Five lines, including UAS-dsh M8-3, UAS-dsh M8-4, UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dshM16-21, acted as strong suppressors and were mapped to region 3-100.7 (Figure 5A). Three of these lines, UAS-dsh M8-13, UAS-dsh M8-66, and UAS-dsh M16-21, were mapped more finely to region 99D4-5 to 100B4 (see materials and methods). Since the Drosophila axin gene maps at 99D4-5 and is an important regulator of Wg signaling, we sequenced axin from these three lines and from the UAS-dsh parental strain. All three lines possess mutations in axin (Figure 5B). The UAS-dsh M16-21 and UAS-dsh M8-66 mutations change a threonine residue into an asparagine residue and an arginine into a lysine residue, respectively. UAS-dsh M8-13 maps to the GSK3-β binding domain of Axin and changes a conserved proline into a serine residue at position 421. These axin mutant lines are homozygous viable and have no phenotypes by themselves. They suppress Dsh misexpression phenotypes but do not suppress Wg or DFz2 misexpression phenotypes (materials and methods; data not shown; see discussion).

Figure 4.

—Positions of dsh and UAS-dsh mutations generated in the above screens. (A) The domain structure of dsh is indicated. The PDZ domain and the DEP domain follow the amino terminal DIX domain. A newly identified domain called PR for proline rich lies between the PDZ and the DEP domains. Mutations that disrupt Wg signaling functions of UAS-dsh are indicated above the schematic diagram and mutations that specifically disrupt planar polarity signaling are indicated below. (B) The UAS-dsh8-12, UAS-dsh8-79, and UAS-dsh8-68 mutations cluster in the proline-rich domain. This region contains a class I core consensus binding site for an SH3 protein. The positions of the mutations UAS-dsh8-12, UAS-dsh8-79, and UAS-dsh8-68 are indicated. (C) Protein sequence alignment of Dsh with murine Dvl1. Numbering of amino acids according to the published sequences is shown. Identical or highly conserved amino acids are indicated in gray. The DIX, PDZ-binding, proline-rich, and DEP domains are boxed. dsh and UAS-dsh mutations are shown above the sequence and amino acids that are altered by these mutations are boxed. The underlined regions designated as ST1, -2, -4, and -5 have been subjected to in vitro mutagenesis, leading to exchange of every serine and threonine residue within each cluster for alanine. Consensus phosphorylation sites for Casein kinase 2 are designated as CK1–4 and are marked by asterisks on top of the alignment. (D) The structure of the DEP domain of DVL1 (the mammalian homolog of Dsh established by Wong et al. (2000); downloaded from NCBI http://www.ncbi.nlm.nih.gov/Structure/mmdb/mmdbsrv.cgi?form=6&db=t&Dopt=s&uid=15530 and visualized using the Cn3D program). The position of the mutations is in white.

DISCUSSION

By using three independent experimental approaches, we have identified a set of new mutations in Dsh that disrupt two distinct signaling events. These new mutations map to specific regions within Dsh and thus provide important information on the function of the different protein domains. We screened for mutations in the endogenous dsh gene that disrupt planar polarity (Figure 1, A and B) and utilized a misexpression phenotype to screen for mutations that affect Wg signaling (Figure 1C). Mutations that disrupt the Wg signaling function of Dsh occur throughout the protein while mutations that disrupt planar polarity signaling are confined to the DEP domain (Figure 4, A and C). However, we also demonstrate that mutation of potential phosphorylation sites positioned between the basic region and the PDZ domain of Dsh (cluster ST4) specifically disrupts the ability of Dsh transgenes to rescue the tissue polarity phenotype. We discuss the mutations uncovered in this work by domains, beginning at the N terminus.

DIX domain mutations: Seven UAS-dsh alleles encode missense mutations that map to the DIX domain and all reduce or abrogate Wg signaling in three separate assays (Table 2). These results clearly demonstrate that the DIX domain is required for Wg signaling and agree with other studies in Drosophila (Yanagawaet al. 1995; Axelrodet al. 1998; Boutroset al. 1998). Furthermore, studies of the mammalian homolog of Dsh, Dvl1, show that the DIX domain of Dvl can interact directly with itself and that it binds to axin (Kishidaet al. 1999; Liet al. 1999a). The N-terminal DIX domain in Dsh shares 37% amino acid identity with the C-terminal DIX domain of Axin. The Axin DIX domain also interacts with itself and is necessary to regulate the stability of β-catenin in SW48O cells (Kishidaet al. 1999). This indicates that DIX domains are important for protein-protein interactions and that the DIX domain of Dsh may mediate Wg signaling by binding to and inhibiting Axin.

The basic region: Dsh possesses a highly conserved basic region (aa 219–228; Figure 4C) of unknown function. We did not isolate any point mutations in this region in our genetic screens. Moreover, deletion of the basic region compromises neither the function of Dsh in Wg signaling nor its function in planar polarity signaling. We note, however, that two of the transgenic lines carrying the DshΔbasic construct under control of the hsp70 heat-shock promoter rescue the lethality of the dshv26 null allele even in the absence of heat shock, in contrast to all other lines we tested (Table 3). This result points to a potential role of the basic region as a negative regulator of the signaling function of Dsh.

PDZ domain mutations: Surprisingly, we recovered only one allele, UAS-dsh8-1, that maps to the PDZ domain. It only mildly attenuates the Dsh eye misexpression phenotype and does not attenuate the Dsh wing misexpression phenotype at all. While this result might suggest that there is only a minor requirement for this domain in Wg signaling, it could also mean that single point mutations have little effect on the function of the PDZ domain. In our hands, deletion of the PDZ domain completely abolishes Dsh function in both Wg signaling and tissue polarity. Our result contrasts with studies by Axelrod et al. (1998), which show that the PDZ domain is not necessary to rescue the embryonic lethality of dsh null embryos. This discrepancy may be due to the fact that different types of rescue assays were used to study the domain requirements of Dsh. We used heat-shock-inducible transgenes that allowed complete rescue to adulthood under nearly physiological conditions, whereas Axelrod et al. overexpressed different mutant versions of Dsh by RNA injection into embryos. The construct we used in our experiments deletes amino acids 287–336, which lie within the PDZ domain, while the construct that Axelrod used removes amino acids 152–333, consisting of the basic domain, CK1–CK4 and ST1–ST5. Thus an alternative explanation is that the construct utilized by Axelrod et al. deletes a region that inhibits the activity of Dsh. This is consistent with our observations implicating the basic domain as a negative regulator of dsh activity. Other studies have found a requirement for the PDZ domain in regulating β-catenin stability and in regulating transcription of LEF reporter constructs (Yanagawaet al. 1995; Liet al. 1999b; Yamamotoet al. 1999). In addition, the PDZ domain binds to Axin, FRAT, CK1, CK2, PP2A, and IDAX, proteins that regulate Wg signaling (Willertet al. 1997; Liet al. 1999a; Peterset al. 1999; Sakanakaet al. 1999; Strovelet al. 2000; Hinoet al. 2001). Thus, while the PDZ domain may be dispensable for the function of Dsh under certain experimental conditions, it appears to be essential for Wg signaling and tissue polarity under physiological conditions.

View this table:
TABLE 3

Rescue of dshv26 and dsh1 phenotypes by Dsh transgenes

Figure 5.

—Novel Daxin alleles suppress UAS-dsh misexpression phenotypes. (A) Daxin16-21 suppresses GMR-Gal4; UAS-dsh phenotypes in the eye (left) and 69B-Gal4; UAS-dsh phenotypes in the wing (right). (B) Protein sequence alignment of DAxin, mouse Axin, and Conductin. Numbering of amino acids according to the published sequences is shown. Amino acids that are identical or conserved with respect to DAxin are high-lighted in gray. The RGS, DIX, putative GSK3-β-binding site and β-catenin-binding sites are boxed. DAxin mutations are indicated and altered amino acids are boxed.

SH3-binding domain mutations: We isolated three new UAS-dsh alleles that carry mutations in a novel domain of Dsh. This region lies between the PDZ domain and the DEP domain, is proline rich, and possesses a consensus sequence for a class I core SH3 protein-binding motif, RTEPVRP at position 352–358 (Figure 4B; Kayet al. 2000). Proline-rich sequences that contain this core domain mediate the binding of these proteins to SH3 proteins. This core motif is conserved in the mammalian homologs of dsh and so are the surrounding prolines. UAS-dsh8-12 mutates the last proline in this core binding motif into a leucine at position 358, and UAS-dsh8-79 and UAS-dsh8-68 mutate an aspartic acid at position 360 to valine and a glycine at position 362 to aspartic acid, respectively (Table 2; Figure 4). The proline and aspartic acid residues are conserved between dsh and its dvl homologs, while the glycine residue is replaced by an alanine residue in the dvl genes. UAS-dsh8-12 and UAS-dsh8-79 disrupt but do not completely abolish Wg signaling while UAS-dsh8-68 possesses more Wg signaling activity than does either UAS-dsh8-12 or UAS-dsh8-79 (Table 2). This could be because UAS-dsh8-68 maps further from the core SH3-binding site motif or because the amino acid that it mutates is not conserved. Although the mutations encoded by UAS-dsh8-79 and UAS-dsh8-68 do not map within the core SH3-binding consensus motif, it is known that amino acids that surround this motif are important for optimal ligand preference. Studies utilizing combinatorial peptide libraries have defined binding sequences for SH3 proteins and show that prolinerich sequences around the core domain are important for binding (Kayet al. 2000).

The identification of mutations in a putative SH3-binding domain is intriguing since these proteins have not been implicated in Wg signaling events. Interestingly, the cytoplasmic tails of Arrow and DFz2 also contain putative SH3-binding domains (Bhanotet al. 1996; Wehrliet al. 2000). In addition, D-Axin contains putative SH3-binding domains in the RGS domain, the β-catenin-binding domain, and immediately amino terminal to the DIX domain (Hamadaet al. 1999; Willertet al. 1999). SH3 proteins act as adaptors linking signaling molecules into complexes and localizing proteins to the cell membrane (Mayer 2001). Hence it is tempting to speculate that proteins with multiple SH3 domains participate in Wg signaling events, perhaps to localize Dsh and Axin in proximity to the Arrow and DFz2 cell surface receptors. Indeed, Dsh is localized to the cell membrane when it is coexpressed with Fz1 in Xenopus oocytes. It has also been suggested that the DEP domain is important for this localization during planar polarity signaling (Axelrodet al. 1998).

DEP domain mutations: The isolation of two planar polarity alleles that encode mutations within the DEP domain agrees with other studies that demonstrate a requirement for this domain in planar polarity signaling (Axelrodet al. 1998; Boutroset al. 1998). The dshA3 mutation maps four amino acids distal to the previously isolated dsh1 allele and encodes an arginine-to-histidine mutation at position 413 (Table 2; Figure 4). The dshA3 and dsh1 mutations replace a positively charged amino acid with a neutral amino acid, suggesting that Dsh may contact negatively charged proteins during planar polarity signaling or bind to membrane phospholipids. dshA21 encodes a cysteine-to-arginine mutation at position 472. Cysteine residues can be palmitoylated and such a lipid modification can be important for cell membrane attachment (although we tried to incorporate labeled palmitate in Dsh and failed to detect any). Thus it is possible that this mutation disrupts the ability of the DEP domain to become membrane localized during planar polarity signaling. On the recently established structure of the DEP domain (Wonget al. 2000), the dshA3 and dsh1 mutations are located fairly close to the cysteine that is mutated in dshA21, suggesting that these three residues are collectively involved in the function of the DEP domain (Figure 4D).

Despite the fact that the DEP domain is viewed to be specific for polarity signaling, we did obtain three new UAS-dsh alleles that disrupted Wg signaling. These mutations clustered to a region of the DEP domain extending from position 440 to 459, away from the dsh planar polarity alleles (Table 2; Figure 4). This indicates that the DEP domain is required for Wg signaling, in agreement with findings that dsh constructs that lack the DEP domain cannot rescue the embryonic lethality of a dsh null allele (Axelrodet al. 1998; Boutroset al. 1998; Liet al. 1999b; this study).

Although all UAS-dsh alleles that were tested lost or attenuated the Wg signaling function, nearly all caused planar polarity defects when expressed in the wing. Both gain and loss of signaling activity alter planar cell polarity. Therefore, it is hard to determine whether these alleles contain mutations that are specific for Wg signaling and that leave planar polarity functions intact, or if they act as dominant negatives for this function. Another possibility is that the Wg signaling function of dsh is more sensitive to perturbations in dsh activity than is the planar polarity function and that these alleles behave as hypomorphs. Indeed we find that lower levels of a dsh transgene are required to rescue planar polarity functions of dsh than to rescue Wg signaling functions (Table 3). Three dsh alleles exist that specifically perturb planar polarity, however, arguing that planar polarity functions and Wg signaling functions are separable. In addition, when constructs that contain the dsh1 mutation are overexpressed they cannot rescue the endogenous dsh1 mutation, arguing that dsh1 is not a hypomorph and that Dsh acts as a modular protein (Boutroset al. 1998).

Phosphorylation mutants: Dsh is a phosphoprotein and Wg signaling generates hyperphosphorylated forms of Dsh, which are enriched in membrane fractions in biochemical assays (Yanagawaet al. 1995). While this finding suggests that the phosphorylation state of Dsh may regulate its activity, we did not isolate any mutations in putative phosphorylation sites in the forward screens. Furthermore, in vitro mutagenesis of putative phosphorylation sites of Dsh in a region spanning amino acids 178–254 did not reveal a requirement for these sites in the Wg signaling function of Dsh. Thus, phosphorylation sites in Dsh may be redundant and more than one site may need to be mutated to produce a phenotype. This conclusion is supported by phosphotryptic mapping experiments, which identified at least three phosphorylation sites in Dsh (Willertet al. 1997).

Surprisingly, however, region ST4 is essential for planar polarity signaling (Figure 4C; Table 3). This region was previously shown to bind the protein kinase PAR-1, which is thought to act in Wg signaling rather than in tissue polarity (Sunet al. 2001). We note, however, that the PAR-1 kinase is implicated in generating cell asymmetry in Caenorhabditis elegans (Guo and Kemphues 1995) and in the Drosophila oocyte (Shulmanet al. 2000; Tomancaket al. 2000; Coxet al. 2001; Huynhet al. 2001). These results point to a potential role of PAR-1-mediated Dsh phosphorylation in planar polarity.

Second-site suppressors; mutations in axin: In addition to mutations in the UAS-Dsh transgene itself, the UAS-dsh misexpression screen yielded second-site modifiers on the third and fourth chromosome. Modifiers on the first and second chromosome could not be recovered due to the strategy of our screen (Figure 1C). Five of the second-site modifiers map near axin and were indeed found to contain mutations within the axin gene (Figure 5B). They behave as dominant suppressors of Dsh misexpression phenotypes in both the wing and eye (Figure 5A) but do not modify Wg or DFz2 misexpression phenotypes (data not shown). In addition, these alleles are homozygous viable and have no phenotype when they are recombined away from UAS-dsh. Since Axin normally suppresses Wg signaling, and null axin alleles do not interact with UAS-dsh, we infer that these alleles specifically suppress overexpressed forms of Dsh but do not affect Dsh that is regulated by Wg signaling. This would imply that overexpressed Dsh works through a mechanism that is different from Dsh that is activated by Wg. For example, overexpressed Dsh may interact with Axin through binding to a domain that is different from the Axin domain that interacts with Wg-activated Dsh.

Conclusions: This work and other studies suggest that Dsh is a modular protein with specific domains dedicated to Wg and planar polarity signaling. How is Dsh activity regulated and how does it mediate Wg signaling? The targeting of Dsh to the cell membrane and the regulation of its phosphorylation state are correlated with Wg and planar polarity signaling. The DEP domain is necessary to localize Dsh to the cell membrane (Axelrodet al. 1998) and we have identified a putative SH3-binding site that may also be important for membrane localization of Dsh. It will be interesting to examine the phosphorylation state of the Dsh protein that is produced from the alleles generated in these screens and to determine if protein from alleles that mutate the potential SH3-binding site are membrane localized. CK1, CK2, and PAR-1 bind to Dsh and phosphorylate it. Moreover, CK1 and PAR-1 are positive regulators of Wg signaling that promote stabilization of β-catenin and induce the expression of Wnt target genes (Willertet al. 1997; Peterset al. 1999; Sakanakaet al. 1999; Sunet al. 2001). Thus, Dsh localization to the cell membrane may be controlled by CK1 and PAR-1, which may lead to changes in Dsh activity.

Acknowledgments

We thank M. Fish and M. Müller-Borg for technical assistance. A.W. was supported by a postdoctoral fellowship of the Deutsche Forschungsgemeinschaft. A.P. was supported by a Walter and Ida Berry Fellowship. R.N. is an investigator of the Howard Hughes Medical Institute.

Footnotes

  • Communicating editor: T. Schüpbach

  • Received December 13, 2001.
  • Accepted March 18, 2002.

LITERATURE CITED

View Abstract