Main

Apoptosis-inducing factor (AIF) was originally characterized as a phylogenetically conserved cell death mediator, a mitochondrial protein that translocates to the nucleus following an apoptotic stimulus and triggers chromatin condensation and DNA degradation.1 In the nucleus, AIF binds to DNA and recruits an endonuclease complex, which includes cyclophilin A,2 to trigger large-scale (10 kb) fragmentation of DNA. More recently, AIF has been implicated as a cell survival factor required for the maintenance of mitochondrial respiratory complex function and/or for scavenging reactive oxygen species. In particular, the Harlequin (Hq) mouse strain, which carries a hypomorphic mutation in AIF, develops late-onset neuronal and retinal degeneration accompanied by increased neuronal expression of oxidative stress markers.3 Other groups have found that inactivation of AIF by gene targeting in mice or RNA interference in cell lines results in decreased protein expression and function of respiratory complex I, and leads to impaired cell growth and survival.4, 5, 6, 7, 8 Thus, AIF has multiple functions, both in cell death execution and in mitochondrial energy homeostasis.

The notion that AIF is an important cell death effector is supported by several loss-of-function studies. In Hq mice, glutamate-induced excitotoxic cell death of hippocampal neurons is attenuated compared to wild-type controls.9 Hq mice also display smaller infarct volumes after cerebral hypoxia–ischemia,10, 11 and Hq neuronal cortical cells exhibit partial resistance to cell death in response to serum deprivation and PARP-1 signaling.3, 12 In Caenorhabditis elegans, knockdown of the AIF ortholog WAH-1 results in a reduction in developmental cell death;13 in Saccharomyces cerevisiae, AIF null mutants exhibit decreased death in response to oxidative stress and chronological aging.14 These findings support the idea of a phylogenetic conservation of AIF-mediated cell death. Importantly, AIF has been shown to exert its cytotoxic effects in a caspase-independent manner.15, 16

There is an ongoing debate as to the contribution of mitochondria to the cell death program of different species. In mammalian cells, the ‘point of no return’ of the intrinsic cell death pathway occurs when the outer mitochondrial membrane permeabilizes and several apoptogenic proteins, including cytochrome c (Cyt c), Smac/DIABLO, AIF and endonuclease G, are released from the mitochondrial intermembrane space.17 In Drosophila melanogaster, several reports have found no indication that Cyt c is released from mitochondria during apoptosis18, 19, 20, 21 and, until recently, there was little evidence for mitochondrial involvement in the fly cell death program. Two recent reports, however, showed that mitochondrial remodeling is an important process in the cell death program in Drosophila, and that Cyt c translocates to the cytosol in Reaper- and Hid-induced, but not ultraviolet (UV)-induced, cell death.22, 23 Moreover, Cyt c is required for caspase activation associated with spermatid differentiation.24 Thus, there is emerging evidence of a critical involvement of mitochondria in apoptosis in Drosophila, implying that cell death mechanisms between flies and mammals are more broadly similar than previously believed.

Genetic studies in flies have shown that most cell death that occurs during development and in response to chemotherapeutic agents can be blocked by genetic inhibition of caspases or activators of caspases.25, 26, 27 One interesting exception is the death of nurse cells during late oogenesis, in which caspase activity is not detectable and transgenic expression of caspase inhibitors p35 and Drosophila inhibitor of apoptosis (Diap)1 has no effect on egg chamber development.28 Nevertheless, in light of the physiologically relevant cytotoxic functions of AIF orthologs in C. elegans and yeast, we asked whether an AIF ortholog in D. melanogaster similarly functions as a cell death effector.

Results and Discussion

Identification of the Drosophila AIF ortholog

The fly genome contains a single open reading frame (SD03428, CG7263) with >65% similarity to mammalian AIF. The predicted D. melanogaster AIF (DmAIF) protein (674 amino acids, aa) exhibits 50% sequence identity and 68% similarity to mouse AIF (mAIF), and 33% identity and 50% similarity to C. elegans AIF (WAH-1) (Figure 1A). In silico analysis revealed a mitochondrial localization sequence (MLS, Supplementary Figure 1a). Expression of cDNA constructs corresponding to C-terminally tagged DmAIF, either full length or lacking the first 176 aa including the MLS (ΔN-DmAIF), yielded proteins of the expected size (Supplementary Figure 1b and c). Subcellular localization of these proteins revealed a mitochondrial distribution for full-length DmAIF and a non-mitochondrial distribution for ΔN-DmAIF (Figure 4b and Supplementary Figure 2a), confirming that the N terminus contains an MLS. Regions of mAIF critical for its oxidoreductase activity are highly conserved in DmAIF, particularly a putative FAD-binding domain (aa 183–323 and 462–539), an NADH-binding domain (aa 324–461), as well as the core consensus of the classical Rossman fold involved in direct binding of NADH and FAD (aa 196–201 and 367–372) (Figure 1A).30, 31 This may suggest that, like mammalian AIF,32 DmAIF is redox active.

Figure 1
figure 1

Cloning and analysis of expression of DmAIF. (A) Sequence alignment of DmAIF, mAIF and WAH-1. Black shading and gray shading indicate identical and similar amino-acid residues, respectively. Sequence alignment was performed using ClustalW and Boxshade programs using translations of SD03428 (DmAIF), AF100927 (mAIF) and AY147006 (WAH-1). Red asterisks represent the NLS identified in mAIF,1 and red circles indicate residues in mAIF required for binding DNA;29 both regions are poorly conserved in DmAIF. (B) Northern blot analysis of DmAIF expression during early (0–13 h) and late (13–22 h) embryogenesis, third-instar larvae (L3), pupae (P), and male (M) and female (F) adults. Expression was detected by northern blotting using a full-length DmAIF cDNA probe. (C) Whole-mount in situ analysis of DmAIF expression during embryogenesis. DmAIF antisense (a–d) and sense (e–h) cDNA probes labeled with DIG were used to detect DmAIF expression in wild-type embryos at various stages of development

Two other salient features of mAIF, a putative nuclear localization sequence (NLS) at aa 377–387 and 445–450 (red asterisks, see legend to Figure 1A) and residues required for binding to DNA (red circles), are absent in DmAIF (Figure 1A),1, 29 suggesting that DmAIF might not translocate to the nucleus. A Myc-tagged full-length DmAIF transgene expressed under the control of the eyeless promoter yielded a cytoplasmic punctate staining that colocalized with mitochondrion-targeted GFP (Supplementary Figure 2a). However, exposure of these flies to γ-irradiation to trigger apoptosis did not result in mitochondrio-nuclear translocation of myc-tagged DmAIF (Supplementary Figure 2b). Similar results were obtained in the SL2 Drosophila cell line: overexpression of the lethal protein GRIM (Supplementary Figure 3a) or UVC irradiation (Supplementary Figure 3b) failed to trigger DmAIF redistribution from mitochondria. Together, these findings suggest that DmAIF remains sequestered in mitochondria, at least under the apoptotic conditions tested here and in contrast to mammalian AIF.2, 13, 16

In mammalian tissues, AIF is ubiquitously expressed.30 Similarly, DmAIF was found expressed throughout all stages of development but appeared downregulated at the pupal stage, as revealed by northern blotting and in situ hybridization of embryos (Figure 1B and C). In summary, we identified the Drosophila ortholog of AIF, which shares high sequence similarity, mitochondrial localization and an expression pattern similar to mammalian AIF.

Developmental and bioenergetic defects due to inactivation of the DmAIF gene

We generated DmAIF-mutant flies by gene targeting, using ‘ends-in’ homologous recombination.33 The targeting construct consisted of the entire DmAIF genomic locus (Figure 2a); base pair insertions designed to disrupt the open reading frame of DmAIF were introduced into exons 3 and 6. One gene-targeted line, designated DmAIFT52, was validated by genomic Southern blotting using DNA probes flanking the 5′ and 3′ end of the DmAIF locus (Figure 2b). Sequencing of a PCR-cloned region of the targeted locus confirmed that the introduced point mutations were correctly targeted (data not shown). Western blotting of lysates from wild-type and mutant larvae showed absence of DmAIF protein in gene-targeted animals (Figure 2c); we thus designate this mutant line DmAIFKO.

Figure 2
figure 2

DmAIF gene-targeted flies exhibit growth arrest at early larval stage and defects in mitochondrial respiration. (a) Targeting scheme for the DmAIF locus. Illustrated are the donor construct, the native DmAIF locus and the mutated locus after gene targeting. Boxes indicate exons. The donor construct consisted of a 5-kb fragment encompassing the entire DmAIF gene (in red) and carrying frameshifting base pair insertions (asterisks) within exons 3 and 6. A germline transformant strain bearing this construct on the X chromosome was used for targeted mutagenesis. The resulting targeted lesion is a tandem duplication of two variants of the DmAIF gene, each bearing the original introduced mutations as confirmed by sequencing of the endogenous locus. The predicted genomic fragment sizes after digestion with XbaI for the native locus and mutated locus are illustrated. HR, homologous recombination; FRT, FLP recombination target; w-hs, whs expression cassette; N, NotI; S, SalI; X, XbaI. (b) Southern blot of genomic DNA from wild-type (WT) and DmAIFKO/+ heterozygous (Het) flies. Genomic DNA was digested with NotI and SalI (left panel) or XbaI only (right panel) and analyzed by Southern blotting using 3′ and 5′ flanking probes, respectively, as indicated in panel a. (c) Lysates from wild-type (WT) and DmAIFKO homozygous mutant (KO) first-instar larvae were immunoblotted with anti-DmAIF and anti-actin antibodies. (d) Gross morphology of wild-type (WT) and DmAIFKO homozygous mutant (KO) larvae on days 4 and 6, respectively, after egg laying. Bar represents 300 μm. (e) Survival of heterozygous (control, squares) and DmAIFKO homozygous mutant (KO, circles) flies. AEL, after egg laying. (f) DmAIFKO mutants are arrested at the first- (L1) and second-instar (L2) larval stages. Four days after egg laying, larvae were staged as L1–3, based on mouth hook structure and size. n>100. In another experiment, larvae were grown at high density to mimic nutrient limitation, and staged on day 4. (g) Selective defects in respiratory complex activities in DmAIFKO mutants. Mitochondria were isolated from first- and second-instar wild-type and DmAIFKO mutants and assessed for respiratory complex activities. Values represent activities of each complex (I, II+III, G3PDH+III, III, IV) from mutant mitochondria as a percentage of control (100%). Values are mean±S.E.M. for two independent experiments. Similar results were obtained in another experiment not assessing complex II activity using only early first-instar larvae. (h) ATP levels from second-instar (day 3 AEL) heterozygous and homozygous DmAIFKO mutants, determined by ATP-dependent luciferase activity normalized to protein content. Means±S.E.M. are shown from two experiments. *P<0.05

Strikingly, DmAIFKO larvae were markedly reduced in size, although their general anatomy (segmental pattern, development of the dorsal vessel, tracheal system and fat body) appeared normal, suggesting normal organogenesis coupled to deficient growth (Figure 2d). Zygotic homozygous DmAIFKO mutant embryos hatch, but larvae die within several days after egg laying (AEL) (Figure 2e). A time-course analysis of lethality showed that by day 6 AEL approximately half of DmAIFKO mutant larvae were dead; by day 8 no viable DmAIFKO mutants were recovered (Figure 2e). Mutant larvae were noticeably smaller in size within hours after hatching (data not shown). By day 4, when most wild-type flies had reached the third-instar larval (L3) stage, the DmAIFKO mutant flies were at the first- and second-instar larval (L1 and L2) stages (Figure 2f). No third-instar larvae (L3) were recovered even in the few surviving mutants at days 7 and 8. Growth arrest at L1 of DmAIFKO larvae was exacerbated in crowded culture conditions (Figure 2f), suggesting an underlying bioenergetic defect. Accordingly, we observed that the enzymatic activity of respiratory chain complexes I and IV from DmAIFKO larvae was strongly reduced as compared to heterozygous controls (Figure 2g). Moreover, the level of ATP was severely reduced in DmAIFKO larvae compared to controls (Figure 2h). In light of previous findings demonstrating that AIF is required for normal oxidative phosphorylation in human cell lines, mice and yeast,5, 6, 7 these data point to an evolutionarily conserved bioenergetic function for AIF.

Selective defects in developmentally regulated cell death in DmAIFKO embryos

We next analyzed developmental programmed cell death (PCD) in mid-stage embryos by acridine orange (AO) staining. DmAIFKO embryos had less incidence of developmental PCD (Figure 3b and d) compared to wild-type (Figure 3a and c) embryos (stage 12). As a consequence of the reduction of PCD, we investigated the pattern and persistence of extra cells in the central nervous system by different neuronal markers. Antibody staining against kruppel and dHb9 was performed. The pattern of neuronal cells appeared unperturbed; however, DmAIFKO embryos had extra cells in both the ventral nerve cord (VNC) and Bolwig's organ (Figure 3f and i) compared to wild-type embryos (Figure 3e and h). The number of VNC cells was comparable in abdominal region A1, but the presence of extra cells increased in a posterior manner in DmAIFKO embryos from abdominal segments A2–A6 compared to wild-type embryos (Figure 3g). DmAIFKO embryos had on average more than three extra cells in the Bolwig's organ (Figure 3j). Since kruppel is a pan-neuronal CNS marker, we looked at a specific subset of neurons in the VNC. As anticipated, antibody staining against dHb9 revealed the presence of extra cells along the lateral edges of the VNC in DmAIFKO embryos (Figure 3l and m) compared to wild-type embryos (Figure 3k).

Figure 3
figure 3

DmAIFKO embryos exhibit extra cells in the central nervous system. (ad) AO staining on wild-type (a, c) and DmAIFKO embryos (b, d). (eg) DmAIFKO embryos (f) show extra cells by α-dHb9 antibody staining compared to wild-type embryos (e). (g) Histogram of extra cells in the VNC by α-Kr antibody staining (n=8). (hj) DmAIFKO embryos (i) have extra cells in the Bolwig's organ compared to wild-type embryos (h) (α-Kr). (j) Histogram of cell counts in the Bolwig's organ (n=6). (km) DmAIFKO embryos (l) have extra cells in the VNC by α-dHb9 antibody staining compared to wild-type embryos (k) (yellow brackets). (m) Histogram of extra VNC cells by α-dHb9 antibody staining (n=6). Error bars represent ±S.D.; *P<0.05; **P<0.005, Student's t-test

Cell death-associated organ disruption mediated by extramitochondrial DmAIF

We next generated transgenic flies carrying epitope-tagged UAS-DmAIF and UAS-ΔN-DmAIF constructs (Figure 4a). Control experiments, in which these constructs were transfected into SL2 cells, revealed that full-length DmAIF-myc co-stained with the mitochondrion-selective cationic dye Mitotracker Red CMXRos, while ΔN-DmAIF-myc (encoding aa 177–674) no longer localized to mitochondria (Figure 4b). We next tested the function of DmAIF in vivo by generating transgenic flies expressing these constructs. While flies expressing full-length DmAIF in the eye appeared normal (Figure 4c), flies expressing ΔN-DmAIF exhibited a marked reduction in eye size (Figure 4c, right panels) and disorganized ommatidial architecture (Figure 4d). The eye phenotype exhibited 100% penetrance but variable expressivity, ranging from a roughened appearance and normal size to a severe (>80%) reduction in size. Eye discs from ey-GAL4;UAS-ΔN-DmAIF third-instar larvae stained positive for activated caspases (Figure 4e) and the vital dye AO (Supplementary Figure 1d), indicating that the eye phenotype was caused by ectopic cell death.

Figure 4
figure 4

Misexpression of DmAIF triggers ectopic cell death. (a) Schematic diagram of Myc-tagged full-length DmAIF (DmAIF-myc) and an amino-terminal deletion mutant lacking amino acids 1–176, including the presumptive mitochondrial import sequence (ΔN-DmAIF-myc). (b) Localization of DmAIF-myc and ΔN-DmAIF-myc in SL2 cells. Expression of constructs was detected using anti-c-Myc antibody (green). Cells were counterstained with Mitotracker Red CMXRos (red) and Hoechst 33342 (blue) to visualize mitochondria and nuclei, respectively. (c) Stereomicroscopic photographs of eyes from control and transgenic flies expressing full-length (DmAIF-myc) and NH2-terminally truncated (ΔN-DmAIF) forms of DmAIF under the control of the ey promoter. Note that the left panel of ΔN-DmAIF is representative of the majority of flies expressing ΔN-DmAIF; the right panel of ΔN-DmAIF is typical of 5% of flies. Genotypes are as follows: ey-GAL4/CyO (control); ey-GAL4,UAS-DmAIF-myc/CyO (DmAIF); ey-GAL4,UAS-ΔN-DmAIF/CyO (ΔN-DmAIF). (d) Scanning electron micrographs of eyes from (left) control ey-GAL4/CyO and (right) ey-GAL4,UAS-ΔN-DmAIF/CyO flies. (e) Immunostaining of eye discs from (left) control ey-GAL4/CyO and (right) ey-GAL4,UAS-ΔN-DmAIF/CyO third-instar larvae with an antibody recognizing cleaved caspase 3 (Asp175) and visualized using Alexa 488-conjugated secondary antibodies

As AIF has been shown to trigger PCD in the absence of caspase function,16, 34 we determined whether DmAIF is a caspase-independent cell death promoter by crossing ey-GAL4;UAS-ΔN-DmAIF to mutant lines genetically deficient in caspases or activators of caspases. Inactivation of the Drosophila Apaf-1 (apoptotic protease-activating factor-1)/CED4 ortholog Dark, the fly apical caspase DREDD, Dmp53, or loss of one copy of the H99 locus (which spans the essential regulators grim, reaper and hid) failed to modulate the ΔN-DmAIF phenotype (Figure 5a, small panels).26, 27 However, expression of the caspase inhibitors Diap1 or Diap2 did significantly, albeit incompletely, ameliorate the reduced eye size (Figure 5a). Interestingly, coexpression of the baculoviral caspase inhibitor p35 and ΔN-DmAIF in the eye failed to abolish the DmAIF-mediated disruption of normal eye architecture, and paradoxically, dramatically enhanced the phenotypic changes promoted by ΔN-DmAIF (Figure 5a). Similar results were obtained when the ΔN-DmAIF transgene was targeted to dorsal and thoracic bristles using pnr-GAL4. In this case, expression of ΔN-DmAIF resulted in a reduction in bristle number and size, which could not be rescued by p35 coexpression; rather, p35 enhanced the phenotype (Figure 5b).

Figure 5
figure 5

Epistasis analysis of DmAIF and several cell death regulators. (a) Large panels show flies expressing ΔN-DmAIF, p35, or both transgenes under control of ey-GAL4, and flies coexpressing ΔN-DmAIF and Diap1, or ΔN-DmAIF and Diap2. The small panels show flies expressing ΔN-DmAIF in genetic backgrounds homozygous for Dark or Dmp53, hemizygous for Dredd (Dredd/Y) or heterozygous for Df(3L)H99 (H99/TM3). Note that female ey-GAL4,UAS-ΔN-DmAIF/+ flies, which display a less severe eye phenotype than male flies (shown in Figure 4c), were used in these epistasis experiments. (b) Flies carrying the pnr-GAL4 transgene were crossed to UAS-ΔN-DmAIF, UAS-p35 or UAS-ΔN-DmAIF;UAS-p35 flies. Note the exacerbated reduction in eye size (a) and loss of thoracic bristles (b) in flies coexpressing ΔN-DmAIF and p35. Genotypes in large panels in (a) are ey-GAL4,UAS-ΔN-DmAIF/+ (ΔN-DmAIF); ey-GAL4,UAS-p35 (p35); ey-GAL4,UAS-ΔN-DmAIF/UAS-p35 (ΔN-DmAIF;p35); ey-GAL4,UAS-ΔN-DmAIF;UAS-Diap1 (ΔN-DmAIF;Diap1); and ey-GAL4,UAS-ΔN-DmAIF;UAS-Diap2 (ΔN-DmAIF;Diap2). Genotypes in panel b are pnr-GAL4,UAS-ΔN-DmAIF/+ (ΔN-DmAIF); UAS-p35/+;pnr-GAL4/+ (p35); and UAS-p35/+;pnr-GAL4,UAS-ΔN-DmAIF/+ (ΔN-DmAIF;p35)

Altogether, these findings suggest that transgene-enforced expression of extramitochondrial DmAIF can kill cells and perturb tissue organization in a manner that may depend at least partially on caspase activity.

Thioredoxin-2 as an obligate mediator of extramitochondrial DmAIF cytotoxicity

To determine whether the cytotoxic effects of ΔN-DmAIF represent a specific, genetically modifiable cell death pathway, we set out to identify binding partners of DmAIF. A yeast two-hybrid screen of a mouse brain cDNA library was performed using mAIF as bait. One major hit (2 among 45 clones) was thioredoxin (txn1). The genome of D. melanogaster encodes several members of a thioredoxin family, including deadhead (dhd) and thioredoxin-2 (DmTrx-2).35 Since DmTrx-2-mutant flies were available in public stock centers, we chose to investigate in detail a potential interaction between DmTrx-2 and DmAIF. We transfected Flag-tagged DmTrx-2 into wild-type and AIF-deficient mouse embryonic stem cells,36 and confirmed that DmTrx-2 immunoprecipitates with endogenous AIF (Figure 6a). To test whether this physical interaction is important for the proapoptotic action of DmAIF in vivo, we expressed an inverted repeat of the DmTrx-2 gene (UAS-DmTrx-2-IR) to silence DmTrx-2 mRNA expression in flies expressing ΔN-DmAIF. Importantly, downmodulation of DmTrx-2 abolished the eye-disruptive effects of ΔN-DmAIF overexpression, and re-established normal eye morphology (Figure 6b and c).

Figure 6
figure 6

Drosophila thioredoxin-2 (DmTrx-2) modulates DmAIF-induced cell death. (a) Flag-tagged DmTrx-2 (Flag-DmTrx-2) was expressed in Aif+/Y or Aif/Y mouse embryonic stem cells.36 Flag-DmTrx-2 was immunoprecipitated from lysates with anti-Flag antibodies and immunoblotted with anti-mAIF or anti-Flag antibodies. Lysates are shown as controls. *IgG light chain. (b) Two independent lines carrying the UAS-DmTrx-2-IR transgene were crossed to da-GAL4. Total RNA from third-instar larvae carrying both transgenes was reverse transcribed, and PCR was performed on cDNA using primers specific for DmTrx-2 and G3PDH. da-GAL4 larvae served as control. Control-RT, non-reverse transcribed da-GAL4 total RNA. (c) ey-GAL4,UAS-ΔN-DmAIF/CyO flies were crossed to w1118 (control, left panel) or to UAS-DmTrx-2-IR (right panel), and the effect of coexpression of ΔN-DmAIF and DmTrx-2 was evaluated. (d) ey-GAL4,UAS-ΔN-DmAIF/CyO male flies were crossed to w1118 or to two independent lines of UAS-DmTrx-2-IR females. Eclosing progenies were tabulated according to whether they carried UAS-ΔN-DmAIF, UAS-DmTrx-2-IR or both transgenes. Note that downregulation of Trx-2 suppresses pupal lethality caused by expression of ΔN-DmAIF, thereby increasing the frequency of eclosing flies. Numbers in brackets refer to number of flies counted. For example, in cross 1, 12% of eclosing progeny received ey-GAL4,UAS-ΔN-DmAIF, indicating that expression of ΔN-DmAIF is semilethal (Mendelian expected frequency is 50%). In contrast, in cross 2, 45% of eclosing progeny received both ey-GAL4,UAS-ΔN-DmAIF and UAS-DmTrx-2-IR (expected frequency was also 50%), suggesting that downmodulation of DmTrx-2 can significantly suppress ΔN-DmAIF-induced pupal lethality

Among newly eclosed F1 generation ey-GAL4 flies, we noticed a relative increase in numbers of those that carry both the UAS-ΔN-DmAIF and UAS-DmTrx-2-IR transgenes compared to UAS-ΔN-DmAIF alone. We had earlier observed that expression of ΔN-DmAIF in the eye results in lethality at the pre-eclosion stage (Figure 6d), likely due to the leaky tissue expression of ΔN-DmAIF under ey-GAL4 control. In contrast, in two independent lines of UAS-DmTrx-2-IR, downregulation of DmTrx-2 ameliorated the relative numbers of eclosing flies expressing ΔN-DmAIF, and almost restored expected Mendelian frequencies (Figure 6d). Thus, these data suggest that the redox protein DmTrx-2 cooperates with DmAIF to facilitate cell killing, and that DmAIF-induced cytotoxicity represents a specific and genetically modifiable cell death pathway.

Conclusion

There has been no previous report of the implications of AIF in Drosophila. In this paper, we demonstrate an evolutionarily conserved function of AIF in organismal growth and mitochondrial respiration. Loss of zygotic expression of DmAIF results in arrested larval growth, loss of viability and defective functioning of the respiratory complexes, a phenotype analogous to mice carrying a null allele of Aif, which results in embryonic growth retardation and lethality by day 11.5 of gestation.5 Similarly, knockdown of the C. elegans AIF ortholog WAH-1 leads to a slower growth rate and smaller brood size, although the authors of the report did not investigate a potential mitochondrial bioenergetic defect.13 Our results, thus, confirm a role for AIF in maintaining energy homeostasis in the cell. However, the precise role of AIF in supporting normal respiratory complex function remains unknown.

Our findings also shed light on a proapoptotic function of AIF in Drosophila. Flies lacking zygotic DmAIF expression exhibit fewer dying cells during embryogenesis relative to wild-type flies and a modest increase in neuronal cells along the VNC. The persistence of extra cells occurred predominantly in the posterior abdominal segments of mutant embryos. The basis for this regional specificity is unclear but may involve redundant expression of AIF homologous genes that are known to exist in mammals yet have not been characterized in Drosophila. In any event, it will be interesting to investigate germline DmAIF mutant clones, as maternal expression of DmAIF protein is substantial (Figure 1C).

Our finding of a proapoptotic function of Drosophila AIF in developmentally regulated cell death raises several interesting issues. Presumably, Drosophila AIF, like its counterparts in C. elegans and mammals, requires translocation to the nucleus to mediate cell killing, where its putative endonuclease and chromatin-condensing activities are activated.2, 13 However, in two different systems in which apoptosis was induced (UV-irradiated SL2 cells and γ-irradiated eye disc cells), we did not observe release of DmAIF from mitochondria, suggesting that stress-induced apoptosis does not involve DmAIF. A number of studies have reported that Drosophila Cyt c also remains localized within mitochondria in apoptotic cells, in apparent evolutionary contrast to the situation in higher organisms.18, 20, 21 On the other hand, it was recently reported that Cyt c translocates to the cytosol in Reaper- and Hid-induced, but not UV-induced, cell death.22, 23 Thus, it remains possible that, like Cyt c, DmAIF translocation from mitochondria depends on the apoptotic stimulus and cellular context. Furthermore, our results do not exclude a role for non-nuclear proapoptotic effects of DmAIF. Indeed, mammalian AIF can exert part of its biochemical effects by arresting protein translation, through a direct interaction with eIF3g,37 and the C. elegans AIF ortholog WAH-1 can enhance phosphatidylserine exposure of dying cells through a direct interaction with a phospholipids scramblase.38 This means that, in different species, AIF does mediate major non-nuclear alterations in cellular physiology (protein synthesis arrest and loss of plasma membrane asymmetry) that are considered to be hallmarks of the apoptotic process, and it remains to be seen whether these changes are also influenced by DmAIF.

Our data provide some clues into the mechanisms through which AIF exerts its cytocidal effect. Overexpression of a truncated form of DmAIF lacking its N-terminal mitochondrial-targeting signal (but not flies expressing full-length DmAIF) triggered ectopic caspase activation and cell death. It is noteworthy that another GAL4 driver tested in combination with UAS-ΔN-DmAIF, namely gmr-GAL4, did not elicit obvious phenotypic changes (N Joza and JM Penninger, unpublished data). This may reflect an intrinsic resistance of certain cell types to DmAIF-mediated death or insufficient expression of the transgene (i.e., DmAIF may require relatively high levels of expression to trigger cell death). In addition, we obtained somewhat conflicting results with regard to whether DmAIF requires caspases to mediate cell killing. While expression of p35 failed to suppress (and paradoxically, enhanced) DmAIF-mediated cell death, a significant inhibition was observed with expression of Diap1 or Diap2. Since p35 and Diap1 are reported to inhibit distinct caspases (in particular, p35 fails to inhibit DRONC39), it seems that a caspase may contribute to the cell death triggered by Drosophila AIF. Indeed, overexpression of the C. elegans AIF ortholog WAH-1 lacking a mitochondrial-targeting signal induces cell death that is partially blocked in a ced-3 (caspase)-mutant background.13 This contrasts with the caspase-independent cell death reported for AIF orthologs in S. cerevisiae and mammals.14, 16 Thus, there may be a species-specific requirement for caspases in AIF-mediated cell killing.

Finally, we have identified a novel interaction between Drosophila AIF and the redox protein thioredoxin-2 (DmTrx-2). Knockdown of DmTrx-2 suppressed DmAIF-induced cell death, suggesting that DmAIF cooperates with DmTrx-2 in execution of cell death. Future work will clarify whether this interaction is relevant to DmAIF-dependent developmental cell death in the fly embryo.

Materials and Methods

Cloning and characterization of DmAIF

A search of the BDGP EST database identified a sequence (SD03428) with significant homology to mAIF and representing the full open reading frame. To construct ΔN-DmAIF (encoding aa 178–674), the primers 5′-TTGCGGCCGCTCCACCATGGCCACGAGTCCGCCCAGTTCTGA-3′ (containing an upstream NotI site, Kozak sequence and in-frame ATG, all underlined) and 5′-GCTTCTCTCGGATCCCGTCCATTG-3′ were used to amplify the 5′ region of DmAIF by PCR. The PCR product was cloned into the NotI and ClaI sites of pBluescript KS II (pBS). Next, LD41427 was digested with ClaI and SalI and cloned into pBS to generate the complete ΔN-DmAIF. ΔN-DmAIF was excised using NotI and XhoI and cloned into the pUAST vector. For construction of DmAIF expression constructs for cell transfections, the Myc epitope was tagged to the N terminus of DmAIF (encoding aa 178–674) or tagged to the C terminus of full-length DmAIF using the full-length DmAIF cDNA clone SD03428 as template. Constructs were cloned into pMT (for the full-length DmAIF) or pPAC (for the N-terminal deletion mutant).

Fly crosses

Flies were maintained on standard medium at 25°C in a temperature-controlled incubator. The full-length and ΔN-DmAIF constructs were cloned in the pUAST vector and injected into w1118 embryos by standard methods to generate P[UAS-DmAIF-myc] and P[UAS-ΔN-DmAIF] transformants. At least six independent transgenic lines of P[UAS-ΔN-DmAIF] were generated, and all showed similar phenotypes when crossed with ey-GAL4 flies. Experiments were carried out with the P[UAS-ΔN-DmAIF2A] line, in which the transgene is inserted on chromosome 2. A stable line was generated by recombining the ey-GAL4 and P[UAS-ΔN-DmAIF] transgenes and balanced over CyO. This line, denoted ey-GAL4,UAS-ΔN-DmAIF, was used for all experiments, except in Figure 5b where a P[UAS-ΔN-DmAIF] on the third chromosome was used. UAS-DmTrx-2-IR flies (line nos. 36297 and 36298, corresponding to gene CG31884) were obtained from the Vienna Drosophila RNAi Center (Vienna, Austria). UAS-Diap1 and UAS-Diap2 flies were kindly provided by Pascal Meier (Institute of Cancer Research, UK). darkCD4 represents a strongly hypomorphic mutation in the dark locus;27 dreddB118 and p53ns were obtained from JM Abrams (University of Texas SWMC, Dallas, TX, USA) and are likely null mutants of their respective genes. Two P[UAS-p35] lines (BH1 and BH2) were used and produced similar results. H99/TM6B (no. 1576), P[UAS-p35] (nos. 5072 and 5073), ey-GAL4/CyO (no. 5535), pnr-GAL4/TM3,Ser (no. 3039) and UAS-mitoGFP/CyO lines were obtained from the Bloomington Stock Center. Staging of larvae was based on mouth hook structure and size.

Gene targeting of Drosophila AIF

The donor construct consisted of an 5-kb region spanning the entire DmAIF gene locus (red lines in Figure 2a), the last 200 amino acids of the neighboring gene BcDNA:LD22320 and the first 21 amino acids of the gene CG15382 (indicated in black in Figure 2a). The region corresponding to nucleotides 257271 and 262278 (AE003584) was PCR amplified from wild-type w1118 genomic DNA using primers 5′-CTTCCTGCAGATCAAGTATGCCTC-3′ and 5′-GTCGGAATCTCTTACTGGCGAGG-3′. A 70I-SceI site was cloned into exon 6 (nucleotide 800 of coding region of AY052083). The construct was digested with SpeI (nucleotide 253), blunted and recircularized, thereby causing a frameshift in exon 3. Similarly, the construct was digested with BclI (nucleotide 1548), blunted and recircularized, causing a frameshift in exon 6. The construct was cloned into the NotI site of targeting vector pTV2.33

Germline transformants bearing the donor construct were generated. Flies carrying the DmAIF donor element on the X chromosome were crossed to flies carrying both 70FLP and 70I-SceI transgenes on the second chromosome. Progenies from this cross were heat shocked during 0–3 days of development to induce recombinants in the germ line. Eye-color mosaic females of the genotype DmAIF donor; +/70FLP, 70I-SceI were crossed to ey-FLP males, and progenies were screened for loss of white+ mosaicism as described.33 In the DmAIF mutant characterized in this paper, DmAIFT52, the 5′ and 3′ variant copies each carry the expected introduced mutations in exons 6 and 3, respectively. In addition, the 5′ copy carries the exon 3 mutation, which is probably due to a gene conversion event. Such ‘extra’ mutations frequently occur in gene targeting in flies.33 Confirmed DmAIFT52 heterozygous females were crossed to w1118 males for >6 generations to eliminate potential second-site mutations.

Correct targeting of the DmAIF locus was confirmed by Southern blot of genomic DNA. cDNA probes flanking the targeted DmAIF locus were generated by PCR using primers 5′-GCCTGTTAAGAGCTCTTCAAGAG-3′ and 5′-ACAAGAGATTC AGTCAGTTGACC-3′ (5′ flanking probe); and 5′-ACCACAGTGCTCTGGAACTGGA-3′ and 5′-GAATTGGTTTCAGGCGGTTATCC-3′ (3′ flanking probe). Fly genomic DNA was digested with the indicated restriction enzymes and Southern blotted.

Northern and western blotting, reverse transcription-PCR and in situ hybridizations

Northern blotting of total RNA from flies at various developmental stages was performed using full-length DmAIF as a probe, essentially as described.27 Whole-mount embryo in situ hybridization assays were conducted with the use of digoxigenin (DIG)-labeled single-stranded antisense and sense DNA probes, as previously described.27 Probes were generated by PCR using the full-length DmAIF cDNA clone SD03428 as template and the primers 5′-AGCCAGTGCAAATCACGATGAG-3′ and 5′-ATCACGCAGCATGGGCAGGTTC-3′, comprising the first 1 kb of the gene. For reverse transcription-PCR, total RNA was extracted from wandering larvae using TRI reagent (Sigma), treated with DNase I and reverse transcribed using RTG You-Prime-First-Strand Beads (GE Healthcare). PCR was performed in a reaction using combined DmTrx-2 primers (5′-ATGGACAGCTGACCAAGGCATC-3′ and 5′-CCCACTTAGATATTGGCCTTGATG-3′) and glycerol-3-phosphate dehydrogenase (G3PDH) primers (5′-CCACTGCCGAGGAGGTCAACTA-3′ and 5′-GCTCAGGGTGATTGCGTATGCA-3′).

Anti-DmAIF antibodies were generated from rabbits immunized with peptides corresponding to DmAIF (aa 661–674), and purified on an affinity column (Pierce). Fly protein lysates were produced by homogenizing larvae in ice-cold RIPA buffer containing protease inhibitors (Roche). Western blotting was performed as previously described.40

Immunohistochemistry

Embryos were collected and treated with AO as described.40 Eye discs were dissected from third-instar larvae, stained for 15 min with 2 μg/ml AO in PBS and imaged immediately. In Supplementary Figure 2b, wandering larvae were γ-irradiated with 45 Gy; 5 h later, eye discs were dissected and stained with AO. For immunostaining, eye discs were fixed in 4% PFA/PBS for 20 min and washed in PBS containing 0.1% Triton X-100 (PBS/Triton). Primary antibodies used were as follows: guinea-pig anti-Kr (1 : 600), anti-dHb9 antibody (1 : 500), anti-cleaved caspase 3 (1 : 50; Cell Signaling) and anti-c-myc (1 μg/ml; Clontech). Antibody labeling was visualized using fluorochrome-conjugated secondary antibodies (1 : 500; Vector Laboratories or 1 : 1000; Molecular Probes). Embryos were imaged by confocal microscopy (Leica TCS SP5) and quantified using Image J software. All cells per segment of the VNC labeled against anti-Kr were quantified, and only the lateral VNC cells peripheral to the horizontal lines labeled against anti-dHb9 antibody were quantified (yellow brackets in Figure 3).

Assays of mitochondrial function

Drosophila mitochondria were prepared by homogenizing approximately 100 first-instar larvae using a similar procedure to that used for preparing mouse tissue mitochondria,7 except that a slightly modified homogenization medium (250 mM sucrose, 5 mM Tris–HCl, 2 mM EGTA, 1% (w/v) BSA, pH 7.4 at 4°C) was used. Measurement of mitochondrial respiratory activity was performed as described.7 For ATP measurements, the luciferase-based ATP bioluminescence kit (Roche) was used. Approximately 50 larvae were homogenized in the lysis buffer provided and centrifuged to remove particulate matter. An aliquot of the supernatant was removed for protein quantitation, and the remainder boiled for 10 min to destroy endogenous ATPase activity before assessing total ATP levels, according to the manufacturer's instructions.