A Genetic Screen for Hedgehog Targets Involved in the Maintenance of the Drosophila Anteroposterior Compartment Boundary

Corresponding author: Konrad Basler, University of Zürich, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland., basler{at}molbio.unizh.ch (E-mail)

Communicating editor: K. GOLIC

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The development of multicellular organisms requires the establishment of cell populations with different adhesion properties. In Drosophila, a cell-segregation mechanism underlies the maintenance of the anterior (A) and posterior (P) compartments of the wing imaginal disc. Although engrailed (en) activity contributes to the specification of the differential cell affinity between A and P cells, recent evidence suggests that cell sorting depends largely on the transduction of the Hh signal in A cells. The activator form of Cubitus interruptus (Ci), a transcription factor mediating Hh signaling, defines anterior specificity, indicating that Hh-dependent cell sorting requires Hh target gene expression. However, the identity of the gene(s) contributing to distinct A and P cell affinities is unknown. Here, we report a genetic screen based on the FRT/FLP system to search for genes involved in the correct establishment of the anteroposterior compartment boundary. By using double FRT chromosomes in combination with a wing-specific FLP source we screened 250,000 mutagenized chromosomes. Several complementation groups affecting wing patterning have been isolated, including new alleles of most known Hh-signaling components. Among these, we identified a class of patched (ptc) alleles exhibiting a novel phenotype. These results demonstrate the value of our setup in the identification of genes involved in distinct wing-patterning processes.

DROSOPHILA limbs are subdivided into distinct sets of cells designated as compartments (GARCIA-BELLIDO et al. 1973 ; LAWRENCE and STRUHL 1996 ; DAHMANN and BASLER 1999 ). Once cells have been allocated to a particular compartment, they remain part of it throughout development. The boundary separating two adjacent compartments cannot be crossed by wild-type cells. Although other models have been discussed in the past, it is widely assumed that cells of opposite compartments are kept separate by distinct cell adhesion properties referred to as "cell affinities" (GARCIA-BELLIDO 1975 ; DAHMANN and BASLER 1999 ).

The Drosophila wing is subdivided by two such lineage boundaries, one between the anterior (A) and posterior (P) compartments and another between the dorsal (D) and ventral (V) compartments. While it was originally assumed that compartment-specific properties are controlled in a compartment-wide manner by selector genes (GARCIA-BELLIDO 1975 ), studies over the past 5 years have indicated a uni- or bidirectional interplay between cells of opposite compartments across the boundary. For example, cells on the dorsal and ventral sides of the D/V boundary express high levels of the Notch ligands Serrate and Delta, respectively. The resulting Notch signaling activity in cells flanking the D/V boundary contributes in an important, yet not entirely understood, manner to its maintenance.

At the A/P boundary, the signaling mechanisms that control compartment-specific adhesion properties are better understood. P cells heritably express the selector gene engrailed (en) (LAWRENCE and MORATA 1976 ). En programs P cells to secrete Hedgehog (Hh) protein, which acts as a short-range signal on A cells located close to the compartment boundary (BASLER and STRUHL 1994 ). En also prevents the expression of the Hh signal transduction component Cubitus interruptus (Ci) in P cells. Hence, only A cells can respond to Hh, and they do so by upregulating the expression of Hh target genes, such as decapentaplegic (dpp) and patched (ptc). dpp encodes a BMP homolog that functions as a long-range morphogen to establish different cell fates along the anteroposterior axis. ptc encodes the receptor for Hh and negatively regulates the signaling activity of Smoothened (Smo). Upregulation of Ptc levels in response to the Hh signal promotes the sequestration of Hh protein, thereby reducing the range of Hh activity to a narrow stripe of A cells (CHEN and STRUHL 1996 ). Clonal analysis has demonstrated a crucial role for both the selector gene en and the Hh signal transduction components in defining compartment-specific cell affinities. For example, if A cells adjacent to the boundary become mutant for smo (BLAIR and RALSTON 1997 ; RODRIGUEZ and BASLER 1997 ), they sort out from other A cells and segregate into P territory. Thus the Hh signal induces in A cells an important aspect of the compartmental cell segregation system. Cells lacking both en and ci functions sort out from cells of both compartments (DAHMANN and BASLER 2000 ). This suggests that the cell adhesion properties of P and A cells close to the boundary depend on the activity of En and the response to Hh, respectively. The requirement of the zinc-finger protein Ci for establishing A-specific cell segregation properties indicates that putative cell adhesion molecules are regulated transcriptionally (DAHMANN and BASLER 2000 ).

In this study we present a genetic screen designed to identify genes required for the maintenance of the A/P compartment boundary. On the basis of the assumption that mutations in such genes would cause phenotypes similar to those observed with mutations in smo or ci, we established a wing-specific, F₁, FRT-FLP screen to create clones of wing cells carrying random mutations (GOLIC 1991 ; XU and RUBIN 1993 ). Wing specificity was achieved by driving the expression of flp with regulatory elements of the vestigial (vg) gene. Efficient coverage of the genome was achieved by the use of "double FRT chromosomes," which carry an FRT sequence on both sides of an autosomal centromere. Approximately 250,000 mutant chromosomes were analyzed. Despite the high number of smo-like mutations found, no gene encoding a novel cell adhesion molecule was identified. However, one complementation group that exhibits a strong smo-like phenotype was identified as a class of gain-of-function ptc alleles that provide new insights into the action of the Hh ligand and receptor. Moreover, we demonstrate that the screen is ideally suited to identifying genes controlling wing patterning.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila genotypes:
We tested the following gal4/UAS-flp combinations for their range of activity (Fig 2):

• dpp::flp: y w; P[mini w⁺, dpp-gal4] P[mini w⁺, UAS-flp]/TM6b;

  
  

  View larger version (97K): In this window In a new window Download PPT slide   Figure 1. Drosophila wings with smo mutant clones induced by means of different flp transgenes. (A) Wild-type wing for comparison with the longitudinal veins designated as L1–L5 and the A/P compartment boundary represented by the solid line. (B–D) When clones have been induced by the use of a heat-shock-driven hsp70-flp transgene, the adult wings exhibit different penetrance of a phenotype that can be correlated with the sorting out of smo mutant clones: duplication and displacement of L4 (marked by the asterisk in B) and the reappearance of anterior margin bristles in a more posterior margin region (indicated by the arrow in B); more severe phenotypes display an increasing anteriorization of the wing (C and D). (E) A wing where UAS-flp was driven by dpp-gal4. Ectopic veins between L3 and L4 are indicated by arrowheads. When spalt-gal4 or ptc-gal4 was used instead, the phenotype was very similar (data not shown). (F–H) Wings of flies with the vgBE-gal4 UAS-flp combination.

  
  

  View larger version (68K): In this window In a new window Download PPT slide   Figure 2. vgBE-gal4 is active in almost all cells of the wing and haltere primordia during larval development. ß-Gal staining was performed on wing (W), leg (L), and haltere (H) discs of the following genotypes: (A) ptc-gal4 UAS-flp/actin5C-promoter-FRT-Draf-stop-FRT-lacZ; (B) spalt-gal4 UAS-flp/actin5C-promoter-FRT-Draf-stop-FRT-lacZ; (C) vgBE-gal4 UAS-flp/actin5C-promoter-FRT-Draf-stop-FRT-lacZ; and (D) vgBE-gal4 UAS-flp/UAS-lacZ. (E) GFP expression in imaginal discs of the genotype hsp70-gfp FRT19/FRT19; vgBE-gal4 UAS-flp. Discs of the genotype dpp-gal4 UAS-flp/actin5C > Draf > lacZ had a similar lacZ pattern in the wing as those in A (not shown).

• ptc::flp: y w; P[mini w⁺, UAS-flp] P[mini w⁺, gal4]ptc/TM6b;

• spalt::flp: y w; P[mini w⁺, spalt-gal4] P[mini w⁺, UAS-flp]/TM6b;

• vg::flp: y w; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp]/CyO.

These combinations were crossed to y w; actin5C-FRT-Draf-stop-FRT-lacZ/CyO flies. Third instar imaginal disc staining of ß-galactosidase (ß-Gal) was carried out according to standard procedures.

For generating interchromosomal recombination on the X chromosome, the following genotype was generated: y w P[y⁺, hsp70-HA-gfp-HA] P[ry⁺, hsp70-neo, FRT]19/y w P[ry⁺, hsp70-neo, FRT]19; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp]/CyO.

We tested the following gal4/UAS-flp combinations for generating smo mutant clones (Fig 1):

• with hsp70-flp: y w; P[ry⁺, hsp70-flp]; smo³ P[mini w⁺, FRT]39/P[ry⁺, hsp70-flp]; P[y⁺, hsp70-cd2] P[mini w⁺, FRT]39;

• with dpp::flp: y w; P[ry⁺, hsp70-flp]; smo³ P[mini w⁺, FRT]39/P[y⁺, hsp70-cd2] P[mini w⁺, FRT]39; P[mini w⁺, dpp-gal4] P[mini w⁺, UAS-flp]/TM6b;

• with spalt::flp: y w; P[ry⁺, hsp70-flp]; smo³ P[mini w⁺, FRT]39/P[y⁺, hsp70-cd2] P[mini w⁺, FRT]39; P[mini w⁺, spalt-Gal4] P[mini w⁺, UAS-flp]/TM6b;

• with ptc::flp: y w; P[ry⁺, hsp70-flp]; smo³, P[mini w⁺, FRT]39/P[y⁺, hsp70-cd2] P[mini w⁺, FRT]39; P[mini w⁺, Gal4]ptc; P[mini w⁺, UAS-flp]/TM6b;

• with vg::flp: y w ; smo³, P[mini w⁺, gal4]vgBE P[mini w⁺, FRT]39/P[y⁺, hsp70-cd2] P[mini w⁺, FRT]39; P[mini w⁺, UAS-flp]/TM6b.

2xFRTs: 2xFRTs were generated by combination of the corresponding single FRTs through meiotic recombination. To test the efficiency of the 2xFRTs in combination with the vg::flp, markers for clonal analysis were recombined onto 2xFRTs, resulting in the following genotypes:

• for the second chromosome: y w hsp70-flp; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp] P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42/P[mini w⁺, arm-lacZ) P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42 P[mini w⁺, hsp70-HA-gfp, smo⁺];

• for the third chromosome: y w hsp70-flp; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp]; P[y⁺, hsp70-cd2] P[mini w⁺, FRT]79 P[ry⁺, hsp70-neo, FRT]82 P[mini w⁺, hsp70-HA-gfp, smo⁺].

Clonal analysis of new ptc alleles ( Fig 6): For the analysis of homozygous mutant clones in wing imaginal discs, flies of the following genotypes were generated:

• for observation of compartmental segregation behavior: y w hsp70-flp; P[ry⁺, hsp70-neo, FRT]42 P[y⁺, hsp70-cd2] P[mini w⁺, lacZ]hh/P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42 ptc^N15;

  
  

  View larger version (57K): In this window In a new window Download PPT slide   Figure 3. 2xFRTs allow the generation of clones with both chromosomal arms. (A) Schematic representation of the recombination events with a 2xFRT: the mother cell in the model is heterozygous for the red marker on one arm and the green marker on the opposite arm. Recombination can lead to nine genetically different daughter cells, depending on which chromosomal arm FLP-induced recombination occurred and which combination of chromosomal arms was paired together. If recombination and pairing resulted in a daughter cell identical to its mother cell, then another nine possible granddaughter cells could be created in the next round of recombination. However, if the daughter cell became homozygous for one arm, only recombination of the other still-heterozygous arm leads to different granddaughter cells. Daughter cells homozygous for both arms result in a recombinatorial dead end. Recombination in such cells no longer creates genetic diversity. Therefore, continuous supply of recombinase would promote the generation of homozygous cells. The combination of vg::flp with the 2xFRTs induces clones efficiently and independently on both left (2L and 3L) and right (2R and 3R) chromosomal arms for the second (B) and the third chromosome (C; see MATERIALS AND METHODS for the genotypes).

  
  

  View larger version (25K): In this window In a new window Download PPT slide   Figure 4. Crossing schemes for the screen on the X and second chromosome. (A) Screening of the second and third chromosomes was basically identical except that vg::flp, which is located on the second chromosome, had to be recombined on the second 2xFRT chromosome of the tester flies. (B) In contrast to the autosomes, recombination of the X chromosome occurs only in females. As females can lose a mutation through meiotic recombination, five independent balanced lines of each mutation were established. One of these five lines was then selected on the basis of male lethality and reproducibility of the initially observed phenotype. FRT, rectangle; centromere, small solid circle; vg::flp, large solid circle; yellow + transgene, y+; and mutation, a cross.

  
  

  View larger version (94K): In this window In a new window Download PPT slide   Figure 5. Representative wings of the complementation groups presented in Table 3. Clones have been induced by vg::flp. smo (A), col/kn (B and C), fu (D), new class of ptc alleles (E), pka (F), ptc (G), en/inv (H), brk (I), sgg (K), 2F26 (L), 2D5 (M), 3N5 (N), and 3F43 (O).

  
  

  View larger version (104K): In this window In a new window Download PPT slide   Figure 6. Clones of cells homozygous mutant for the novel class of ptc alleles cross the anteroposterior compartment boundary and lose expression of Hh target genes. (A and B) Anterior clones homozygous for the ptcP83 allele (marked by the absence of CD2; green in A and B) migrate into territory of the posterior compartment (A; lack of hhZ expression in red is marked by arrow) and fail to upregulate the expression of dppZ (red in B, arrowhead). The novel ptc alleles are partly viable. Wings of the following genotypes are shown: ptcQ67/Df(2R)44CE (C), ptcQ67/ptcIIW (D), ptcQ67/Df(2R)H3D3 (E), and ptcP83/Df(2R)H3D3 (F).

• for effect on Hh target genes: y w hsp70-flp ; P[mini w⁺, lacZ]dpp P[ry⁺, hsp70-neo, FRT]42 P[y⁺, hsp70-cd2]/P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42 ptc^N15.

Overexpression of Ptc ( Fig 7): For overexpression analysis of mutant and wild-type ptc alleles, we tested the following gal4/UAS-ptc combinations: with en-gal4:

• y w; P[mini w⁺, lacZ]ptc P[mini w⁺, gal4]en/P[mini w⁺, UAS-ptc^wt-myc];

  
  

  View larger version (60K): In this window In a new window Download PPT slide   Figure 7. The novel ptc alleles fail to sequester Hh but retain the ability to repress Smo. Third instar wing imaginal discs were costained with anti-Myc (green in A–F) and anti-ß-Gal (red in A–D) or anti-DE-cadherin (red in E and F), respectively. Wing discs were of the following genotypes: en-gal4 ptcZ/UAS-ptcwt-myc (A), en-gal4 ptcZ/UAS-ptcmut-myc (B), apt-gal4 ptcZ/UAS-ptcwt-myc (C and D), and apt-gal4 ptcZ/UAS-ptcmut-myc (E and F).

• y w; P[mini w⁺, lacZ]ptc P[mini w⁺, gal4]en/P[mini w⁺, UAS-ptc^mut-myc];

with apt-gal4:

• y w; P[mini w⁺, lacZ]ptc P[mini w⁺, gal4]apt/P[mini w⁺, UAS-ptc^wt-myc];

• y w; P[mini w⁺, lacZ]ptc P[mini w⁺, gal4]apt/P[mini w⁺, UAS-ptc^mut-myc].

Mutagenesis:
Males isogenic for the corresponding FRT chromosome were fed a 25-mM ethyl methanesulfonate (EMS), 1% sucrose solution. Mutagenized males were then mated in a 1:4 ratio to the tester virgins. The F₁ progeny were scored using a stereomicroscope for alteration of wing patterning. Flies exhibiting an interesting phenotype were backcrossed to the tester stock for a rescreen to check for germline transmission and reproducibility of the mutation. If possible, several phenotypic males were then mated to a balancer stock to establish a stable mutant line. Scoring F₁ progeny and the rescreen was different in the screen for the X chromosome. Here, only F₁ females could be scored and five individual stocks of every mutation were set up from single phenotypic virgins. Out of these five individual stocks one was selected on the basis of male lethality or the presence of viable males exhibiting a phenotype and by reproducibility of the phenotype when crossed to vg::flp. The genotypes of the mutagenized males were as follows: y w P[y⁺, hsp70-gfp] P[ry⁺, hsp70-neo, FRT]19/Y for the X chromosome; y w hsp70-flp; P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42 P[y⁺] for the second chromosome; and y w hsp70-flp; P[mini w⁺, FRT]79 P[ry⁺, hsp70-neo, FRT]82 P[y⁺] for the third chromosome. The tester stocks that were crossed to the mutagenized males and used for rescreening were y w P[ry⁺, hsp70-neo, FRT]19; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp] for the X chromosome; to y w hsp70-flp; P[mini w⁺, gal4]vgBE P[mini w⁺, UAS-flp] P[ry⁺, hsp70-neo, FRT]40 P[ry⁺, hsp70-neo, FRT]42 for the second; and y w hsp70-flp; P[mini w⁺, Gal4]vgBE P[mini w⁺, UAS-flp] P[mini w⁺, FRT]79 P[ry⁺, hsp70-neo, FRT]82 for the third chromosome. The following stocks were used for balancing: FM7 for mutations on the X and y w hsp70-flp; CyO/Sp and y w hsp70-flp; TM6b/MKRS for second and third chromosomal mutations.

Mapping of mutations:
Mutations on the second and third chromosome were first mapped to one chromosomal arm by reproducing the phenotype with single FRT chromosomes. Mutations conferring similar phenotypes and mapping to the same chromosomal arm were then grouped by complementation analysis. Complementation groups exhibiting a known phenotype were tested for complementation of known null alleles of the candidate gene. The alleles used in this study were as follows: smo³, a null allele of smo; fu^A, a kinase dead allele of fu; col¹, an amorphic allele of collier/knot; ptc^IIW, a null allele of ptc; Df(2R)enE, a deficiency that removes en and the closely related invected (inv) gene; pka-C1^E95, a null allele for the catalytic subunit of protein kinase A; and cos2⁵, a null allele of costal-2. Other mutations were mapped by using the Bloomington deficiency kit.

Immunohistochemistry:
Imaginal discs dissected from late third instar larvae were fixed and stained with the appropriate antibodies to mark clones and monitor reporter and transgene expression, respectively. If required, a heat shock for 1 hr at 38° followed by a recovery for 1 hr at 25° was given to have clonal marker genes expressed. Antibodies were rabbit polyclonal anti-green fluorescent protein (GFP; CLONTECH, Palo Alto, CA), mouse monoclonal anti-ß-Gal (Cappel), mouse monoclonal anti-Myc 9E10, and Alexa 488 and 594 secondary antibodies (Molecular Probes, Eugene, OR).

Construction of UAS-ptc:
Ptc cDNA derived from an available UAS-ptc construct (JOHNSON et al. 2000 ) was N-terminally fused with an myc epitope and reinserted into a pUAST vector. For generation of mutated ptc, site-directed mutagenesis was performed using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The construct presented here contains two missense mutations producing a protein with both R111W and G276D amino acid exchanges in the first extracellular loop.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

smo mutant clones cause characteristic defects in the venation and bristle pattern of the wing:
smo mutant cell clones located along the A/P compartment boundary position themselves in P territory even if they originate from A compartment cells (BLAIR and RALSTON 1997 ; RODRIGUEZ and BASLER 1997 ). Apparently, the transduction of the Hh signal programs Hh-receiving cells to sort out from cells not transducing Hh, be these wild-type P cells that lack Ci or experimental A cells mutant for smo. Mutations in genes encoding essential mediators or effectors of this Hh-induced segregation behavior should cause similar phenotypes; i.e., mutant A clones should also enter posterior territory. The situation in which such clone behavior can best be observed is fixed preparations of third instar imaginal discs in which the clones and the position of the A/P boundary are independently marked (BLAIR and RALSTON 1997 ; RODRIGUEZ and BASLER 1997 ). This assay is not suitable, however, for a high-throughput screen. We searched therefore for a reliable adult phenotype that is caused by the sorting-out phenomenon of smo mutant boundary cells.

Due to the maintenance of some A compartment properties, smo mutant clones originating from anterior cells can affect the pattern of the third (L3) and fourth (L4) longitudinal veins as well as the identity of wing margin bristles (BLAIR and RALSTON 1997 ; RODRIGUEZ and BASLER 1997 ; Fig 1, B–D). We observed that L4 was often disrupted, or partially duplicated, and shifted posteriorly. In addition, double-row wing margin bristles, typically present at the wing margin between L2 and L3, occasionally reappear more posteriorly. In some cases, even completely anteriorized wings could be observed (see also CHEN and STRUHL 1996 ). Often, ectopic veins between L3 and L4 were detected in conjunction with small duplications of L3. L3 is formed in A cells that are positioned next to Hh-receiving cells (BIEHS et al. 1998 ). smo mutant clones that are not in immediate contact with P cells do not move into P territory. Instead, they form ectopic boundaries of Hh target gene expression, which consequently result in ectopic or defective L3 veins.

We concluded therefore that a screen for mutations affecting vein L4 or posterior wing margin bristles should lead to the identification of genes coding for Hh signal transduction components or downstream effectors responsible for the Hh-dependent A/P cell segregation system.

The vg::flp system induces recombination specifically in the wing:
To set up a wing-specific F₁ screen we sought to modify the commonly used FRT-FLP system (GOLIC 1991 ; XU and RUBIN 1993 ). In most previous cases FLP is driven by a heat-shock-inducible promoter that is transiently activated at a particular time during development. However, without spatially controlling the expression of the FLP recombinase, clones may be absent in the tissue of interest, or even worse, clones may be present in other tissues where they can be harmful. The latter problem is of particular concern for an F₁ screen, as an individual exhibiting the desired phenotype has to be propagated to recover the underlying mutation.

Following the approach chosen by NEWSOME et al. 2000 who expressed FLP in an eye-specific manner by using the eyeless enhancer, we tested four enhancers [ptc, dpp, spalt, and vg boundary enhancer (vgBE)] for their ability to drive high levels of FLP expression in the wing. ptc and dpp are both Hh target genes, so their enhancers are active in a stripe of A cells along the anteroposterior compartment boundary, i.e., in the region critical for A affinity (Fig 2A). In contrast, the enhancer of the Dpp target gene spalt is active in a broader stripe centered on the ptc/dpp stripe and comprising both A and P cells (LECUIT et al. 1996 ; NELLEN et al. 1996 ; Fig 2B). Thus the spalt enhancer in addition would permit the generation of mutant P clones potentially moving into A territory. The fourth candidate, the vestigial D/V boundary enhancer, is activated by the Notch-Su(H) pathway and is thus entirely independent of Hh signaling (KIM et al. 1996 ). In the wing pouch of third instar larvae, this enhancer drives expression in a thin stripe along the dorsoventral compartment boundary corresponding to the presumptive wing margin cells (Fig 2D). Thus, the relevant region of activity for our screen would be the distal tip of the wing where the anteroposterior and dorsoventral axes intersect.

As we have previously observed that high levels of FLP recombinase are required for interchromosomal recombination, we used the Gal4 system to amplify the activities of the above-mentioned enhancers. Each of the four corresponding Gal4 drivers were used in combination with a UAS-flp transgene to induce smo mutant clones. ptc, dpp, and spalt-gal4 all caused very similar wing phenotypes with ectopic veins appearing between L3 and L4 (arrowheads in Fig 1E), but no displacement of L4 or defects of the bristle pattern was observed. Examination of third instar imaginal discs of such animals revealed that the smo clones were small and probably arose late in development (data not shown). Despite the rather mild wing phenotypes, animals with the ptc-gal4 and dpp-gal4 transgenes exhibited further defects on thorax, head, and legs that weakened these flies significantly.

In contrast to the above-described genotypes, the vgBE-gal4 UAS-flp combination (hereafter called vg::flp) was expected to generate smo clones only at the distal tip of the wing. However, a wide variety of adult wing phenotypes was observed, very similar to those associated with smo clones generated by a heat-shock-induced hsp70-flp transgene (Fig 1, F–H). Moreover, despite the severe disruption of wing pattern, these flies were fully viable and did not exhibit defects in other tissues. The severity of the wing phenotype, however, was unexpected since the vgBE enhancer shows a spatially restricted expression pattern during third instar. To test whether this enhancer drives flp expression elsewhere at earlier stages we used an actin5c > stop > lacZ transgene to irreversibly mark cells experiencing FLP activity. A UAS-lacZ transgene was used as a control to monitor the current state of Gal4, and thus FLP, activity. Whereas wings and halteres of the control animals showed a thin lacZ-expressing stripe along the dorsoventral boundary (Fig 2D), the actin5c > stop > lacZ animals exhibit ß-Gal activity in the entire wing disc (Fig 2C). To confirm and extend this observation we also tested whether interchromosomal recombination occurs throughout the wing primordium with vg::flp and used an X chromosomal FRT with a hsp70-gfp reporter to mark such clones. Again we observed recombination to occur throughout the entire wing disc (Fig 2E). Moreover, the size of the clones suggests that many of them were induced at early larval stages, which is in accordance with the strong smo phenotypes observed with vg::flp. We conclude from these experiments that all cells of the wing and haltere discs, but not those of other discs, must exhibit an early transient vgBE enhancer activity. For these reasons the vg::flp system was considered to be the most suitable source of recombinase for our purpose.

The use of 2xFRTs to screen entire autosomes:
A major disadvantage of FRT screens is that only a small fraction of the genome can be screened at once, i.e., one chromosomal arm. In an attempt to overcome this drawback we used meiotic recombination to construct chromosomes with FRTs on both sides of the centromere (referred to as 2xFRTs). We tested their use by combining them with vg::flp and appropriate imaginal disc marker genes. For both the second (Fig 3B) and third (Fig 3C) autosomal 2xFRTs, we observed efficient and independent recombination on both sides of the centromere. Importantly, no significant preference of one FRT over the other could be detected. In theory, exchange of chromosomal arms can occur as long as recombinase is present. Continuous supply of recombinase eventually approaches a state of complete loss of heterozygous cells and a concomitant presence of homozygous cells, i.e., "twinspots" and "clones" (Fig 3A). In the case of the first and third chromosomes, this state was nearly reached. However, since the vg::flp components are located on the left arm of the second chromosome, recombination of this arm can result in daughter cells that have lost the recombinase and are therefore no longer able to exchange the right arms of the second chromosome. Hence, they will remain heterozygous for the right arm if no recombination event has occurred there previously. Even though such cases were indeed found, we observed a high efficiency of clone induction for both arms of the second chromosome. We conclude from these experiments that the combination of 2xFRTs and vg::flp is ideally suited for a high-throughput F₁ screen for genes required for the segregation of A and P cells in the wing.

Identification of mutations conferring smo-like phenotypes:
The 2xFRT chromosomes were mutagenized in males with EMS and crossed to vg::flp females with the corresponding 2xFRTs (Fig 4A). Approximately 100,000 mutant F₁ animals were screened for each autosome and 50,000 for the X chromosome (Table 1). Adults with interesting phenotypes were individually backcrossed to vg::flp flies to assess the reproducibility of the phenotype. In many cases the observed phenotype did not recur in the F₂ generation. However, many mutations did breed through, in which case several males displaying the same phenotype were then used to establish a balanced stock. The genetic setup for the X chromosome mutagenesis was complicated by the fact that clones could be generated only in females, and mutant females were required to establish stable, balanced stocks (see MATERIALS AND METHODS; Fig 4B). Finally, balanced stocks of second and third chromosome mutants were retested with 1xFRT chromosomes to assign the mutations to single chromosomal arms. Mutations exhibiting similar phenotypes and mapping to the same chromosomal arm were grouped and subjected to complementation analysis. Some complementation groups were then further mapped by noncomplementation of deficiencies or candidate genes (see MATERIALS AND METHODS).

Four complementation groups that exhibit smo-like phenotypes were identified (Table 2 and Fig 5, A–E and H). Whereas the first and fourth of these complementation groups exhibited defects representing the entire spectrum of smo-like phenotypes (Fig 5A and Fig E), the other two did not show any alterations in L4 and margin bristles but displayed ectopic veins between L3 and L4 and a partial or complete fusion of these two longitudinal veins (Fig 5, B–D). Complementation analysis with a smo null allele revealed that the first group represents new alleles of smo itself, supporting the validity of the screen. The second and third complementation groups were identified as new alleles of collier and fused. collier, also known as knot, is a Hh target gene encoding a transcription factor required for the formation of the L3/L4 intervein region (VERVOORT et al. 1999 ). Fused is a serine/threonine kinase that acts positively in Hh receiving cells (PREAT et al. 1990 ). The fourth complementation group behaved like the first, yet mapped to the right arm of the second chromosome and could therefore not be allelic to smo. To test whether this gene was involved directly in compartmental affinity or whether it functioned in Hh signal transduction, we examined mutant clones in wing imaginal discs for their sorting-out behavior and for alterations of Hh target gene expression, respectively. Consistent with the adult wing phenotype, mutant clones did indeed sort out into P territory (Fig 6A).However, mutant clones also failed to upregulate the Hh target gene dpp (Fig 6B). Hence, the new complementation group appeared to identify a locus required for the transduction of the Hh signal and was further analyzed as described below.

In addition to the above-described complementation groups, we isolated a further single mutation that caused an L4 phenotype. Surprisingly, analysis in discs revealed that mutant clones migrated from the P to the A compartment (data not shown). In agreement with this observation we found, however, that the mutation failed to complement a small deficiency removing the two neighboring genes en and inv and hence represents an allele of one of these two genes.

A new class of ptc alleles with properties of smo loss-of-function mutations:
We focused our attention on the fourth complementation group. Apart from smo, only two other genes are known to be positive regulators of Hh signal transduction in the Hh receiving cells, fu and ci. Since fu is on the X and ci on the fourth chromosome, they could be excluded as candidate genes. We identified a deficiency, Df(2R)44CE, which failed to fully complement these new alleles (Table 3). Intriguingly, this deficiency is deleted for the ptc gene. Clones lacking ptc function display a gain rather than a loss of Hh signaling, rendering it unlikely that the new complementation group represented ptc alleles. However, since Ptc represses Smo function upon Hh binding, it is possible that the new complementation group coded for Ptc proteins that can no longer be repressed by Hh and therefore would display smo-like phenotypes. To investigate this possibility, we tested those putative new ptc alleles for complementation of an additional deficiency, Df(2R)H3D3, and of two known ptc alleles, ptc^IIW and ptc^S2 (Table 3 and Fig 6, C–F). All three new alleles fully complemented ptc^S2. ptc^S2 carries a missense mutation in the sterol-sensing domain of Ptc and encodes a protein that can bind and sequester Hh, but is unable to repress Smo (CHEN and STRUHL 1998 ; MARTIN et al. 2001 ; STRUTT et al. 2001 ). None of the three new alleles, however, was fully viable over the protein null allele ptc^IIW. Rare escapers with such a genotype often exhibited a fusion of L3 and L4, similar to the fused phenotype (Fig 6D).

To determine unambiguously whether the new alleles form a novel class of ptc alleles, we sequenced the entire ptc locus of all three mutants and identified missense mutations in all three alleles. Two mutations mapped to the first of the two large extracellular loops, R111W in ptc^P83 and G276D in ptc^N15, while the third (N936Y in ptc^Q67) mapped to the second such loop. The three alleles differed in strength. According to their viability over the deficiency or over the ptc null allele, they could be ranked as ptc^N15 > ptc^P83 > ptc^Q67 with ptc^N15 being the strongest allele.

The molecular nature and the clonal phenotype of these new ptc alleles suggested that the receptors encoded by these alleles cannot be repressed by the Hh ligand. To test the ability of the three mutant Ptc proteins to sequester Hh protein, we introduced their mutations into UAS-ptc transgenes. Wild-type or mutant ptc was then ectopically expressed in P compartment cells with an en-gal4 driver, and Hh signaling was monitored with a ptc-lacZ reporter. While the wild-type ptc transgene caused a strong reduction or even ablation of ptc-lacZ expression (Fig 7A), presumably by sequestration of Hh protein, Hh signaling was unaffected by the expression of the mutant Ptc proteins (Fig 7B). To verify the functionality of the mutant ptc constructs, wild-type and mutant ptc transgenes were expressed in dorsal compartment cells using apterous-gal4. Both wild-type (Fig 7C) and mutant (Fig 7D) Ptc completely repressed Hh signaling in the dorsal compartment. To verify that the failure to sequester Hh was not due to a mislocalization of the mutant Ptc proteins, these proteins were localized in situ with respect to E-cadherin. No differences could be detected between the staining patterns of wild-type vs. mutant forms of Ptc (Fig 7E and Fig F). We conclude therefore that these novel mutant Ptc proteins are not repressed by Hh because they fail to bind Hh efficiently, and they therefore inactivate Smo constitutively.

After completion of these studies a gain-of-function allele of ptc, ptc^con, was reported to also complement ptc^S2 (MARTIN et al. 2001 ). This allele leads to an amino acid exchange in the first extracellular loop and hence is equivalent to the above-described class of ptc alleles.

Other mutations affecting wing patterning:
In addition to the smo-like phenotypes we also scored a number of other phenotypes. A selection of mutants of these other phenotypic groups were kept for further analysis (Fig 5 and Table 4). Some of these were found to affect loci encoding negative regulators of Hh signaling, such as pka (Fig 5F), cos2, and ptc (Fig 5G). Others included negative modulators of other signaling pathways, such as brk (Fig 5I) and sgg (Fig 5K). The categories of observed phenotypes covered a wide range: notches, excessive or broadened veins, ectopic veins or loss of veins (Fig 5M), displaced or duplicated veins, blisters, bent wings, narrower or broader wings, smaller or larger wings, loss of margin bristles or ectopic margin bristles, axis duplications, ectopic bristles covering the wing blade or along veins (Fig 5O), hinge to wing transformations, crumpled wings, deformed compartments (Fig 5N), and outgrowths. The number and diversity of observed phenotypes validate the approach of a wing-specific FRT/FLP screen.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The aim of our screen was to identify genes involved in the formation of the anteroposterior compartment boundary. Specifically, we were interested in finding genes that directly confer A/P compartment-specific cell affinity, and we therefore set up a screen for phenotypes similar to those caused by mutations in smo. Cells lacking Smo activity do not exhibit proper A affinity because they cannot respond to Hh. Hence we expected to identify not only putative cell adhesion molecules but also positive regulators of the Hh signaling pathway.

Several screens based on the FRT-FLP method have been successfully implemented for genes affecting wing patterning (JIANG and STRUHL 1995 , JIANG and STRUHL 1998 ) and for genes required for integrin-mediated cell adhesion between dorsal and ventral wing surfaces (PROUT et al. 1997 ; WALSH and BROWN 1998 ). We chose to limit the FLP-induced mosaicism to the wing and found the vg::flp transgene combination ideally suited for this. To our surprise and advantage, vg::flp activity was not restricted to the region of the D/V boundary. It is possible that the vg boundary enhancer is broadly activated by Notch signaling at early stages of wing disc development (K. IRVINE, personal communication).

As a further improvement we generated chromosomes with FRTs flanking both sides of the centromere (2xFRT). The individual FRTs of these chromosomes appear to operate independently of each other, allowing efficient recombination of both chromosomal arms in the same animal. We assumed that the chance of a mutation on one chromosomal arm masking the phenotype caused by a mutation on the other arm would be very low. Indeed, when mutations were retested with single FRT-carrying chromosomes, we found only very few cases where wing phenotypes occurred independently with both arms. An interesting modification of our setup could be to use recessive cell-lethal mutations on both chromosomal arms to eliminate homozygous wild-type cells for one or both arms (NEWSOME et al. 2000 ). We found that 2xFRT chromosomes represent an efficient tool for screens based on the FRT-FLP method.

A total of 250,000 mutant chromosomes covering the X chromosome and both major autosomes were screened. Four complementation groups were identified that affected wing patterning similar to mutations in smo. The largest of these groups represented alleles in smo itself. Two groups exhibiting a subset of smo phenotypes represented new alleles of fused and collier/knot. Fused is a positive regulator of Hh signaling, and collier/knot is an Hh target gene required for the formation of the L3/L4 intervein region. Surprisingly, the remaining complementation group turned out to consist of novel ptc alleles with striking characteristics. Molecularly, they represent point mutations causing an amino acid substitution in either the first or the second large extracellular loop. In contrast to ptc null alleles, homozygous mutant clones failed to upregulate Hh target genes even in the presence of Hh. Together these findings suggest that the mutant proteins repress Smo constitutively, most likely because they fail to bind Hh. Animals mutant for trans-heterozygous combinations of these new ptc alleles with ptc^S2 were fully viable. The ptc^S2 product lacks the ability to repress Smo but is able to sequester, and hence bind to, Hh (CHEN and STRUHL 1996 ). The intragenic complementation we observed suggests that both functions of Ptc, binding of Hh and repression of Smo, can be provided by individual proteins that possess only one of each. Recently, it was shown that a combination of two proteins, one consisting of the N- and the other the C-terminal half of Ptc, reconstitutes Ptc function (JOHNSON et al. 2000 ). Although these experiments cannot be directly compared with our findings, together they do suggest that Ptc function can be separated intramolecularly into independent modules of N- vs. C-terminal and extra- vs. intracellular domains. One possible scenario that could explain the intragenic complementation would be if Ptc proteins act in a multimeric complex.

Besides those genes exhibiting a smo-like phenotype, many other mutations that affected the patterning of the wing, as well as its growth, were collected. We identified not only new alleles of most known components of the Hh signaling pathway, but also components of other signal transduction pathways. The vast variety of wing phenotypes observed in this screen suggests that its current setup could be useful to identify other genes involved in distinct processes of wing patterning.

The initial goal of our screen, the identification of a compartment-specific cell affinity gene, was not accomplished. All mutations conferring the phenotype we screened for could be classified as new alleles of known genes. Possible explanations for why no cell affinity genes were found are: the screen did not reach saturation; such genes act redundantly; the phenotype of loss of a compartment-specific cell adhesion molecule differs from that caused by the loss of Hh signaling or En activity; such a gene is essential for cell viability; the model by which transcriptional regulation of a modulator of cell adhesion is responsible for the segregation of A and P cells is incorrect.

Regarding the saturation issue, it must be considered that the entire genome could not be screened. Due to technical limitations, genes proximal to the FRTs and genes on the fourth chromosome are not accessible by the FRT-FLP method. The high number of alleles of certain known genes that we identified indicates that the majority of the genome was screened to saturation. Intriguingly, we found many more complementation groups on the first and second chromosomes compared to those on the third. The spectrum of phenotypes also differed among the chromosomes. Many mutations affecting vein positioning were found on the second chromosome, while the third chromosome revealed a high number of loci affecting growth and size. Uneven distribution of mutations for a specific phenotype, however, is not unusual and was observed in another genome-wide FRT-FLP screen (WALSH and BROWN 1998 ).

A serious concern is the possibility that the sought-after cell adhesion function is provided by a redundant set of genes. Duplicate genes or the contribution of several loci to the A/P affinity system would prevent the discovery of loss-of-function mutations by our assay. Experimental evidence that overexpression of a single cell adhesion molecule is sufficient to disrupt the A/P compartment boundary (DAHMANN and BASLER 2000 ) does not rule out this possibility.

It is also possible that the loss of compartment-specific cell affinity would be manifested in phenotypes that differ from those caused by aberrant Hh or En activity. In particular, it is conceivable that the gene coding for the compartment-specific cell adhesion property is essential for the survival of wing cells. Ci and En activities may merely modulate its transcription above a certain threshold required for the segregation of A and P cells. The total loss of this function may cause cell lethality, thereby resulting in a phenotype different from the smo-like defects.

In the light of these caveats the outcome of this screen does not rule out the model of differential compartment-specific cell affinity in the establishment of the lineage restriction boundary. However, within the limits of the saturation discussed above, our results do not support a model in which a single dedicated cell adhesion molecule defines the segregation behavior of A and P cells.

	ACKNOWLEDGMENTS

We thank B. Dickson for the FRT40/42 chromosome, G. Struhl for UAS-flp lines, the Bloomington Stock Center for numerous alleles and deficiencies, R. Burke for comments on the manuscript, and the Swiss National Science Foundation and the Kanton of Zürich for support.

Manuscript received November 7, 2002; Accepted for publication January 9, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

BASLER, K. and G. STRUHL, 1994  Compartment boundaries and the control of Drosophila limb pattern by hedgehog protein. Nature 368:208-214.[Medline]

BIEHS, B., M. A. STURTEVANT, and E. BIER, 1998  Boundaries in the Drosophila wing imaginal disc organize vein-specific genetic programs. Development 125:4245-4257.[Abstract]

BLAIR, S. S. and A. RALSTON, 1997  Smoothened-mediated Hedgehog signalling is required for the maintenance of the anterior-posterior lineage restriction in the developing wing of Drosophila. Development 124:4053-4063.[Abstract]

CHEN, Y. and G. STRUHL, 1996  Dual roles for patched in sequestering and transducing Hedgehog. Cell 87:553-563.[Medline]

CHEN, Y. and G. STRUHL, 1998  In vivo evidence that Patched and Smoothened constitute distinct binding and transducing components of a Hedgehog receptor complex. Development 125:4943-4948.[Abstract]

DAHMANN, C. and K. BASLER, 1999  Compartment boundaries: at the edge of development. Trends Genet. 15:320-326.[Medline]

DAHMANN, C. and K. BASLER, 2000  Opposing transcriptional outputs of Hedgehog signaling and engrailed control compartmental cell sorting at the Drosophila A/P boundary. Cell 100:411-422.[Medline]

GARCIA-BELLIDO, A., 1975  Genetic control of wing disc development in Drosophila. Ciba Found. Symp. 0(29):161-182.

GARCIA-BELLIDO, A., P. RIPOLL, and G. MORATA, 1973  Developmental compartmentalisation of the wing disk of Drosophila. Nat. New Biol. 245:251-253.[Medline]

GOLIC, K. G., 1991  Site-specific recombination between homologous chromosomes in Drosophila. Science 252:958-961.[Abstract/Free Full Text]

JIANG, J. and G. STRUHL, 1995  Protein kinase A and hedgehog signaling in Drosophila limb development. Cell 80:563-572.[Medline]

JIANG, J. and G. STRUHL, 1998  Regulation of the Hedgehog and Wingless signalling pathways by the F-box/WD40-repeat protein Slimb. Nature 391:493-496.[Medline]

JOHNSON, R. L., L. MILENKOVIC, and M. P. SCOTT, 2000  In vivo functions of the patched protein: requirement of the C terminus for target gene inactivation but not Hedgehog sequestration. Mol. Cell 6:467-478.[Medline]

KIM, J., A. SEBRING, J. J. ESCH, M. E. KRAUS, and K. VORWERK et al., 1996  Integration of positional signals and regulation of wing formation and identity by Drosophila vestigial gene. Nature 382:133-138.[Medline]

LAWRENCE, P. A. and G. MORATA, 1976  Compartments in the wing of Drosophila: a study of the engrailed gene. Dev. Biol. 50:321-337.[Medline]

LAWRENCE, P. A. and G. STRUHL, 1996  Morphogens, compartments, and pattern: Lessons from Drosophila? Cell 85:951-961.[Medline]

LECUIT, T., W. J. BROOK, M. NG, M. CALLEJA, and H. SUN et al., 1996  Two distinct mechanisms for long-range patterning by Decapentaplegic in the Drosophila wing. Nature 381:387-393.[Medline]

MARTIN, V., G. CARRILLO, C. TORROJA, and I. GUERRERO, 2001  The sterol-sensing domain of Patched protein seems to control Smoothened activity through Patched vesicular trafficking. Curr. Biol. 11:601-607.[Medline]

NELLEN, D., R. BURKE, G. STRUHL, and K. BASLER, 1996  Direct and long-range action of a DPP morphogen gradient. Cell 85:357-368.[Medline]

NEWSOME, T. P., B. ASLING, and B. J. DICKSON, 2000  Analysis of Drosophila photoreceptor axon guidance in eye-specific mosaics. Development 127:851-860.[Abstract]

PREAT, T., P. THEROND, C. LAMOUR-ISNARD, B. LIMBOURG-BOUCHON, and H. TRICOIRE et al., 1990  A putative serine/threonine protein kinase encoded by the segment-polarity fused gene of Drosophila. Nature 347:87-89.[Medline]

PROUT, M., Z. DAMANIA, J. SOONG, D. FRISTROM, and J. W. FRISTROM, 1997  Autosomal mutations affecting adhesion between wing surfaces in Drosophila melanogaster. Genetics 146:275-285.[Abstract]

RODRIGUEZ, I. and K. BASLER, 1997  Control of compartmental affinity boundaries by hedgehog. Nature 389:614-618.[Medline]

STRUTT, H., C. THOMAS, Y. NAKANO, D. STARK, and B. NEAVE et al., 2001  Mutations in the sterol-sensing domain of Patched suggest a role for vesicular trafficking in Smoothened regulation. Curr. Biol. 11:608-613.[Medline]

VERVOORT, M., M. CROZATIER, D. VALLE, and A. VINCENT, 1999  The COE transcription factor collier is a mediator of short-range hedgehog-induced patterning of the Drosophila wing. Curr. Biol. 9:632-639.[Medline]

WALSH, E. P. and N. H. BROWN, 1998  A screen to identify Drosophila genes required for integrin-mediated adhesion. Genetics 150:791-805.[Abstract/Free Full Text]

XU, T. and G. M. RUBIN, 1993  Analysis of genetic mosaics in developing and adult Drosophila tissues. Development 117:1223-1237.[Abstract]

This article has been cited by other articles:

				D. J. Casso, S. Liu, D. D. Iwaki, S. K. Ogden, and T. B. Kornberg A Screen for Modifiers of Hedgehog Signaling in Drosophila melanogaster Identifies swm and mts Genetics, March 1, 2008; 178(3): 1399 - 1413. [Abstract] [Full Text] [PDF]

				N. Link, P. Chen, W.-J. Lu, K. Pogue, A. Chuong, M. Mata, J. Checketts, and J. M. Abrams A collective form of cell death requires homeodomain interacting protein kinase J. Cell Biol., August 9, 2007; 178(4): 567 - 574. [Abstract] [Full Text] [PDF]

				A. G. de Ayala Alonso, L. Gutierrez, C. Fritsch, B. Papp, D. Beuchle, and J. Muller A Genetic Screen Identifies Novel Polycomb Group Genes in Drosophila Genetics, August 1, 2007; 176(4): 2099 - 2108. [Abstract] [Full Text] [PDF]

				J. Mathieu, H.-H. Sung, C. Pugieux, J. Soetaert, and P. Rorth A Sensitized PiggyBac-Based Screen for Regulators of Border Cell Migration in Drosophila Genetics, July 1, 2007; 176(3): 1579 - 1590. [Abstract] [Full Text] [PDF]

				X. Lu, S. Liu, and T. B. Kornberg The C-terminal tail of the Hedgehog receptor Patched regulates both localization and turnover Genes & Dev., September 15, 2006; 20(18): 2539 - 2551. [Abstract] [Full Text] [PDF]

				T. D. Tayler, M. B. Robichaux, and P. A. Garrity Compartmentalization of visual centers in the Drosophila brain requires Slit and Robo proteins Development, December 1, 2004; 131(23): 5935 - 5945. [Abstract] [Full Text] [PDF]

				J. D. Zirin and R. S. Mann Differing strategies for the establishment and maintenance of teashirt and homothorax repression in the Drosophila wing Development, November 15, 2004; 131(22): 5683 - 5693. [Abstract] [Full Text] [PDF]

				C. Thomas and P. W. Ingham Hedgehog Signaling in the Drosophila Eye and Head: An Analysis of the Effects of Different patched Trans-heterozygotes Genetics, December 1, 2003; 165(4): 1915 - 1928. [Abstract] [Full Text] [PDF]

				J. Jia, C. Tong, and J. Jiang Smoothened transduces Hedgehog signal by physically interacting with Costal2/Fused complex through its C-terminal tail Genes & Dev., November 1, 2003; 17(21): 2709 - 2720. [Abstract] [Full Text] [PDF]

				F. Janody, Z. Martirosyan, A. Benlali, and J. E. Treisman Two subunits of the Drosophila mediator complex act together to control cell affinity Development, August 15, 2003; 130(16): 3691 - 3701. [Abstract] [Full Text] [PDF]