Chronic sleep loss sensitizes Drosophila melanogaster to nitrogen stress

SUMMARY Chronic sleep loss profoundly impacts metabolic health and shortens lifespan, but studies of the mechanisms involved have focused largely on acute sleep deprivation. 1,2 To identify metabolic consequences of chronically reduced sleep, we conducted unbiased metabolomics on heads of three adult Drosophila short-sleeping mutants with very different mechanisms of sleep loss: fumin ( fmn ), redeye ( rye ), and sleepless ( sss ). 3–7 Common features included elevated ornithine and polyamines, with lipid, acyl-carnitine, and TCA cycle changes suggesting mitochondrial dysfunction. Studies of excretion demonstrate inefﬁcient nitrogen elimination in adult sleep mutants, likely contributing to their polyamine accumulation. Increasing levels of polyamines, particularly putrescine, promote sleep in control ﬂies but poison sleep mutants. This parallels the broadly enhanced toxicity of high dietary nitrogen load from protein in chronically sleep-restricted Drosophila , including both sleep mutants and ﬂies with hyper-activated wake-promoting neurons. Together, our results implicate nitrogen stress as a novel mechanism linking chronic sleep loss to adverse health out-comes—and perhaps for linking food and sleep homeostasis at the cellular level in healthy organisms. RESULTS


RESULTS
Sleep mutant, but not acutely sleep-deprived, fly heads have remodeled nitrogen metabolism We conducted metabolomics on $10-day-old iso31 control and sleep mutant fmn, rye, and sss fly heads. Thirty metabolites were commonly regulated across sleep mutants, which we attribute to chronic sleep loss ( Figure 1A). Seven ($23%) of these are primarily linked to nitrogen metabolism or polyamine synthesis: ornithine, acetylarginine, a-acetylornithine, d-acetylornithine, acetylisoputreanine, diacetylspermidine, and pipecolate (Figure 1B). The polyamines putrescine and spermidine were also elevated in rye and sss, and putrescine trended high in fmn (p = 0.0535) ( Figure S1A). Other commonly regulated metabolites, such as taurine, sarcosine, and guanine, have links to nitrogen metabolism but also other pathways ( Figure S1B). 8,9 We then asked whether acute sleep deprivation (SD) similarly alters nitrogen metabolism. Targeted metabolomics in iso31 heads at morning zeitgeber time (ZT)2 (control) compared with evening ZT14, ZT2 after overnight SD (ZT2 SD), and ZT2 after overnight SD with morning rebound sleep (ZT2 SR), showed minimal effects on nitrogen metabolism. Ornithine alone trended higher at ZT2 SR ( Figure S1C). Thus, nitrogen metabolome remodeling in sleep mutants requires chronic sleep loss.
Mitochondrial, lipid, and other metabolome changes in heads of Drosophila sleep mutants Five metabolites commonly regulated across sleep mutants were carnitine or acyl-carnitines, suggesting defects in b-oxidation ( Figure 1C). Many other acyl-carnitines were elevated idiosyncratically in particular sleep mutants (Data S1). Elevated 2-methylcitrate and methylmalonate, and lower carboxymethyllysine and aconitate, suggest mitochondrial stress across sleep mutants ( Figure 1C). [10][11][12][13] In rye and sss, mitochondrial defects likely contribute to lipid loss that skews remaining lipid ratios; fmn has similar, weaker trends ( Figures 1D and 1E). Cholesteryl esters are also down in rye and sss, with a similar trend in fmn (Data S1). Other commonly regulated metabolites included smaller clusters of threonine and erythrosine derivatives ( Figure S1B).
Chronic, but not acute, sleep restriction reduces efficiency of nitrogen excretion Polyamine synthesis requires considerable nitrogen, 9 implying sleep mutants are nitrogen stressed. To test this, we assayed total protein, urate, urea, and NH 4 in $10-day-old whole flies at dawn and dusk. Urea was unquantifiable, while total protein, NH 4 , and urate showed either no effect, or idiosyncratic effects in individual sleep mutants ( Figures S2A-S2F). We next assayed hemolymph total protein and NH 4 in sleep mutant and control flies at dawn and dusk to assess circulating nitrogen stress. Both were markedly elevated in rye; total protein was modestly elevated and NH 4 trended up (p = 0.06) in fmn (Figures S2G and S2H). Although sss lacked these differences (p > 0.53), their elevated whole-fly and head urate likely indicates that they accumulate nitrogen stress in a different form ( Figure S2; Data S1; discussion). These results encouraged us to examine sleep mutant nitrogen excretion.
Because excretion is an active behavior, we assayed NH 4 and urate excretion at the beginning and end of the day/waking phase, and late at night in $10-day-old flies. Both metabolites showed a genotype main effect (p < 0.001), driven by consistently decreased excretion in all sleep mutants, albeit with some mutant-specific variation in time(s) of day driving the effect (Figures 2A and 2B). To determine whether this reflected constipation, we pre-fed FCF brilliant-blue-laced food for 24 h and measured dye excretion ( Figure 2C). A genotype main effect (p < 0.0001) was driven by increased excretion in rye and sss, across all times for sss and mostly in the morning for rye (Figure 2C). fmn excretion volume was similar to control, but this likely reflects depletion of gut contents during the assay; excrement deposition is increased in fmn pre-feeding vials ( Figure S2I). Decreased nitrogen metabolites in increased-to-unchanged excrement volume demonstrates inefficient nitrogen excretion in all adult sleep mutants tested. All metabolomic data are from iso31 (gray), fmn (orange), rye (purple), and sss (red) pools of $10day-old, mixed-sex, mated fly heads collected at $ZT6. All statistical comparisons shown are colormatched sleep mutant versus iso31 control. (A) Venn diagram of the number of metabolites consistently up-or down-regulated in one or more sleep mutants compared with the iso31 control.
(B and C) Line graphs of scaled metabolite levels, grouped by involvement in nitrogen metabolism (B) or mitochondrial oxidative metabolism (C). Data shown are individual pools of lysate; n = 5; Welch's t tests (p values) with FDR correction for multiple comparisons (q values); +p < 0.05 but q > 0.05, *p/q < 0.05, **p/q < 0.01, ***p/q < 0.001, ****p/q < 0.0001. (D) XY graph of total lipid content. Data shown are individual pools of lysate overlaid with median ± interquartiles; n = 5; Dunnett test; NS, not significant; ****p < 0.0001. (E) Pie graphs showing % of major lipid families in the head lipidome. Changes in sleep mutant lipid composition are largely driven by lost lipids (D), mostly triacylglycerols, especially in rye and sss. Related to Figure S1 and Data S1.
We next repeated these studies $1 day after eclosion, when sleep mutants are short-sleeping but have experienced less lifetime sleep loss. 14 No young sleep mutant had deficient NH 4 or urate excretion, while a genotype main effect on dye excretion (p < 0.0001) was driven by decreased excretion volumes in fmn and rye ( Figure 2F). These findings suggest that loss of nitrogen excretion efficiency requires chronic sleep loss.
The lack of an effect on nitrogen excretion might explain the failure of overnight SD to elevate head polyamines, in our study and others ( Figure S1). 2 We tested this by measuring NH 4 , urate, and blue dye excretion in $10day-old iso31 flies after 12 h overnight mechanical SD (mechSD) versus unshaken controls. We found increased NH 4 and blue dye and up-trending urate (p = 0.08) after mechSD ( Figures 2G-2I). The increased excretion volume suggests that nitrogen excretion efficiency is not necessarily increased by mechSD, but rather that nitrogen excretion efficiency remains largely intact with acute sleep loss.
We next sought to compare excretion after acute and chronic sleep restriction in parallel, independently of sleep-altering mutations. As flies adapt to extended mechanical shaking, we restricted sleep by activating wake-promoting neurons using 60D04-Gal4>TrpA1 (D>Trp) and 11H05-Gal4>TrpA1 (H>Trp) lines. This sleep loss is heat-gated and resists homeostatic rebound, allowing acute and chronic sleep restriction. 15,16 Excretion assays after 1-and 10-day sleep restriction at 29 C were inconclusive, with mostly negative results that may reflect no effect of sleep restriction, perhaps due to a lack of sleep-need buildup, 16 or may instead stem from confounding effects of duration at 29 C on excretion in genetic controls ( Figures S2J-S2O). Urate excretion in D>Trp-the one case lacking time at 29 C confounds in genetic controls-was elevated at 1 day of 29 C but not 10 days at 29 C, driven by decreased urate excretion only in D>Trp ( Figure S2K). This supports our overall results suggesting the impairment of nitrogen excretion by chronic sleep restriction.

Blocking terminal polyamine synthesis increases sleep in Drosophila melanogaster
Because sleep loss promotes somnogen accumulation, we conducted RNAi screens to test how polyamine metabolism and linked pathways regulate sleep in Drosophila. Drug-inducible geneswitch (GS) drivers allowed adult-specific knockdown. Screening whole-fly knockdown with actinGS>dicer yielded hits for argininosuccinate lyase (asl), spermidine synthase (spds), and spermine synthase (sms) (Figures 2J and 3A; Data S2). We also screened pan-neuronal knockdown with nsybGS>dicer, but no hits validated ( Figure S3A; Data S2). Neither screen recapitulated sleep gain previously reported with constitutive pan-neuronal knockdown of oat (Data S2). 17 Possible explanations include knockdown timing and strain effects.
Because single-beam confounds can mask sleep-gain phenotypes, we re-tested the four spds and sms alleles above and conducted all subsequent sleep experiments on multi-beam sleep monitors. Adult whole-fly asl RNAi#1 decreased total and day sleep, and fragmented sleep (Figures 3B and 3E; Data S2). Adult whole-fly spds RNAi#1 increased total and night sleep, and consolidated sleep, in both sleep assays ( Figures 3C, 3F, and 3H; Data S2). Adult whole-fly sms RNAi#4 increased and consolidated sleep, and decreased latency to sleep at ZT12, in both sleep assays ( Figures 3D, 3G, and 3I; Data S2). Consistent with its more robust knockdown, on multi-beam monitors, spds-RNAi#5 reproduced all sleep phenotypes of spds-RNAi#1, increased day sleep, and decreased sleep latency ( Figure 3J; Data S2). Consistent with its weaker knockdown, sms-RNAi#5 did not replicate most sleep phenotypes of sms-RNAi#4 on multi-beam monitors ( Figure S3I; Data S2).
In sum, even a relatively modest loss of spds in adulthood increases sleep. Loss of sms in adulthood, if near-complete, also increases sleep. Further, our data are inconclusive on whether asl loss in adulthood decreases sleep. Together, our data show that blocking terminal polyamine synthesis (i.e., putrescine conversion into spermidine or spermine) is sleep promoting ( Figure 2J).
Putrescine promotes sleep in control flies, but polyamines are toxic to sleep mutants We next tested sleep in iso31 flies on food laced with vehicle or 16 mM L-ornithine, putrescine, or spermidine. Putrescine The same datasets were used to compute panels shown here and corresponding auxiliary sleep and activity metrics in Figure S3 and Data S2. For (B)-(Q), all line graphs show averaged sleep behavior over time and all dot graphs show individual fly values overlaid with median ± interquartiles. For all panels, NS, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Related to Figure S3 and Data S2. increased total sleep, driven primarily by increased and consolidated day sleep, with decreased ZT0 latency to sleep (Figures 3L-3O and S3O-S3R; Data S2). Females may also sleep more at night with shortened ZT12 latency on putrescine, though a small decrease in the night AI leaves it unclear as to whether this reflects nocturnal lethargy ( Figure S3P; Data S2). Regardless, most female and all male sleep gain came from day sleep, and neither sex has decreased day AI (p > 0.8) (Data S2). Together, this indicates that putrescine is a novel somnogen. This view is reinforced by the weak, day-sleep-promoting effects of spermidine and ornithine, which can convert to putrescine either directly (from ornithine) or via acetylated intermediates ( Figure 2J). Spermidine increased day sleep in males and decreased ZT0 sleep latency in both sexes, with no AI changes (Figures 3L-3O and S3O-S3R; Data S2). Surprisingly, 16 mM ornithine did not affect sleep in either sex (Figures 3L-3O and S3O-S3R; Data S2). 17 Re-testing at the maximal 50 mM dose, in females, we replicated shorter ZT0 sleep latency but observed only a modest uptrend in total sleep (p = 0.0579), driven by increased and consolidated day sleep with no day AI change (Figures 3P, 3Q, and S3S-S3V; Data S2). Together, we observe much weaker, but directionally consistent, effects of ornithine on mated female sleep than previously reported. 17 Position confounds on single-beam monitors may partially explain the weaker phenotype; we observe dose-dependent chemo-repulsion on ornithine, driving increased time spent by females at the tube midpoint at 16 mM ( Figures S3W-S3Z). Strain differences may also contribute.
The strong sleep-promoting effects of putrescine; weaker, sexually dimorphic sleep-promoting effects of spermidine and high-dose ornithine; and sleep-promoting effects of adult spds and sms knockdown, all point to putrescine as the primary polyamine somnogen ( Figure 2J). This model also accommodates other results, including possible sleep decrease with adult asl knockdown and previously reported sleep gain with constitutive neuronal oat knockdown. 17 We next tested whether polyamines could rescue sleep in sleep mutants. However, this proved impossible-sleep mutants were markedly locomotor impaired and rapidly killed by 16 mM polyamine food. This prompted us to conduct an acute 6-day survival study of sleep mutant and control flies on single-beam sleep monitors, comparing within-genotype toxicity on our sleep-study metabolites. Spermidine reduced survival (Surv) only in fmn and sss females, and putrescine decreased survival severely in female sleep mutants, but only modestly in iso31 ( Figures 4A-4D). In males, putrescine reduced survival in all three sleep mutants but not iso31, while spermidine reduced survival only in sss ( Figures 4E-4H). Ornithine had no effect ( Figures 4A-4H). Polyamine effects on hazard ratio (HR) mirrored survival.

Chronic sleep restriction renders flies sensitive to dietary nitrogen
Polyamine toxicity is consistent with dietary nitrogen sensitivity. Indeed, in previous reports fmn flies had normal lifespans on standard food but short lifespans on a nitrogen-rich, high-calorie diet. 4,18 However, sss (and chronic short sleep generally) are associated with a short lifespan. 1,6,15 In our hands, all three sleep mutants (including fmn) have short lifespans on standard food ( Figures S4A and S4B). 4 Thus, general ill-health might explain sleep mutant sensitivity to polyamines. To assess specific sensitivity to nitrogen stress, we tested iso31 and sleep mutant lifespans on (1) high-protein, (2) high-sugar, and (3) all-sugar diets. In wild-type Drosophila, nitrogen-rich high-protein and nitrogenstarvation all-sugar diets both shorten lifespan relative to a highsugar diet with low protein. 19,20 Although all-sugar narrowed survival differences between sleep mutants and iso31, sleep mutants remained short-lived on all three diets ( Figures S4C-S4H). However, within-genotype diet comparisons suggested nitrogen sensitivity in sleep mutants. In iso31, lifespan and hazard were generally high-sugar > highprotein > all-sugar ( Figures 4I and 4M). Sleep mutants performed better on all-sugar versus high-protein, with (1) a narrowed ''penalty'' on all-sugar in fmn, (2) no lifespan and blunted hazard differences in rye, and (3) longer lifespan/lower hazard on all-sugar in sss ( Figures 4I-4P). The lifespan/hazard penalty on all-sugar versus high-sugar was also blunted or lost across sleep mutants, especially males ( Figures 4I-4P). However, high-protein versus high-sugar effects varied ( Figures 4I-4P).
Differential feeding could impact lifespan directly, and protein both promotes post-prandial sleep and reduces sleep depth, which could impact sleep restriction. [21][22][23] To test whether either factor contributes to our complex lifespan results, we simultaneously measured awake time by food and sleep on multibeam sleep monitors (Data S3). This approach avoids climbing proficiency and aversive shock confounds inherent in alternative methods, such as Caf e and FLIC assays, 24 though chemotaxis is a potential confound. Sleep mutants spend more awake time near food than iso31 on all diets, suggesting that overfeeding may contribute to their diet-independent, baseline lifespan reduction ( Figures S4A-S4J). Importantly, diet-independent overfeeding is consistent with increased excretion volume from adult sleep mutants ( Figures 2C and S2I), corroborating the For all panels, asterisk color codes which group in each comparison had a longer lifespan by Wilcoxon analysis (Surv) or a lower hazard by likelihood tests of Cox proportional hazard ratios (HRs). Bonferroni-corrected significance threshold for all panels is p = 0.0167. *p < 0.0167, **p < 0.01, ***p < 0.001, ****p < 0.0001, NS, not significant. Related to Figure S4 and Data S3. idea that awake time by food grossly tracks feeding behavior in our flies. Sleep mutants are short-sleeping on all diets, and presumptive overfeeding is driven primarily by increased awake time (Data S3).
We next compared within-genotype awake time by food on different diets. iso31 dwelled at high-protein > high-sugar > allsugar. rye and sss also dwelled at high protein > both other diets, while fmn dwelled similarly at all diets ( Figure S4I). Thus, rye and sss mutants with fewer ''iso-like'' diet effects on lifespan had more iso-like diet effects on presumptive feeding time, while fmn had the most iso-like effects of diet on lifespan and the least iso-like effects of diet on presumptive feeding time ( Figures 4I-4P and S4I). This misalignment suggests that feeding differences do not account for the differential effects of diet on the longevity of sleep mutants versus iso31.
iso31 also slept less on high-protein and high-sugar versus allsugar diets, fmn slept equivalently on all 3 diets, rye slept more on high-protein versus all-sugar, and sss slept more on high-protein versus both other diets (Data S3). Diet-induced sleep increases generally do not coincide with AI decreases (Data S3). Relative sleep gains in sleep mutants from dietary nitrogen are unlikely to enhance nitrogen toxicity and may instead mitigate it, contributing to the complexity of our sleep mutant longevity results.
Genotype-specific metabolic differences in sleep mutants, including uricotelic pathways that couple sugar intake to nitrogen metabolism, may also contribute to this complexity (Figures 1, 2, S1, and S2; Data S1). 25 To side-step this, we next tested lifespans on the same diets at 29 C in chronically sleeprestricted D>Trp and H>Trp flies. In DGal4, HGal4, and TrpA1 genetic controls, lifespan on high-sugar > high-protein > allsugar, much like iso31 at 25 C ( Figures 4Q, 4R, 4T, 4V, 4W, and 4Y). In contrast, D>Trp flies, while the longest-lived on high-sugar, had equivalent lifespan and hazard on high-protein and all-sugar ( Figures 4S and 4X). H>Trp flies showed even more robust effects; both sexes lived longer on all-sugar than high-protein, and males had a comparable lifespan and hazard on all-sugar versus high-sugar ( Figures 4U and 4Z). The all-sugar diet also rescued lifespan and hazard compared with one or both genetic controls in D>Trp males and all H>Trp flies ( Figures S4J-S4U). These results are consistent with enhanced toxicity of dietary nitrogen and reduced toxicity-or outright protective effects-of nitrogen starvation during chronic sleep loss.
These results are also not attributable to feeding or sleep. Gal4 control, D>Trp, and H>Trp groups all spent awake time at highprotein > high-sugar > all-sugar; TrpA1 controls dwelled similarly at all-sugar and high-protein, but TrpA1 lifespan on high-protein was > all-sugar, like Gal4 controls (Figures S4V and S4W). Like sleep mutants, D>Trp and H>Trp sleep-restricted flies dwelled longer at all diets versus control genotypes, suggesting that fasting does not drive all-sugar rescue of sleep-restricted lifespan and hazard (Figures S4J-S4W). D>Trp and H>Trp flies were also sleep-restricted at 29 C on all diets, with no within-genotype diet effects on total sleep (Data S3).

DISCUSSION
In this study, we sought metabolic changes common across sleep mutants as candidate effectors of chronic sleep loss on health. Lipid and mitochondrial dysregulation likely contribute to sleep mutant ill health, given redox factors that link sleep to lifespan. 15,26 But lipid metabolism is also widely reported to be affected by acute SD across species. 2,27-31 High methylcitrate and low aconitate after acute SD have also been reported, as have effects of lipid metabolism on sleep. 27,32-34 Thus, we pursued our findings of altered nitrogen metabolism.
Consistent with reports of its elevation during or after acute SD, 35,36 ornithine alone trended up during recovery sleep after mechSD ( Figure S1) and may be a leading indicator of nitrogen stress during acute sleep loss. To our knowledge, our finding that polyamines accumulate in sleep mutant heads is the first report of this chronic sleep loss effect in the tissues of any organism. 2 However, indirect evidence suggests this extends to humans. Sleep apnea patients excrete elevated ornithine and polyamines 37,38 ; we found no similar reports with acute SD. 2 Further, acute SD increases putrescine in hyperammonemic but not control rat brain dialysate, consistent with accumulated nitrogen stress being required for polyamine elevation. 39 Terrestrial insects generally lack Otc for urea cycling and excrete mostly uricotelic metabolites and raw NH 4 , suggesting that sleep mutants with inefficient nitrogen excretion should be hyperammonemic 8 ( Figure 2J). Consistent with this, hemolymph total protein and NH 4 tend to be high in adult fmn and rye, but puzzlingly not in sss (Figures S2G and S2H). This likely reflects a P-element in the sss mutant that provides the only functional white cassette in our sleep mutant studies; a white homolog regulates urate metabolism in silkworm. 40 Accordingly, sss alone among our sleep mutant and iso31 control genotypes has very high urate (Figures S2C and S2F; Data S1), which causes nitrogen stress in excess. 41 Thus, all three sleep mutants build up nitrogen stress in some form. Polyamine accumulation may buffer this, soaking up nitrogen equivalents, which would reduce but not eliminate nitrogen stress. In fact, polyamines switch from being protective to themselves driving nitrogen stress in excess, especially putrescine and acetylated polyamines most enriched in sleep mutant heads (Figures 1 and S1). 9 The role of sleep in nitrogen homeostasis may explain the putrescine soporific function demonstrated by our supplementation and RNAi studies ( Figures 2J, 3, and S3). Polyamine levels are fine-tuned in cells due to their myriad critical roles, including those in sleep-relevant pathways such as redox balance and autophagy. 1,9,15,26,42 Homeostasis is achieved by elaborate and well-conserved synthetic, trafficking, and degradation mechanisms. 9 Putrescine's somnogenic role may complement these cellular mechanisms, dialing-up sleep to help bring polyamine levels down if systemic putrescine rises too high. Sleep remains low in short-sleeping mutants despite high polyamines because homeostatic effectors are impaired.
Finally, we report a novel interaction between chronic sleep loss and diet that regulates longevity (Figure 4). High-protein and all-sugar toxicity relative to high-sugar is well known for wild-type flies. 19,20 We show that sleep-restricted flies live longer on all-sugar than high-protein compared with control flies, suggesting sensitivity to dietary nitrogen ( Figures 2J and 4). This was particularly prominent with thermogenetic sleep loss in D>Gal4 and H>Gal4 flies, which showed outright protective effects of all-sugar diet ( Figure S4). Sleep-restricted males also appeared somewhat more nitrogen-sensitive than females, perhaps because egg-laying offloads nitrogen.

OPEN ACCESS
Our findings have several implications for health. Pathologically, redox balance couples sleep loss to lifespan. 15,26 Nitrogen stress is oxidizing, with even generally anti-oxidant species like polyamines and urate becoming oxidizing in excess. 9,41 Likely, nitrogen stress drives a subset of metabolic onramps converging on systemic oxidation and short lifespan during chronic sleep loss. Nitrogen stress may also be relevant for sleep-loss-associated disease. Kidney disease is associated with chronic sleep loss in humans, 43,44 and elevated putrescine and polyamine degradation products contribute to kidney failure. 45 Chronic sleep loss is also associated with Alzheimer's disease 46 ; animal models suggest that nitrogen stress precedes cognitive decline 47 and drives beta-amyloid pathology via polyamine synthesis, 48 while human Alzheimer's brains compensate by remodeling nitrogen metabolism away from polyamine synthesis. 48,49 Finally, our work has teleological implications for behavior observed in healthy humans, such as increasing sugar intake when sleep-restricted. 50,51 Going for the cookies instead of protein-rich food when tired may constitute surprisingly adaptive interactions of the food and sleep homeostasis systems, avoiding nitrogen intake during a period of heightened sensitivity.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

AUTHOR CONTRIBUTIONS
J.L.B. and A.S. conceived the study; J.L.B., A.K., and P.P. developed and validated methodology as needed and performed and analyzed most experiments; J.L.B. and A.K. collected material for all metabolomics studies; D.M. developed and validated methodology for our acute sleep loss metabolomics study and performed and analyzed the study; C.T.H. developed and validated novel MATLAB scripts used for awake time by food analysis; and J.L.B., A.W., and A.S. supervised other authors and secured funding for this work. J.L.B. and A.S. wrote and revised the manuscript. All authors contributed intellectually to this work.

DECLARATION OF INTERESTS
The authors declare no competing interests.

Materials Availability
This study entirely used previously published and/or publicly available fly lines.
Data and Code Availability d Our HD4 global metabolomics dataset is deposited on Metabolights at https://www.ebi.ac.uk/metabolights/MTBLS3318. d Previously unreported scripts for time-at-position (multibeamPositionAnalysis_minutesPerDay.m) and awake time-atposition analysis (posWhileAwakeToExcel.m) were deposited to the repository on GitHub (https://github.com/cthsu86/ damSleepConverter). Other code used has been previously reported. 57 d Other data and additional information required to reanalyze data reported in this paper is available from the lead contact upon request.

Drosophila melanogaster
Heavily used fly lines in the manuscript include iso31 control; sleep mutants fmn, rye, and sss on an iso31 background (minimum 5X generations); and geneswitch/dicer lines that were well-established in the lab prior to this study. 42 Figures 2D-2F, many of which were likely virginal given their youth at the time of collection. Sex and age varied by experiment, as specified in our method details (next) and figure legends.

METHOD DETAILS
Metabolon global metabolomics and lipidomics Ten total pools of $200-250 heads from mixed sex flies aged $1-2 weeks post-eclosion ($10 days on average) were collected for each genotype, split evenly between HD4 global metabolomics and CLP lipidomics assays. Sexes were pooled to mitigate the large numbers required for the study. Collections were done at $ZT6; mid-day timepoint was chosen to enrich for metabolites dysregulated by chronic, as opposed to acute, sleep loss. Heads were collected by vortexing whole flies snap frozen on dry ice and separating heads from bodies by size on dry ice-cooled grates. Samples were stored at -80C and shipped to Metabolon on dry ice. Sample preparation, control procedures, and analysis were carried out at Metabolon Inc as described elsewhere. [58][59][60][61][62] Both HD4 and CLP procedures are briefly outlined below.

HD4 Global Metabolomics
Samples extracted and spiked with recovery standards using a MicroSTAR System (Hamilton Company) and methanol-precipitated under vigorous shaking Genogrinder 2000 (Glen Mills). Samples were fractionated, dried, resuspended in appropriate solvents, and analyzed using four distinct modes on a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35,000 mass resolution. Metabolites were identified by the Laboratory Information Management System, an automated system that identified ion features in our head lysate samples using a reference library of known metabolites defined by retention time, molecular weight (m/z), preferred adducts, in-source fragments, and associated MS spectra. The data was curated by visual quality control using software developed at Metabolon. Raw data for each metabolite was scaled to its internal median, after imputing the smallest non-zero value for that metabolite for any zeroes.

CLP Lipidomics
Lipids were extracted using a modified Bligh-Dyer extraction method with deuterated internal standards. Samples were then subjected to infusion-MS analysis in both positive and negative modes on a Shimadzu LC with nano PEEK tubing and a Sciex SelexIon-5500 QTRAP in MRM mode (>1,100 MRMs). Individual lipids were quantified as signal / internal standard and summed into class and total lipid concentrations.
Acute SD targeted nitrogen metabolomics Twenty total pools of $90 mixed sex iso31 flies aged $1.5 weeks post-eclosion were divided evenly among four conditions: (1) collected at ZT2, (2) collected at ZT14, (3) collected at ZT2 after a 14hr mechanical sleep deprivation, and (4) collected at ZT2 after a 12hr sleep deprivation followed by a 2hr sleep rebound. Heads were collected by vortexing whole flies snap frozen on dry ice and separating heads from bodies by size on dry ice-cooled grates. Samples were stored at -80C until they were processed. Samples were extracted and prepared for LC-MS analysis as previously described. 63,64 Briefly, a stainless steel bead and 300 ml of 2:1 Methanol:Chloroform were added to each sample. Samples were homogenized for a total of 4 minutes at 25 Hz in a tissue homogenizer. Next, 100 ml of water and chloroform were added to each sample. Samples were vortexed and then centrifuged for 10 minutes at 13,300 rpm at 4 C. 170 ml of the upper fraction containing polar metabolites was collected from each sample and dried in a speed vacuum for 2.5 hours. Dried fractions were resuspended in 100 ml of acetonitrile:water, vortexed for 20 seconds and centrifuged for 10 minutes at 13,300 rpm at 4 C prior to transferring to MS vials. Samples were analyzed in analytical triplicates and pooled quality control samples were run at the beginning and end of the run as well as after every 6th injection. For each sample, 2 ml were injected onto an Acquity UPLC BEH Amide column (1.7 mm, 2.1 mm x 150 mm) with a 0.2 mm inline precolumn filter using an Acquity H-Class UPLC system (Waters Corporation) coupled to a Xevo TQ-S micro mass spectrometer operating in a positive ion polarity mode. Initial chromatographic conditions consisted of 100% Solvent D (90:  (15/20)). Data was processed using TargetLynx (Waters) to obtain ion counts for further analysis using an in-house R-script. Spermine is excluded from our results because of low signal / high noise that rendered the signal suspect. Urea is excluded from our results because we were unable to corroborate its relevance through biochemical methods.
Lysate collection from whole flies for biochemistry Whole bodies were used in lieu of heads because of difficulty consistently detecting many target metabolites in heads using biochemical methods. Pools of 5 female or 5 male $10-day post-eclosion flies were anesthetized with CO2 and quickly sorted into 2mL Safelock Tubes, weighed, then snap frozen on dry ice. Blocks were pre-chilled to 4C, and the tubes were transferred to wet ice, where 200uL of 1X PBS supplemented with 1 cOmplete EDTA-free protease inhibitor tablet / 2.5mLs PBS (PBS-PI) and a stainless steel bead were added. Samples were quickly loaded into chilled blocks and lysed on a TissueLyser II (Qiagen) at 25.0m/sec in two 2-min bursts. Beads were removed and solid detritus was pelleted by spinning at 15,000rpm for 15min at 4 C. Supernatant was carefully transferred to a clean tube and diluted 1:4 in PBS-PI before being used for total protein, ammonia, and urate (uric acid) biochemical assays.

Hemolymph collection from female flies for biochemistry
Only female flies were run for these experiments because their typical hemolymph volume is much higher than males. Groups of 25 $10-day post-eclosion flies were anesthetized with CO 2 , rapidly pricked in the thorax with tungsten probes (Ted Pella 13570), and loaded into 0.5mL Ependorfs perforated at the base with 22-gauge syringe tips, which were nested inside of 1.5mL Ependorfs containing 45uL of PBS-PI. Nested tubes were then centrifuged at 5000rpm for 5min at 4C. 25uL of hemolymph + PBS-PI was removed for total protein analysis, and the remaining $20uL was diluted with a further 60uL of PBS-PI for ammonia analysis. Flies were weighed after hemolymph harvest.
Excrement collection from male flies for biochemistry Only male flies were run for these experiments to eliminate egg-laying as a confound. Groups of 12 flies were anesthetized with CO2 and sorted into 1.5mL Ependorf tubes perforated twice through each cap with an 18-gauge needle. The flies were returned to their home incubators for 2 hours, then anesthetized with CO2 and flipped into fresh 1.5mL Ependorf tubes to be weighed. The excrement in the first set of Ependorfs was resuspended by vortexing into 150uL of PBS-PI, which was used at this concentration to run ammonia and urate (uric acid) biochemical assays. For sleep mutant studies, $10 day or $1 day post-eclosion flies were tested at ZT0-2, ZT10-12, or ZT17-19. For acute mechanical SD studies, $10 day post-eclosion flies were sleep deprived for 12hr overnight and tested at ZT0-2. For thermogenetic SD studies, $1-4 day post-eclosion flies raised in an 18-20C room were (i) stored at 18C for $9 days, then moved to 29C for $1 day of acute sleep restriction before testing, or (ii) stored at 29C for $10 days of chronic sleep restriction before testing. Thermogenetic flies were entrained to 12:12LD during the $10 day variable temperature period, and excretion was tested at 29C from ZT10-12 (zeitgeber time chosen for most consistent excretion effects in sleep mutants).
Biochemical Assays: Total Protein, Ammonia, Uric Acid, Urea Total protein assay was conducted using an Abcam 207003 total protein assay kit according to manufacturer instructions: absorbance measured at 540nm. Ammonia assay was conducted using a Sigma-Aldrich MAK310 kit according to manufacturer instructions: fluorescence measured at excitation 355nm / emission 460nm. Uric acid assay was conducted using a Sigma-Aldrich MAK077 kit according to manufacturer instructions: fluorescence measured at excitation 535nm / emission 595nm. Signal from all biochemical assays was normalized to body weight of the pools of flies that provided the material. Fluorescence and absorbance were measured using a Victor-3V plate reader (Perkin-Elmer) or a Cytation 5 (BioTek).

Blue-Poo Assay
Male flies were pre-fed on our lab's standard yeast-molasses food supplemented with 2.5mg/mL of FD&C blue 1 / FCF brilliant blue dye for 24 hours. Excrement was then collected and resuspended into 150uL of MilliQ water. Absorbance was measured at 620nm and normalized to body weight to calculate fecal volume.

Sleep Experiments
For polyamine supplementation experiments, $3-5 days post-eclosion flies of both sexes were loaded into locomotor tubes with 5% sucrose / 2% agar food supplemented with water vehicle, 16mM L-ornithine monohydrochloride (Sigma-Aldrich 2375), 16mM putrescine dihydrochloride (Sigma-Aldrich P7505), or 16mM spermidine trihydrochloride (Sigma-Aldrich S2501). Sleep was measured from movements on DAM5H multibeam monitors (Trikinetics), averaged across the 2 nd -4 th full days of recording. A follow-up study with vehicle vs 50mM L-ornithine monohydrochloride was later conducted separately. Supplement doses were chosen based on a pilot dosing curve assaying sleep and toxicity with DAM2 monitors (data not shown) and published work supporting 16mM as a reasonable dose for screening sleep effects with amino acids and chemically similar polyamines. 17,65 In addition to standard sleep and activity metrics, average time spent / day at each of the 15 possible beam positions is reported for these datasets.
For nitrogen metabolism RNAi screen, $3-5 days post-eclosion female flies were loaded into locomotor tubes with 5% sucrose / 2% agar food supplemented with 500uM mifepristone (RU+ food) (Sigma-Aldrich M8046). Only females were used for this screen and all follow-up behavior and qPCR experiments, because lower female baseline sleep allows more reliable detection of sleep gain phenotypes, especially with the relatively small n's used for behavioral screening. Sleep was recorded from counts of beam breaks on single-beam DAM2 monitors (Trikinetics), and the 4 th -5 th full days of exposure to RU+ food were averaged to determine sleep. Geneswitch(GS)>Dicer,RNAi crosses with mean sleep at least 60 min higher or lower than both GS>Dicer and RNAi controls were considered potential hits and validated. Validation of promising crosses was carried out similarly to the screen, but included both RU+ and ethanol vehicle (RU-) food conditions to assess whether effects were acute. If the first validation experiment was inconsistent with the initial screen result for a given RNAi, further validation of that RNAi was terminated. At least two independent validation experiments were run for each nitrogen RNAi cross identified as having a legitimate sleep phenotype. Hits showing sleep gain were subsequently re-validated by analyzing movements collected on DAM5H multibeam sleep monitors, to rule out possible position confounds. Later follow-up RNAi and Crispr studies were setup similarly to the description above, but used DAM5H multibeam monitors from the outset.
For awake-time-by-food and sleep analysis of sleep mutant and thermogenetic sleep-restriction flies and their genotypic controls, only male flies were used to avoid oviposition as a confound. $3-5 days post-eclosion males were loaded onto DAM5H multibeam sleep monitors in locomotor tubes containing high-protein, high-sugar, or all-sugar diets used for lifespan analysis (below).
Movements and position distribution were recorded from full Days 2-4 post-loading, averaged, and used to calculate standard sleep and activity metrics, as well as absolute and % awake time / day spent adjacent to food.
DAMfilescan and previously reportedcustom MatLab scripts were used to calculate sleep metrics for both sets of experiments. 57 Awake time-by-sleep scripts are novel, and archived at GitHub (see above).

qPCR Validation of Nitrogen Pathway RNAis
To confirm knockdown of target transcripts with RNAi and Crispr tools, we drove expression with actinGS>dcr (RNAi) or actinGS>-cas9 (Crispr) and followed the same whole-fly RNA collection, cDNA synthesis, and qPCR method we published previously. 42 The following qPCR primer sets were used for each target transcript: asl forward: TCGACAAGCTGTCCCAAGTG reverse: CACCAGATAGTAGGCCCAGTC spds forward: GAAACACGCGCTGAAGGATG reverse: GGCATAGGCCACCTTAGCAA sms forward: GAGCTGCAGAACATTGCTGA reverse: GTACAACAAGGCGCCATCAC a-tubulin forward: CGTCTGGACCACAAGTTCGA reverse: CCTCCATACCCTCACCAACGT

Lifespan Experiments
For all lifespan studies, mated flies of both sexes that had eclosed within the preceding $2 days (sleep mutant studies) or $3-4 days (thermogenetic studies) were collected under CO2 anesthesia and housed single-sex on standard or special food at a maximum density of 30/vial. For classical lifespan studies, flies were flipped to fresh vials and dead were tallied every 3 days (experiments comparing sleep mutant lifespans to control on standard food) or every 2 days (all other experiments). Lifespan vials were co-housed in their incubators with water dishes, to provide supplemental humidity. The w/v of nourishing solutes of roughly similar caloric density were equal for each diet, to minimize differences in caloric density among the diets. Each vial was longitudinally maintained on the same assigned food condition, and dead flies were counted at each flip, tracking survival in this way until all flies were dead. Flies were never again anesthetized after initial collection. Flies that escaped during flips were excluded from analysis.
Sehgal For survival on nitrogenous metabolite supplementation studies, flies were initially maintained on standard food, then loaded into locomotor tubes containing 5% sucrose / 2% agar food drugged with either water vehicle or 16mM ornithine, putrescine, or spermidine at $3-4 days post-eclosion. Behavior was recorded using single-beam DAM monitors until the end of the sixth complete day on supplemented food. Sleep records were analyzed similarly to starvation-challenge, except that all flies surviving past the sixth full day of recording were censored.

QUANTIFICATION AND STATISTICAL ANALYSIS
For molecular studies where lower n's precluded reliable normality pre-testing, a parametric distribution was assumed. For sleep and feeding behavior experiments, higher n's allowed reliable Shapiro-Wilkes pre-testing of all groups' normality to determine whether parametric or non-parametric testing was appropriate for each metric, in each individual experiment. Lifespan studies used specialized non-parametric and semi-parametric statistical approaches. Prism or JMP software was used to carry out all statistical analyses. The following tests were used in the indicated conditions: -Two groups, or multiple control groups vs one experimental group 66 : Welch's t test (parametric) or Mann-Whitney test (nonparametric).
-Single control group vs multiple experimental groups: Dunnett test (parametric) or Steel test (non-parametric). Where multiple Dunnett tests were required, they were run as posthoc tests of a two-way ANOVA model.
-All-to-all comparisons, or complex subsets of all comparisons: Tukey HSD test (parametric) or Steel-Dwass test (non-parametric) -% awake time near food (Data S3) is reported as raw percentages for more intuitive presentation of the data, but statistical tests were conducted on arcsin-transformed percentage values, to more closely approximate an unbounded distribution.
-qPCR validation of RNAi knockdown efficiency: one-tailed t tests -Lifespan: Wilcoxon tests were used to measure differences in survival. Likelihood tests of differences in Cox proportional hazard ratios were used to measure differences in hazard. Posthoc Bonferroni correction was applied to the significance threshold for survival and hazard comparisons where familywise error was cumulative. 67 Additional details for individual experiments (n definition and number, summary statistics, etc) can be found in corresponding Figure Legends.