O-GlcNAcylation of PERIOD regulates its interaction with CLOCK and timing of circadian transcriptional repression

Although O-GlcNAcylation has been shown to influence PER function, specific residues that are modified by O-GlcNAcylation have not been identified [23–24]. This represents a critical barrier to understand the function of site-specific O-GlcNAcylation events and the interplay between O-GlcNAcylation and phosphorylation to regulate the phase-specific functions of PER. We therefore sought to identify PER O-GlcNAcylation sites and obtain temporal data on their occupancy. We purified FLAG-tagged PER from heads of wper⁰; p{3XFLAG-per(WT)} flies at specific time-points over the circadian cycle using FLAG affinity purification and performed quantitative mass spectrometry (MS) to obtain a circadian profile of PER phosphorylation and O-GlcNAcylation using fly tissues as starting materials. Although we attempted to perform this study using both fly head and body tissues, we were only able to identify PER O-GlcNAcylation sites in head tissues since we were unable to pull down sufficient PER proteins in bodies for comprehensive PTM identification. Nevertheless, we postulate that PER residues that are O-GlcNAcylated in head tissues will also be O-GlcNAcylated in peripheral tissues due to the ubiquitous expression of ogt [28]. To enable quantitation of PTM sites, flies were fed with ¹⁵N-labeled (heavy) yeast or ¹⁴N (light) yeast for two generations to ensure complete labeling in flies (herein termed ¹⁵N and ¹⁴N flies) [29]. To ensure that ¹⁵N-fed and ¹⁴N-fed flies show similar behavioral rhythms, we examined their locomotor activity rhythms using Drosophila activity monitoring [30]. Both types of flies displayed strong behavioral rhythms with periods close to 24-hr in constant darkness (S1A and S1B Fig). Furthermore, we measured PER daily abundance in ¹⁵N and ¹⁴N flies to confirm that temporal expression of PER was not altered as a result of the diets and observed no difference (S1C Fig). To profile PER PTMs in vivo, ¹⁵N and ¹⁴N flies were collected at six time-points over the circadian cycle (ZT1, 3, 12, 16, 20, 24), and protein extracts from heads of ¹⁵N and ¹⁴N flies were separately subjected to FLAG affinity purification (AP) prior to sample preparation for LC-MS/MS analysis (S2A and S2B Fig). Relative PTM quantification was achieved by pooling purified PER from all ¹⁴N samples, and aliquoting equal amounts of pooled ¹⁴N fractions to each of the six ¹⁵N purified PER samples at a 1:1 ratio prior to MS analysis (S2C Fig).

Despite the large amount of fly head tissues we used for our protein extraction and FLAG-PER purification, our ¹⁵N/¹⁴N quantitative MS did not yield satisfactory temporal resolution of PER PTM cycling. We therefore consolidated our MS data from multiple time-points with the goal of identifying PER O-GlcNAcylation sites qualitatively (Fig 1 and Table 1). We observed that multiple residues located in the CBD and within the CLK:CYC inhibition domain (CCID) [31, 32] are O-GlcNAcylated. These include PER(S942) as well as a potential sites at S951, T952, or T954. Our MS analysis yielded O-GlcNAc-modified peptides in which only PER(S942) is modified, providing unambiguous identification of PER(S942) as an O-GlcNAcylated residue (Fig 1 and Table 1). In addition, O-GlcNAc-modified peptides spanning S942 to T954 were also identified, but we were not able to narrow down the single modified residue within these peptides. It is possible that only S942 is O-GlcNAcylated within this region. Alternatively, either S951, T952, or T954 may represent a second O-GlcNAcylated residue within this region. Taken together, we hypothesize that O-GlcNAcylation in the CBD may be important in modulating PER-CLK interaction and PER repressor activity. In addition to the O-GlcNAcylated residues in the CBD, a number of other potential sites were identified in other parts of the PER protein, but most of them are not located in characterized functional domains.

In addition to identifying PER O-GlcNAcylation sites, we took this opportunity to identify PER phosphorylation sites in fly tissues for the first time and confirm phosphorylation sites that have previously been identified in Drosophila S2 cells [17–19, 21]. Our analysis provides evidence that the majority of phosphorylation sites identified in Drosophila S2 cells are bona fide phosphorylation sites in fly tissues (Table 1). During the process of optimizing PER affinity purification and MS analysis using fly tissues as starting materials, we generated additional qualitative label-free MS datasets by analyzing PER phosphorylation sites at multiple time-points over the circadian cycle (ZT3, 16, 20, 24). S1 Table summarizes the label-free data and compares it to the phosphorylation sites identified in the ¹⁵N/¹⁴N-labeled MS analysis. PER O-GlcNAcylation site identification was not performed as part of the label-free MS analysis. We observed a high level of congruence between the PER phosphorylation sites identified in our two fly head tissue data sets, as well as those generated using Drosophila S2 cells as starting materials [17–19, 21]. In summary, our MS analysis sets the stage for future studies to understand the functional interplay between PER O-GlcNAcylation and phosphorylation.

Subsequent to the identification of PER O-GlcNAcylation sites, we proceeded to analyze the function of site-specific PER O-GlcNAcylation by mutating one or a cluster of S/T residues to non-O-GlcNAcylatable alanine on the per gene and evaluated PER repressor function on CLK activity using the per-luciferase (per-luc) reporter assay in Drosophila S2 cells [3, 37]. To prioritize, we analyzed PER O-GlcNAcylation residues with a Mascot score >50 (Table 1). We compared per-luc activity in S2 cells expressing per wild type or mutant variants (Fig 2A). In order to take into account the varying expression level of the PER variants (Fig 2B top panel, S3A and S3B Fig), we normalized CLK-activated per-luc activity observed for PER(WT) and PER mutant variants to their respective protein expression levels to more accurately assess the impact of blocking PER O-GlcNAcylation at specific residues on repressor activity (Fig 2C). We observed that all but one PER O-GlcNAcylation site mutant (PER(S174A)) exhibited significant increase in repressor activity. Interestingly, PER(S942A) and PER(T951A-S954A) are the two O-GlcNAc site mutants that exhibited the strongest repressor activity, further strengthening the rationale for testing the hypothesis that O-GlcNAcylation in the CBD may be important in modulating PER-CLK interaction and PER repressor activity. For this reason, this current study will focus on investigating the function of PER O-GlcNAcylation in the CBD, while analysis of other PER O-GlcNAcylation residues will be pursued in the future.

To confirm this finding in whole animals, we generated transgenic flies expressing p{per(WT)-HA10HIS} (herein referred to as per(WT) and mutant variants) in the per⁰ genetic background [38] so that only transgenic per (WT or mutant) was expressed. First, we performed quantitative RT-PCR to measure clock gene expression (per and tim) in heads of per(S942A) flies to examine the function of PER(S942) O-GlcNAcylation in central clock oscillators. All flies used for molecular analysis are homozygous for the per transgene. As predicted from the elevated PER repressor activity observed in S2 cells, we found that cycling of per and tim mRNAs was significantly dampened in per(S942A) flies as compared to per(WT) flies in both LD cycles and constant darkness (DD) (Fig 2D and 2E). Furthermore, per and tim mRNAs exhibited earlier initiation of repression phase in DD (Fig 2E). Since the per(S942A) mutant exhibited lowered levels of clock gene mRNAs, PER protein level is expected to decrease. As expected, peak PER abundance was significantly reduced in the heads of per(S942A) flies than in per(WT) flies in both LD and DD conditions (Fig 2F and 2G). To ensure that lower level of PER protein is a result of increased PER(S942A) repression on CLK-dependent transcription of per rather than reduced PER stability, we monitored the rate of PER degradation by cycloheximide (CHX) chase assay in S2 cells. The PER(S942A) mutant degraded at a similar rate as PER(WT) in the presence of OGT, demonstrating that the S942A mutation has no significant effect on PER stability (S3C and S3D Fig).

Since peripheral tissues are known to be more sensitive to metabolic fluxes [39, 40] and O-GlcNAcylation is a nutrient-sensitive PTM, we also analyzed the effect of blocking PER(S942) O-GlcNAcylation in oscillators of peripheral tissues, specifically the fat body. The fat body, which is analogous to mammalian liver and adipose tissue, plays an essential role in regulating energy metabolism in insects [39, 40]. We assayed per and tim mRNA levels in the fat body of per(WT) and per(S942A) flies on DD1 after LD entrainment to evaluate PER-dependent repression of CLK activity. Consistent with what we observed in fly heads, per(S942A) flies displayed significantly dampening of per and tim mRNA cycling as compared to per(WT) flies in the fat body (Fig 2H and 2I). This suggests that PER(S942) O-GlcNAcylation normally weakens the activity of PER to repress CLK-dependent transcription in both head and fat body oscillators.

As in PER(S942A) mutant, PER(S951A/T952A/T954A) mutant also exhibited a significant difference in repressor activity as compared to PER(WT) in per-luc reporter assay (Fig 2C). To evaluate the effects of these residues, which are also located in the CBD, we generated wper⁰; per(S951A/T942A/T954A) transgenic flies to confirm the effects of the PER(S951A/T952A/T954A) mutations in whole animals. We measured temporal cycling of per and tim mRNAs in heads of per(S951A/T952A/T954A) flies and observed dampened expression of per and tim mRNA as compared to per(WT) flies in both LD and DD conditions (S4A and S4B Fig), although not to the extent observed in per(S942A) flies. However, this did not translate into significant differences in PER protein abundance and cycling (S4C and S4D Fig). The differential effects observed when blocking O-GlcNAcylation at PER(S942) and PER(S951/T942/T954) suggest that PER(S942) O-GlcNAcylation is a key event in regulating PER repression within the CBD. It is important to stress that PER(S942) has never been identified as a phosphorylation site in all previous comprehensive mapping studies [17–19, 21], so it is highly unlikely that the effects for S942A mutation is due to disruption of PER(S942) phosphorylation.

Given that several of the non-O-GlcNAcylatable per mutants exhibited elevation in PER repressor activity in S2 cells and in flies, we proceeded to investigate if this molecular phenotype can translate to alterations in output behavioral rhythms. We first evaluated daily locomotor activity rhythms of per(WT) and mutant flies. Activity rhythm is a robust readout that reflects the function and speed of the central oscillators located in fly heads [30]. Flies were entrained for 3 days in LD cycles followed by 7 days in DD. As expected, heterozygous per(WT) flies manifested robust rhythms with ~24-hr periods, indicating full rescue of arrhythmic per⁰ mutation (Fig 3A). Homozygous per(WT) flies displayed somewhat shorter behavioral rhythms due to the extra copy of per. Interestingly, as compared to per(WT), per(S942A) and the triple per(S951A/T952A/T954A) mutants homozygous for their respective transgenes exhibited shorter periods by 1.4 and 0.7 hrs. The earlier initiation of repression previously observed in per(S942A) mutant likely contributes to its short-period phenotype. In comparison, per(S951A/T952A/T954A) mutants displayed a smaller though significant change in repression activity, which likely accounts for the smaller change in period length.

In addition to monitoring locomotor activity rhythms of per(S942A) flies, we also examined the effects of blocking PER(S942) O-GlcNAcylation in peripheral oscillators by measuring feeding rhythms. Feeding assays were only performed to compare per(WT) and per(S942A) flies as per(S951A/T952A/T954A) mutants displayed minor changes in activity rhythms and clock gene repression. Feeding activity rhythms are governed by oscillators of metabolic tissues (i.e. in the fat body) [40], which are more sensitive to nutrient flux than central oscillators in the brain. Since we demonstrated earlier that clock gene cycling was dampened in the fat body of per(S942A) flies as a result of PER being a stronger repressor, we expected that feeding rhythms will also be disrupted. We entrained per(WT) and per(S942A) flies in 12h:12h LD cycles and assayed their feeding activity rhythms for two consecutive days either in LD or DD conditions using the CAFE assay [41]. The amount of food consumed over a 2-hr period was determined at 4-hr intervals over the circadian cycle. As control, per⁰ flies were measured in parallel, and as expected they exhibited arrhythmic feeding activity in LD and DD (Fig 3B and 3C) as determined by JTK-cycle (P = 1) [42].

per(WT) flies displayed robust rhythms of feeding behavior (JTK-cycle (P < 0.01)); they feed during daytime and fast at night in both LD and DD conditions (Fig 3B and 3C). Robust feeding rhythms were also observed in per(S942A) flies in LD (JTK-cycle (P < 0.01)), but the peak was phase advanced. This is consistent with the shorter period length in locomotor activity observed in the per(S942A) mutant. The consequence of blocking O-GlcNAcylation at PER(S942) on feeding behavior is even more severe in DD in the absence of light cues as rhythmic feeding was abolished in DD (JTK-cycle (P = 1)). Together, our data demonstrates that O-GlcNAcylation at PER(S942) modulates the function of central and peripheral oscillators by regulating PER activity.

Premature PER nuclear entry during the night could explain why per(S942A) mutant exhibits stronger and/or premature initiation of repression on CLK activity. A previous study showed that PER O-GlcNAcylation is involved in regulating nuclear entry [23]. To examine this possibility, we monitored timing of PER nuclear entry in adult clock neurons in LD cycles. In the fly brain, rhythmic expression and nuclear localization of PER in lateral clock neurons (LN_vs) are necessary for proper clock function and maintenance of rhythmic locomotor activity (reviewed in [2]). The lateral neurons (small and large; l-LN_vs and s-LN_vs) express pigment dispersing factor (PDF), which has been commonly used to label the cytoplasm of the LN_vs [43]. Despite having shorter behavioral rhythms, we found that per(S942A) mutants did not exhibit a significant difference in accumulation of nuclear PER as compared to per(WT) flies at time-points when nuclear entry is most prominent, ZT18 to ZT22 [4] (S5A and S5B Fig). Although a previous report demonstrates that global knockdown or overexpression of OGT in clock neurons affects the timing of PER nuclear entry [23], it is not surprising that site-specific non-O-GlcNAcylatable per mutants, in this case per(S942A), may not show the same phenotype. As such, our results suggest that the stronger and premature initiation of CLK repression observed in heads and fat bodies of short-period per(S942A) flies cannot be explained by altered timing of PER nuclear localization in circadian oscillators.

Given that PER does not enter the nucleus prematurely in the LN_vs of per(S942A) mutant flies, we speculate that the PER(S942) may have a higher affinity to CLK that results in stronger and/or premature initiation of repression. We therefore examined PER-CLK interactions by performing co-immunoprecipitations (coIP) using head extracts from per(WT) and per(S942A) flies. These experiments were performed in head tissues as IP reactions of clock proteins are more efficient in head tissues and produce more robust results. Indeed, the per(S942A) flies showed significantly higher PER-CLK interaction at ZT16 than per(WT) flies (Figs 4A, 4B and S6A). To validate our results, we performed coIPs using S2 cells coexpressing per(WT)-cmyc or per(S942A)-cmyc with clk-V5, in the presence or absence of pMT-ogt-FLAG. Consistent with the fly data, we observed that PER(S942A) proteins exhibited significantly stronger binding to CLK as compared to PER(WT) (Fig 4C and 4D). Interestingly, when PER(S942A) mutant was coexpressed with CLK in the presence of OGT, the level of PER-CLK binding was lower than the levels when PER(S942A) mutant was coexpressed with CLK in the absence of OGT. This suggests that O-GlcNAcylation of CLK or perhaps at other PER residues may negatively impact PER-CLK interaction. Nevertheless, our results suggest that stronger and/or premature initiation of repression phase in per(S942A) flies is due to higher affinity between CLK and PER(S942A).

In comparison to per(S942A) mutant flies, per(S951A/T952A/T954A) flies only displayed minor changes in activity rhythms and clock gene expression in flies. Based on these observations, the PER(S951A/T952A/T954A) mutant protein is predicted to exhibit minor changes on PER-CLK interactions, despite that these three residues are also localized within the CBD. As expected, no significant changes in PER-CLK interactions were observed between per(WT) and per(S951A/T952A/T954A) flies at the indicated time-points (S7A and S7B Fig). Thus, our results suggest that O-GlcNAcylation events at PER(S951/T952/T954) only have minor modulatory effects on PER-CLK interactions. Alternatively, it is possible that only PER(S942) is O-GlcNAcylated within the CBD (Fig 1).

Since part of PER repressor function is to remove CLK from clock gene promoters, we would expect that CLK may be removed prematurely when PER(S942) O-GlcNAcylation is blocked. We therefore measured CLK occupancy at the E-box elements of per and tim promoters by performing chromatin immunoprecipitation (CLK-ChIP) using extracts from adult fly heads. Consistent with our hypothesis, we found that flies expressing per(S942A) showed significantly reduced CLK occupancy on per and tim promoters at ZT18 and ZT20 as compared to per(WT) flies (Fig 4E). To rule out the possibility that decreased CLK occupancy observed at ZT18 and ZT20 in per(S942A) flies was due to decreased CLK levels, we examined CLK levels after the ChIP reactions and found that CLK was not limiting in flies expressing per(WT) or per(S942A) (S6B Fig).

To further support our hypothesis that dynamic O-GlcNAcylation at PER(S942) is critical for regulating the strength and/or timing of PER-dependent clock gene repression, we overexpressed 3XFLAG-ogt in tim-expressing clock neurons using a tim(UAS)-gal4 driver (referred to as TUG) [44] and examined PER-CLK interactions in head extracts of these flies. First, we verified that ogt overexpressors (TUG>FLAG-ogt) exhibited a significant ~2-hr period-lengthening of behavioral rhythms, as previously observed [23, 24] (Fig 5A). All 4 independent ogt overexpressor fly lines we generated showed similar period-lengthening phenotypes as compared to parental controls. To confirm that the overexpressed OGT enzyme is functional, we optimized O-GlcNAc chemoenzymatic labeling to measure levels of O-GlcNAcylated proteins in fly head extracts.

Currently, detecting protein O-GlcNAcylation remains to be a challenge when using traditional protein analytical techniques because the addition of a sugar group does not influence the migration of a polypeptide in gel electrophoresis (reviewed in [45]). Additionally, commercially available O-GlcNAc-specific antibodies or lectin yield non-specific signals in our hands. Thus, we chose to use a more sensitive and specific chemoenzymatic labeling method to examine in vivo protein O-GlcNAcylation status in this study [46–48].

We first tested the specificity of this approach by detecting O-GlcNAcylated PER and OGT in Drosophila S2 cell culture. S2 cells were transiently transfected with V5 tagged per with or without pMT-ogt-FLAG. Immunoprecipitated PER and OGT were labeled with resolvable mass tags prior to SDS-PAGE and western blotting. Results showed successful labeling of O-GlcNAcylated PER with 20 kD PEG mass tag in the presence of OGT, leading to clear mobility shift of O-GlcNAcylated PER isoforms (S8A Fig). As OGT is known to be auto-O-GlcNAcylated, addition of 10kD PEG mass tag in the reaction resulted in O-GlcNAcylated OGT isoforms that appeared as slower migrating bands as detected by Western blots (S8B Fig). The use of PEG mass tag labeling provides two advantages. First, both modified and non-modified isoforms can be detected, providing information on stoichiometry. Second, epitope-tagged antibodies or target-specific antibodies can be used for detection of all isoforms. Alternatively, O-GlcNAcylated proteins can also be labeled with biotin to facilitate detection with α-streptavidin antibodies [46–48]. Our success in detecting and visualizing in vivo O-GlcNAcylation status of proteins in S2 cells by chemoenzymatic labeling allowed us to proceed and confirmed that PER O-GlcNAcylation was elevated in head extracts of ogt overexpressor flies, an indication that overexpression of ogt led to increase in enzyme activity (Fig 5B and 5C).

Since blocking PER(S942) O-GlcNAcylation promotes PER-CLK interaction, we expect that PER-CLK interaction should be reduced in ogt overexpressors assuming PER(S942) is hyper-O-GlcNAcylated. As anticipated, we observed that control TUG flies exhibited significantly higher PER-CLK interaction at ZT18 as compared to ogt overexpressor flies in coIP assays using head extracts of flies (Fig 5D and 5E). For ogt overexpressor flies, the reduction in PER-CLK interaction observed in fly heads may partially account for the physiological effect of OGT overexpression, i.e. period-lengthening of clock-controlled locomotor activity rhythms (Fig 5A).

We speculate that PER(S942) O-GlcNAcylation occurs either during the daytime or early night to prevent de novo PER from prematurely binding to CLK or at late night to facilitate CLK dissociation from PER after PER-dependent repression. Both of these scenarios would result in stronger clock gene repression in per(S942A) flies. Since we observed that the differences in PER-CLK interactions between per(WT) and per(S942A) flies as well as between WT TUG control and ogt overexpressor flies occurred at around the start of the circadian repression phase (i.e. ZT16 and ZT18 respectively; Fig 4B and 5E), we postulated that the former scenario may be more likely. Temporal data on PER(S942) O-GlcNAcylation status would certainly help to rule out one of the two scenarios. It is therefore unfortunate that we are not able to determine the timing of PER(S942) O-GlcNAcylation by MS. Kim et al. [23] and Kaasik et al. [24] have previously examined global PER O-GlcNAcylation status over the circadian cycle using anti-O-GlcNAc antibodies in combination with Western blotting, but their results were incongruent. Whereas Kim et al. observed that PER O-GlcNAcylation peaks at ZT20 [23], Kaasik et al. observed a peak in PER O-GlcNAcylation at ZT10 [24]. We therefore opted to profile global PER O-GlcNAcylation status by chemoenzymatic labeling to gain insights into the timing and function of PER(S942) O-GlcNAcylation.

We examined the temporal profile of PER O-GlcNAcylation in head extracts of per(WT) flies in LD condition by labeling PER with either 20kD PEG or a biotin tag [46–48]. In both cases, maximal PER O-GlcNAcylation occurred at around ZT4 to ZT8 with subsequent decline from ZT12 to ZT24 (Fig 6A and 6B). Our results were more in line with the temporal PER O-GlcNAcylation profile observed in Kaasik et al. [24]. Daily PER O-GlcNAcylation cycle was found to be significantly rhythmic (JTK cycle; P < 0.0005) (Fig 6C). Our results reveal that PER is more highly O-GlcNAcylated during the day. This suggests that PER(S942) is likely O-GlcNAcylated during the day and may persist into early night to prevent de novo PER from prematurely interacting with CLK to initiate the repression phase.

O-GlcNAcylation of cellular protein is sensitive to nutrient input [25, 49]. Since we observed that PER O-GlcNAcylation is higher during the day and gradually decreases over the circadian cycle, we expect that this temporal pattern may correlate to daily feeding activity. We measured feeding rhythms in per(WT) flies fed ad libitum using the CAFE assay [41]. The flies used for these assays were entrained simultaneously with flies used for PER O-GlcNAc labeling to better assess correlation between feeding activity and PER O-GlcNAcylation. We found that mixed sexes of per(WT) flies displayed rhythmic feeding that peaked during early day in LD condition (S9A Fig). Similar results were obtained when male and female per(WT) flies were housed and tested separately (S9B Fig). Furthermore, a separate experiment comparing feeding activity rhythms of per(WT) and per(S942A) flies also showed higher daytime feeding activity in per(WT) flies (Fig 3B). Taken together, our data suggest that nutrient flux via feeding activity provides time-of-day metabolic signals to the circadian oscillator via temporal O-GlcNAcylation of PER. Specifically, PER(S942) O-GlcNAcylation, which is expected to occur during the day and persist into early night, may prevent interaction of PER and CLK prematurely to regulate timing of circadian repression.

To profile PER PTMs, we generated transgenic flies that expressed a 13.2 kb genomic clone of per in w¹¹¹⁸per0 (wper⁰) background. A previously characterized vector that contains a 13.2kb per genomic fragment tagged with HA and 10X histidine at the carboxyl terminal (pCaSpeR-per(13.2WT)-HA-10HIS) [55] was used as the template for inserting 3XFLAG at the amino terminal before the starting Methionine to facilitate FLAG Affinity Purification. Transformants were generated by P-element transformation (BestGene Inc., Chino Hills, CA), and the 3XFLAG-per(13.2WT)-HA-10HIS transgene was tested for functionality by determining whether it can rescue wper⁰ flies in behavioral and molecular assays.

To generate transgenic flies carrying wild-type (WT) or O-GlcNAcylation site mutants of per, we opted to use PhiC31 site-directed recombination [56]. The genomic per(13.2WT)-HA-10HIS was excised from pCaSpeR-per(13.2WT)-HA-10HIS using the restriction sites XhoI and BamHI and subcloned into pattB vector (kind gift from Amita Seghal) to yield pattB-per(13.2WT)-HA-10HIS. The pattB vector was modified so that the sites KpnI and XbaI were removed, the BgIII site was replaced by a BamHI site, and the BamHI site was replaced by a BgIII site. For generating flies expressing non-O-GlcNAcylatable mutants on per, a 3.4 kb XbaI-BamHI or 1kb BamHI-KpnI genomic fragment was excised from the pattB-per(13.2WT)-HA-10HIS plasmid and subcloned into a pGem7 vector [20]. The resulting pGem7-per(XbaI-BamHI) or pGem7-per(BamHI-KpnI) plasmid served as a parent template for site-directed PCR mutagenesis (Agilent Technologies, Santa Clara, CA) depending on the location of the O-GlcNAc site (S2 Table for mutagenic primer sequences). After mutagenesis and confirmation by Sanger sequencing (UC Davis Sequencing), the mutant variants of 3.4 kb or 1 kb per subfragments were used to replace the WT fragments in pattB-per(13.2WT)-HA-10HIS. Plasmids were injected into w¹¹¹⁸ fly embryos carrying attP sites on chromosome 3 (attP2) (BestGene, Inc.). Transformants were crossed with wper⁰ flies to remove endogenous copies of per prior to behavioral and molecular analyses.

To generate flies overexpressing ogt, ogt was amplified and subcloned into pUAST-attB vector [57]. 3XFLAG was added to the N-terminus of ogt during the cloning process. Plasmids were injected into w¹¹¹⁸ fly embryos carrying attP sites on chromosome 3 (attP2) (BestGene, Inc.). To express 3XFLAG-ogt in clock neurons, transgenic flies carrying the UAS-FLAG-ogt transgene were crossed to w; tim-(UAS)-GAL4 (referred to as TUG) driver line [44].

CAFE assay was performed as described [41] with modifications. Mixed-sex population of five male and five female wper⁰, wper⁰; per(13.2WT), and wper⁰; per(S942A) flies or separately housed male or female (10 per group) flies were fed Bloomington Drosophila Stock Center standard fly food during entrainment in 12 h light/12 h dark day-night cycles and food consumption were measured starting on the third day of LD or the first day of DD. Prior to the day of measurement, grouped flies were transferred to a vial containing 2% agar as the medium with 5% sucrose solution maintained in calibrated glass micropipettes (VWR). After 24 hours of training, old micropipettes were replaced by fresh experimental micropipettes filled with 5% sucrose solution approximately 2 hours before each of the indicated time-points. After 2 hours, the amount of liquid consumed from the experimental micropipette was recorded, and the evaporation effect was evaluated by measuring the change in liquid volume in a micropipette placed in a vial without flies. Food consumption for each group/vial was determined by subtracting the amount of liquid consumed from the experimental micropipette with the amount of evaporated liquid. These values were normalized to the amounts of flies in the vial that survived until the end of the experiment. These experiments were performed in biological triplicates (one group/vial represents one independent experiment). Error bars = SEM. Rhythmicity of feeding was determined by JTK Cycle [42].

Proteins from S2 cells and fly heads were extracted using modified RIPA buffer as previously described [58]. Extracts were quantified and either directly analyzed by immunoblotting or incubated with 15μ α-V5 resin (Sigma) or 20μl of α-HA resin (Sigma) at 4°C for 4 hours. Beads were washed once in M-RIPA and twice with reaction buffer (20mM HEPES pH 7.9, 50mM NaCl, 1μM PUGNAc, 25mM NaF, 0.5mM PMSF, and 5mM MnCl₂) supplemented with 1x protease inhibitor (Sigma) [59]. Procedures for chemoenzymatic labeling with biotin or PEG (Polyethylene Glycol) 10kD or 20kD mass tag were performed as described [48] with modifications. Attachment of biotin or PEG mass tag to O-GlcNAc group requires a two-step derivatization process [59]: (1) a mutant galactosyltransferase, GalT1 (Y289L), utilizes UDP-azidogalactose (UDP-GalNAz) as substrate to add an azide onto the O-GlcNAc group; (2) Biotin alkyne or an alkyne-functionalized PEG mass tag indirectly attaches to the O-GlcNAc group via azide-alkyne cycloaddition chemistry. Briefly, after immunoprecipitation followed by washes, immune complexes were resuspended in 20 μl of reaction buffer containing 2 μl of Gal-T1 Y289L (Invitrogen) and 2μl of 0.5mM UDP-GalNAz (Invitrogen) [59] following overnight incubation at 4°C with gentle rotation. Azide-labeled beads were washed twice with reaction buffer and subsequently resuspended in 50μl of labeling buffer (1% SDS and 50mM Tris-HCl pH 8.0). The samples were reacted with biotin alkyne (Invitrogen) or an alkynyl-functionalized poly(ethylene glycol) (10-kDa or 20-kD) (Creative PEGWorks, Chapel Hill, NC) according to manufacturer’s protocol or a previously described protocol by [59] respectively. Samples were eluted in 50μl of 1XSDS sample buffer. For proteins labeled with a mass tag (PEG), PER or OGT was resolved using SDS-PAGE (5% or 8% minigel with a 40:1 acrylamide/Bis-acrylamide solution, Bio-rad, Hercules, CA). For proteins labeled with biotin, PER was resolved using SDS-PAGE (5% Criterion gels, Bio-rad). Antibody dilutions to detect O-GlcNAcylated PER or OGT proteins are as follows: α-V5 (1:5000), α-HA (1:1000), α-FLAG (1:7000), α-streptavidin (Cell Signaling Technologies, Danvers, MA) (1:5000), and α-PER (GP5620; RRID:AB_2747405) (1:2000).

wper⁰; p{3XFLAG-per(13.2WT)-HA10HIS} flies were fed with an ¹⁵N diet (0.2g ¹⁵N yeast, 1% Bacto agar, 15% unsulfured molasses, phosphoric and propionic acid mix, and tegosept). Saccharomyces cerevisiae were metabolically ¹⁵N-labelled as described [29, 60]. As control, flies were also fed with ¹⁴N diet. The adult progenies of ¹⁵N- or ¹⁴N-fed parental flies were reared in ¹⁵N- or ¹⁴N diet and entrained for 3 days in 12hr light:12hr dark at 25°C and collected every four hours over a period of 24 hours on the fourth day. Upon collection, flies were immediately frozen on dry ice until protein extraction. For each time-point, 4ml of fly heads were homogenized into fine powder by a chilled mortar and pestle and resuspended in Lysis Buffer (20mM HEPES pH 7.9, 5% glycerol, 350mM NaCl, 0.1% Triton X-100, 1mM DTT, 1mM MgCl₂, 0.5mM EDTA, 25mM sodium fluoride, 1x protease inhibitor (Sigma, St. Louis, MO), 1x PhosSTOP (Roche, South San Francisco, CA). Homogenate was dounced for 15 strokes using a 50ml tissue grind tight pestle (Kimble-Chase, Vineland, NJ) and were filtered using a 70μm cell strainer. Samples were spun at 300 rcf for 1 minute and then incubated at 4°C on a nutator for 30 minutes. Additional Lysis Buffer was added to dilute the sample from 350mM to 150mM NaCl before centrifuging at 15,000 rpm for 15 minutes at 4°C. Supernatant was collected and incubated with 200μl α-FLAG resin (Sigma) overnight over a nutator at 4°C. Next day, beads were washed twice for 15 mins with Lysis Buffer without EDTA, DTT, or PhosSTOP. Samples were eluted in 300ul R+A buffer (30% glycerol, 3% SDS, 6mM EDTA, 0.15M Tris-HCl pH 6.8) at 95°C. Eluate was reduced with 15μl 1M DTT for 10 minutes at 65°C and then alkylated with 35μl 1M IAA in room temperature for 20 minutes in the dark. Eluates were then flash frozen using liquid nitrogen immediately. We used the pooled standard approach to enable more accurate comparisons between different time points. ¹⁴N eluate from six time-points were pooled together and split evenly to mix with each ¹⁵N eluate at a 1:1 ratio on ice. For each time-point, 600μl of cold acetone was added to the ¹⁴N/¹⁵N eluate mixture and placed in -20°C overnight. Precipitate were spun at 14,000 rpm for 10 minutes at 4°C and the resulting pellet were washed briefly with 1ml cold acetone. Precipitated eluate was resuspended in 80μl R+A sample buffer containing 3μl of 4X SDS sample buffer. The ¹⁴N/¹⁵N eluate was resolved in 12% SDS-PAGE and the excised PER band was used for protease digestion and analysis by mass spectrometry.

For analysis of ¹⁴N/¹⁵N-labeled samples, proteins were digested in-gel with trypsin and elastase in separate reactions to result in overlapping peptides, such that individual modified sites can be determined. We have previously used this multi-protease approach [61] to maximize high sequence coverage when mapping PER phosphorylation sites [17]. For in-gel digestion the excised gel bands were destained with 30% ACN, shrunk with 100% ACN, and dried in a Vacuum Concentrator (Concentrator 5301, Eppendorf, Hamburg, Germany). Digests with trypsin and elastase were performed overnight at 37°C in 0.05 M NH₄HCO₃ (pH 8). About 0.1 μg of protease was used for one gel band. Peptides were extracted from the gel slices with 5% formic acid.

NanoLC-MS/MS analyses were performed on an LTQ-Orbitrap Velos Pro or an Orbitrap Fusion (Thermo Fisher Scientific, Waltham, MA) equipped with an EASY-Spray Ion Source and coupled to an EASY-nLC 1000 (Thermo Fisher Scientific). Peptides were loaded on a trapping column (2 cm x 75 μm ID. PepMap C18, 3 μm particles, 100 Å pore size) and separated on an EASY-Spray column (25 cm x 75 μm ID, PepMap C18, 2 μm particles, 100 Å pore size) with a 90-minute linear gradient from 3% to 30% acetonitrile and 0.1% formic acid. For the Oribtrap Velos MS scans were acquired in the Orbitrap analyzer with a resolution of 30,000 at m/z 400, MS/MS scans were acquired in the Orbitrap analyzer with a resolution of 7,500 at m/z 400 using HCD fragmentation with 30% normalized collision energy. A TOP5 data-dependent MS/MS method was used; dynamic exclusion was applied with a repeat count of 1 and an exclusion duration of 30 seconds; singly charged precursors were excluded from selection. Minimum signal threshold for precursor selection was set to 50,000. Predictive AGC was used with AGC target a value of 1e6 for MS scans and 5e4 for MS/MS scans. Lock mass option was applied for internal calibration in all runs using background ions from protonated decamethylcyclopentasiloxane (m/z 371.10124). For the Orbitrap Fusion, both MS and MS/MS scans were acquired in the Orbitrap analyzer with a resolution of 60,000 for MS scans and 15,000 for MS/MS scans. HCD fragmentation with 35% normalized collision energy was applied. A Top Speed data-dependent MS/MS method applying HCD and ETD fragmentation from the same precursor with a fixed cycle time of 3 seconds was used. Dynamic exclusion was applied with a repeat count of 1 and an exclusion duration of 120 seconds; singly charged precursors were excluded from selection. Minimum signal threshold for precursor selection was set to 50,000. Predictive AGC was used with AGC a target value of 5e5 for MS scans and 5e4 for MS/MS scans. EASY-IC was used for internal calibration.

Mascot Distiller 2.5 was used for raw data processing and for generating peak lists, essentially with standard settings for the Orbitrap (high/high settings). Mascot Server 2.5 was used for database searching with the following parameters: peptide mass tolerance: 7 ppm, MS/MS mass tolerance: 0.02 Da, enzyme: “semi-trypsin” for tryptic digests and “none” for elastase digests; fixed modifications: carbamidomethyl (C); variable modifications: Gln->pyroGlu (N-term Q), oxidation (M), acetyl (protein N-term), phosphorylation (STY), HexNAc (ST). Searches containing both HCD and ETD spectra (Fusion) were searched separately for either b and y ions (HCD) or c and z ions (ETD). For ETD-searches different modifications definitions (without neutral losses) for phosphorylation and HexNAc were applied. Separate Mascot searches were performed for light peptides (quantitation: “none”) and heavy peptides (quantitation: ¹⁵N-labeling). Database searching was performed against a small custom database containing 187 of the most abundant proteins identified in these samples before in a first round search (without PTMs) against UniProt Drosophila database. This was necessary to limit search space and processing times.

The results from the different Mascot searches (different time-points, proteases and fragmentation techniques) were merged (separately for light and heavy peptides) and filtered for phosphorylated and HexNAc-modified peptides using a custom software tool (A. Schlosser). A Mascot score cut-off of 15 and a delta score cut-off of 10 [62] were applied, and only “rank 1” peptides were accepted. For one modification site and one type of modification, only the peptide spectrum match (PSM) with the highest score was kept, all other PSMs were filtered out. All remaining spectra were verified manually, e.g. by checking the presence of modification specific marker ions. HexNAc-modified peptides were only accepted when at least one of the HexNAc-specific fragment ions (204, 186 and 168) [63] was present in the corresponding HCD spectra. After manually filtering, all remaining peptides were exported to generate a summary of the results (Table 1). The N¹⁴/N¹⁵ MS data have been submitted to the Chorus repository (project ID 1424): (https://chorusproject.org/anonymous/download/experiment/e47a30f7f2c749aba438652d7d88ef04↗) and (https://chorusproject.org/anonymous/download/experiment/e6d6163b31bf40288606f827c6f18371↗).

All flies were reared on standard Drosophila medium (Bloomington Drosophila Stock Center standard recipe). Entrainment and collection of flies at the appropriate time-points were described as above. Roughly 4ml of fly heads were grinded into fine powder using chilled ceramic mortar and pestle and mixed in 30ml of lysis buffer (20mM HEPES pH 7.5, 1mM DTT, 1x protease inhibitor). Homogenate was dounced and poured over a cell strainer as described above prior to centrifugation at 7000xg for 45 minutes at 4°C to separate nuclear and cytoplasmic lysates, repeated once. Lower layer (pellet) as the nuclear fraction from both spins was resuspended in 10ml Nuclear Extraction buffer (20mM HEPES pH 7.5, 10% Glycerol, 350mM NaCl, 0.1% Triton X-100, 1mM DTT, 1mM MgCl₂, 0.5mM EDTA, 1x protease inhibitor, 10mM NaF) with the addition of MG132 (Sigma) and DNAse (Promega). Upper layer (supernatant) as the cytoplasmic fraction was supplemented with Lysis buffer with the addition of MG132 (Sigma) and DNAse (Promega). Nuclear and cytoplasmic fractions were incubated at 4°C for 30 minutes over a nutator. After incubation, nuclear fraction was diluted to 150mM NaCl with Lysis buffer. Nuclear and cytoplasmic fractions were centrifuged at 27,000rpm for 15 minutes at 4°C. Supernatant of nuclear and cytoplasmic samples was recovered before incubation with 200μl α-FLAG M2 resin (Sigma) at 4°C overnight. Beads were washed three times with Lysis buffer for 15 minutes each and subsequently eluted in 200μl of 3XFLAG peptide (Sigma) at a dilution of 250μg/ml at room temperature for 15 minutes. Eluates were resolved on a Tris-Tricine gel and PER bands were excised for protease digestion and mass spectrometry as described in [64]. The label-free MS proteomics data for PER phosphorylation site mapping have been deposited into ProteomeXchange (PXD008281) (ProteomeXchange: http://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD008281↗), MassIVE repository (MSV000081736) (MassIVE: https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=384c7750b3004b7eac91054935a4e038↗), and Chorus repository (Project ID 1424) (https://chorusproject.org/anonymous/download/experiment/1e0023a15da84e51bb18c55146104b32↗).

Locomotor activity rhythms were measured as previously described [30]. 3-4-day old male flies were collected and subjected to 12hr light:12hr dark (LD) cycles at 25°C for four days followed by seven days of constant darkness (DD) to measure free-running period using the Drosophila Activity Monitor System (DAMS) (TriKinetics, Waltham, MA). Data analysis was performed using FaasX as described in [30].

pAc-per-V5, pAc-3XFLAG-per-6Xcmyc, and pAc-clk-V5 were previously described [20, 58]. For generating Serine/Threonine (S/T) to Alanine (A) O-GlcNAc site mutants (S2 Table for primer sequences), pAc-per-V5 or pAc-3XFLAG-per-6Xcmyc served as the template for site-directed PCR mutagenesis using QuikChange site-directed PCR mutagenesis (Agilent Technologies). All O-GlcNAc mutations were verified by Sanger sequencing. ogt cDNA (described above) was subcloned into a pMT-3XFLAG-6XHIS vector described in [20], with the epitope at the C-terminus of the ORF. pCopia-renilla luciferase and per-E-luc constructs were described previously [37]. S2 cells and DES expression medium were obtained from Life technologies (Carlsbad, CA), and transient transfections were performed using Effectene (Qiagen, Valencia, CA) according to manufacturer’s instructions and as previously described [17,20].

Luciferase reporter assays were performed as described [3, 37]. Measurements of luciferase activity were performed using the Dual-Glo luciferase assay system (Promega, Madison, WI) according to the manufacturer’s recommendation on a TriStar LD 941 microplate reader (Berthold Technologies, Oak Ridge, TN).

pAc-per-V5(WT) or mutant variants were transfected into S2 cells with pMT-FLAG-ogt or an empty plasmid. 20 hours after transfection, ogt expression was induced for 16 hours. Cycloheximide was then added to a final concentration of 10 μg/ml. Cells were harvested and lysed with EB2 [20] at the times indicated. Proteins were analyzed by western blotting as detailed below and in [65].

S2 cell and adult fly head protein extractions, western blotting, and image analysis, were performed as previously described [58, 65] with modifications. Primary antibodies α-V5 (Invitrogen, Carlsbad, CA) (1:5000) was used to detect CLK-V5 and PER-V5, α-cmyc (9E10, Sigma, St. Louis, MA) (1:5000) to detect PER-CMYC, α-FLAG (Sigma) (1:7000) to detect FLAG-OGT, α-HA 3F10 (Roche, Indianapolis, IN) (1:1000) to detect PER-HA, α-PER (GP5620; RRID:AB_2747405) [57] (1:3000) to detect PER, and α-HSP70 (Sigma) (1:10,000) was used for normalization. Secondary antibodies conjugated with HRP were added at final dilution as follows: α-mouse IgG at 1:5000 for α-V5 detection, 1:7000 for α-FLAG detection, or 1:10,000 for α-HSP70 detection, α-guinea pig IgG at 1:2000 for α-PER detection, and α-rat IgG (1:1000) for α-HA detection of PER-HA. Membranes were imaged and protein levels were quantified using the ChemiDoc MP system with Image Lab software (Bio-Rad). To calculate PER degradation rate, PER intensity was normalized to HSP70 intensity at each time-point, and was then converted to a fraction of the peak value (peak = 1). For quantifying PER levels from fly heads, PER values were normalized against HSP70 intensity at each time-point, and subsequently expressed as a fraction of the peak PER levels.

Co-IP assays using protein extracts from S2 cells and fly heads were performed as described [58, 65] with modifications. Proteins were extracted using modified RIPA buffer with the addition of 100μM PUGNAc to preserve O-GlcNAcylation of proteins prior to input analysis by western blotting or Co-IP with appropriate antibodies. Samples were pre-cleared using sepharose beads (Sigma) to reduce nonspecific binding. For co-IP in S2 cells, CLK IP samples were incubated with 15μl α-V5 resin (Sigma) and negative non-specific control IP samples were incubated with 15μl α-HA resin (Sigma). For co-IP in fly heads, CLK IP samples were incubated with 4μl α-CLK antibody (Santa Cruz Biotechnology, Dallas, TX) for 3 hours prior to incubation with 20μl gamma sepharose beads (GE, Pittsburgh, PA) for 1 hour; PER IP samples were incubated with 20μl α-HA resin (Sigma). Immune complexes were resolved by SDS-PAGE as described [58, 65]. IP signal intensity was normalized to intensity of the bait protein. These values were then converted as relative to the peak value of the dataset (peak = 1). Representative data shown are averages of normalized PER or CLK interactions from at least three independent experiments.

Total RNA was extracted from fly heads and abdominal fat bodies using TRI-Reagent (Sigma). cDNA synthesis from total RNA and real-time PCR analysis was performed as previously described [58, 65]. For isolating abdominal fat bodies, flies were collected at the indicated time-points on the first day of DD and immediately transferred in TRI-Reagent (Sigma) for 40 minutes at room temperature with agitation following dissection in RNAlater buffer (Thermo Fischer Scientific). At least 16 flies were dissected for each genotype and time-point. After dissection, fat bodies were rinsed with nuclease-free water twice and resuspended in TRI-Reagent prior to RNA isolation.

ChIP assays were essentially performed as described [58]. qPCR of an intergenic region (FBgn0003638) of the Drosophila genome representing background CLK binding was subtracted from input samples. Technical triplicates for the qPCR step were performed for each of the three biological replicates. Two-tailed t-tests were used to determine statistical differences between genotypes at each time-point.

Adult fly brain immunohistochemistry was performed as described previously [65] with modifications. Briefly, adult flies were entrained in LD cycles for 3 days and collected at the appropriate time-point following incubation with fixative solution (4% paraformaldehyde, 0.2% Triton X-100 in PBS) for at least 40 minutes in the dark at room temperature with gentle rotation. Fixative solution was removed, and then wash solution (0.2% Triton X-100 in PBS) was added to transfer flies into an embryo dish. Brains were dissected using #5 Rubis nano tweezers (Electron Microscopy Sciences, Hatfield, PA). Approximately 10 brains were dissected for each time-point. After dissection, brains were incubated in fixative solution at room temperature for 40 minutes with gentle rotation. Brains were rinsed quickly with wash solution three times prior to three slow washes in wash solution for 10 minutes each with gentle rotation. Wash solution was removed, and blocking solution (0.2% Triton X-100, 5% horse serum in PBS) was added to the brains for 40 minutes at 4°C with gentle rotation. Brains were then incubated with new blocking solution at 4°C overnight with primary antibodies at the following dilutions: α-HA 3F10 antibody (Roche) (1:100), and α-PDF antibody (Developmental Studies Hybridoma Bank, Iowa City, IA) (1:100). After ~18 hr, brains were rinsed quickly with wash solution three times prior to four slow washes with wash solution for 10 minutes each with gentle rotation. Brains were then incubated in secondary antibodies in blocking solution overnight. Secondary antibodies used at the following dilutions were Dylight88-conjugated α-rat (1:100) and Alexa647-conjugated α-mouse (1:100) (Thermo Fischer Scientific). After ~18 hr, brains were rinsed quickly with wash solution three times prior to four slow washes with wash solution for 10 minutes each with gentle rotation. Brains were rinsed quickly with PBS following incubation with 85% glycerol for 15 minutes. Brains were mounted on microscope slides in VectaShield mounting medium (Vector Laboratories, Burlingame, CA) under a #1.5 (17-mm) coverslip. Confocal images were obtained with an Olympus FV1000 Confocal Microscope (Olympus Life Science, Center Valley, PA) and processed with the FV1000 software (Olympus Life Science). Scoring of PER subcellular localization was performed as previously described [66, 67]. At least five brains were used for each genotype and time-point. For statistical analysis, scored LN_vs from each brain served as one biological replicate. Two-tailed t-tests were used to determine statistical differences between genotypes at each time-point.