Atg8 licenses adipokine nuclear exit

equal contribution. * corresponding author: akhila@fredhutch.org Abstract Adipokines released from adipocytes function as a systemic ‘adipometer’ and signal satiety1, 2. For organisms to accurately sense surplus and scarcity, adipocytes must switch adeptly between adipokine release and retention3. Despite the central requirement for adipocytes to retain adipokines to enable organismal adaptation to nutrient deprivation1, how fasting induces adipokine retention remains to be fully characterized. Here we investigated how Unpaired2 (Upd2), a fruit fly ortholog of the human adipokine Leptin4, is retained during fasting. Unexpectedly, we observe that on fasting Upd2 accumulates in the nucleus and discover that fasting-induced Upd2 nuclear accumulation is regulated by Atg8. Atg8 is a ubiquitin like protein which conjugates to a lipid moiety5. We find that, Atg8 based on its own lipidation status promotes adipokine nuclear exit in fed cells and adipokine retention in starved cellswe term this ‘adipokine licensing’. Then, we show that Atg8 lipidation is the rate-limiting step in adipokine licensing. We then illustrate how organisms use adipokine licensing to survive nutrient deprivation. Additionally, we show that adipokine nuclear retention, controlled by Atg8, stimulates post-fasting hunger; thus, Atg8-mediated nuclear retention sensitizes adipokine signaling after fasting. Hence, our findings point to a new mechanism to tackle adipokine resistance6, 7, an underlying cause of common obesity. Collectively, we have identified a novel cell-intrinsic mechanism that orchestrates systemic response to energy flux and uncovered an unexpected role for Atg8 in context-dependent protein localization.


Introduction
Organisms constantly evaluate their basal nutritional reserves to decide on whether to allocate resources to expensive functions that enhance fitness or to conserve energy. 5 Adipokines, secreted by fat cells provide an accurate assessment of nutrient-reserve information. Leptin in mammals 1 and its functional ortholog in fruit-flies, Unpaired2 (Upd2) 4 , are primary adipokines that are released in proportion to fat stores [8][9][10] . They impinge on brain circuits that control energy expenditure, appetite and overall metabolism 4,11,12 . In surplus nutrient states, adipokines convey a permissive signal, indicating that energy resources can be devoted 2 to costly activities like immunity and reproduction. During periods of scarcity, their circulation is reduced, signaling an energy deficit, and enabling organisms to conserve energy.
Starvation-induced reduction in Leptin/Upd2 levels is crucial for an organism's neuroendocrine response to reduced energy reserves; this has been well-documented in humans 8,9 , mice 1, 6 and flies 3,4 . Specifically, Leptin injections, during fasting, dysregulates 5 neuroendocrine physiology, and decreases survival capacity in mice 1 . Based on this study 1 , Flier and colleagues have proposed that "the primary physiologic role of Leptin is to provide a signal of energy deficit to the CNS" 6,13 . Similarly, flies with reduced levels of Upd2 display increased starvation survival 4 , suggesting that reduced Upd2 levels enable flies to develop resilience to starvation. 10 Despite the central requirement for adipocytes to retain adipokines during starvation in order to allow organismal adaptation to nutrient deprivation, how fasting induces adipokine retention remains to be fully characterized. Importantly, hyperleptinemia, i.e., high levels of circulating leptin, has been identified as a primary cause of 'leptin resistance'-a state wherein Leptin is unable to effectively signal satiety state to the brain 14 , resulting in dysfunctional feeding 15 behavior and decreased energy expenditure. Hence, understanding how adipokine retention is regulated in fat cells will provide mechanistic insights that can be applied to intervene in hyperleptinemia-induced adipokine signaling resistance.
We previously showed that in Drosophila cells, both Upd2 and human Leptin adopt a unconventional secretion route for release into circulation 3 . We also showed that during nutrient 20 deprivation, increased circulating levels of the starvation hormone Glucagon impinges on fat cells to inhibit adipokine release via this route, ultimately enabling fat cells to retain adipokines 3 . This mechanistic insight is consistent with the idea of using glucagon like peptides (GLPs) to treat obesity and sensitize Leptin signaling 15  A primary mechanism by which organisms adapt to metabolic stress is the nutrient scavenging pathway called autophagy 16,17 . In adipocytes, autophagy controls lipid metabolism by a process known as lipophagy 18,19 . A central player in the autophagy pathway is a ubiquitin- 30 like protein family member 20 called LC3/GABARAP in mammals, and Atg8 in flies and yeast 21 . 3 Atg8 is widely distributed in the nucleus 22 and during starvation, it translocates to the cytoplasm 23 where its gets conjugated to the lipid moiety phosphatidylethanolamine (PE) 5 . Lipidated Atg8 regulates fusion of autophagosomes to lysosomes, resulting in cargo degradation 24 . It is now widely recognized that many core autophagy proteins, including Atg8, play central roles in nondegradative cellular activities 25 . These include, but are not limited to: LC3-associated 5 phagocytosis (LAP) 26 , viral replication 27 , extracellular vesicle secretion (LDELS) 28 , and LC3 associated endocytosis (LANDO) 29 . These non-degradative roles for Atg8 have been documented in the cytosol. But what remains unclear is whether Atg8/LC3 proteins localized to the nucleus 22 play a specific role other than to provide a pool of Atg8 available for lipidation 23 , and to help cells cope with oncogenic stress by degrading the nuclear lamina 30 . 10 While investigating cell intrinsic mechanisms which regulate acute adipokine retention, we unexpectedly found a direct, non-degradative role for Atg8 in licensing adipokine nuclear exit in fed cells and promoting adipokine retention during fasting. Furthermore, we found that this mechanism is central to systemic energy balance as it provides a resilience mechanism for flies during fasting and sensitizes adipokine signaling post-fasting. Given, the deep conservation of 15 autophagy machinery between flies and mammals 31,32 , and also the processes governing energy homeostasis, particularly lipid biology, between flies and mammals [33][34][35][36][37][38] , we believe that our findings in the Drosophila model, regarding Atg8's role in licensing adipokine release, will be of relevance more broadly. 20

Fasting induces Upd2 nuclear accumulation:
To explore whether cells have intrinsic mechanisms to retain adipokines during fasting, we probed whether cells retain Upd2/Leptin on amino-acid (AA) deprivation. Note, we chose the AA-starvation regime, because it provided the most consistent results, as opposed to serum 25 starvation or culturing cells in buffers including PBS and EBSS. Upd2, a JAK/STAT pathway agonist, is secreted into media of cultured Drosophila S2R+ cells 39 . In previous work, a quantitative assay (ELISA) for Upd2 release was established in the Drosophila S2R+ cell culture system 3 . For these experiments a C-terminal GFP-tagged version of Upd2 (Upd2::GFP) is transiently transfected into S2R+ cells 3 . Upd2::GFP can rescue upd2Δ mutants 4 , strongly 30 suggesting that GFP tagged version of Upd2 likely recapitulates the function of untagged wildtype (WT) Upd2. Using this previously established quantitative secretion assay (see methods) 4 we asked whether culturing S2R+ cells in amino acid (AA)-free media affected Upd2::GFP release (hereafter referred to as Upd2-WT). We observed a -76% reduction (p<0.001) in Upd2 secretion within 6 hours of AA-starvation. Furthermore, 20 hours post-starvation, Upd2 secretion decreased by -96% (p=0.0004) (Fig.1A). These results suggest that Drosophila S2R+ cells sense a fasted state and acutely retain a significant portion of the adipokine Upd2 (~76%) within 5 a brief period of AA deprivation (6 hours).
Fasting induced transcriptional 40 and translational 41 controls, are likely to work in concordance with acute mechanisms to retain already transcribed and translated adipokines.
Acute adipokine retention mechanisms possibly include a "brake" on the trafficking machinery involved in Upd2 release, and/or degradation of adipokine primed for release. To begin to 10 explore these acute mechanisms, we documented Upd2's cellular localization during a fasting time-course (imaging fixed cells -every 2-4 hours on AA-fasting). When cells are cultured in complete medium ('Fed' state), Upd2 exhibits a punctate cytoplasmic localization, with diffuse and faint nuclear localization (Fig.1Ba). Within 4-8 hours of AA fasting, increased Upd2 nuclear accumulation is observable (Fig.1Bb-Starved (Stv), 8hr). Upd2 nuclear accumulation is detected 15 even at 16 hours of AA-starvation ( Fig.1Bc-Stv, 16hr). We also noted that Upd2, at time points assayed (4-24hr), exhibits a punctate cytoplasmic localization (Fig.1Bb, Bc). While it is possible that this cytoplasmic pool of Upd2 is retained or degraded, we did not further define or characterize the fasting-induced acute adipokine retention mechanisms in the cytoplasm (See Discussion). 20 Instead, intrigued by fasting-induced Upd2 nuclear accumulation, we sought to define and characterize it. To this end, we performed independent fasting-time series experiments with 50-100 cells per time-point (these were blinded experiments-see methods) and quantified the ratio of Upd2 nuclear intensity versus the whole cell at each time point (Fig.1B'). Across such independent experiments, we observed a similar dynamic in Upd2 nuclear accumulation. No 25 significant increase in Upd2 nuclear accumulation was observable at 2 hours of AA-fasting.
However, within 4 hours we observed, on average, a +25% increase (p=9.9E-7) in Upd2 nuclear accumulation. This upward trend in Upd2 nuclear accumulation peaks at 8-12 hours post-AA starvation with a 30-50% increase in Upd2 nuclear accumulation (p= 1.2E-8). At 16 hours, Upd2 nuclear accumulation while higher than fed state (+17%; p=5.1E-3), was lower than the peak at 30 8-12 hours (Fig.1B'). 5 Next, we asked whether refeeding restores Upd2 release. On performing a quantitative ELISA for Upd2-WT in cells from three different nutrient states--fed, starved, and re-fed--we observed that by 5 hours after refeeding (Fig.1C), the amount of Upd2-WT released into the media was 104% higher than the starved state (p=1. 2E-6). This indicates that cells can toggle between retention and release within a 5-hour period; strikingly, this mirrors the temporal 5 dynamics of adipokine retention (Fig.1A). Next, we examined the correlation between refeeding and Upd2 nuclear accumulation. As observed previously (Fig.1B, B'), within 8 hours of starvation, significantly higher Upd2 nuclear accumulation can be seen (+43%; p=6.8E-14; with Upd2 release within 5-hours on refeeding. Taken together these observations indicate that fasting-induced nuclear accumulation is an acute adipokine retention mechanism, and that refeeding reverses Upd2 nuclear accumulation within a similar timeframe (4-6 hours). For the rest of this study, we focused on understanding the mechanisms that regulate the early phase 15 (4-8 hours) of acute adipokine Upd2 nuclear accumulation.

Atg8 regulates nucleocytoplasmic localization of Upd2:
We then sought to identify mechanisms that link nutrient state to adipokine nuclear retention.
AA-starvation triggers autophagy, a nutrient scavenging pathway 42 , of which Atg8/LC3, a ubiquitin-like protein, is a primary component 21 . In fasted cells, Atg8 undergoes a ubiquitin-like 20 conjugation to a lipid moiety phosphatidylethanolamine (PE) 5 . Subsequent to its lipidation Atg8 regulates fusion of autophagosomes to lysosomes, resulting in cargo degradation 24 . It is now widely recognized that many core autophagy proteins, including Atg8, play central roles in nondegradative cellular activities 25 . For instance, Atg8 is well-documented for its involvement in the unconventional secretion of a wide variety of cargoes including, but not limited to, yeast Acb1, 25 mammalian interleukins and RNA binding proteins [43][44][45][46][47] .
In a previous study, our lab demonstrated that the adipokine Upd2 and its human ortholog Leptin (hLeptin), adopts an unconventional secretory route 3 . Given Atg8's role in AA-deprivation triggered nutrient-scavenging 21 , and its documented role in unconventional secretion [43][44][45]47 , we specifically examined the effect of removing Atg8 on Upd2 release. In fed Drosophila S2R+ cells, 30 dsRNA mediated knockdown of Atg8 (Atg8-KD) results in an -85% reduction (p=0.048) in Upd2 secretion ( Fig.2A), phenocopying the effect of fasting on Upd2 release. Moreover, this 6 observation, that Atg8 is required for Upd2 release, is consistent with both Atg8's role in unconventional secretion 28,48 and our prior report that Upd2 is a cargo of the unconventional secretion pathway 3 .
To better understand this reduced secretion phenotype, we examined cellular localization of Upd2 following Atg8-KD (Fig.2Bb) compared to a control dsRNA (LacZ-KD) (Fig.2Ba). While 5 punctate cytoplasmic Upd2 is still present, nuclear Upd2 accumulation is clear in Atg8-KD cells by confocal microscopy. On quantifying, Atg8-KD resulted in a +34% (p=0.0035) to +60% (p=3.9E-5) increase in Upd2 nuclear accumulation (Fig.2B'). Collectively, these observations are consistent with Atg8 reduction, even in fed cells, correlating with increased Upd2 nuclear accumulation. 10 As Atg8 itself is degraded by autophagy 49 , we considered that in Atg8-KD, reduced Atg8 levels perhaps signify a nutrient scavenging state; hence an increase in Upd2 nuclear accumulation could be a 'passive' outcome. An alternate, but not mutually exclusive, hypothesis is that Atg8 plays an 'active'/ direct role that ultimately results in fasting-induced adipokine localization. This latter hypothesis is consistent with the idea that Upd2 might interact in a 15 complex with Atg8, directly or via adapter molecules. To test this, we co-transfected Myc::Atg8-WT and Upd2-WT::GFP and performed GFP and Myc pull downs, we found that Upd2-WT coimmunoprecipitates (Co-IPs) with Atg8-WT (Fig.2C, lane 1).
Next, we wanted to test whether Upd2's interaction with Atg8 is reliant on a short linear protein domain known as the AIM (Atg8-interacting motif), which contains the consensus 20 sequence W/F/Y-X-X-L/I/V, and is present in bona fide Atg8 interaction proteins 50 . On analyzing protein sequences, we found that Upd2 has AIM-like sequences ( Figure. 2D). Mutating Upd2's AIM (using an established approach of mutating W and L to A 51 ;here after referred to as Upd2-AIM -) impaired the interaction between Upd2 and Atg8, as assayed by a co-immunoprecipitation experiment (Fig.2C, lane 2). Upd2 is usually found at very low levels in the lysate of S2R+ cells 25 because of its constitutively secreted (see GFP-IP lane 1 in Fig.2C). We noted that Upd2-AIMlevels appeared higher than Upd2-WT in lysate (GFP-IP lane 2 in Fig.2C). Despite being present at a higher level it does not effectively complex with Atg8 (RFP-IP lane 2 in Fig.2C). Furthermore, the higher levels of Upd2-AIMin the lysate is suggestive of a defect in release of Upd2-AIM -.
Altogether, these datasets suggest that when Atg8 levels are reduced (Atg8-KD) or when Upd2's interaction with Atg8 is significantly impaired (Upd2-AIM-), Upd2 is retained in the 15 nucleus and is not readily available for secretion. Hence, our results are suggestive of Atg8 playing a central role in Upd2's nucleocytoplasmic localization.

Nucleocytoplasmic shuttling of Upd2 is required for its subsequent release:
Upd2 is a 47 kDa protein. Given the structural restraint of nuclear pore complexes (NPC) to only permit proteins <40kDA to pass through 52 , Upd2 is likely to require active nucleocytoplasmic 20 (NC) transport mechanisms that enable transit through the NPC 53 . Hence, we sought to explore whether Upd2's NC localization is regulated by classical NC transport pathways. In such a pathway, cargoes >40kDa (but even some smaller cargoes such as histones) rely on their ability to interact with a set of soluble nuclear transport receptors-'importins' and 'exportins'-also known as Karyopherins 54,55 . 25 First, we analyzed Upd2's sequence using a bioinformatic tool-classic NLS (cNLS) mapper- [56][57][58] to ask whether Upd2 has a predicted NLS. cNLS mapper predicted that Upd2 contains an NLS at residues 141-150 (GRVIKRKHLE) with a score of 10.5. According to cNLS mapper, a GUS-GFP reporter protein fused to an NLS with a score of 8, 9, or 10 is exclusively localized to the nucleus. Next, we asked whether Upd2 is predicted to contain an NES using 30 another tool NetNES 59 . While NetNES predicted an NES like sequence for Upd2 residues 248-8 254 (LCEIELTI), residues 251 and 253 do not broach the threshold, thus making it difficult to say if this whole sequence is indeed an NES (see Discussion).
To clarify if Upd2 relies on a classical nuclear export factors, we genetically manipulated nuclear exportin CRM1 60 , also known as Embargoed (Emb) in flies 61 . We asked whether Emb knock-down (Emb-KD) affects Upd2 localization in a fed state. We found that transfection of fed 5 S2R+ cells with Emb dsRNA, compared to control LacZ dsRNA (Fig.3Aa), recapitulated the increase in Upd2 nuclear accumulation (Fig.3Ab) that we had observed in either an AA-starved state (Fig.1Bb, Bc), or following Atg8-KD (2Bb). On quantifying (Fig.3A'), we observed that Emb-KD, using two independent dsRNAs, resulted in +18% (p=0.0053) and +16% (p=0.024) increase in Upd2 nuclear accumulation (levels akin to 16-hour starvation time point Fig.1B'). We also 10 noted punctate Upd2 localization at the nuclear periphery (blue arrows in Fig.3Ab). We next analyzed the effect of reducing CRM1/Emb on Upd2's release from cells. We observed that, consistent with our previous findings on the inverse correlation between Upd2 nuclear accumulation and its release, that in Emb-KD cells, as assayed by ELISA, Upd2 release is reduced by -33 to -49%; p<0.01 (Fig.3B). Altogether, irrespective of whether Upd2 directly 15 interacts with CRM1/Emb or via an adapter due to its "cryptic" NES (See Discussion), these data are indicative of CRM1/Emb export being involved in Upd2 nuclear exit. These observations support the hypothesis that manipulations which impede nuclear export Upd2 nuclear accumulation decreases its subsequent release.
Given the inverse correlation between Upd2's nuclear accumulation and its secretion, we 20 wondered about the opposite condition., i.e., whether manipulations that impede nuclear import increase Upd2 release. In Drosophila there are at least four encoded Importin-α genes and at least eight Importin-β genes 62 . We were concerned that individual knock-down of these 12 genes may not produce a strong effect, should there be redundancy. Our goal was to blunt nuclear import in general and observe its effect on Upd2 release. Therefore, we adopted a 25 pharmacological approach by using a small molecule Importin-β inhibitor, 2,4diaminoquinazoline, commonly named Importazole 63 . Heald and colleagues have shown that Importazole specifically blocks Importin-β-mediated nuclear import; it does so by impairing Importin-β's interaction with RanGTP, and importantly does not disrupt transportin-mediated nuclear import or CRM1-mediated nuclear export 63 . This drug was established for blocking 30 nuclear import using Xenopus egg extracts and cultured cells 63 . Hence, to block Importin-β based nuclear import, we incubated Drosophila S2R+ cells for 6 hours with 40-100μM 9 Importazole and assayed its effect on Upd2 release using the quantitative sandwich ELISA assay when compared to control treatment (DMSO). Unlike what we predicted, we observed that 40μM Importazole reduced Upd2 secretion by -23% (p=0.15) and that 100μM Importazole reduced Upd2 secretion by -48% (p=0.01) (Fig.3B). We observed the similar reductions in Upd2 release with Importazole treatment, in three independent experiments, with 6 biological 5 replicates per condition per experiment. We were surprised that Importin-β inhibition impaired Upd2 release and wondered whether it affected Upd2 nuclear entry. Therefore, we examined cells treated with Importazole and documented Upd2 NC localization by confocal imaging. As predicted, based on Upd2's NLS sequence, impeding Importin-β-based nuclear transport by treatment with Importazole, reduced Upd2 nuclear localization for both the 40μM Importazole  also noted that compared to control treatment (Fig.3Da), Upd2-WT in Importazole-treated cells ( Fig.3Db and 3Dc), was hard to detect in distinct punctate localization. Instead, Upd2-WT in Importazole-treated cells appeared to 'fill' majority of the cytoplasmic compartment (Fig.3Da,   3Dc). Hence, increased Upd2 cytoplasmic accumulation correlates with reduced release. Taken together, these observations suggested to us that Upd2 uses an Importin-β based mechanism 20 for its nuclear import.
It is possible that impairing NC shuttling broadly, including blocking nuclear import by pharmacological means, impacts Upd2 release in an indirect manner. Hence, we sought to precisely manipulate Upd2 so that it remains cytosolic and then assay how such a 'forced' cytosolic localization affects its release. In addition, we also wondered whether forced nuclear 25 export, by appending a strong NES sequence, could 'override' the Upd2-AIM-nuclear accumulation phenotype. Studies in Drosophila models have shown that appending a particular NES sequence (LDELLELLRL) to the N-terminal of proteins that localize to the nucleus, reliably and actively shuttles those proteins out to the cytosol 64 , thus rendering nuclear proteins cytosolic. Hence, we added this strong nuclear export signal (NES + : LDELLELLRL) in the N- 30 terminal (See methods) of Upd2-WT (referred to here-in as Upd2-NES+) and Upd2-AIM -( referred to here-in as Upd2-NES+AIM-). We then performed confocal imaging of cells transfected with: Upd2-WT (Fig.3Ea, Eb), Upd2-AIM- (Fig.3Ec, Ed), Upd2-NES+ (Fig.3Ee, Ef) and Upd2-NES+AIM- (Fig.3Eg, Eh). As previously documented in Fig.1, Upd2-WT in fed cells exhibits a punctate cytosolic localization (Fig.3Ea) and displays significant nuclear accumulation on AA-starvation (Fig.3Eb). Again, as reported in the prior section (Fig.2E), Upd2-AIMis nuclear with almost no detectable punctate cytosolic expression in the fed or starved state (Fig.3Ec, Ed). 5 Upd2-NES + is cytosolic in fed cells (Fig.3Ee) and continues to be cytosolic even in starved cells ( Fig.3Ee, see 3E' for quantification). Moreover, like Upd2-WT localization in Importazole treated cells (Fig.3Db, Dc), Upd2-NES+ 'fills' the cytoplasm (Fig.3Ee), instead of exhibiting the punctate localization observed in Upd2-WT fed cells (Fig.3Ea). Like Upd2-NES + , Upd2-NES + AIMis cytosolic in fed and starved cells (Fig.3Eg, Eh). Furthermore, the NES + appended to Upd2-AIM -10 overrides the Upd2-AIMnuclear accumulation phenotype (Compare Fig.3Ec, Ed to Eg, Eh), indicating that by providing an ectopic active export mechanism, Upd2 can exit the nucleus. But again, just like Upd2-NES+ (Fig.3Ee, Ef), and Importazole treatment Fig.3Db, Dc, the cytosolic Upd2 appears more diffuse in Upd2-NES + AIM - (Fig.3Ee, Ef). Hence, we asked if engineered nuclear export, which results in Upd2's nuclear export despite being AIM-is sufficient to promote 15 Upd2 release. On performing quantitative ELISA on Upd2-NES + and Upd2-NES + AIM -, however we observed that both versions exhibit secretion defects ( Fig.3F; -64% reduction p=0.0013), suggesting that forced nuclear export of Upd2 while can override the AIM-nuclear retention phenotype, it is unable to direct Upd2's release. This concurs with our observation that interfering with either Upd2's nuclear export (Fig.3B) or its nuclear import (Fig.3C), both affect Upd2 20 release.
In conclusion, these results suggest that Upd2 needs to be shuttled into the nucleus prior to being secreted. Also, these results are consistent with a model that after translation in the cytoplasm, Upd2 enters the nuclear environment in order to obtain 'licensing' for release.
Upd2 nucleocytoplasmic shuttling licenses its release via Atg8-GRASP pathway: 25 We surmised that Atg8 could potentially serve as Upd2's 'licensing' factor, by both enabling Upd2 nuclear exit, and then subsequently targeting Upd2 to an appropriate secretion route. To investigate this possibility, we asked whether, in fed cells, the Upd2 cytosolic puncta co-localized or associated with Atg8. First, we performed live imaging of fed cells transfected with Upd2::GFP and mCherry::Atg8 ( Fig.4A and Extended Data Fig.1), to gather preliminary information on the 30 dynamics of the Upd2 puncta and Atg8. We observed that Atg8 and Upd2 co-localize in nuclear periphery (Fig.4A, see blue arrows in 0s, 90s and 135s. Similarly, we also observed the two 11 proteins co-localize in puncta at the cell periphery (Fig.4A, white arrows in 0s-180s). We found that this co-localization was not a stable one, i.e., a complete overlap changed into an association within 45 seconds (see blue arrows in nuclear periphery Fig.4A-135s vs. 180s).
Altogether, these imaging datasets indicated that the Upd2 puncta in the cytosol were positive for Atg8, but that the co-localization of Atg8 to Upd2 punctum was dynamic and unstable, and 5 suggesting that in fixed imaging we were likely to observe them overlap partially.
Once we obtained this baseline for dynamics of Atg8 and Upd2 co-localization, we wondered whether colocalization between Upd2 and Atg8 would be different for Upd2-NES+ compared to Upd2-WT. In fed cells co-transfected with Atg8 and either Upd2-WT (Fig.4Ba) or Upd2-NES+ ( Fig.4Bb), we examined and quantified the co-localization (see Methods) between 10 these two proteins. In >10 cells per condition, we observed that Upd2-WT exhibited partial to complete (yellow arrows in Fig.4Ba) co-localization with Upd2 (quantified in B' : (0-90% colocalization) whereas Upd2-NES+ was almost mutually exclusive to Atg8 (Fig.4Bb, B'; note difference between Upd2-WT and NES+ is significant p=0.006) . This suggested to us that if Upd2 is forcibly shuttled out of the nucleus then, its ability to localize to Atg8 puncta in the cytosol 15 is compromised. Significantly, this observation lends credence to the idea that Upd2 nuclear localization and subsequent exit is key to Upd2's ability to access Atg8 appropriately.
Since Upd2 release is impaired in both Upd2-NES+ (Fig.3F), and Upd2-AIM-( Fig.2A), we reasoned that Upd2's ability to interact with Atg8 is an important step both for its nuclear exit and subsequent secretion. Our lab previously reported that Upd2 uses a GRASP-secretion route 20 3 . Therefore, we examined the co-localization of GRASP with respect to Upd2-WT and Upd2-AIM-. Imaging fixed cells co-transfected with GRASP and Upd2-WT ( Fig.4Ca) versus Upd2-AIM-( Fig.4Cb), showed that GRASP is largely a cytosolic protein and that Upd2-WT puncta colocalize with GRASP, whereas Upd2-AIM-renders Upd2 unable to access the cytosol.
Consistent with this, we also observed that while Upd2-WT co-immunoprecipitates with GRASP 25 ( Fig.4D; Lane 2), despite being present in much lower amount in the lysate of the cells due to its higher secretion. By contrast, Upd2-AIM-, which is not secreted (Fig.2D), does not co-IP with GRASP ( Fig.4D; Lane 2).
Taken together these observations (Fig. 3, 4) are consistent with a model that Upd2's nuclear entry provides Upd2 a 'venue' to have a productive interaction with Atg8 (note Upd2-30 NES+ co-localization with Atg8 is impaired: Fig.4B, B'). Atg8 then enables Upd2's nuclear exit and provides the context required for Upd2 to be targeted to the GRASP secretory route.
Atg8's lipidation state is the rate-limiting step in Upd2 nuclear exit: The most well-characterized function of Atg8 is its degradative role in autophagy, where in undergoes a ubiquitin-like conjugation to a lipid moiety phosphatidylethanolamine (PE) 5 and enables the fusion of autophagosomes to lysosomes. Therefore, we asked whether Atg8's lipidation state has an impact on Upd2 nuclear exit. 5 First, we examined what effect, if any, autophagy induction and its subsequent induction of Atg8 lipidation, has on Upd2 nuclear retention. Torin1, is a highly potent and selective ATPcompetitive mTOR inhibitor 65 , and is one of the most widely used methods to pharmacologically activate autophagy 66 . Torin1 activated autophagy induces Atg8-lipidation 67 . We treated wellfed Drosophila S2R+ cells, cultured in complete media, with Torin1 (50-750nM; see Methods) 10 and assayed its effect on Upd2 nuclear accumulation relative to control (DMSO) treatment Altogether, this suggested to us that induction of autophagy by Torin1 treatment, like AAwithdrawal, induced Upd2 nuclear accumulation.
Atg8-lipidation levels increase during autophagy induction. Atg8 is widely distributed in the nucleus 22 . Whether this nuclear pool of Atg8 has any specific role remains to be fully 20 understood. A study by Liu and colleagues 23 , presented compelling evidence that the nuclear pool of Atg8 translocates to the cytosol on starvation, and is the primary source of lipidated Atg8 for autophagosome formation 68 . However, it remains possible that Atg8 has a yet to be uncovered, non-degradative, functions in the nucleus via its actions on AIM/LIR containing proteins such as the adipokine Upd2. Therefore, going a step further, we considered the 25 possibility that AA-starvation induced Atg8-lipidation may result in a depletion of the nuclear pool of Atg8 in Drosophila S2R+ cells, a state that would ensure Upd2 remains localized to the nucleus.
To test this directly, we examined the localization of 'normal' wild-type Atg8 (Atg8-WT) versus a lipidation-defective Atg8 [Atg8-PE -; as done in prior studies which defined Atg8 30 lipidation, 5 we mutated the penultimate glycine, the lipidation acceptor amino acid of Atg8, to an 13 alanine; see methods]. We transfected N-terminally mCherry tagged-Atg8-WT (Fig.5Ba, Bb) or Atg8-PE- (Fig.5Bc, Bd) into Drosophila S2R+ cells, and examined localization in fed and AAstarved states. We note that these experiments are all performed in WT S2R+ cells that do contain an endogenous Atg8 pool, however our goal was to document the distribution of mCherry-tagged versions of Atg8-WT and Atg8-PE-in the presence of endogenous Atg8. We 5 observed that mCherry::Atg8-WT is distributed diffusely in the nucleus and the cytoplasm ( Fig.5Ba) in fed cells, but during a short AA-starvation most of mCherry::Atg8-WT is localized to punctate structures in the cytoplasm (Fig.5Ba). This pattern of Atg8-WT localization in S2R+ cells concurs with what has been shown previously by Baehrecke and colleagues 69 , indicating that our Atg8 transgenes and starvation regimes work as would be expected for this cell line These results strongly suggest that starvation, because it induces Atg8's lipidation and causes 20 it to re-localize to punctate structures in the cytosol, depletes the Atg8 nuclear pool, while a lipidation-defective Atg8 molecule continues to be present in the nucleus even during starvation, at levels comparable to the fed state. We extended this observation to Drosophila adult fat cells (Extended Data Fig.2B), in which we found that on starvation Atg8-WT localizes to punctate cytosolic structures (Extended Data Fig.2Ba; white arrows) and is depleted from the nucleus 25 (Extended Data Fig.2Ba; blue arrows). Lipidation-defective Atg8 (Atg8-PE-), while present in the cytosol, continued to show strong nuclear enrichment, even on starvation (Extended Data Thus far, our data show that events which induce Atg8-lipidation increase Upd2 nuclear accumulation, including AA-starvation (Fig.1B) and Torin1 treatment (Fig.5A). Hence, we 30 hypothesized that lipidation-induced depletion of Atg8's nuclear pool is a rate-limiting step in Upd2's nuclear exit during starvation. To test this, we co-transfected Upd2-WT, with either 14 control (empty mCherry vector), mCherry::Atg8-WT (which can undergo lipidation and hence move into the cytosol on starvation; Fig.5Bb), or mCherry::Atg8-PE-(lipidation-defective Atg8 which remains in a nuclear pool even during starvation; Fig.5Bd). See schematic in the top panel of Fig.5C for experimental design. We then quantified (in 65-150 cells per condition) the ability of starvation to induce nuclear accumulation of Upd2, in the presence of these Atg8 transgenes. 5 As documented in the prior section (Fig.1Ba), Upd2 is localized to cytosolic puncta in fed cells.
Co-expression of either mCherry empty vector (Fig.5Ca), mCherry::Atg8-WT (Fig.5Cc) Atg8, that continues to be nuclear during starvation, interferes with fasting-induced Upd2 nuclear accumulation, and is sufficient to allow Upd2 to exit the nucleus during fasting. Consistent with 15 this, co-IP experiments show that Upd2-WT can complex with Atg8 despite its lipidationdefective state (Extended Data Fig.2C; lane 2). In sum, these observations support a model in which the nuclear pool of Atg8 licenses Upd2's nuclear exit, and during fasting, it is the depletion of the nuclear Atg8 pool due to lipidation that triggers Upd2 nuclear accumulation. 20 We next wondered whether our studies of adipokine nuclear accumulation in Drosophila S2R+ cells, which are derived from macrophage like cells, are of any relevance to adult fat cells, the source tissue of adipokine signaling. We first examined whether fasting induces adipokine nuclear accumulation like we observed in Drosophila S2R+ cells, is visible in fat cells. To do this, we generated endogenous CRISPR engineered tag knock-ins into the Upd2 locus with two 25 different tags (HA and GFP-See Methods) to follow Upd2's endogenous localization. Both these genomic tagged versions of Upd2 were able to rescue the fat storage defects (Extended Data Fig.3 A, B) that have been reported for a upd2-deletion allele (upd2Δ) 4 . In addition to the genomic tagged Upd2 alleles, we also used flies that expressed an mCherry-tagged Upd2 cDNA transgene in fly fat (Lpp-Gal4> UAS-Upd2::mCherry). We acutely fasted flies with either tissue-30 specific expression of Upd2-WT::mCherry (Fig. 6Ab, 6A'a) or the endogenously tagged Upd2::HA ( Fig.6A'b) and Upd2::GFP (Fig.6A'c) flies. In all these different over-expressed and 15 endogenously tagged versions of Upd2, we observed that Upd2, which is usually barely detectable in fat cells, and especially their nuclei, exhibited a perceptible increase nuclear localization (Fig.6Ab) of Upd2 levels (Fig.6A'; +41-51% (p<1.1E-6)). This increase is comparable to the acute fasting induced Upd2 accumulation in S2R+ cells (Fig.1Bb), providing evidence for conservation of our findings in S2R+ cells. 5 To ask whether Atg8 controls adipokine accumulation in adult fly fat cells, we first performed fat-tissue specific acute (Upd2crGFP; ppl-Gal4tubGal80t s ; Fig. 6Ba) and 'mild'chronic (Upd2crGFP; Lpp-Gal4; Fig. 6Bb) and assessed endogenous Upd2 nuclear accumulation. Note for the 'mild-chronic' Atg8-KD crosses were maintained at 18c (a temperature at which RNAi based KD is weak in fruit-flies). Then when the F1s were 7 days old, 10 they were shifted to 29c for a 5-7day Atg8 knock down. Tissue-specific knock down of Atg8 using RNAi, using such an acute 5-7 day period, resulted in a moderate (+16%-26%) but statistically significant (p=0.03 or p=0.05) increase in Upd2 nuclear accumulation. Given we only performed Atg8-KD in adult fat for an acute period, these moderate effects are likely due to a weaker Atg8-KD. Nevertheless promisingly, these results (Fig. 6B) of in vivo adult fat tissue-specific Atg8- 15 KD's effect on Upd2 accumulation are consistent with our findings in Drosophila S2R+ cells that
Then, we reasoned that examination of the Upd2-AIM-transgenes was likely to provide the most clarifying result. Atg8-KD in fat tissue, even in clones, can have numerous effects that might complicate interpretation. Furthermore, the AIM-mutation represents a "Atg8-null" situation 20 for Upd2, and correlates strongly with Atg8-KD phenotype in S2R+ cells. To recap, the Upd2-AIMmutation does not co-IP with Atg8 ( Fig.2C) and displays strong nuclear accumulation ( Fig.2Eb, E'). Hence, we examined the effect of the point-mutation to Upd2's AIM, on Upd2's localization in adult fly fat in a upd2Δ background (upd2Δ; Lpp-Gal4> UAS-Upd2-AIM-). We observed a striking increase in nuclear accumulation of the Upd2-AIM-transgene, even in adult 25 fat cells from well-fed flies (Fig.6Cb), compared to Upd2-WT (Fig.6Ca) which, at the same imaging settings, was barely detectable. In addition, the Upd2-AIM-transgenes displayed similar nuclear accumulation even in a control background (Lpp-Gal4> UAS-Upd2-AIM-) where endogenous Upd2 is intact (Extended Data Fig.5Ab). These results suggest that the mechanisms that we uncovered for Atg8's control of Upd2 nuclear exit in Drosophila S2R+ cells, 30 are conserved in adult fly fat. Furthermore, they suggest that by examining the effect of Upd2- 16 AIM-which accumulates in fly fat, we will be able to uncover whether Atg8-mediated regulation of Upd2's nuclear exit, and subsequent release, has an effect on systemic physiology.
In a starved state, insulin (Dilp2, Dilp5) accumulates in the fly insulin producing cells (IPCs) 70 . One of the primary functions of Upd2 in adult fly fat is to remotely signal to the brain to create a permissive environment for insulin release 4,12 . As expected, based on a prior study  78 . We found, as expected, that upd2Δ mutants (upd2Δ; Lpp-Gal4>UAS-Luc) showed increased expression of InR, Bmm, and 4ebp compared to yw 25 controls, indicating that systemic Insulin signaling is reduced in the absence of Upd2 (Fig.6F).
Moreover, we found that over-expression of Upd2-WT restored expression of all three FOXO targets to control levels, consistent with Upd2-WT's ability to promote Dilp release from upd2Δ mutant IPCs 4 . Following Upd2-AIM-over-expression, however, expression levels of InR, Bmm, and 4ebp remained high, underscoring that Upd2-AIM-cannot rescue Dilp release in upd2Δ 30 mutants (Fig.6F). Interestingly, we observed that Bmm, which encodes a lipase required for the break-down of stored fats into fatty acids, and 4ebp, which has been shown to function as a 18 metabolic brake during starvation 79,80 , were expressed at higher levels in the upd2Δ flies with over-expressed Upd2-AIM-, compared to the upd2Δ flies expressing Luc-suggesting that the presence of Upd2-AIM-may signal a state of chronic starvation to the organism, it also implies that perhaps that the adipokine which is retained in the nucleus has a direct or indirect gene regulatory role (See Discussion).

5
Atg8-mediated adipokine nuclear accumulation increases resilience to starvation and promotes post-fasting hunger: We wondered whether Atg8 controlled, fasting-induced adipokine nuclear accumulation (Fig.5C) affords an organism advantages during fasting. Consistent with this idea, we had documented in the prior section that expression of Upd2-AIM-signals a chronic starvation state, at least based 10 on its gene expression profile (Fig.6E) and insulin accumulation phenotype (Fig.6C). Given that Leptin is a satiety signal in mammals 11 , therefore we reasoned that Upd2 overexpression could indicate a satiety state. Hence, we set out to test feeding behavior (see Methods) using a widely used 81 feeding behavior assay in flies, the FLIC assay 82   Upd2-WT (upd2Δ; Lpp-Gal4> UAS-Upd2-WT) showed reduced feeding events (-50 to -80% ) compared to control transgene (Extended Data Fig.6B). We next assayed hunger-driven feeding 19 motivation (Fig.7B) and events (Fig.7C) by subjecting flies to an over-night starvation regime and measuring their feeding behavior in relation to their ad libitum fed siblings. We observed that flies expressing a control transgene in a upd2Δ background (upd2Δ; Lpp-Gal4>UAS-Luc) exhibited increased motivation to feed ( Fig.7Ba; +197%; p=0.05) and increased feeding events (Fig.7Ca), and that this increase in hunger-driven feeding events for the control was consistent we illustrate that adipokine retention on fasting has beneficial effects, promoting resilience on fasting and driving post-fasting hunger response. Therefore we reveal that Atg8 lipidation not 30 only enables cells to intrinsically manage nutrient stress 16,17 , but because Atg8 controls adipokine release or retention, it enables an entire organism to adapt its neuroendocrine axes 20 to respond to nutrient stress. We term this ability of Atg8, to promote adipokine nuclear exit in fed cells and to conversely cause adipokine nuclear retention in starved cells, 'adipokine licensing'. Altogether, we have identified a cell-intrinsic mechanism of Atg8-mediated adipokine licensing that coordinates nutrient state with organismal physiology and behavior.
Critically, our report provides a specific answer to an outstanding question: whether in fed 5 cells Atg8/LC3's localization in the nucleus has a functional role. A recent study by Liu and colleagues 23 , presented compelling evidence that the nuclear pool of Atg8, translocates to the cytosol on starvation and is primary the source of lipidated Atg8 for autophagosome formation 68 . Also, nuclear Atg8 is known to associate with lamin-associated domains and degrade nuclear lamina during oncogenic stress 30 as well as act in repression of several genes in Drosophila fat 10 during starvation 83 . However, it remains possible that Atg8 has other, yet to be uncovered functions in the nucleus of fed cells. By uncovering Atg8's ability to serve as a context-dependent 'licensing factor' to promote adipokine Upd2's nuclear exit, we provide an insight into the functional significance of nuclear Atg8 in fed cells.
More broadly, our findings indicate that Atg8 is likely to perform a context-dependent role 15 in protein localization of other AIM/LIR bearing client proteins. Atg8 is a ubiquitin-like protein 20 .
Specifically, SUMO (small ubiquitin-like modifier) is conjugated to proteins and modulates protein localization 87 , in particular NC transport 88 . Interestingly, SUMO conjugation is not essential for NC transport itself, but positions its target protein (RanGAP1) to the NPC, therefore 20 increasing the efficiency of its translocation. Similarly, Atg8 might position adipokine Upd2, and other such 'client' proteins, to enable efficient export via the classical nuclear export pathway or may serve to direct its larger client proteins to other mechanisms of nuclear export such as nuclear budding 89 .
In support of this possibility, we observed that CRM1/Emb plays a role in Upd2 nuclear 25 export (Fig.3Ab). Our evidence comes from a Emb-KD test, which is a genetic approach, and leaves open the possibility that Upd2 may need an adapter protein to access Emb/CRM1 export pathway. This idea is consistent with our analysis that Upd2 has a strong predicted NLS(sequence GRVIKRKHLE with score 10.5: a GUS-GFP reporter protein fused to an NLS with a score of 8, 9, or 10 is exclusively localized to the nucleus) allowing it to enter the nucleus. 30 However, Upd2's predicted NES is 248-254 (LCEIELTI) and the Glutamate residues in position 251 and 253 do not broach the threshold, making it difficult to say if this whole sequence is 21 indeed an NES. Thus, perhaps Upd2 requires an adapter to access the CRM1 export pathway. This is reminiscent of a mechanism where-in a nuclear import importin-α requires a specific nuclear export factor, CAS, to access the cytosol 90 . Whether Atg8 serves as such an adapter itself, or if Upd2's association with Atg8 changes its conformation allowing the cryptic NES to be exposed so that it can bind CRM1/Emb, and/or Atg8 enables Upd2 to associate with other 5 proteins that then serve to link up Upd2 to a classical nuclear export pathway are all possibilities.
Preliminary evidence that our lab gathered, using mass-spectrometry (MS) analysis of immunoprecipitated (IP) Atg8-WT, Upd2-WT and Upd2-AIM-indicate that while Atg8 and Upd2-WT, both pull-down Emb/CRM1 as a MS interactor, the Upd2-AIM-IP-MS interaction list does not contain Emb. This supports the idea that Atg8 may serve as an adapter to enable Upd2 10 access Emb/CRM1. However, the IP-MS interactors will need further biochemical validation that is being actively pursued. Importantly, despite having a putative NES, Atg8/LC3 itself does not seem to require CRM1 for its nuclear export 22 . This is consistent with the idea that it can diffuse freely across the NPC barrier gives its size (<15kDa). In sum, how Atg8 interacts with the nuclear export machinery to enable context-dependent nuclear exit of adipokine Upd2, and whether 15 other AIM/LIR containing proteins are regulated in such a fashion, will be developed in future work.
Another unexpected observation is that Upd2 requires both nuclear import, and then subsequent nuclear export, in order to be secreted (Fig.3). Impeding its nuclear entry by either pharmacological means, using Importazole (Fig. 3C, D, D'), or using an ectopic NES to shuttle 20 Upd2 out of the nucleus (Fig. 3E, F), results in an abnormal cytosolic localization that is more diffuse, as opposed to its WT punctate localization (Compare Fig. 3Ea to 3Ee), and results in a significant reduction in secretion. Furthermore, the ectopic nuclear export of Upd2 interferes with its ability to co-localize with Atg8 (Fig. 4B). This is consistent with the idea that Upd2 requires the nuclear environment for a productive interaction with Atg8. We surmise that on entry into the 25 nucleus, Upd2, via its AIM sequence, gains access to Atg8, which then routes it not only to a nuclear exit pathway, but to an appropriate secretory route (Schematic in Fig. 7D, top left panel).
Consistent with this, Upd2-AIM-mutant protein is unable to effectively complex with GRASP ( Fig. 4D) and our lab has previously shown that Upd2 uses a GRASP-mediated unconventional secretory route 3 . A tantalizing possibility is that Atg8 leads Upd2 through the outer nuclear 30 envelope, which is continuous with the endoplasmic reticulum (ER) and, thus, with the cellular secretory route 91,92 . However, further characterization is needed to define precisely how Atg8 22 targets Upd2 to the correct cytosolic location for secretion. It will also be interesting to test whether Atg8 enables Upd2 secretion via a mechanism akin to the one described by Debnath, Leidal and colleagues for RNA binding proteins, in which in a novel secretory process, called LDELS, the Atg8/LC3 conjugation pathway controls extracellular vesicle (EV) cargo loading and secretion 28, 45, 93 . 5 While much is known about Atg8 lipidation being a key node of regulation during autophagic degradation 94 , we reveal new insight on the a non-degradative outcome of fasting induced lipidation. Our results (Fig. 5B) show that adipokine nuclear retention is an outcome of fasting-induced lipidation. We show that lipidation induced depletion of the Atg8 nuclear pool (Fig. 7D, top panel) causes Upd2 retention. Consistent, with this we show that the presence of productive Upd2 secretion and its subsequent access to the blood-brain barrier by following an 15 LDELS like route 45 . But we show here that lipidation of Atg8 is not required for Upd2 nuclear exit, and in fact, we show the opposite. We show that Atg8 lipidation is the molecular switch which controls 'adipokine licensing' and determines whether Upd2 is targeted to the cytosol or is retained in the nucleus. While The Upd2-AIM-allele, which is impaired in its ability to interact with Atg8 (Fig. 2C), not only phenocopies Atg8-KD's effects on Upd2 localization and secretion (Fig. 2D), but displays an almost exclusive nuclear accumulation phenotype with barely detectable cytosolic puncta 25 (Fig. 2Eb,b'), coupled with a -96% reduction in Upd2 secretion (Fig. 2D). These observations suggest that a Upd2-AIM-mutation represents an "Atg8-null"-like state for Upd2. By expressing the UAS-Upd2-AIM-transgene in adult fly fat in an upd2Δ background, we were able to investigate how the Atg8-mediated adipokine nuclear licensing mechanism impacts an organism's ability to cope with nutrient deprivation. We identified that adipokine nuclear retention 30 promotes resilience to nutrient stress and increased hunger-driven feeding behavior (Fig. 7B, C and Extended Data Fig. 6C), despite having reduced basal feeding rate (Extended Data Fig.6B).
Thus, expression of Upd2-AIM-displays a dominant effect, even though it does not interfere with Upd2-WT localization (Extended Data Fig.5C). These results suggest that nuclear accumulation of Upd2 not only prevents its release, but also affects gene regulation, reminiscent of SREBP, a master-regulator of lipogenesis that activates transcription of lipogenic genes 74,75 . This regulation of gene expression could occur either indirectly, through regulation of Insulin 5 signaling, or directly via Upd2-dependent regulation of the nuclear landscape; thus, meriting further investigation.
In sum, we have uncovered an unexpected role for Atg8 in licensing the nuclear exit of a fly adipokine, Upd2, in fed cells, and in nuclear retention during fasting. This mechanism provides resilience during nutrient scarcity and sensitizes the post-fasting hunger response of an

5
Requests for further information, reagents, and resources should be directed to and will be fulfilled by the corresponding author Akhila Rajan (akhila@fredhutch.org).

Materials Availability
Drosophila strains generated in this study are available from the corresponding author, Akhila Rajan (akhila@fredhutch.org).

Data and Code Availability
The datasets generated in this study are available from the corresponding author, Akhila Rajan (akhila@fredhutch.org).  The following UAS lines were generated for this study: UAS-Upd2::mCherry, UAS-Upd2-AIM-::mCherry, UAS-GFP::Atg8; UAS-GFP::Atg8-PE-. To control for expression levels, all transgenes pertaining to the same gene were inserted at the same att site. Either attp40 or attP2.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
For RNAi experiments, in Fig.6B, flies were generated with the following genotype: Upd2-crGFP; Ppl-Gal4, TubGal80 ts and Upd2-crGFP; Lpp-Gal4. This strain was used for crossing to Atg8- 5 RNAi lines for analysis. The parents of this cross (F0 generation) were maintained at 18c for both the strains, F1s were allowed to eclose at 18c. A week after eclosion, the F1s were moved to 29c for 5-7 days prior to dissection and staining.
Ability of UAS-GFP::Atg8 and UAS-GFP::Atg8-PE-to rescue loss of Atg8 function was tested by introducing Atg8 variant transgenes into the Atg8-Trojan-Gal4 line 97 , in which insertion of the 10 Trojan-Gal4 into the Atg8 locus generates a lethal allele.

Cell lines
Drosophila S2R+ cells were used for all cell culture related experiments. This cell line was chosen because previous studies have validated their applicability to study autophagy 69,98,99 and protein secretion 100

METHOD DETAILS:
Cloning and Transgenic Flies 20 All cloning was done using the Gateway® Technology. For Atg8, entry cDNA was PCR amplified from Atg8 cDNA (clone LD05816 DGRC-Gold collection) and cloned into pENTR-D/TOPO using BP reaction (Gateway® BP Clonase II enzyme mix, Cat#11789-020, Invitrogen). pENTR-GRASP was previously described 3 . For Drosophila Upd2 variants, pENTR-Upd2 3 made for previous work in our lab was used. For site directed mutagenesis of putative AIM sites, pENTR-

25
Upd2 was mutagenized to convert tryptophan and leucine encoding codons to alanine. For addition of a canonical NES+, sequence corresponding to amino acids LQELLELLRL 64 was inserted after the start codon in either pENTER-Upd2-WT or p-ENTR-Upd2-AIM-. For site directed mutagenesis of Atg8 lipidation site, pENTR-Atg8 was mutagenized to convert its penultimate Glycine to Alanine. All mutagenesis was done using the Q5 ® Site-Directed Mutagenesis Kit from NEB (Cat # E0554S). The sequence of oligonucleotides used for this mutagenesis reaction are available on request. The entry vectors were then moved using LR clonase reaction (Gateway® LR Clonase® II Enzyme mix, Cat#11791-020, Invitrogen) into destination vectors compatible with fly transformation, or cell culture and with the appropriate Cterminal tags for Upd2 and N-terminal tags for Atg8. Transgenic flies were generated by 5 Bestgene.

Treatment of cells with drugs:
For drug treatment experiments, the media was replaced with media containing the drug on day 3 after transfection with upd2::GFP. 4 hours later the conditioned media was used for ELISA or imaging. Drugs used in this study include Torin1 (Cat# ab218606, Abcam) and Importazole sulphuric acid and absorbance was measured at 450nm. The TMB readings were normalized to transfection efficiency as measured from Renilla Luciferase assays (see below).

Renilla Luciferase assay:
On day 2 of the ELISA assay, after the conditioned medium was transferred for use in ELISA assays, cells were re-suspended in 50μl of PBS, and incubated with 50μl/well Renilla-Glo® Luciferase reagent (Cat# E2710, Promega) for 10 minutes and read using a multiwell luminometer.

Immunoprecipitation and Western blots:
For Immunoprecipitation (IP) from S2R+ cells, protein for each condition was prepared by lysing  For live imaging (Extended Data Fig.1), cells were incubated with ProLong Live antifade reagent (Cat# P36974, ThermoFisher), LysoView™ -633 (Cat# 70058, Biotium), and NucBlue (Cat# R37605, Thermofisher) as per manufacturer's instructions, prior to imaging in appropriate media condition. Images were captured in time lapse mode on Zeiss LSM800 in the sequential scan setting.

FLIC assay:
The Fly Liquid Interaction Counter (FLIC) assay was performed as detailed previously in Ro et al., 82 and provides a measure of continuous feeding behavior and motivation. The Digital feeding monitor (DFM) used to run FLIC assays was purchased from Sable System Inc. and instructions 10 provided in the manufacturers model were followed for set up. Flies used in experiment were aged 7 days and un-sedated flies were loaded into each DFM chamber using mouth pipettes as sedation affected the assay. For recording basal feeding behavior well-fed flies (6 replicates per condition) on standard lab-food were recorded in individual chambers for their interactions and feeding events for 3 hours. For recording hunger-driven feeding behavior flies were fasted for 15 16hrs at 29 o C on 0% sucrose, 1% agar and then loaded into chambers. The feeding behaviors of fed versus fasted (6-replicates per condition) loaded onto the same DFM were used for analysis. Analysis was performed using R. Macros used for R analysis were kindly provided by TAG assays were carried out as previously described (Rajan et al., 2017). In brief: Flies were homogenized in PBST (PBS + 0.1% Triton X-100) using 1mm zirconium beads (Cat#ZROB10,

Image quantification and statistical analysis:
Excel or GraphPad Prism 7 software was used for data quantification and generation of graphs. (C) Upd2-WT coimmunoprecipitates with Atg8 and this interaction relies on Upd2's AIM. GFP-and Myc-IPs were prepared from S2R+ cells transiently transfected with the indicated cDNAs. GFP, Myc IPs and 2% of the input were analyzed by immunoblotting for the indicated proteins. Note Upd2-AIMalways runs slightly higher than Upd2 (see asterisk) and is always more abundant in the lysate as the Upd2-AIMis not secreted (see D below), but Upd2-AIM-it is not readily detectable in the Myc-IPs (see red asterisk), despite being more abundant in input. Atg8-Myc is always more abundant in Upd2-AIMinput than in Upd2 WT input, despite loading same amount of protein in input. Normalized percent fold change in secreted GFP signal detected by GFP sandwich ELISA assay performed on conditioned media of S2R+ cells transiently transfected with Upd2-WT::GFP dsRNA for control (LacZ) or two independent Emb dsRNAs. ELISA for Upd2-WT::GFP was performed after 5 days of knockdown, and the amount of Upd2-WT::GFP in a 24-hour period (Day 5) was assessed. Statistical significance is quantified by unpaired two-tailed t-test on 6 biological replicates per condition.
(C) Effect of pharmacological inhibition of nuclear import on Upd2's release in fed cells. Quantification of relative normalized secreted GFP signal detected by GFP sandwich ELISA assay performed on conditioned media of S2R+ cells transfected with Upd2-WT::GFP and treated with drug Importazole, an Importin-beta inhibitor for 6 hours. Percent change in Upd2::GFP secretion relative to DMSO is depicted. Error bars represent %SD. Statistical significance quantified by t-test on 6 biological replicates per condition.

(D)
Effect of pharmacological inhibition of nuclear import on Upd2's nuclear localization in fed cells. Confocal micrographs of single optical-slices of Drosophila S2R+ cells transiently transfected with Upd2-WT::GFP (green) stained with Lamin (red) and treated with DMSO (a) or indicated concentrations of importazole (b, c). Scale bar is 5um and right most panel shows DIC image merge. In D', the ratio of Upd2::GFP nuclear/whole cell intensity is plotted. Note that in starved state Upd2 nuclear accumulation does not significantly increase when treated with 40uM Importazole. Each dot represents a cell, 15-30 cells were counted per condition. Statistical significance is quantified by unpaired two-tailed t-test. nuclear/whole cell intensity is plotted. Note that an NES + addition prevents Upd2 from accumulating in the nucleus even on starvation (E') and is epistatic to the AIMmutation (E''). Each dot represents a cell, 15-80 cells were counted per condition. Statistical significance is quantified by unpaired two-tailed t-test.
(F) Effect of nuclear export signal (NES) addition to Upd2-WT or AIM -(Upd2-NES + ::GFP or Upd2-NES + AIM -::GFP ) on its release. Normalized percent fold change in secreted GFP signal detected by GFP sandwich ELISA assay performed on conditioned media of S2R+ cells transiently transfected with Upd2::GFP variants: WT; NES + ; NES + AIMthe amount of Upd2-WT::GFP in a 24-hour period was assessed. Statistical significance is quantified by unpaired two-tailed t-test on 6 biological replicates per condition. (B) Effect of nuclear export signal (NES) addition to Upd2-WT's colocalization with Atg8. Confocal micrographs of single optical-slices of fed Drosophila S2R+ cells transiently transfected with Upd2-WT::GFP (green), mCherry-Atg8-WT (red) stained with Lamin (magenta). In WT state, Upd2 shows partial co-localization with Atg8. Addition of NES alters Upd2's punctate distribution in the cytoplasm and its co-localization with Atg8 is significantly diminished as quantified in B' using the Mander's coefficient. Statistical significance is quantified by unpaired two-tailed t-test (C) Effect of mutating Upd2's AIM on its colocalization with GRASP. Confocal micrographs of single optical-slices of Drosophila S2R+ cells transiently transfected with cDNAs expressing Upd2-WT::GFP or Upd2-AIM -::GFP (green) and GRASP::RFP (red) stained with DAPI (white). In fed state, Upd2-WT::GFP is localized to cytosolic puncta enriched for GRASP (white arrows). When Upd2's AIM is mutated (Upd2-AIM -::GFP) Upd2 is nuclear (white arrows), and the cytosolic puncta are not readily detectable. GRASP is exclusively cytosolic. Scale bars are 5μM.

(D)
Effect of mutating Upd2's AIM on its interaction with GRASP. GFP-and RFP-IPs were prepared from S2R+ cells transiently co-transfected with GRASP-RFP and Upd2-WT::GFP or Upd2-AIM-::GFP. GFP, RFP IPs and 2% of the input were analyzed by immunoblotting for the indicated proteins. Note Upd2-AIMalways runs slightly higher than Upd2 and is always more abundant in the lysate as the Upd2-AIMis not secreted, but it is not readily detectable in the RFP-IPs, despite loading more input.  Torin1 (b, c, d). Scale bar is 5um and right most panel shows DIC image merge. In B', the ratio of Upd2::GFP nuclear/whole cell intensity is plotted. Each dot represents a cell, 50-100 cells were counted per condition. Statistical significance is quantified by unpaired two-tailed t-test.
(B) Effect of starvation induced changes in the nuclear pool of mCherry-tagged Atg8 wildtype (Atg8-WT) versus lipidation-defective Atg8 (Atg8-PE-). Confocal micrographs of single optical-slices of Drosophila S2R+ cells transiently transfected with (a, b) mCherry::Atg8-WT (red) or (c, d) mCherry::Atg8-PE-(red) stained with Lamin (blue) in either well-fed state (a, c) or AA-starved (b, d). Scale bar is 2um and right most panel shows DIC image merge. In C', the ratio of mCherry::Atg8 nuclear/whole cell intensity is plotted. While Atg8-WT pool in the nucleus is reduced on starvation, lipidation-defective pool of Atg8 is similar between fed and starved states. Each dot represents a cell, 50-80 cells were counted per condition. Statistical significance is quantified by unpaired twotailed t-test.
(C) Effect of providing exogenous mCherry-tagged Atg8 wild-type (Atg8-WT) versus lipidation-defective Atg8 (Atg8-PE-) on Upd2 nuclear localization during starvation: Schematic, in the top panel, shows experimental design; an assay for the effect of increasing the dose of lipidation-defective Atg8 on Upd2 nuclear accumulation. Below, confocal micrographs of representative single optical-slices of Drosophila S2R+ cells transiently co-transfected with Upd2-WT::GFP and (a, b) mCherry or (c, d) mCherry::Atg8-WT or (e, f) mCherry::Atg8-PE-stained with GFP (green) and Lamin (red) in either well-fed state (a, c, e) or AA-starved (b, d, f). Scale bar is 2um and right most panel shows DIC image merge. In D', the ratio of Upd2::GFP nuclear/whole cell intensity is plotted. While Atg8-WT and control show that Upd2 nuclear accumulation increases on starvation, cells co-transfected with lipidation-defective Atg8-PE-, do not show Upd2 nuclear accumulation increase on starvation. Each dot represents a cell, 80-130 cells were counted per condition. Statistical significance is quantified by unpaired two-tailed ttest.
Also see companion Figure S2, that assays Upd2's ability to immunoprecipitate with lipidation-defective Atg8-PE-.    Figure 6A. (B, C) Effect of fat-specific expression of Upd2-AIM-on post-fasting hunger driven feeding: Quantification of feeding behavior in transgenic flies, expressing indicated cDNAs (UAS-X) specifically in fat-tissue (Upd2∆; Lpp-Gal4>UAS-X), measured using the Fly Liquid Interaction Counter (FLIC-see methods). Each dot represents a fly. 10-12 flies were used per experiment per genotype. The flies were starved for 16 hours prior to testing. The difference between fed and starved flies of the same genotype were assayed as licks i.e., interactions with food, which is a measure of feeding motivation (B) and feeding events (C). Error bars represent percent standard deviation and statistical significance was calculated using 2-tailed unpaired t-test. See companion Extended Data Figure 6 B, C.
(D) Working model: Atg8 mediates adipokine nuclear exit and its subsequent release: In a fed state, the ortholog of human Leptin in flies, Upd2, enters the nucleus; via its Atg8interaction motif (AIM) Upd2 complexes with Atg8. Atg8 then enables Upd2 nuclear exit via an Exportin/CRM1 based mechanism. Atg8 subsequently targets Upd2 to the GRASP secretory pathway. On fasting, Atg8 lipidation and recruitment to autophagy decreases the Atg8 nuclear pool. In turn, this increases Upd2 nuclear accumulation and causes cells to retain the adipokine on fasting. When Upd2's interaction with Atg8 is disrupted (AIM-), Upd2 localizes to the nucleus even in a fed state and is not secreted. In flies, constitutive expression of Upd2-AIM-in fat cells, results in a chronic starvation phenotype. Upd2-AIMflies are starvation resistant and exhibit higher hunger-driven feeding motivation.
Altogether, this suggests that fasting-induced Upd2 accumulation facilitates organismal adaptation to nutrient deprivation and drives post-fasting hunger response. When cells are dosed with lipidation-defective nuclear Atg8, Upd2 export from the nucleus occurs even on fasting. Future studies, indicated by question marks, are required to evaluate whether Upd2 which is exported from the nucleus by lipidation-defective Atg8 can be targeted to the appropriate secretory compartment. If indeed Upd2 can be subsequently released via a lipidation-defective route, then ensuing investigations should explore whether such a lipidation-defective Atg8 mediated secretion allows Upd2 to reach its appropriate neuronal target.   Figure 4A) Confocal micrographs of time-lapse shots, acquired every 45 seconds(s), of Drosophila S2R+ cells cultured in complete media and transiently transfected with cDNAs for tagged Upd2::GFP (green) and mCherry::Atg8 (magenta); live cell dyes mark nucleus (NucBlue™ -dark blue). Atg8 and Upd2 exhibit discrete punctate localization. The Atg8 and Upd2 puncta show dynamics of both colocalization as well as association in: i) nuclear periphery (blue arrows), ii) cell periphery (white arrows). Cell outlines and nuclear outlines are marked for clarity and based on DIC images of the cells. Scale bars, 2μM.  Anti-Atg8