Alternative oxidase confers nutritional limitation on Drosophila development

Abstract The mitochondrial alternative oxidase, AOX, present in most eukaryotes apart from vertebrates and insects, catalyzes the direct oxidation of ubiquinol by oxygen, by‐passing the terminal proton‐motive steps of the respiratory chain. Its physiological role is not fully understood, but it is proposed to buffer stresses in the respiratory chain similar to those encountered in mitochondrial diseases in humans. Previously, we found that the ubiquitous expression of AOX from Ciona intestinalis in Drosophila perturbs the development of flies cultured under low‐nutrient conditions (media containing only glucose and yeast). Here we tested the effects of a wide range of nutritional supplements on Drosophila development, to gain insight into the physiological mechanism underlying this developmental failure. On low‐nutrient medium, larvae contained decreased amounts of triglycerides, lactate, and pyruvate, irrespective of AOX expression. Complex food supplements, including treacle (molasses), restored normal development to AOX‐expressing flies, but many individual additives did not. Inhibition of AOX by treacle extract was excluded as a mechanism, since the supplement did not alter the enzymatic activity of AOX in vitro. Furthermore, antibiotics did not influence the organismal phenotype, indicating that commensal microbes were not involved. Fractionation of treacle identified a water‐soluble fraction with low solubility in ethanol, rich in lactate and tricarboxylic acid cycle intermediates, which contained the critical activity. We propose that the partial activation of AOX during metamorphosis impairs the efficient use of stored metabolites, resulting in developmental failure.

example, lay their eggs on rich nutrient sources, which sustain rapid growth during larval development. Metabolites that are synthesized and stored during the larval stages are then used to drive the subsequent processes of cellular proliferation, migration, and differentiation. The program of metamorphosis during pupal stage is then executed in an environment where further feeding is not possible.
The metabolic processes that characterize the larval and pupal stages may be crudely described as anabolic and catabolic, respectively. Nevertheless, the biosynthesis that occurs during larval development depends on a supply of biological energy, while pupae rely on converting some of their stored resources into new biomolecules, as well as fueling and regulating the morphogenetic program. Importantly, most of the energy generation, as well as the production of relevant metabolites for biosynthesis and signaling, originate within the mitochondria. The tricarboxylic acid (TCA) cycle, in particular, is central to catabolism, anabolism, and signaling, and is in turn reliant on the mitochondrial respiratory chain.
The model organism Drosophila melanogaster has been extensively studied with regard to the molecular and cellular processes that underpin this developmental program. However, the metabolic events which accompany development have received less attention.
D. melanogaster is considered a cosmopolitan species (Markow, Nazario-Yepiz, & Ramirez Loustalot-Laclette, 2017;Matzkin, Johnson, Paight, Bozinovic, & Markow, 2011), able to grow on a wide variety of food sources. These commonly include glucose and other sugars from decaying fruit, as well as yeast, which represents a rich source of amino acids and other nutrients. The major metabolic fuel for the pupal stage, accumulated during larval development, is triglycerides (Church & Robertson, 1966;Kühnlein, 2012;Merkey, Wong, Hoshizaki, & Gibbs, 2011). Lactate, the major glycolytic end-product, may also be important as a fuel, at least during the onset of metamorphosis. Lactate dehydrogenase (LDH) is required for both the synthesis and remobilization of lactate. It is highly expressed in larvae and is also induced by steroid signaling (Abu-Shumays & Fristrom, 1997). LDH activity declines during metamorphosis (Rechsteiner, 1970), reflecting the drop in its messenger RNA late in larval development (Graveley et al., 2011), with the accumulated lactate being mostly used up during the prepupal stage (Li et al., 2017). In larvae, glycolysis serves the needs of adenosine triphosphate (ATP) production and supplies carbon skeletons for biosynthesis via the TCA cycle (Tennessen, Baker, Lam, Evans, & Thummel, 2011), while in the pupa, triglycerides are catabolized mainly in mitochondria. Efficient mitochondrial respiration is therefore crucial at both stages.
In an earlier series of experiments, we found that flies expressing the alternative oxidase AOX from the tunicate Ciona intestinalis failed to complete development when reared on a low-nutrient agar medium containing only yeast and glucose (Saari et al., 2018). AOX branches the mitochondrial respiratory chain, bypassing complexes III and IV in a non-proton-motive reaction that oxidizes ubiquinol directly by molecular oxygen (Rogov, Sukhanova, Uralskaya, Aliverdieva, & Zvyagilskaya, 2014). The gene for AOX is present in most groups of eukaryotes, including animals (McDonald & Gospodaryov, 2018), but has been lost from specific lineages during the course of evolution, notably from vertebrates and advanced insects.
The reasons for its evolutionary loss or retention are unclear. In lower eukaryotes and plants it confers resistance against stresses or metabolic disruption resulting from overload, inhibition, or damage to the standard mitochondrial respiratory chain (Dahal, Martyn, Alber, & Vanlerberghe, 2017;Dufour, Boulay, Rincheval, & Sainsard-Chanet, 2000). Such stresses include the excess production of reactive oxygen species (ROS), limitations on ATP synthesis, restraints on metabolic flux, and disturbances to cellular redox and ionic homeostasis. AOX is believed to play a similar protective role in animals (McDonald & Gospodaryov, 2018;Saari et al., 2018).
Since similar metabolic stresses arise in humans experiencing pathological dysfunction of mitochondria, we reasoned that AOX could be developed as a potential wide-spectrum therapeutic (El-Khoury et al., 2014).
As a first step, we have established the transgenic expression of Ciona AOX in model organisms, including both Drosophila (Fernandez-Ayala et al., 2009) and the mouse (El-Khoury et al., 2013;Szibor et al., 2017), to evaluate its effects on development, physiology, and pathology. In plants, AOX is enzymatically active only under conditions where the quinone pool becomes highly reduced (Castro-Guerrero, Krab, & Moreno-Sanchez, 2004;Hoefnagel & Wiskich, 1998). Thus, it contributes negligibly to electron flow under standard physiological conditions, while being available as a stress buffer whenever required. The same appears to be so for Ciona AOX expressed in the mouse (Dogan et al., 2018). In accordance with this, the ubiquitous expression of Ciona AOX in both flies (Fernandez-Ayala et al., 2009) and mammals (Szibor et al., 2017) has almost no detectable physiological effect under nonstressed conditions. However, if the standard respiratory chain is dysfunctional, for example, due to toxic inhibition (El-Khoury et al., 2013;Fernandez-Ayala et al., 2009;Szibor et al., 2017), overload (Mills et al., 2016), or genetic damage Rajendran et al., 2018), AOX is able to compensate the resulting phenotypes to a significant degree. However, since AOX bypasses two of the proton-pumping steps of the standard respiratory chain, it is not able to fully restore ATP production. Thus, it cannot compensate null mutations or even profound knockdown of core subunits of cytochrome c oxidase (complex IV), nor mutations that abolish the synthesis of a vital prosthetic group of the enzyme (Dogan et al., 2018;Fernandez-Ayala et al., 2009;Kemppainen et al., 2014).
The developmental failure of AOX-expressing flies under nutritional stress (Saari et al., 2018) implies that transgenic AOX becomes enzymatically activated during at least part of the life-cycle and/or in some crucial tissue(s), under low-nutrient conditions. In addition, since activated AOX is known to facilitate TCA cycle reactions and decrease ROS production, but also to result in less ATP synthesis, developmental failure could result from multiple, nonexclusive causes. This might explain why the addition of either glucose or yeast to the low-nutrient medium failed to improve the developmental outcome for AOX-expressing flies.
To investigate the phenomenon further, we embarked on a study to identify the missing dietary components that would enable AOXexpressing flies to complete development, as well as to pin down more accurately the timing of developmental failure and its correlates in terms of stored nutrient accumulation.
Our findings unambiguously ascribe AOX-dependent developmental failure to the pupal stage. We also found that L3 (wanderingstage) larvae reared on the low-nutrient medium are deficient in specific stored nutrients, regardless of genotype. Together, these results imply that AOX-expressing flies are unable to make proper use of this depleted metabolic store, so as to complete development.
Finally, we identified a specific set of metabolites as the most likely candidates to rectify this deficiency, thus enabling AOX-expressing flies to fully undergo metamorphosis.

| Drosophila strains and culture
Drosophila strains used in the study were the transgenic line UAS-AOX F6 for C. intestinalis AOX (on chromosome 2), constructed inhouse and described previously (Fernandez-Ayala et al., 2009), and the standard ubiquitous driver line daGAL4 (insertion on chromosome 3). Both lines were maintained as homozygotes and crossed together as indicated in figures. In addition, RNAi lines 30282 (Vienna Drosophila Resource Centre, GD library, chromosome 3, maintained over the TM3Sb balancer) and 65175 (Bloomington, TRiP line, chromosome 2, maintained over the CyO balancer), both for ATPCL, were crossed with daGAL4 to test the effects of ATPCL knockdown, using an internal control. Flies were maintained and cultured on standard high-sugar medium (Fernandez-Ayala et al., 2009), described here as a complete medium, with 12 hr cycles of light and darkness at 25°C, except where indicated in specific experiments. Low-nutrient medium consisted of (w/v) 1% agar, 3% glucose, and 3.5% yeast (Instant SD; Algist-Bruggerman NV, Gent, Belgium), which were boiled together then cooled to 65°C before addition of standard antibiotics nipagin (to 0.1%) and propionic acid (to 0.5%), before dispensing into food vials. All other food supplements were added to the mixture during boiling, except for doxycycline which was dispensed as a stock solution onto food plugs into which it was allowed to penetrate and dry in a fume hood (final concentrations as shown in figure legends). Food supplements that were added to the low-nutrient medium in specific experiments, were as follows: sucrose (VWR, 1.5% w/v), with or without fructose The effects of different media on eclosion were tested by mating batches of 10 virgin females and five males overnight in food vials containing complete medium, then transferring them daily to fresh food vials containing the medium under test, for egg laying.
Pupae per vial and the number of eclosed adults were recorded.

| Mammalian cells and culture
Flp-In™ T-REx™ 293 cells transformed with C. intestinalis AOX and their parental cell-line were cultured as previously (Hakkaart, Dassa, Jacobs, & Rustin, 2006). To induce transgene expression 1.5 × 10 6 cells were seeded in 10 cm 2 plates and cultured for 72 hr in medium containing 1 μg/ml doxycycline (Sigma). The medium containing doxycycline was replaced after 48 hr. AOX transgene expression was verified by western blot analysis using a customized antibody, as previously (Dassa et al., 2009).

| Respirometry
Respirometry on permeabilized cells was conducted essentially as described previously (Cannino et al., 2012), with the sequential addition of substrates and inhibitors as indicated in figure legends.

| Fractionation of treacle
Treacle was fractionated into aqueous and nonaqueous components as follows. In a fume hood, 1.8 g treacle was diluted in water to 40 ml by gentle heating. After cooling, the solution was poured into a separation funnel, mixed with 40 ml of diethyl ether (Sigma), and extracted by gentle inversion approximately 10 times. After the phases had fully separated they were isolated and each extracted twice more with the opposite solvent. A further 40 ml of water was added to the ether fraction and the two fractions were then left overnight in open beakers for all ether to evaporate. Each fraction was then made up into 40 ml fly food by adding the low-nutrient medium ingredients, for testing in the eclosion assay, or else analyzed further by mass spectrometry, as described below, or used in respirometry. For further fractionation by ethanol precipitation, 10 ml of the aqueous fraction were decanted into 50 ml centrifuge tubes, and 99% ethanol added to bring the final ethanol concentration to the desired level (between 40% and 75%). After overnight precipitation at −20°C and centrifugation at 14,000g max for 30 min at 4°C, supernatants were decanted, and the ethanol evaporated, while pellets were air-dried and resuspended in 10 ml of water. Individual fractions were then made up into fly food for testing or analyzed by mass spectrometry, as described below. SAARI ET AL.

| Metabolite assays
To assay triglycerides, batches of 10 L3 (wandering-stage) larvae were homogenized in 100 μl phosphate buffered saline (PBS)-0.05% Tween (Medicago, Uppsala, Sweden), using a disposable plastic pestle. Samples were heated at 70°C for 5 min, cooled to room temperature, and vortexed. Aliquots of the homogenate (5 μl) were added to 100 μl of Triglyceride Reagent (Thermo Fisher Scientific) in transparent 96-well plates. After incubation at 37°C for 30 min, absorbance at 540 nm was measured using a plate reader (Plate Chameleon™ V; Hidex) and normalized for protein content based on the Bradford assay (Bradford Reagent; Sigma). For lactate and pyruvate assays, batches of 10 L3 larvae were homogenized similarly, in 6M guanidine hydrochloride on ice, then incubated at 95°C for 5 min. Supernatants were transferred to fresh vials and stored at −80°C, then thawed on ice and diluted 1:10 with water. Standards and reaction master mix were prepared according to the manufacturer's protocol (L-Lactate assay kit/Pyruvate assay kit; Sigma). The reactions were performed in a black 96-well microplate by mixing samples 1:5 with the reagent master mix and incubation at room temperature for 30 min. Fluorescence (excitation 535 nm and emission 590 nm) was measured at 1 s intervals using the same plate reader and normalized for protein content.

| Mass spectrometry
Ten microliter aliquots of treacle fractions were mixed vigorously with 300 µl of methanol containing internal standards (0.5 ppm d8valine, 0.5 ppm d4-succinic acid, 0.5 ppm d5-glutamic acid, 2.4 ppm heptadecanoic acid) and dried under constant nitrogen flow at room temperature. The resulting residual metabolites were converted into methoxime and trimethylsilyl derivatives by a two-step procedure.

| Statistical analysis
Student's t test (Microsoft Excel) was used to assess significance when performing pairwise comparisons, implementing the Bonferroni correction where more than two groups were compared. When comparing multiple levels of a single factor to each other, one-way analysis of variance (ANOVA) followed post hoc by the Tukey honestly significant difference (HSD) test (http://astatsa.com/) was applied. To test which of two factors was a significant determinant of a numerical outcome, two-way ANOVA (GraphPad Prism) was used.
Note that post hoc analysis (e.g., via the Tukey HSD test) is only appropriate in cases where a significant interaction between the factors is detected by ANOVA, or where more than two levels of a given significant factor are compared. χ 2 tests to compare observed and expected outcomes were performed online (https://www. graphpad.com/). Details are given in figure legends, as appropriate.  Figure S1A). AOX expression also had no effect on the proportion of eggs that developed on lownutrient medium as far as the pupal stage (Figure 1b and S1B). In contrast, the proportion of pupae that finally eclosed was dramatically decreased in AOX-expressing flies when cultured on the lownutrient medium (Figure 1c).

| Specific food additives do not compensate for the developmental defect conferred by AOX
Next, we tested whether the addition of specific vitamins, minerals or sugars was able to restore eclosion competence to AOX-expressing flies. We selected concentrations of the various additives that have either been revealed previously to impact the phenotype of flies with relevant auxotrophies, or that are deemed effective in delivery to humans suffering an equivalent deficiency.
In several cases, we tested a range of potentially effective concentrations, as well as combinations of additives. All additives and combinations of additives were found to be ineffective, including iron supplementation using two concentrations of

| Antibiotic treatment does not modify the developmental phenotype of AOX-expressing flies
Next, we tested whether the developmental failure of AOX-expressing flies on low-nutrient medium could be related to the growth of commensal bacteria. To address this issue, we tested AOX-expressing and control flies on complete and low-nutrient medium supplemented with a wide-spectrum antibiotic, doxycycline. We used two concentrations of the drug, a stringent dose (100 μg/ml) that had a significant, , GFP (UAS-GFP Stinger , control strain transgenic for nuclear-localized GFP, also GAL4-dependent), daGAL4 (ubiquitously expressing GAL4 driver). (a) A number of eggs laid in the indicated crosses, on low-nutrient medium containing 3.5% yeast and 5% glucose (means + SD of 3-6 individual vials in each case). There were no significant differences between strains or crosses (one-way ANOVA). (b) Percentage of eggs laid from the different crosses on low-nutrient medium (3.5% yeast and 5% glucose), reaching the pupal stage (means + SD of 3-6 individual vials in each case). There were no significant differences between strains or crosses (one-way ANOVA). (c) Proportion (%) of eclosing adult flies in three replicate crosses (3-6 vials per strain in each cross) on low-nutrient medium (3.5% yeast and 5% glucose). Means + SD; **Significant difference (p < .01) from all other classes (one-way ANOVA followed by Tukey's post hoc HSD test). See also Figure S1. ANOVA, analysis of variance; AOX, alternative oxidase; GFP, green fluorescent protein; HSD, honestly significant difference; SD, standard deviation SAARI ET AL.

| Specific treacle fractions contain the active components rescuing developmental failure
Since the addition of specific dietary supplements to low-nutrient medium failed to rescue the AOX-associated developmental failure, we attempted to isolate, or at least enrich for, the active material present in the complex food additives. This approach was simplest for treacle, because it contained only very low levels of particulates. The published composition of treacle (Table S1) gives only an approximate description of its biologically active ingredients, insufficient for and also comparing flies of the two different genotypes on each medium with each other (using Student's t test), revealed no significant differences. (b) Low-nutrient (3.5% yeast and 3% glucose) medium to which the indicated component from the standard high-sugar medium was added, at the concentrations indicated in Section 2. * and **Significant differences (p < .05 and .01, respectively) in eclosion frequencies of flies of each given genotype on different media (by one-way ANOVA followed by Tukey's post hoc HSD test). Pairwise comparisons of eclosion frequencies of flies of the two different genotypes on each given medium (using Bonferroni-corrected Student's t test), also revealed a significantly decreased eclosion rate in AOX-expressing flies for most individual added ingredients (for clarity, not shown on the figure): p < .001 for sucrose (or for minimal medium with no additions), p < .05 for wheat-germ, treacle, and soy flour. (c) Dose-response to different amounts of yeast or glucose added to, or subtracted from, the low-nutrient medium. Note that a single data set was used for the "3% glucose" and "3.5% yeast" conditions, as denoted by the boxes, this also being the standard composition of the low-nutrient medium used elsewhere in the study. *Statistically significant differences between different nutrient conditions within a genotype (one-way ANOVA followed post hoc by Tukey's HSD test, p < .05); # Statistically significant differences in a pairwise comparison of genotypes at each given nutrient condition (Student's t test, p < .001). (d) Dose-response to different amounts of two active ingredients from the standard medium, treacle, and maize flour (% w/v as shown), for the UAS-AOX F6 × daGAL4 cross only. * and **Significant differences between concentrations of a given nutrient (one-way ANOVA followed post hoc by Tukey's HSD test, p < .05 and .01, respectively). All flies were cultured at 25°C, except for the experiment of the panel (c), where 26°C was used, for technical reasons. ANOVA, analysis of variance; AOX, alternative oxidase; HSD, honestly significant difference; SD, standard deviation the present study. Therefore we used simple chemical fractionation to focus attention on the particular, active components. Ether extraction was used to separate treacle into aqueous and nonaqueous fractions which were then tested for their ability to support the development of AOX-expressing flies. We reproducibly found the complementing activity only in the aqueous fraction (Figure 5a). It was still active after being combined with the nonaqueous fraction, showing that the latter did not contain an inhibitor of the former. controls always studied in parallel in each case). Horizontal lines annotated with symbols denote significant differences between groups within a genotype and a given experiment (one-way ANOVA with Tukey post hoc HSD test). Symbols above individual bars represent significant differences in pairwise comparisons between genotypes, for a given additive in a given experiment (Student's t test). (b) Low-nutrient (3.5% yeast and 3% glucose) medium to which multivitamins were added at the indicated dilutions from the standard amount of one tablet per 300 ml of fly food, in combination with 3% fructose and 1.5% sucrose, for the UAS-AOX F6 × daGAL4 cross only. There were no significant differences between the groups (one-way ANOVA). (c) Low-nutrient medium plus CuSO 4 at the indicated concentrations. Within each genotype, there were no significant differences between the groups (one-way ANOVA). Symbols denote significant differences between the genotypes in pairwise comparisons, at each CuSO 4 concentration tested (Student's t test). (d) Low-nutrient medium plus DMEM. Note that preliminary trials were conducted to determine a concentration of DMEM that was completely nontoxic to control flies. As for other mixed additives that had no measurable effect, it cannot be excluded that DMEM contains both positively and negatively acting components that cancel each other out. Within each genotype, there were no significant differences between the groups (Student's t test). Symbols denote significant differences between the genotypes in pairwise comparisons, with and without DMEM (Student's t test). Statistical significance in all panels is denoted by ¤ , # , and *: p < .05, .01, and .001, respectively. ANOVA, analysis of variance; AOX, alternative oxidase; DMEM, Dulbecco's modified Eagle medium; HSD, honestly significant difference; SD, standard deviation SAARI ET AL.

| Treacle does not contain an inhibitor of AOX
To explain the effect of treacle and other complex supplements, we considered the possibility that it contained an inhibitor of AOX, thus negating the effects of AOX expression on development. To test this possibility, we conducted respirometry on permeabilized HEK293derived cells expressing Ciona AOX under the control of a doxycyclineinducible promoter (Hakkaart et al., 2006). After confirming AOX expression by western blot analysis ( Figure S3A), we conducted respirometry according to a standard protocol (see Figure 6a), in which we compared oxygen consumption before and after the inhibition of complex III with antimycin. The role of AOX in antimycin-resistant respiration was confirmed by subsequent treatment with n-propyl gallate, a specific inhibitor of alternative oxidases. The addition of the watersoluble fraction of treacle to the permeabilized cells at a concentration equivalent to that in fly food, as well as 10% and 1% of this amount, had no effect on the AOX-driven respiration (Figure 6a,b). Instead, treacle addition modestly stimulated respiration in both AOX-expressing and control cells ( Figure S3B).

| Flies reared on low-nutrient medium show decreased levels of triglycerides and lactate
The above findings strongly suggest one (or both) of two scenarios, relating nutrient storage and AOX. Either F I G U R E 4 Doxycycline does not influence the development of AOX-expressing flies. Proportion of pupae eclosing on the indicated media, of the genotypes or crosses (female × male) shown (means + SD, n ≥ 4). (a) Complete medium or low-nutrient medium with and without 100 μg/ml doxycycline. (b) Low-nutrient medium with and without supplementation by treacle and/or doxycycline at the indicated doses, alongside daGAL4 control on low-nutrient medium. (c) Low-nutrient medium with and without supplementation by maize flour and/or 100 μg/ml doxycycline. In each panel, horizontal lines annotated with asterisks (*, **, ***) denote significant differences between groups within a genotype or doxycycline concentration (one-way ANOVA with Tukey post hoc HSD test, p < .05, .01, and .001, respectively). Symbols above individual bars represent significant differences in pairwise comparisons between (a, c) genotypes or (b) doxycycline concentrations (Student's t test, ¤ and # : p < .05 and .001, respectively). ANOVA, analysis of variance; AOX, alternative oxidase; HSD, honestly significant difference; SD, standard deviation genotype, but those resources are insufficient to enable AOX pupae to complete development, due to AOX activation during metamorphosis. To test these possibilities, we assayed the relative levels of triglycerides ( Figure 7a) and lactate (Figure 7b) in wandering-stage AOX-expressing L3 larvae and controls grown on complete versus low-nutrient medium. Growth on the lownutrient medium significantly decreased the relative concentration of triglycerides (two-way ANOVA, p < .01) and lactate (two-way ANOVA; p < .001), while AOX expression had no significant effect on these metabolites. Pyruvate, being interconvertible with lactate via LDH was also decreased in larvae grown on low-nutrient medium ( Figure S4, p < .01), but not by AOX expression. These findings support the hypothesis that the low-nutrient diet restricts the accumulation of stored nutritional resources and that it is the inability to mobilize these diminished resources that causes developmental failure in AOX-expressing pupae.

| Treacle fractionation reveals a list of candidate nutrients for compensating AOX
Treacle and its various fractions that were tested earlier in the developmental assay were analyzed further by mass spectrometry, alongside standards for 48 common metabolites. Of these, the majority was detectable in treacle, and most of these were present in the aqueous fraction thereof (Table 1). However, only four were clearly enriched in the pellet after precipitation in 75% ethanol.
Since fructose, by far the most abundant of them, had already been tested in the developmental assay (Figure 3b), we proceeded to test the next most enriched, citrate, because of its metabolic role in the generation of cytosolic acetyl-CoA, the main precursor for fatty acid synthesis. However, citrate was unable to alleviate the developmental failure of AOX-expressing flies on low-nutrient medium (Figure 8), even though its conversion to acetyl-CoA by ATP citrate lyase appears to be essential for Drosophila development ( Figure S5).

| DISCUSSION
In this study, we established that the developmental failure of AOX-expressing flies on low-nutrient medium occurs specifically during the pupal stage (Figure 1). Several key metabolites regarded as nutrient stores for metamorphosis were at diminished levels in flies cultured on low-nutrient medium, regardless , Trcl 1/100, 1/10, and 1/1 (treacle aqueous fraction equivalent to 1%, 10%, and 100% of the standard concentration in fly food), Aa (60 ng/ml), and nPG (0.2 mM). Numbers opposite the trace represent oxygen consumption rate before and after treacle addition (nmol/min per 5 × 10 6 cells), extrapolated from slope. (b) Respirometry data on permeabilized cells (averaged data from two experiments), with and without induction by doxycycline as shown. Extent of antimycin resistance with and without the addition of treacle (aqueous fraction, added to equivalent level as in fly food). See also Figure S3. Aa, antimycin A; ADP, adenosine diphosphate; AOX, alternative oxidase; Dig, digitonin; nPG, n-propyl gallate; PGM, sodium pyruvate, glutamate, and malate; Succ, sodium succinate the use of stored resources during metamorphosis. This, in turn, strongly suggests the enzymatic activation of AOX during metamorphosis, at least in some tissues. This could be verified by using an enzymatically incapacitated version of the enzyme expressed at the same high level as daGAL4-driven UAS-AOX. A previously generated control line transgenic for a mutated (inactive) AOX variant would not suffice, since the transgene was expressed at a much lower level (Andjelković et al., 2015). However, the expression of neither cytosolically expressed GFP (Saari et al., 2018) nor the GAL4 transcription factor caused developmental failure in low-nutrient medium, making it unlikely that a general proteotoxic effect is responsible.
The metabolic event that putatively switches on AOX during metamorphosis remains unknown. As indicated above, the enzyme should be catalytically inert except when the quinone pool becomes highly reduced (Dogan et al., 2018). The depletion of respiratory substrates, as observed in larvae cultured on lownutrient medium, should logically lead to the opposite outcome, that is, to the respiratory-chain electron carriers becoming completely oxidized. However, respiratory chain shutdown might occur under such circumstances via the complex V inhibitor IF1 (Esparza-Moltó, Nuevo-Tapioles, & Cuezva, 2017), creating the conditions for AOX activation. AOX activation could, in principle, result from a number of metabolic consequences of low-nutrient diet that are hard to predict, given that rather little is known about the nutrients (other than triglycerides) that are stored in the pupa. A shortage of vitamins (e.g., folic acid, tocopherol) or mineral components of prosthetic groups such as heme is unlikely, given that these additives did not rescue the phenotype ( Figure 3). Although pupal development is considered to be primarily catabolic, the precise extent and stage-specificity of substrate utilization and the involvement of different catabolic pathways (glycolysis, pentose phosphate shunt, OXPHOS) are not known in full detail. The capacity of the respiratory chain appears to be crucial both at the start of pupation and in the period immediately before eclosion (Fourche, 1967). Therefore, respiratory overload could be a factor in the activation of AOX, for example, if the CoQ pool becomes over-reduced. This may combine with a low level of stored nutrients leading eventually to ATP depletion. AOX activation could also, for example, involve partial inhibition of complex IV (Srinivasan & Avadhani, 2012) or damage to complexes I, II, and/or III (Mena, Urrutia, Lourido, Carrasco, & Núñez, 2015), resulting from oxidative or nitrosative stress, or to a relaxation of constraints on complex I that are seen in high-glucose conditions (Cannino et al., 2012).
Since the decreased larval accumulation of triglycerides and lactate/pyruvate on low-nutrient medium does not appear to pose any problem for wild-type flies and is not affected by AOX expression, it is reasonable to assume that it does not trigger a starvation response impacting ecdysteroid and insulin-like growth factor signaling. In late larval development, the fly passes through a series of checkpoints (see Tracy & Baehrecke, 2013), the most important of which is the critical weight checkpoint (Nijhout, 2003). Once passed, further growth is curtailed except under high-nutrient conditions , under the control of the insulin receptor (InR) responding to the level of a crucial insulin-like peptide, dilp8 (Garelli, Gontijo, Miguela, Caparros, & Dominguez, 2012). Ecdysteroid counteracts this signal by inducing, instead, the autophagic events that occur at the onset of pupariation (see Tracy & Baehrecke, 2013).
Assuming this system works normally in larvae cultured under low-nutrient conditions, the hormonal signals that initiate pupariation should not be altered due to AOX expression, which had no significant effect on nutrient accumulation. Note also that because AOX-expressing pupae die at different times during pupal stage (Saari et al., 2018) an interference with hormonal signaling is unlikely to be an explanation for the developmental failure.
In addition to its hormonal induction at the onset of metamorphosis, autophagy occurs as a response to starvation.
In larvae, this is triggered via the target of rapamycin (TOR) (a) (b) F I G U R E 7 Triglyceride and lactate content of L3 larvae grown on different media. Relative amounts of (a) triglycerides and (b) lactate in L3 wandering-stage larvae from the indicated crosses, after normalization to values from control larvae (transgenic recipient strain w 1118 , cultured on complete medium); means + SD of ≥ 6 batches of 10 larvae of each genotype and culture medium. Two-way ANOVA ( Figure S4B, all controls vs. AOX-expressing larvae, i.e., UAS-AOX F6 × daGAL4 cross) found no significant differences based on genotype, but significance for diet as determinant of both lactate (p < .001) and triglyceride (p < .01) levels, as well as for pyruvate (see Figure S4A). ANOVA, analysis of variance; AOX, alternative oxidase; AU, arbitrary units; SD, standard deviation  Continues) pathway, driven by an insufficient supply of amino acids (Scott, Schuldiner, & Neufeld, 2004). Conceivably, this may accelerate developmental progression, exacerbating the vulnerability of AOX-expressing pupae to subsequent developmental failure.
However, such a scenario appears unlikely, since dietary supplements rich in amino acids did not alleviate the phenotype. Finally, a more prosaic explanation of AOX activation in pupae would invoke stage-specific activity of the daGAL4 driver, although this has not been explored experimentally.
AOX limits excess ROS production (Cannino et al., 2012;Sanz, Fernández-Ayala, Stefanatos, & Jacobs, 2010) and attenuates responses to ROS (Dogan et al., 2018;El-Khoury et al., 2016;Hakkaart et al., 2006 Compounds listed in the following order: first those compounds enriched in the 75% EtOH fraction in order of their enrichment factor (in cases where the enrichment factor is similar, listing is in order of their abundance in unfractionated treacle); next (in alphabetical order) those compounds detected in unfractionated treacle, but which were not detected in the aqueous fractions, and finally (again in alphabetical order) those compounds which were not reproducibly detected in unfractionated treacle, but which were found in one or more specific fractions. Valine was at such a low level that it was reliably detected in only one aqueous fraction.
F I G U R E 8 Citrate addition does not rescue the developmental failure of AOX-expressing flies. Proportion of pupae of the genotype or cross (female × male) shown, eclosing on low-nutrient medium supplemented with the indicated levels of citrate (means + SD, n ≥ 4). Horizontal lines annotated with asterisks (* and **) denote significant differences between groups within a genotype (one-way ANOVA with Tukey post hoc HSD test, p < .05 and .01, respectively). Symbols "#" above individual bars represent significant differences in pairwise comparisons between genotypes for a given citrate concentration (Student's t test, p < .001). Note that this experiment was conducted at 26°C for technical reasons, also giving a more robust phenotype (Saari et al., 2018). ANOVA, analysis of variance; AOX, alternative oxidase; HSD, honestly significant difference; SD, standard deviation SAARI ET AL.

| 353
antioxidants. However, before this hypothesis can be discarded, the metabolic changes produced by low nutrient diet will need to be profiled in detail, including those impacting ROS.

| Identity of the crucial nutrients required for the completion of development by AOX-expressing flies
Although we were able to rule out from consideration a large number of nutrients acting individually, as well as several when added together, it remains plausible that some of those already tested may act in combination to restore full developmental potential to AOXexpressing flies. A role for a combination of nutrients is also suggested by the observation that, at low ethanol percentages (40% or 60%), full activity was not recovered in either the pellet or supernatant fractions (Figure 5b). Based on the mass spectrometry findings, the active fraction of treacle is enriched in fructose, citrate, ascorbate, and the ketone body 3-OH-butanoic acid. However, this "short-list" does not include metabolites for which standards were not included, nor those that were not identifiable from the mass spectrogram. Furthermore, although these compounds were the ones that were enriched in the 75% ethanol pellet, it is entirely possible that whichever of them is the crucial one(s) is only active in combination with some other compound also present in the 75% ethanol pellet, if not actually enriched there. The candidate metabolites listed in Table 1 include other TCA cycle intermediates, among them succinate, which AOX may deplete (Mills et al., 2016), several amino acids and some glycolytic intermediates, notably lactate, which we found to be systematically decreased in L3 larvae cultured on the low-nutrient medium. Almost all of these are strong candidates to supplement those metabolites that would normally be supplied cataplerotically, thereby ensuring that a sufficient level of stored nutrients is available throughout metamorphosis, enabling AOX-expressing pupae to complete development. The importance of cataplerotic pathways is illustrated by the developmental requirement for ATPCL ( Figure S5), which applies even in wild-type flies grown on standard high-sugar medium. ATPCL uses citrate, shuttled into the cytosol, to synthesize acetyl-CoA, the raw material for fatty acid, and hence triglyceride production. Given the many possible permutations of compounds and their concentrations that might be crucial in replacing such cataplerotic pathways, the development of a high-throughput approach is clearly required, to narrow down further the identity of the active substance(s).
Importantly, the effect of the missing nutrients need not be solely on the actual synthesis of triglycerides and other stored metabolites for subsequent catabolism. Some of them are known to be involved in the generation of epigenetic signals, which may program metabolism during metamorphosis according to the prevailing conditions. For example, specific catabolic enzymes may be required in higher amounts under conditions of dietary stress, enabling the pupa to survive on a lower repository of stored triglycerides. In combination with partially activated AOX, this could lead to a catastrophic depletion of resources. For example, the list of detected compounds in the active treacle fraction included succinate and several other TCA cycle intermediates that could modulate the activity of alphaketoglutarate-dependent histone demethylases (Berry & Janknecht, 2013), or promote histone methylation via fumarate and ROS signaling (Sullivan et al., 2013).

| Implications for use of AOX in therapeutic and biotech applications
Mice ubiquitously expressing C. intestinalis AOX showed no deleterious phenotypes, even under the moderate stress of treadmill exercise (Szibor et al., 2017). Metabolic profiles, at least in the heart, were not distinguishable from wild-type littermates, and there was no evidence pointing to enzymatic activation of the enzyme in vivo. AOXexpressing flies also move and behave normally (Andjelković et al., 2015;Fernandez-Ayala et al., 2009). However, possible activation of the enzyme at specific developmental stages or under nutritional stress has not been studied in the mammalian model. The mice were always fed ad libitum on standard chow diet, which may be considered equivalent to the rearing of flies on standard high-sugar medium, where AOX expression also had no impact on development. To conduct a parallel study on the mammalian model would require the mice to be fed a specific diet with restricted calorie intake and/or nutritional composition. In at least one case where we attempted to rescue the phenotypic consequences of respiratory chain deficiency in the mouse (Cox15 knockout in skeletal muscle), AOX expression exacerbated rather than alleviated the phenotype (Dogan et al., 2018).
In this case, interference with ROS-based stress signaling was inferred, which may apply also to the present case of nutritional stress in Drosophila. Clearly, before AOX is to be developed as a therapeutic, the conditions for, and consequences of, its activation need to be evaluated in detail. However, if judiciously exploited, such properties could also be beneficial, for example in allowing better regulation of the enzyme in a therapeutic setting, or even using AOX to improve adaptation to nutritional stresses.

| Implications for evolution and animal development
As discussed in Section 1, the life cycle of holometabolous insects provides a "best of all worlds" scenario, in which maximum advantage can be taken of food resources for growth and dispersal to maintain genetic fitness and adaptability, while minimizing vulnerability to predation.
In an insect species that would be naturally endowed with AOX, there would be strong selective pressure to evolve adjustments to the ways in which the onset of metamorphosis is regulated, so as to take account of potentially decreased bioenergetic efficiency during the pupal stage. This pressure might also favor the evolutionary loss of AOX itself if this inherent vulnerability to starvation outweighs the stress-resistance AOX may confer. In practice, the natural environment is likely to be much more similar to what we here describe as a low-nutrient medium than the complete medium on which we and almost all other Drosophila researchers culture flies in the laboratory.
Since AOX has indeed been lost during insect evolution, our findings could offer one possible explanation.

ACKNOWLEDGMENTS
We thank Eveliina Teeri for technical assistance and Andreas Bergmann