A collective form of cell death requires homeodomain interacting protein kinase

Elimination of cells by programmed cell death (PCD) is a universal feature of development and aging (Jacobson et al., 1997; Vaux and Korsmeyer, 1999). In both vertebrates and invertebrates, dying cells often progress through a stereotyped set of transformations referred to as apoptosis. In this form of PCD the nucleus condenses, and the collapsing cell corpse fragments into “apoptotic bodies” that are engulfed by specialized phagocytes or neighboring cells (Kerr et al., 1972; Wyllie et al., 1980; Kerr and Harmon, 1991). Apoptosis requires autonomous genetic functions within the dying cell, and extrinsic cues that elicit apoptosis have been investigated in numerous experimental models (Danial and Korsmeyer, 2004; Salvesen and Abrams, 2004). Other forms of death are also thought to contribute during development and differ from apoptosis with respect to cellular morphology, mechanism, or mode of activation. These may include necrosis, characterized by swelling of the plasma membrane, or autophagic cell death, which is linked to extensive vacuolization in the cytoplasm (Kroemer et al., 2005). These forms of cell death can be caspase dependent or independent and may or may not be under deliberate genetic control (Kroemer et al., 2005).

Two conserved protein families comprise central elements of the apoptotic machinery (Salvesen and Abrams, 2004). Orthologous proteins represented by Ced4 in the nematode, Apaf1 in mammals, and Drosophila Ark (Dark) function as activating adaptors for CARD-containing apical caspases. During apoptosis, Ced4/Apaf1/Dark adaptors associate with pro-caspase partners (Ced3, Caspase 9, and Dronc) in a multimeric complex referred to as the “apoptosome”. This complex is regulated by Bcl2 proteins, but apparently through different mechanisms (for review see Kornbluth and White, 2005).

Previously, we and others genetically examined components of the Drosophila apoptosome (Rodriguez et al., 1999, 2002; Chew et al., 2004; Daish et al., 2004; Xu et al., 2005; Akdemir et al., 2006; Mills et al., 2006; Srivastava et al., 2006). dark and dronc are recessive, lethal genes. Both exert global functions during PCD and in stress-induced apoptosis. However, their roles in apoptosis are not absolute because rare cell deaths occurred in embryos lacking maternal and zygotic product of either gene (Xu et al., 2005; Akdemir et al., 2006). Elimination of dronc in the wing caused a unique, age-dependent phenotype associated with late-onset blemishing throughout the wing blade (Chew et al., 2004). Here, we show that this progressive phenotype is characteristic for wing epithelia that lack apoptogenic functions and is caused by defects in a communal form of PCD where epithelial cells are collectively and rapidly eliminated. We leveraged these findings to discover additional genes required for PCD and recovered a limited set of loci, many of which were previously unknown to function in cell death. Here, we establish that homeodomain interacting protein kinase (HIPK) is essential for coordinated death in the wing epithelium and, consistent with PCD functions in earlier developmental stages, regulates proper cell number in diverse tissue types.

Wings mosaic for dronc⁻ tissue exhibit normal morphology at eclosion but develop progressive, melanized blemishes with age (Chew et al., 2004). We applied similar methods to determine whether lesions in other apoptogenic genes present a similar phenotype. After eclosion, wings mosaic for Df(H99), a deletion removing the apoptotic activators reaper (rpr), grim, and hid (Abrams, 1999), were morphologically normal at eclosion, but over 3–7 d, melanized blemishes appeared at random throughout the wing (Fig. 1 C). Likewise, homozygous drice^Δ1 adult “escapers” deficient for the effector caspase Drice (Muro et al., 2006) also presented normal wings at eclosion but developed blemishes with age (Fig. 1 D). Wings mosaic for dark⁸², a null allele of dark (Akdemir et al., 2006), were indistinguishable from wild-type (WT) at eclosion (Fig. 1 A), but within 4 d developed wing blemishes (Fig. 1 B). These late-onset blemishes became markedly more severe as animals aged. Similar yet less severe wing blemishes occurred in adults homozygous for dark^CD4, a hypomorphic allele of dark (Chew et al., 2004). Together, these observations establish that late-onset progressive blemishing in mosaic wings is a characteristic phenotype shared among mutants in canonical PCD pathways.

In the wing of newly eclosed adults, PCD removes the epithelium that forms the dorsal and ventral cuticles (Johnson and Milner, 1987; Kimura et al., 2004). To determine whether the cause of the blemish phenotype might trace to defective death in the wing epithelium, we examined this tissue in dark mutants. For these studies, wings of dark^CD4 adults were prepared for light and electron microscopy. Histological analyses at the light level showed that on the first day of eclosion, the dorsal and ventral cuticles of WT animals became tightly merged with no intervening tissue evident between these layers (Fig. 1 E). However, even 14 d after eclosion, cells and cell remnants remained situated between the dorsal and ventral cuticles in dark mutants. This “undead” tissue was most easily visualized in lateral sections through melanized blemishes (Fig. 1, F–H). Further examination of the persisting epithelium at the EM level showed evidence of intact cells soon after eclosion (Fig. 1 P) and ectopic cellular material 24 h after eclosion (Fig. 1, I–K).

To directly examine the death of wing epithelial cells in vivo, we adapted a transgenic nuclear DsRed reporter driven by vestigial-Gal4 (vg:DsRed) (Vegh and Basler, 2003; Barolo et al., 2004), which permits visualization of the fate of these cells soon after eclosion. Observations with this pan-epithelial marker in the wing confirmed earlier studies (Kimura et al., 2004; Xu et al., 2005). Fig. 1 L shows that within 1 h of eclosion, intact epithelial cells are clearly present and regularly patterned throughout the wing. 1–2 h later (2–3 h after eclosion), the entire intervein epithelium disappears, manifested here by the abrupt loss of DsRed throughout the wing blade (Fig. 1 M). Live, real-time imaging of the wing in newly eclosed adults revealed unexpected features associated with elimination of the intervein epithelium (Fig. 2 and Videos 1–3). Epithelial cells, labeled by nuclear fluorescence, were arranged in a regular, predictable pattern throughout the wing. Then, consistent with nuclear breakdown, fluorescence became redistributed throughout the cell followed by indications of blebbing and the appearance of fragmenting cells. Occasionally, weak fluorescence enclosed in cell corpses condensed to bright punctate bodies. This series of apoptogenic changes spread extremely rapidly throughout the epithelium, appearing here as a collective wave initiating from the peripheral edge and moving across the wing blade (Fig. 2, top panels). Within just 4 min, virtually all nuclei (∼450 cells) within a space of ∼114 mm² converted from viable to apoptotic morphology. The process involved tight coordination at the group level because the likelihood of a single cell apoptosing was clearly linked to similar behaviors by nearest neighboring cells over short time frames (Video 1). Also, the direction and size of the cell death wave may not be fixed in every region of the wing, but centrally located cell groups were generally eliminated earlier (Fig. 1 L).

Unlike conventional examples of PCD in development, we found no indication that overt engulfment of apoptotic corpses occurred at the site of death. Instead, DsRed-labeled cell remnants were passively swept en masse toward the nearest wing vein (Fig. 2, bottom panels) where, apparently under hydrostatic pressure, cell debris streamed proximally toward the body through the wing or along the wing vein (Fig. 2, bottom panels; and Video 3). Together, these observations describe a communal form of PCD that rapidly eliminates the wing epithelium through coordinated group behavior.

We used the vg:DsRed reporter to track the fate of mosaic wing epithelia where mutant clones were induced. In sharp contrast to WT wings, abnormally persisting cells could be readily detected as patches of DsRed in the nuclei of epithelial cells in mosaic tissues. For example, wings mosaic for dronc⁻ clones retained extensive patches of persisting DsRed-labeled cells (Fig. 1 N). Here, cells and nuclei were readily detected 4 d after eclosion (Fig. 1 N), and even at 11 d post-eclosion, extensive evidence of cell debris was seen (not depicted). We found that wings mosaic for the H99 deletion gave identical results (Fig. 1 O). Likewise, adults mutated for dark exhibited persisting cells throughout the wing blade (Fig. 1 Q). Consistent with this, rare drice^Δ1 escapers also showed evidence of persisting cells after eclosion (Muro et al., 2006). These observations link failures in PCD to progressive melanized wing blemishes, raising the possibility that other apoptogenic mutants might also produce this phenotype.

Unlike previously described wing defects, which are congenital and evident at eclosion (Lawrence, 1992; Lindsley and Zimm, 1992), the age-dependent phenotype described in Fig. 1 (A–D) is characteristic of mutations in genes that function in canonical PCD pathways. Moreover, when the dosage of dronc was reduced by half in dark^CD4 adults (Chew et al., 2004) or if WT Dmp53 was removed from these same animals, melanized blemishes became far more severe. These genetic interactions are highly specific because wing defects were never observed in Dmp53⁻ homozygotes or in dronc⁵¹ heterozygotes (unpublished data). Numerous other mutants showed no such effects in combination with a dark hypomorph. We reasoned that, if genetically eliminated, additional regulators and effectors in PCD pathways should phenocopy wings mosaic for dark⁻ or dronc⁻ tissue. We screened a collection of preexisting transposon mutants to capture insertions that exhibit normal wings at eclosion but develop melanized blemishes with age. Our strategy exploits the FLP/FRT system together with wing-specific drivers to interrogate animals bearing wing genotypes mosaic for clones of P element–derived lethal mutations. Progeny with mosaic wings were examined for late-onset wing blemishes at 1, 7, and 14 d post-eclosion. We screened over 1,000 lethal insertions, representing 356 2nd chromosome mutations and 707 3rd chromosome mutations.

The majority of insertions (87%) produced no visible defects as wing mosaics (see online supplemental tables). 13% of insertions tested produced abnormalities, and these were scored for the phenotypic categories shown in Fig. S1. Congenital defects including notched, blistered, or wrinkled wings occurred alone or occasionally as compound phenotypes (Fig. S1, A–D). The candidate strains that developed wing blemishing were further subdivided based on phenotypic severity. Insertions in class A developed pronounced blemishes within a week of eclosion (Fig. S1 E), whereas those in class B developed relatively light-colored patches between 1 and 2 wk after eclosion (Fig. S1 F). Mutant lines exhibiting class A phenotypes were rare (∼2%). All members of this class lacked blemishes at eclosion and displayed progressive blemishing occasionally associated with fragile and sometimes broken wings (Fig. S1 E). A new allele of dark (l(2)SH0173) was recovered in this class (Table S2), providing reassuring validation for our screening strategy. Some members among these classes exhibited congenital notches or blisters, but congenital blemishes present at eclosion were not found.

We applied inverse PCR to map or confirm insertion sites of many class A and B strains (Table S2). In addition to dark^l(3)SH0173, we isolated several mutations associated with genes previously implicated in PCD (Table S1). For example, l(2)SH2275 contains an insertion 2 kb upstream of mir-14, a microRNA capable of modulating Rpr-induced cell death (Xu et al., 2003). Likewise, l(3)S048915 maps to the first intron of DIAP1 and may represent a hypermorphic allele at this locus. l(3)S055409 maps near misshapen (Su et al., 1998), a gene implicated in cell killing triggered by Rpr or Eiger, the fly counterpart of TNF (Igaki et al., 2002; Kuranaga et al., 2002). Several insertions map in or near transcriptional or translational regulators that might alter the expression of cell death genes. For example, grunge (l(3)S146907), an Atrophin-like protein (Erkner et al., 2002), functions as a transcriptional repressor (Zhang et al., 2002), while belle (l(3)S097074) belongs to the DEAD-box family of proteins (Johnstone et al., 2005) often implicated in translational regulation and RNA processing.

A portion of the class A and B hits were also directly examined for defective PCD by applying the vg:DsRed reporter in mosaic wings. Of the 29 strains tested, 14 showed obvious evidence for persisting cells in the wing epithelium (Table S2), which include dark^l(3)SH0173 (Fig. 1 Q).

Mutants identified above that exhibit both blemishing and persisting cells are likely candidates for PCD genes. One strain, l(3)S134313, produced severe late-onset blemishing (Fig. 3 D) and a persisting cell phenotype (Fig. 3 F). After mapping this insertion to the first intron of the HIPK, we produced null alleles at this locus by a customized deletion strategy illustrated in Fig. 3. Two FRT-containing P element insertions flanking the coding region of HIPK (Fig. 3 A) were used to generate a novel deletion depicted in Fig. 3 C (see Materials and methods). PCR verified recombination between P elements (Fig. 3 B), and 8 deletion strains were recovered. These validated alleles eliminate exons 4–12, removing over 92% of coding sequence in the predicted HIPK open reading frame. Deletions at the HIPK locus were uniformly lethal before the 3rd instar stage. However, zygotic HIPK is not essential to complete embryogenesis because ∼70% of HIPK homozygotes hatch to 1st instar larvae. HIPK^D1 was recombined on the FRT79 chromosome to generate adult wings mosaic for this allele, and like the original insertion, these animals also developed robust progressive blemishes (Fig. 3 E) and a persisting cell phenotype (Fig. 3 G). Both phenotypes were more severe than the original P insertion, suggesting that the l(3)S134313 allele is hypomorphic for HIPK. These findings link loss of HIPK function to our query phenotypes, establishing that the action of HIPK is essential for post-eclosion PCD in the wing epithelium.

Using general stains (acridine orange) or TUNEL methods, embryonic PCD was not overtly disturbed in HIPK mutants. To investigate the possibility of more subtle or specific phenotypes, we examined the nervous system using antibodies that label specific populations of neurons affected by the H99 deletion (White et al., 1994). Using α-Kruppel antibody (Kosman et al., 1998), we confirmed that stage 14–15 WT embryos contained 9–12 Kruppel-positive cells in the Bolwig's Organ (Fig. 4, A and C). However, a portion of animals lacking maternal HIPK contained as many as 15 cells per organ (Fig. 4, B and C) at a penetrance comparable to H99 animals, which are completely cell death defective (Fig. 4 C). We also examined neurons expressing dHb9, a homeodomain protein marking a subset of cells that persist in cell death–defective H99 embryos (Rogulja-Ortmann et al., 2007). In germline clones, distinct classes of dHb9 staining patterns emerged. A subset of animals exhibited extreme patterning defects. Other animals displayed a striking increase in dHb9-positive cell numbers (Fig. 4, E–G) when compared with WT embryos of the parental strain (Fig. 4 D). These data establish that HIPK fundamentally regulates cell numbers in the nervous system, and because the same subpopulation of cells are affected by the H99 mutation, they implicate HIPK as a more general regulator of PCD.

The pupal eye undergoes reorganization involving cell death of interommatidial cells after pupation (Wolff and Ready, 1991). To determine if HIPK regulates cell death in the retina, we generated whole eye clones and used the α-Dlg (discs large) antibody to outline cell borders in dissected pupal eyes after pupation. The WT pattern of interommatidial cells is represented in Fig. 4 H. In contrast, extra interommatidial cells were frequently retained in whole eye HIPK⁻ clones (Fig. 4 I). This phenotype is overtly similar to animals lacking the apical caspase Dronc (Xu et al., 2005) and consistent with an essential role in retinal PCD.

Elimination of the wing epithelium in newly eclosed adults is predictable, easily visualized, and experimentally tractable. The major histomorphologic events involve cell death, delamination, and clearance of corpses and cell remnants. Recent studies established that post-eclosion PCD is under hormonal control and involves the cAMP/PKA pathway (Kimura et al., 2004). While dying cells in the adult wing present apoptotic features (e.g., sensitivity to p35 and TUNEL positive), elimination of the epithelium is distinct from classical apoptosis in several important respects. First, unlike most in vivo models, overt engulfment of cell corpses does not occur at the site of death (Johnson and Milner, 1987; Ashkenas et al., 1996). Instead, dead or dying cells and their remnants are washed into the thoracic cavity via streaming of material along and through wing veins (Fig. 2, Videos 1–3; and Kimura et al., 2004). Second, extensive vacuolization is seen in ultrastructural analyses, which could indicate elevated autophagic activity (for review see Johnson and Milner, 1987; Kimura et al., 2004; and Fig. 1, I–K). Third, widespread and near synchronous death that occurs in this context defines an abrupt group behavior. The process affects dramatic change at the tissue level, causing wholesale loss of intervein cells and coordinated elimination of the entire layer of epithelium. Rather than die independently, these cells die communally, as if responding to coordinated signals propagated throughout the entire epithelium, perhaps involving intercellular gap junctions. This group behavior contrasts with canonical in vivo models where a single cell, surrounded by viable neighbors, sporadically initiates apoptosis.

One study proposed that an epithelial-to-mesenchymal transition (EMT) accounts for the removal of epithelial cells after eclosion (Kiger et al., 2007). Although our results do not exclude EMT associated changes in the newly eclosed wing epithelium, compelling lines of evidence, presented here and elsewhere, establish that post-eclosion loss of the wing epithelium occurs by PCD in situ—before cells are removed from the wing (Kimura et al., 2004). First, before elimination, wing epithelial cells label prominently with TUNEL. Second, every mutation in canonical PCD genes so far tested failed to effectively eliminate the wing epithelium (Fig. 1), and at least two of these were recovered in our screen. Third, elimination of the wing epithelium was reversed by induction of p35, a broad-spectrum caspase inhibitor (Kimura et al., 2004). Fourth, using time-lapse microscopy, we clearly detected condensing or pycnotic nuclei, followed by the rapid removal of all cell debris in time frames (minutes) not consistent with active migration. Instead, removal of cell remnants occurred by a passive streaming process, involving perhaps hydrostatic flow of the hemolymph.

Here, we sampled over one fifth of all lethal genes and nearly 10% of all genes in the fly genome for the progressive blemish phenotype, a reliable indicator of PCD failure in the wing epithelium. Nearly half of the mutants that produced melanized wing blemishing also displayed a cell death–defective phenotype when examined with the vg:DsRed reporter. The precise link between these defects is unclear, but a likely explanation suggests that as the surrounding cuticle fuses, persisting cells, now deprived for nutrients and oxygen, become necrotic and may initiate melanization. Mutants could arrest at upstream steps, involving the specification or execution of PCD, or they might affect proper clearance of cell corpses from the epithelium. We recovered new alleles of dark (l(2)SH0173) and a likely hypermorph of thread (l(3)S048915), which provides reassuring validation of this prediction.

By leveraging this distinct phenotype, we captured novel cell death genes, including the Drosophila orthologue of HIPK. Though first identified as an NK homeodomain binding partner (Kim et al., 1998), we found this gene to be an essential regulator of PCD and cell numbers in diverse tissue contexts. Of the four mammalian HIPK genes, HIPK2, the predicted orthologue of Drosophila HIPK, has been placed in the p53 stress-response apoptotic pathway (D'Orazi et al., 2002; Hofmann et al., 2002; Di Stefano et al., 2004, 2005), but whether the Drosophila counterpart similarly impacts this network is not yet known.

The l(3)Sxxxxxx(Bellotto et al., 2002) and l(2)SHxxxx(Oh et al., 2003) FRT stocks were obtained from Szeged Stock Center. vg-Gal4, UAS-FLP; FRT79, FRT82 and vg-Gal4, UAS-FLP, FRT40, FRT42 were provided by K. Basler (University of Zürich, Zürich, Switzerland). MS1096-Gal4, UAS-FLP flies are from J. Jiang (UT Southwestern Medical Center, Dallas, TX). FRT-Df(H99) stocks were provided by A. Gould (National Institute for Medical Research, London, UK) and J. O'Tousa (University of Notre Dame, Notre Dame, IN). To generate wing clones, 4 males of the genotype l(3)Sxxxxxx-FRT82/TM3(6) were crossed to 3 females of vg-Gal4, UAS-FLP; FRT79, FRT82. F1 flies were examined at eclosion and at 1 and 2 wk of age for appearance of “melanized blemishes” on the wing. l(3)Sxxxxxx-FRT80/TM3(6) and l(2)SHxxxx-FRT40/Cyo flies were similarly screened by crossing to vg-Gal4:UAS-FLP; FRT80 and vg-Gal4:UAS-FLP, FRT40, FRT42 flies, respectively. MS1096-Gal4:UAS-FLP; FRT42B was used for mutations on 2R. Adult wings were removed at different ages and fixed in paraformaldehyde for standard histology. Electron micrographs were generated using an electron microscope (TEMP2 1200 EX II; JEOL).

The FLP/FRT system was used to generate mutant wing clones, and persisting cells were visualized using DsRed. UAS-RedStinger/Cyo; dronc⁵¹-FRT79/TM6 was crossed to vg-Gal4:UAS-FLP; FRT79, FRT82. After eclosion, adults were aged from 1 to 14 d. Wings were removed, mounted on glass slides, and visualized using a fluorescent DLM (Axioplan; Carl Zeiss MicroImaging, Inc.) and a monochrome digital camera (Hamamatsu). Df(H99)-FRT80, dark⁸²-FRT42D, or HIPK^D1-FRT79 lines with or without UAS-RedStinger were also crossed to their respective FLP/FRT lines and imaged as stated above. Epithelial cell death was recorded in time-lapse experiments using the previous crosses to image w¹¹¹⁸; UAS-RedStinger/vg-Gal4, UAS-FLP; TM3(6)/FRT79 adults at 1–2 h after eclosion using a stereomicroscope (SteREO Discovery V.12; Carl Zeiss MicroImaging, Inc.) with Pentafluar S. Adults were glued on their dorsal surface to glass slides and imaged while alive. UAS-RedStinger lines were from Bloomington Stock Center.

P-element insertion sites were mapped by inverse PCR according to protocols from BDGP (http://www.fruitfly.org/about/methods/inverse.pcr.html). Genomic DNA of insertion lines containing the PlacW insertion element was extracted using Wizard Genomic DNA Extraction kit (Promega) and digested with HhaI, HpaII, and MboI restriction enzymes for 2.5 h at 37°C. Resulting digestions were diluted into 400 μl T4 DNA ligase (Roche) reactions and incubated overnight at 4°C. DNA was ethanol precipitated and used in Expand Long Template (Roche) PCR reactions with primers specific to the PlacW insertion. Unique PCR products were gel purified and sequenced, and insertion locations were confirmed using genomic PCR with primer sets specific to PlacW and surrounding genomic sequences.

Deletions were generated using the Exelixis collection of P elements as described previously (Parks et al., 2004; Thibault et al., 2004). To delete the HIPK locus, insertions f03158 and d10792 were placed in trans together with hs-FLP and heat-shocked to generate a FRT-mediated deletion. PCR primers directed to the remaining P elements and the surrounding genomic locus were used to identify deletion alleles (5′-TACTATTCCTTTCACTCGCACTTATTG-3′ and 5′-TAGATGAGGAAGTTCTGCGTGCAAGA-3′, 5′-CCTCGATATACAGACCGATAAAAC-3′ and 5′-CGACCTTCACCGACTGATCCTGGAT-3′). Two additional primer pairs, one producing a novel PCR product spanning the deleted HIPK locus (5′-GTGTCACTCGAAATTCGCCAGTGACT-3′ and 5′-GACGACTGACTCGGTAGCCTACTTCG-3′) while another specific to the deleted locus producing a negative result (5′-CGCTACTATCGTGCTCCCGAAATCAT-3′ and 5′-CGGATGCCTTGACATTGTTGCAGT-3′), were used to confirm deletions.

Germline clones were generated using the dominant female sterile technique described previously (Chou et al., 1993; Chou and Perrimon, 1996).

HIPK^D1-FRT79/TM6 was crossed to ey-FLP/Cyo; FRT79 GMR-Hid/TM6y+, and pupated animals were removed and aged for 48–55 h. After aging, pupal eyes were dissected and fixed in 4% formaldehyde in PBS. Aged matched siblings carrying TM6y+ were used as controls.

Immunohistochemistry was performed as described in Chew et al. (2004). Guinea pig α-Kruppel was used 1:600 (Kosman et al., 1998), rabbit α-dHb9 was used 1:500 (Broihier and Skeath, 2002), and α-Dlg was used 1:500 (Developmental Studies Hybridoma Bank) at 4°C overnight. Secondary antibodies used were labeled with Texas red or Fluorecein from Vector Laboratories (1:250) or Alexa 568 from Invitrogen (1:500). Genotyping was done using anti-GFP (1:1,000) from Invitrogen recognizing GFP-labeled balancers. Confocal z-series were taken using a confocal microscope (TCS SP5; Leica) and used for counting. Z-series were stacked for presented images.

Adult wings were dry mounted, and images were acquired using a microscope (Stemi V6; Carl Zeiss MicroImaging, Inc.) equipped with a 1.0× lens using a digital camera (Coolpix5000; Nikon) or a stereomicroscope (SteREO Discovery V.12; Carl Zeiss MicroImaging) with PentafluarS using 0.63× or 1.5× PlanApoS lenses and an MRm or MRc5 digital camera (Axiocam) and Axiovision Release 4.6 software. Additional fluorescent wing images were acquired with a microscope (Axioplan 2E; Carl Zeiss MicroImaging, Inc.) and a monochrome digital camera (Hamamatsu) using Plan Neofluar 10×/0.30, Plan Apochromat 20×/0.60, and Plan Neofluar 40×/0.75 objectives and OpenLab software. Confocal images of tissues stained with Fluorescein and Alexa 568 were mounted in Vectashield (Vector Laboratories), and images were acquired on a confocal microscope (TCS SP5; Leica) with Leica LAS AF software. The following lenses were used: HC PL APO 20×/0.70, HCX PL APO 40×/1.25-0.75 oil, and HCX PL APO 63×/1.40-0.60 oil objectives. All images were taken at room temperature and were processed in ImageJ or Photoshop 7.0. Occasionally, images were linearly rescaled to optimize brightness and contrast uniformly without altering, masking, or eliminating data.

Fig. S1 displays the wing phenotypes characterized in the mosaic screen. Videos 1–3 are time-lapse experiments showing PCD of wing epithelial tissue taken as described in Materials and methods. Table S1 displays a summary of the P insertion wing mosaic screen, Table S2 lists loci implicated in coordinated death in the wing epithelium, and Table S3 is a summary of all strains screened.

We thank Drs. Konrad Basler, Jin Jiang, Helmut Kramer, Joseph O'Tousa, Alex Gould, Szeged Stock Centre, and Bloomington Stock Center for fly strains; Jim Skeath, John Reinitz, and Developmental Studies Hybridoma Bank for antibodies; Taarini Vohra for technical assistance; Margaret Hickson for administrative excellence; and the Live Cell Imaging Facility at UT Southwestern for microscope assistance.

This work was supported by the National Institutes of Health (grant RO1GMO72124).