The four aldehyde oxidases of Drosophila melanogaster have different gene expression patterns and enzyme substrate specificities

Aldehyde oxidases (AOX; EC 1.2.3.1) belong to the family of molybdo-flavoenzymes together with the structurally related xanthine oxidoreductase (XOR) enzyme, which serves a key role in the catabolism of purines, metabolizing hypoxanthine into xanthine and xanthine into uric acid (Enroth et al., 2000). The structure of mouse AOX3 was shown to be highly identical to the homodimeric butterfly-shaped XOR enzymes, consisting of two identical 150 kDa subunits, with differences in the substrate binding pocket (Coelho et al., 2012). The structure of each monomer of these enzymes includes a 20 kDa N-terminal 2x[2Fe-2S] domain, a 40 kDa central FAD-containing domain and an 85 kDa C-terminal molybdenum cofactor (Moco) domain (Garattini et al., 2003; Mahro et al., 2011). The Moco of eukaryotic AOXs contains an equatorial sulfur ligand required for the catalytic activity of the enzymes (Wahl and Rajagopalan, 1982; Bray, 1988; Hille, 1996).

Most animal and plant genomes contain AOX gene clusters that arose from a series of duplication events from a common XOR ancestor (Forrest et al., 1956; Nash and Henderson, 1982; Rodríguez-Trelles et al., 2003; Garattini et al., 2008). Rodents and marsupials contain the largest number of AOX functional genes (AOX1, AOX3, AOX4 and AOX3L1), whereas the human genome contains one single and functional AOX1 gene (hAOX1) and two non-functional pseudogenes, representing the vestiges of the mouse AOX3 and AOX3L1 orthologs (Garattini et al., 2003; Garattini et al., 2008; Garattini et al., 2009; Kurosaki et al., 2013).

Mammalian AOXs are characterized by a broad substrate specificity, catalyzing the oxidation of various types of aldehydes and aza- and oxo-heterocycles of pharmacological and toxicological relevance (Schumann et al., 2009; Pryde et al., 2010; Garattini and Terao, 2011; Mahro et al., 2011; Garattini and Terao, 2012). The presence of some highly conserved residues in the catalytic centers and the high structural similarity between AOXs and XOR suggest common reaction mechanisms (Schumann et al., 2009; Coelho et al., 2012). Nevertheless, AOXs exhibit a broader substrate specificity than XOR (Pauff et al., 2009; Coelho et al., 2012). The physiological role and the endogenous substrates of human and mammalian AOXs are unknown.

In Drosophila melanogaster, the rosy (ry) locus encodes XOR (Chovnick et al., 1976; Keith et al., 1987). Loss-of-function ry mutants accumulate xanthine and hypoxanthine and are devoid of urate (Morita, 1958; Glassman and Mitchell, 1959), which makes them hypersensitive to oxidative stress (Hilliker et al., 1992). Drosophila melanogaster ry mutants are characterized by a dull reddish-brown eye colour because of the lowered concentrations of the red eye pigment drosopterin (bright red), and the increased chromogenic oxidation of the eye pigment dihydroxanthommatin (yellow brown) to xanthommatin (dull dark brown) (Phillips and Forrest, 1980). This suggests that the brownish eye color of the ry mutant arises from the progressive oxidation and degradation of eye pigments occurring in the absence of the antioxidant urate (Hilliker et al., 1992; Vorbach et al., 2003; Glantzounis et al., 2005). Metabolic profiling of the Drosophila ry and maroon-like (ma-l, suppressing Moco sulfuration and inactivating both XOR and AOXs) mutants confirmed the biochemical changes in xanthine, hypoxanthine and urate and revealed unsuspected changes in each of the tryptophan, arginine, pyrimidine and glycerophospholipid metabolic pathways (Kamleh et al., 2008; Kamleh et al., 2009).

Two loss-of-function mutant alleles, low pyridoxal oxidase (Po^lpo) and aldehyde oxidase-1 (Aldox-1ⁿ¹), have been previously described to control AOX enzymatic activity (Collins and Glassman, 1969;

Dickinson, 1970; Dickinson and Gaughan, 1981). On the basis of the data available, it was proposed that Po^lpo and Aldox-1ⁿ¹ map to two distinct and unidentified AOX genes belonging to the cluster of four AOX genes present on Drosophila chromosome 3 (Adams et al., 2000; Garattini et al., 2008). At present, the identities of the AOX genes corresponding to Po^lpo and Aldox-1ⁿ¹ are unknown. In addition, it remains to be established whether all of the four predicted Drosophila AOX genes are functionally active. Finally, the physiological significance of the Drosophila AOX genes needs to be ascertained. With respect to this, pyridoxal, the dephosphorylated form of the active vitamin B6 metabolite, seems to be a specific substrate of the Po encoding enzyme pyridoxal oxidase (PO). In insects, the levels of pyridoxal are considered to be of particular physiological relevance (Stanulović and Chaykin, 1971; Browder and Williamson, 1976; Cypher et al., 1982).

In this study we used a native in-gel activity assay to analyze extracts from a variety of D. melanogaster strains, where AOX activities had been disrupted by mutation or RNA interference (RNAi). We cloned the classical Po^lpo allele and showed that PO is the product of the Drosophila AOX1 gene. We also report experiments suggesting that the AOX2 activity is only present during metamorphosis and that two further AOX activities are associated with the AOX3 gene. We show that AOX1, AOX2 and AOX3 are conserved in most Drosophila species. Finally, we provide evidence that AOX4 is evolutionarily more recent and that the activity of the encoded enzyme is undetectable in D. melanogaster extracts using a wide variety of substrates.

Sequencing of the D. melanogaster genome led to the annotation of four genes, CG18522, CG18519, CG6045 and CG18516, coding for an equivalent number of predicted proteins having all the structural characteristics of the AOX enzymes. In this study, we adopted the same nomenclature given to the four genes by Garattini et al. (Garattini et al., 2008): AOX1 (CG18522), AOX2 (CG18519), AOX3 (CG6045) and AOX4 (CG18516). Fig. 1 illustrates that these genes are clustered on chromosome 3R, 2.5 Mbp downstream from the XOR/ry locus (Bridges and Brehme, 1944; Keith et al., 1987; Chovnick et al., 1990). Using appropriate cytogenetic markers, the Po^lpo and Aldox-1ⁿ¹ alleles were mapped to the cytogenetic region 88F-89A, 0.08 cM apart from each other (Nelson and Szauter, 1992). This mapping is consistent with the location of the AOX gene cluster. Fig. 1 also shows the transposon insertions and genomic deletion mutants (deficiencies) that were used in this study and PCR primers used for molecular analysis.

To reconstruct the evolutionary history, we performed a phylogenetic analysis on the predicted structures of the AOX genes and corresponding protein products in Drosophila using the available genomic sequencing data covering a total of 12 distinct lineages. Our phylogenetic reconstruction suggests that Drosophila AOX1 originated from an ancient XOR duplication, distinct from the one giving rise to vertebrate AOXs (Fig. 2). It is likely that this duplication resulted in the appearance of AOX1, which, subsequently, duplicated into AOX2. A further duplication event involving AOX2 gave rise to AOX3. Drosophila willistoni and Drosophila mojavensis are endowed with a single XOR and three active AOX genes that are predicted to be the orthologs of D. melanogaster AOX1, AOX2 and AOX3 (see also supplementary material Fig. S1). These species are representative of the most ancient complement of AOXs in Drosophila. The subsequent evolutionary history of AOXs in the Drosophilidae is characterized by four distinct AOX3 duplications (Fig. 2). Drosophila virilis is endowed by a species-specific AOX3-D in addition to XOR, AOX1, AOX2 and AOX3 (Fig. 2, supplementary material Fig. S1). In Drosophila grimshawi, two distinct duplications giving rise to AOX2-D and AOX3-D2 are evident (Fig. 2, supplementary material Fig. S1).

A major evolutionary event is represented by a third AOX3 duplication, which led to the origin of AOX4 and is conserved among downstream lineages. The duplication led to the extant complement of the predicted molybdo-flavoenzymes present in Drosophila simulans, Drosophila sechellia, D. melanogaster, Drosophila yakuba and Drosophila erecta, which consists of XOR, AOX1, AOX2, AOX3 and AOX4 genes (Fig. 2, supplementary material Fig. S1). In Drosophila ananassae, a further AOX3 duplication gave rise to AOX3-D3 (Fig. 2, supplementary material Fig. S1). Finally, one deletion and three AOX duplications are predicted in the Drosophila obscura group, before the branching of Drosophila pseudobscura and Drosophila persimilis. In these lineages, AOX3 is deleted while AOX1 and AOX4 are duplicated into AOX7 and AOX5, respectively. In turn, AOX5 is duplicated into AOX6 (Fig. 2, supplementary material Fig. S1).

To evaluate whether the AOX genes predicted in D. melanogaster code for active proteins, we performed experiments on fly extracts. Proteins were separated by native PAGE and AOX activity was determined by an in-gel staining assay using various substrates and thiazolyl blue tetrazolium bromide (MTT) as the electron acceptor. With benzaldehyde as a substrate, four bands were detected (Fig. 3A). The top two bands constitute an incomplete resolved doublet (bands a and b). The enzyme-specific substrates pyridoxal, acetaldehyde and hypoxanthine indicate that the four bands are the products of the activity of distinct AOX isoenzymes. In addition, the results show that band a is stained exclusively in the presence of pyridoxal, while band b is stained only in the presence of acetaldehyde as a substrate (Fig. 3A). Band c is detectable only with acetaldehyde. Band d corresponds to XOR, as it is stained after incubation with hypoxanthine.

To relate each of the bands to the activity of the AOX1–4 gene products, we evaluated the complement of bands in three fly strains lacking distinct genomic regions of the chromosome 3R affecting the AOX gene cluster (Fig. 1). Df(3R)BSC515[AOX1-4] removes all four AOX genes. Only AOX1 is deleted in Df(3R)Exel6174[AOX1], whereas in Df(3R)Exel7326[AOX2-4], AOX2–4 are deleted without affecting AOX1. We analyzed the trans-heterozygotes between the Deficiency (Df) chromosomes and the triple mutant recombinant ry², Po^lpo, Aldox-1ⁿ¹ that removes all detectable AOX and XOR activity (Dickinson and Gaughan, 1981). Fly extracts of these genotypes were screened with 20 different substrates (Fig. 3B, supplementary material Fig. S2). Extracts of the homozygote white mutant were used as controls in our experiments, as they show the same four AOX/XOR bands observed in wild-type flies (see Fig. 3A). Out of the 20 substrates considered, salicylaldehyde, 2,4-dihydroxybenzaldehyde (DHB), 4-nitrobenzaldehyde and 2,4,5-trimethoxybenzaldehyde (TMB) are the most effective in distinguishing the various AOX bands. Similar to benzaldehyde (Fig. 3A), salicylaldehyde shows four bands (Fig. 3B). DHB highlights only the top three a–c bands, 4-nitrobenzaldeyhde only the two b and c bands, and TMB only the top a band (Fig. 3B). These results confirm that different D. melanogaster AOXs are characterized by different substrate specificities (Fig. 3B, supplementary material Fig. S2).

No bands stainable with the AOX substrates considered are detectable in Df(3R)BSC515[AOX1-4]/ry², Po^lpo, Aldox-1ⁿ¹ mutant extracts. In contrast, the Df(3R)Exel6174[AOX1]/ry², Po^lpo, Aldox-1ⁿ¹ mutant extracts only lack band a, suggesting that the AOX1 gene codes for the classical PO enzyme. Instead of pyridoxal, TMB, previously shown to be specific for PO (Cypher et al., 1982), was used. In the Df(3R)Exel7326[AOX2-4]/ry², Po^lpo, Aldox-1ⁿ¹ mutant extracts, bands b and c are not detectable. In the same mutant, the fastest migrating band d is visible when salicylaldehyde is used, contrary to what is observed in the ry mutant. This observation confirms previous studies that indicate that XOR can utilize certain AOX substrates (Dickinson and Gaughan, 1981). Finally, our results show that bands b and c have the same substrate specificity, which might imply that these two bands are generated by the same protein (Fig. 3, supplementary material Fig. S2).

To validate and extend the findings of the substrate screen, we further analyzed the AOX activity profiles in the deletion mutant Df(3R)Exel6174[AOX1]/ry², Po^lpo, Aldox-1ⁿ¹ (Fig. 4A) and following ubiquitous AOX1 silencing with a specific RNAi (Fig. 4B). Extracts from the control flies Df(3R)Exel6174[AOX1]/TM6B,Tb¹, white and ry⁵⁰⁶ show a specific AOX1 band following incubation with TMB. The band is absent in extracts of Df(3R)Exel6174[AOX1]/ry², Po^lpo, Aldox-1ⁿ¹ flies (Fig. 4A). In contrast, the bands b and c observed with acetaldehyde as a substrate are present in the extracts of all control and mutant flies (Fig. 4A). Similarly, silencing of AOX1 with a specific RNAi driven by the ubiquitous act-Gal4 promoter in white flies annihilates the TMB-dependent band a, while it leaves unaffected the acetaldehyde-dependent bands b and c (Fig. 4B).

To further confirm that AOX1 is indeed the enzymatic activity missing in Df(3R)Exel6174[AOX1]/ry², Po^lpo, Aldox-1ⁿ¹ flies, the wild-type D. melanogaster AOX1 gene was cloned and the corresponding recombinant protein was expressed and purified in Escherichia coli. Recombinant AOX1 was used to compare the migration behavior and substrate specificity to that in the whole-fly extracts using in-gel staining analysis (Fig. 4C). As expected, recombinant AOX1 and whole-fly-extract band a co-migrate and are stained strongly with benzaldehyde/TMB and more weakly with acetaldehyde (Fig. 4C). Cloning and sequencing of the AOX1 gene from ry², Po^lpo, Aldox-1ⁿ¹ flies demonstrated an 11-bp deletion, which results in a frame-shift mutation that eliminates the Moco domain of AOX1 (supplementary material Fig. S3). These results unequivocally demonstrate that the AOX1 gene codes for the classical PO enzyme, which is inactivated in the Po^lpo mutant allele.

Acetylaldehyde-dependent bands b and c are absent in Df(3R)Exel7326[AOX2-4]/ry², Po^lpo, Aldox-1ⁿ¹ mutant flies, confirming that the two bands must be the products of AOX genes other than AOX1 (Fig. 5A). To identify the AOX gene responsible for the AOX activity associated with bands b and c, we used the Minos insertion mutant of the AOX3 gene (Mi-AOX3; Fig. 1). Unlike extracts obtained for the white, ry⁵⁰⁶ and parental strains, the Df(3R)Exel7326[AOX2-4]/Mi-AOX3 counterparts show no AOX band upon incubation with acetaldehyde (Fig. 5B). Furthermore, silencing of the AOX3 gene with a specific RNAi in adult white flies causes a strong reduction in the levels of the band b and the disappearance of band c (Fig. 5C). The intensity of the signal associated with the AOX1 band after incubation with TMB is comparable in Df(3R)Exel7326[AOX2-4]/Mi-AOX3 and parental flies, demonstrating the specificity of the AOX3 RNAi gene knock-down. The results are consistent with the idea that bands b and c are due to two separable protein products originating from the AOX3 gene. At present, it is unclear whether the two forms of AOX3 are the products of alternative splicing events or the result of post-translational processing.

The data produced by the modENCODE and FlyAtlas projects are accessible on FlyBase (http://flybase.org) and provide comprehensive information about the tissue- and stage-specific expression profiles of almost all the open reading frames identified in the D. melanogaster genome (Celniker et al., 2009; Robinson et al., 2013). Comparing the expression patterns of the various AOX1, AOX2 and AOX3 mRNAs (Fig. 6A), it is evident that AOX1 is expressed throughout development, with peaks in the late embryonic stages, all the larval stages and during adulthood. AOX3 mRNA is also measurable throughout development, with complementary peaks of expression relative to AOX1, i.e. during the early embryonic stages, in second and third instar larvae, in older pupae and in adult flies. AOX2 shows low levels of expression in 10- to 16-h-old embryos and in third instar larvae, while large amounts of the transcript are expressed in 2-day-old pupae. AOX4 is undetectable throughout development, while measurable levels of the transcript are present in adult head samples (FlyAtlas dataset).

To confirm some of the mRNA data at the protein level, we analyzed the profiles of AOX activity in extracts of different developmental stages (Fig. 6B). AOX1 was assayed with TMB, acetaldehyde was used for the detection of AOX3 and cinnamaldehyde was used as a common substrate to visualize all the AOX bands. Consistent with the mRNA data, high levels of AOX1 activity are detected in larvae and adult flies but low activities are seen in early embryos, white pre-pupae and pupae (2 to 3 days old). In contrast, AOX3 activity is high in early embryos, larvae, older pupae and adult flies, as expected from the AOX3 mRNA profile.

Cinnamaldehyde reveals an additional band, which is observed in the pupae stages of the control white flies but not in the homozygous and viable Minos-AOX2 insertion mutant (Mi-AOX2; Fig. 6B). This additional band is absent in the 2-day-old pupae of the ry², Po^lpo, Aldox-1ⁿ¹ mutant strains (Fig. 6C). The observation suggests that the Aldox-1ⁿ¹ locus impairs the expression of both the AOX3 and AOX2 genes.

We evaluated whether the expression of the AOX4 mRNA in the head of the adult fly, as suggested by the FlyAtlas data, translates into the synthesis of catalytically active enzyme. To this purpose, we compared the AOX enzymatic profile in extracts obtained from the whole body, heads and thorax/abdomen of white flies. The results demonstrate no enrichment in AOX4 activity at the level of head samples (Fig. 6D). Interestingly, we observed that AOX1 activity is lower in the head and stronger in the abdomen/thorax. Taken together, these data indicate that there is a general correlation between the tissue- and stage-specific expression of the AOX mRNAs and corresponding catalytically active AOX iso-enzymes.

In previous studies, two genetic loci affecting D. melanogaster AOX activity were identified using the two loss-of-function mutant alleles Po^lpo and Aldox-1ⁿ¹, which are located in close proximity on the chromosomal region 88F-89A (Collins and Glassman, 1969; Dickinson, 1970; Dickinson and Gaughan, 1981; Nelson and Szauter, 1992). However, the available in-gel staining assays permitted the identification of only two AOX enzymes separable from XOR (Dickinson, 1970; Dickinson and Gaughan, 1981), (Warner et al., 1980; Warner and Finnerty, 1981). However, genome sequencing of D. melanogaster predicts a cluster of four AOX genes (AOX1–4) located ~2.5 Mbp downstream from the rosy locus on the right arm of the third chromosome (Fig. 1) (Adams et al., 2000; Garattini et al., 2008). This finding indicates that the initial characterization of AOX activities was incomplete and raises a number of questions regarding the physiological significance of all of these proteins in different tissues, their substrate specificities and their expression profiles during the development of the insect. In this study we reconstruct the phylogenesis of the AOX genes in Drosophila (Figs 2, 7). In addition, we demonstrate that the AOX1 gene encodes for PO (Fig. 4), and that the AOX3 activity is disrupted in the classical Aldox-1ⁿ¹ mutant (Fig. 5). Finally, we establish that the AOX2 gene is expressed during metamorphosis (Fig. 6), while no enzymatic activity could be associated with the AOX4 gene. To reach these conclusions, we used a combination of genetic and biochemical tools, including an optimized in-gel activity staining method, new transposon-induced alleles, deficiency chromosomes and the RNAi methodology.

Previous phylogenetic studies have indicated that the evolution of the AOX genes is the result of a series of gene duplication, pseudogenization and gene deletion events (Garattini et al., 2008; Kurosaki et al., 2013). The reconstruction of the evolutionary history of this family of genes in Drosophila indicates that AOXs have originated from an ancient duplication of XOR. This duplication is distinct from the one giving rise to vertebrate AOXs, suggesting that the Drosophila AOX genes are not directly related to the vertebrate counterparts. In contrast to the human genome, which contains only one functional AOX gene (Berger et al., 1995; Garattini et al., 2003; Garattini et al., 2008), the Drosophilidae genomes include up to six active AOX genes. The minimal and probably the most ancient complement of AOX genes, consisting of the AOX1, AOX2 and AOX3 loci, is observed in D. virilis, D. mojavensis and D. willistoni (Fig. 7). With the exception of D. obscura species group, which lacks AOX3, but has acquired AOX5, AOX6 and AOX7, all the species analyzed are characterized by the presence of the AOX1, AOX2 and AOX3 genes. Out of the four D. melanogaster AOX genes, AOX1, AOX2 and AOX3 are expressed in an active form during different developmental stages of the insect. In contrast, AOX4 is not present in all Drosophila species (Fig. 7). Furthermore, in D. melanogaster, whose genome is predicted to contain an active AOX4 gene, no catalytically active AOX4 enzyme was detected in any of the organ and developmental stages considered. Whether AOX4 has a specialized function in D. melanogaster and why AOX3 was duplicated into AOX3-D (D. virilis species group), AOX4 (D. melanogaster species group, and D. ananassae), AOX3-D2 (Hawaiian Drosophila species group) and AOX3-D3 (D. ananassae) are questions that remain open.

While we were able to show that the Po^lpo allele maps to the AOX1 gene, the molecular nature of the Aldox-1ⁿ¹ allele remains unknown. Further observations of the ry², Po^lpo, Aldox-1ⁿ¹ mutant flies suggest that this mutant allele affects also AOX2, as 2-day-old white pupae are endowed with an additional AOX band, which is absent in the ry², Po^lpo, Aldox-1ⁿ¹ and Mi-AOX2 mutant flies (Fig. 6). Moreover, consistent with a former study on the Aldox-1ⁿ¹ locus (Dickinson and Gaughan, 1981), we also detected a major and minor migrating band for AOX3 (indicated as band b and c in Fig. 3, supplementary material Fig. S2). Dickinson and Gaughan (Dickinson and Gaughan, 1981) suggested that two Aldox-dependent activities existed as modified minor band and unmodified major band forms. The cDNA sequences available in modENCODE support the presence of two alternative AOX3 transcripts. One of them (NCBI accession AY094721) has a deletion of 226 amino acids from in the N terminus (Celniker et al., 2009). Therefore, we suggest that two AOX3 isoforms with distinct sizes are generated from the AOX3 gene, which is consistent with the existence of two enzymatically active bands characterized by identical substrate specificities (Fig. 3, supplementary material Fig. S2). It is also possible that the faster band is a degradation product of AOX3. In this case, however, greater inter-sample variations in the intensity of the minor band to those observed would be expected.

The 20 substrates used in this study were selected among the molecules known to be oxidized by mammalian AOXs, although pyridoxal has been reported to be oxidized by fly extracts as well (Dickinson and Sullivan, 1975). The results obtained with these compounds indicate that D. melanogaster AOX1, AOX2 and AOX3 have different substrate specificities. It was previously shown that the homozygote Aldox-1ⁿ¹ extracts (containing AOX1) have less than 2% of the wild-type AOX activity when acetaldehyde is used as substrate. In contrast, Po^lpo extracts (containing AOX3) present with levels of AOX activity that are close to those determined in wild-type flies (Dickinson and Gaughan, 1981; Cypher et al., 1982). We confirm these observations, demonstrating that acetaldehyde oxidation is detectable only in the presence of high amounts of purified recombinant AOX1 (Fig. 4C). In fly extracts acetaldehyde is specifically oxidized by AOX3. In contrast, AOX1 is the only enzyme capable of oxidizing pyridoxal or the benzaldehyde derivate TMB (Browder and Williamson, 1976; Cypher et al., 1982). We hypothesize that AOX1 plays a role in the decomposition of vitamin B6 from pyridoxal to 4-pyridoxic acid, which then might be excreted by the gut system (Forrest et al., 1961), as abundant amounts of the enzyme AOX1 are measurable in the midgut and Malphigian tubules (Dickinson and Gaughan, 1981) (FlyAtlas). Pyridoxal is one of the three natural forms of vitamine B6, along with pyridoxamine and pyridoxine. Pyridoxal is converted by the pyridoxal kinase into pyridoxal-5-phosphate (PLP), which acts as a cofactor for many enzymes involved in amino acid metabolism (Toney, 2005). Excess of vitamin B6 in the food might be countered by the action of AOX1, representing a competitive reaction to the pyridoxal kinase for the regulation of homeostatic levels of the PLP cofactor in the organism.

In the case of AOX1 and AOX3, we confirm the overlapping substrate specificities for a number of aldehydes, and aza- and oxo-heterocycles (Fig. 3, supplementary material Fig. S2) (Dickinson and Gaughan, 1981; Cypher et al., 1982). Interestingly, AOX2 activity is detectable with cinnamaldehyde but not with benzaldehyde (Fig. 6). These data support the idea that different AOX isoforms recognize a unique set of substrates and carry out different physiological tasks in vivo. In previous studies it was shown that the AOX enzymes in Aldox-1ⁿ¹ and Po^lpo mutants have tissue-specific and overlapping expression patterns, which is also consistent with the modENCODE and FlyAtlas data presented in FlyBase (Dickinson and Gaughan, 1981; FlyAtlas). It was shown that the tissue distributions of AOX1 and AOX3 are overlapping and the two enzymes are synthesized predominantly in the midgut and Malpighian tubules. In contrast to AOX3, AOX1 is weakly expressed in the head (Fig. 6D) (Dickinson and Gaughan, 1981; FlyAtlas). In D. melanogaster whole-body extracts, a zymographic band corresponding to AOX4 cannot be identified, leading us to hypothesize that the AOX4 gene exerts a tissue-specific function in the head of adult flies, as suggested by the mRNA expression data available in FlyAtlas (Robinson et al., 2013). However, our attempts to find a head-specific AOX4 have been inconclusive. Thus, we propose that AOX4 is either not expressed in an active form or its synthesis is restricted to sub-structures of the head, such as the antennae, precluding determination of the enzyme with our methodology. Reports demonstrating the presence of AOX enzymes in the antennae of several moth species are available (Rybczynski et al., 1989; Rybczynski et al., 1990; Merlin et al., 2005; Pelletier et al., 2007). In these studies, AOX activity has been detected in the antennae of both sexes, although larger amounts of the enzyme are present in male insects. Based on these observations, it has been proposed that antennae-specific AOXs may play a role in pheromone degradation. In this respect, it is interesting to note that the mammalian AOX3l1 enzyme may serve a similar function in rodents and dogs, as AOX3l1 is restricted to the olfactory mucosa and may function in the degradation of odorants and pheromones that are often aromatic aldehydes (Kurosaki et al., 2004; Terao et al., 2006; Kurosaki et al., 2013). It is remarkable that the AOX3l1 ortholog is a functionally inactive pseudogene in humans and primates, taxa whose olfactory system is less developed relative to rodents, dogs and insects.

In conclusion, we were able to distinguish the activities of AOX1, AOX2 and AOX3 in D. melanogaster and to confirm their complementary expression patterns during the different stages of the insect's development. We showed that the AOX gene cluster is under continuous evolutionary change in other species of Drosophila. However, the actual physiological functions of the Drosophila AOXs in different species, tissues and developmental stages remain unresolved.

AOX substrates were obtained from Fluka, Sigma-Aldrich, Merck and Reachim: benzaldehyde, acetaldehyde, allopurinol, butyraldehyde, 2,4-dihydroxybenzaldehyde, 3-ethoxy-4-hydroxybenzaldehyde, heptaldehyde, 2-hydroxypyrimidine, hypoxanthine, N′-methylnicotinamide, 2,3-naphtalenedicarbaldehyde, 4-nitrobenzaldehyde, phenanthridine, phthalazin, pyridoxal-HCl, phenylacetaldehyde, salicylaldehyde, 2,4,5-trimethoxybenzaldehyde, valeraldehyde, vanillin, xanthine and trans-cinnamaldehyde.

All flies used in this study were retrieved from public stock centers. Specifically, from the Vienna Drosophila RNAi Center (VDRC) (Dietzl et al., 2007), we used second chromosome insertion for UAS-AOX1-RNAi and an insertion on the X chromosome for UAS-AOX3-RNAi (Table 1). To induce the RNAi, a ubiquitous w*, Actin5C-Gal4/CyO (act-Gal4) driver line was used as described previously (Missirlis et al., 2003) (Table 1). From Bloomington Drosophila Stock Center (BDSC), we used Minos insertions disrupting AOX2 and AOX3, respectively (Metaxakis et al., 2005): the stocks were here referred to as Mi-AOX2 and Mi-AOX3 (Table 1, Fig. 1A). In addition, three deficiency chromosomes were obtained from BDSC that had the following characteristics: (1) Df(3R)Exel6174[AOX1] uncovers specifically AOX1, (2) Df(3R)Exel7326[AOX2-4] uncovers AOX2-4 without affecting AOX1, and (3) (Df(3R)BSC515[AOX1-4] uncovers all four AOX genes (Table 1, Fig. 1A). The classic mutations Po^lpo (Collins and Glassman, 1969) and Aldox-1ⁿ¹ (Dickinson, 1970) were used from the compound, homozygous viable chromosome ry², Po^lpo, Aldox-1ⁿ¹, Sb^sbd-2 obtained from BDRC (Table 1). Wild-type TAN3 (Sadraie and Missirlis, 2011) and white (Gutiérrez et al., 2013) were used as controls.

The cDNA clone LD37006 containing the cDNA of D. melanogaster AOX1 (CG18522) was obtained from the Berkeley Drosophila Genome Project (BDGP) Gold cDNA collection of the Drosophila Genomics Resource Center, and the forward primer CATATGGCTGGAAGAATTACAATCAACG and the reverse primer GTCGACCTAATTGAGCTTGAACTGGCTG were used, which allowed cloning into the NdeI-SalI sites of the expression vector pTrcHis (Temple et al., 2000). The resulting plasmid was designated pZM171 and expresses AOX1 with an N-terminal His₆-tag fusion. The plasmid pZM171 was transformed into E. coli TP1000 cells and His₆-AOX1 was heterologously expressed. The recombinant protein was purified by affinity chromatography using nickel-nitrilotriacetic acid (Ni-NTA) resin (QIAGEN, Valencia, CA, USA) and by a Superose 12 size exclusion chromatography column (GE Healthcare).

For each sample, 10 flies, 10 abdomens, 50 heads, 200 embryos, 100 first instar larvae, 20 s instar larvae, 10 third instar larvae or 10 pupae were collected in a 1.5 ml Eppendorf tube, shock-frozen in liquid N₂ and stored at −80°C until further usage. For the preparation of fly extracts, a micro grinder (Sigma), composed of a battery-driven Pellet pestles cordless motor and Pellet pestles blue polypropylene sticks, was used. A volume of 8 μl per adult fly, larvae or pupae (except: 2 μl per embryo; 4 μl per first larvae; 3 μl per head; 5 μl per abdomen) of 0.1 mol l⁻¹ Tris-HCl (pH 8.0) lysis buffer additionally containing 100 μmol l⁻¹ EDTA, 1 mmol l⁻¹ DTT and 2× protease inhibitor cocktail (Roche) was added to the flies and homogenization was performed for 30 s. The crude extract was centrifuged at 14,000 g for 10 min, the supernatant was transferred into a new 1.5 ml Eppendorf tube and a second centrifugation step was performed at 17,000 g for 10 min to remove additional cell debris. The homogenization procedure was carried out at 4°C.

To detect the activity of AOX enzymes in fly extracts, we performed a colorimetric in-gel activity assay. Here, 40 μg total protein of the whole fly extracts were separated by 6% native PAGE and the gels were subsequently stained for AOX activity using 5 mmol l⁻¹ of a AOX substrate (see above), 0.4 mmol l⁻¹ 3-(4,5-dimethylthiazolyl-2)-2,5-diphenyltetrazoliumbromide (MTT) and 0.1 mmol l⁻¹ phenazine methosulfate (PMS). Upon AOX reaction, the reducing agent precipitates into a purple-black color, making it visible as band in the native gel.

Using the tBlastn algorithm, we searched for exonic sequences coding for polypeptides with similarity to the D. melanogaster XOR, AOX1, AOX2, AOX3 and AOX4 proteins or an evolutionarily closer AOX, when necessary. Reconstruction of the AOX genes was performed, taking into account conservation of the coding exons in terms of sequence, length, and phase of exon/intron junctions. A major problem was sorting of genes coding for an active AOX protein (active genes) from pseudogenes, which may be transcribed, but do not encode a complete protein. AOX genes were considered as pseudogenes by the following criteria: (1) the presence of one STOP codon located in coding sequence; (2) the absence of one or more identifiable exons resulting in disruption of the expected downstream open reading frame; and (3) the lack of a recognizable intron/exon junction maintaining the open reading frame. To assign names to new previously unknown AOX isoenzymes, the following criteria were adopted: if duplications were detected in no more than one species, the ‘-D’ suffix was added to the ‘AOX’ species that was likely to have originated it (i.e. D. ananassae AOX3 and D. ananassae AOX3-D); if the same AOX duplication event was detected in more than one lineage, a new name was given (i.e. AOX4 duplication in both D. persimilis and D. pseudoobscura was named AOX5).

To determine the evolution of the AOX genes, a phylogenetic analysis of the corresponding protein products was performed (12 taxa and 59 sequences, see supplementary material Table S1). Exonic structure, genomic loci and chromosomal position (if available) of the predicted genes were annotated. Multiple sequence alignments were prepared using Muscle (http://www.ebi.ac.uk/Tools/msa/muscle/) and visualized with CLC Main Workbench (http://www.clcbio.com/). The phylogenetic tree containing all the available Drosophila AOX and XOR protein structures was generated through the distance-based method BioNJ (http://www.atgc-montpellier.fr/bionj/) (Gascuel, 1997) and 100 bootstrap replicates were calculated. A comprehensive phylogenetic tree spanning from bacteria to vertebrates, generated through BioNJ, was based on publicly available prokaryotic as well as eukaryotic AOX and XOR protein sequences (see supplementary material Table S2) (Kurosaki et al., 2013).