Organisms have endogenous circadian clocks that maintain daily rhythmic activities despite environmental alterations. The circadian clock system consists of three main components: an input pathway, a central oscillator, and an output pathway (Agostino et al., 2011). The input pathway transmits external timing information (e.g., light or temperature cycles) to the brain. In the central oscillator, the timing information cues 24 h rhythm generators (e.g., circadian gene expression and/or neural activities) to govern circadian output pathways. The output circuitry is comprised of downstream genes and neurons that relay clock signals to overt behaviors and physiology, such as locomotor activity and metabolism.

The central molecular oscillators in Drosophila run in a network of approximately 150 pacemaker neurons, which are organized into several groups (Beckwith et al., 2015; Peschel et al., 2011). The cell-autonomous molecular clocks comprise transcription-translation feedback loops (TTFLs) of circadian transcription factors that are conserved among fruit flies and other species. The first TTFL in Drosophila is driven by PERIOD (PER) and TIMELESS (TIM) (Konopka et al., 1971; Sehgal et al., 1994), which accumulate in the cytoplasm as mRNA levels of per and tim increase throughout the day, reaching the maximum level in the early evening. TIM stabilizes PER (Price et al., 1995; Suri et al., 1999; Vosshall et al., 1994), and the PER-TIM heterodimer complex enters the nucleus at around midnight where it inhibits the CLOCK (CLK) and CYCLE (CYC) complex, the transcriptional activator of per and tim (Allada et al., 1998; Rutila et al., 1998). This negative autoregulatory feedback is sustained until TIM is degraded by the photoreceptor CRYPTOCHROME (CRY) in the early morning (Emery et al., 1998; Stanewsky et al., 1998). The heterodimeric CLK-CYC complex drives additional TTFLs by binding E-box sequences in genomic loci of several clock genes and promoting their transcription, such as PAR-domain protein 1 (Pdp1 ), encoding an activator of Clk transcription (Cyran et al., 2003); vrille (vri ), encoding a repressor of Clk transcription (Blau et al., 1999), and clockwork orange (cwo), encoding a transcriptional repressor that competes with the CLK-CYC complex for E-box binding (Kadener et al., 2007; Lim, Chung, et al., 2007; Matsumoto et al., 2007; Richier et al., 2008; Zhou et al., 2016).

Although the central oscillators running endogenous clocks have been thoroughly studied at the transcription level, the significance of their post-transcriptional control has begun to emerge (Brunner et al., 2006; Kojima et al., 2011; Lim et al., 2013). Nonsense-mediated mRNA decay (NMD) is a post-transcriptional regulatory mechanism via which mRNAs with premature translation termination codons (PTCs) are recognized and removed. The precise mechanisms that initiate the assembly of NMD effectors differ very little among species (Behm-Ansmant et al., 2007; Eulalio et al., 2007; Muhlemann, 2008; Popp et al, 2014). The conventional model involves an exon junction complex (EJC) that is removed by ribosomes in translation of normal transcript. However, PTCs generated from DNA defect, transcription errors, or alternative splicing, cause the ribosomes to stall upstream of EJC. The distance between the ribosomes and EJC triggers the recruitment of a NMD complex, which consists of a number of UP-FRAMESHIFT (UPF) and SUPPRESSOR WITH MORPHOGENETIC EFFECT ON GENITALIA (SMG) proteins. UPF1 is the core RNA helicase that composes the complex together with UPF2 and UPF3 and is phosphorylated by SMG1, a phosphoinositide-3-kinase-related protein kinase. Dephosphorylation of UPF1 is facilitated by SMG5/7 or SMG6 nucleases.

While it has been shown that NMD regulates 5–10% of the transcriptome (Chan et al., 2007; Mendell et al., 2004; Yepiskoposyan et al., 2011), only a few target genes implicated in circadian rhythms have been identified. In Arabidopsis, splice variants of the genes encoding timing of cab expression 1 and early flowering 3 are degraded via NMD (Kwon et al., 2014). Other clock-regulated glycine-rich RNA-binding proteins, GRP7 and CCR1, form a feedback loop in which unproductive splicing of each transcript is coupled to NMD (Schoning et al., 2007; 2008). Alternative splice variants of a Jumonji C domain-containing gene, JMJC5, in Medicago truncatula also show NMD sensitivity, though the strength of the sensitivity varies depending on the tissue and other exogenous conditions (Shen et al., 2016). Neurospora crassa has a shortened periodicity as a result of Upf1 mutation (Morgan et al., 1997), and NMD was recently found to regulate its major clock gene, frequency, for the transcriptional control of white collar genes (Aronson et al., 1994; Cheng et al., 2001; Merrow et al., 2001; Wu et al., 2017).

There is abundant experimental evidence implicating ATF/CREB (activating transcription factor/cAMP response element-binding protein) and its transcriptional coactivators in the generation of behavioral rhythms through cell-autonomous molecular clocks (Belvin et al., 1999; Ginty et al., 1993; Kim et al., 2016; Koyanagi et al., 2011; Lee et al., 2010; Lim et al., 2007; O’Neill et al., 2008; Obrietan et al., 1999; Scheving et al., 1998; Sun et al., 2015; Travnickova-Bendova et al., 2002). However, few studies have demonstrated the regulatory mechanism of ATF/CREB expression for clock function (Hendricks et al., 2001; Kako et al., 1996; Maurer et al., 2016; Williams et al., 2001). In this study, we provide strong genetic evidence that NMD acts in Drosophila circadian pacemaker neurons to sustain circadian locomotor behaviors and identify CrebB as a rate-limiting target of the clock-relevant NMD.

Fly stocks

All flies were reared at 25°C and 60% humidity on a 12 h light/dark (LD) cycle and with standard cornmeal-yeast food. The wild-type controls were iso31 (Bloomington Drosophila Stock Center #5905). Smg5 RNAi (#31090, JF01554), Smg5^e04233 (#25167), PBac{IT.GAL4}Smg5^1409-G4 (#65708, Smg5-GAL4), Df(2L)BSC345 (#24369, Smg5 deficiency), Df(3R)BSC494 (#24998, Smg6 deficiency), UAS-Upf2.A (#60579), UAS-Upf2.FVRR (#60583), UAS-Upf2.E801R (#60581), UAS-CrebB-a (#9232), and UAS-CrebB-c (#7219) were all from the Bloomington Drosophila Stock Center. Smg6^RNAi#1 (6369R-1) and Smg6^RNAi#2 (6369R-3) were from Fly Stocks of National Institute of Genetics, Japan. Smg6^2a and UAS-nlsDsRed2:: SV40-3′UTR were kindly provided by Dr. Metzstein (Frizzell et al., 2012), and UAS-Upf1^D45B was provided by Dr. Singh, who received it from the original generator, Dr. Palumbo (Micale et al., 2009). Pdf-GeneSwitch-GAL4 was a gift from Dr. Ceriani (Depetris-Chauvin et al., 2011), and UAS-CrebB-b was from Dr. Davis (Yu et al., 2006). The UAS-Smg6 -V5 transgenic line was made by cloning the full cDNA of Smg6 at the EcoRI and XhoI sites of the 8.5 kb pUAST-AttB-V5 vector.

Fly behavioral assay

The food for behavioral testing was 5% sucrose (31365-1201; Junsei Chemical Co.) and 2% agar (24440-1201; Junsei Chemical Co.). Pdf-GeneSwitch was activated by feeding the flies food containing 62.5 μM RU486 (H110-01; Sigma-Aldrich) dissolved in ethanol. The vehicle control was ethanol. The circadian rhythms of the flies were analyzed with the Drosophila Activity Monitoring system (TriKinetics) using MATLAB-based ClockLab (Actimetrics). The rhythm peak was calculated from the first to the seventh dark/dark (DD) cycle with a confidence interval of 0.05. The power of rhythmicity was calculated by subtracting the value of “significance” from the value of “power.” Morning and evening anticipation indices were calculated using an Excel-based Counting Macro (Pfeiffenberger et al., 2010), as follows: morning index = total activity 3 h prior to lights on/total activity 6 h prior to lights on – 0.5; evening index = total activity 3 h prior to lights off/total activity 6 h prior to lights off – 0.5.

Immunofluorescence assay

Flies were entrained to the LD cycle for at least 72 h before dissection between Zeitgeber time 2 (ZT2) and ZT3 for PDF staining or DsRed detection. Briefly, a 30 min fixation in 3.7% formaldehyde and washes in phosphate-buffered saline with 0.5% Tween 20 were followed by two nights of primary antibody staining, as described previously (Lim et al., 2011; Park et al., 2014), with mouse anti-PDF (a-PDF C7-s; Developmental Studies Hybridoma Bank) at a dilution of 1:600. The secondary antibodies were either Alexa Fluor 488 (A21202; Invitrogen) or Alexa Fluor 594 (A21203; Invitrogen) donkey anti-mouse IgG. The samples were scanned at a pixel resolution of 1,024 × 1,024 with an LSM780 confocal microscope (Carl Zeiss) using LSM image browser ZEN software (Carl Zeiss).

RNA isolation and quantitative real-time PCR

Flies were entrained to a 12 h LD cycle for at least 72 h before they were harvested at 4 h intervals (at ZT3, ZT7, ZT11, ZT15, ZT19, and ZT23). The heads of 10–30 flies per genotype were fresh frozen at −80°C. RNA was extracted with TRIzol (15596018; Ambion), digested with RQ1 DNase to remove genomic DNA, and purified with phenol-chloroformisoamyl alcohol (25:24:1, M3803; Sigma-Aldrich). Purified RNA was reverse transcribed with M-MLV reverse transcriptase (M1705; Promega) and random primers (C1181; Promega). Quantitative PCR was performed with the synthesized cDNA using Prime Q-Mastermix (9200; Genetbio). Real-time PCR results were analyzed with the 2^−ΔΔCT method (Livak et al., 2001). Relative mRNA levels were normalized to the peak value of the respective control and then calculated as a percentage of the peak value (set to 100%).

Smg5 depletion in circadian pacemaker neurons dampens circadian locomotor rhythm

To identify post-transcriptional regulators with circadian clock function, we conducted a behavioral screen of transgenic flies harboring RNAi against genes encoding individual components of processing bodies (Decker et al., 2012; Zheng et al., 2011). We carried out a primary genetic screen using a tim-GAL4 line that drives the overexpression of each RNAi transgene in all clock-relevant cells expressing tim and depletes its target gene expression. Any phenotype in circadian behaviors was then confirmed using a Pdf-GAL4 driver that limits the RNAi expression to PDF-expressing neurons. PDF is a neuropeptide expressed in small and large ventral lateral neurons (s-LNvs and l-LNvs, respectively) of the adult fly brain and is crucial for synchronizing clock phases among different groups of circadian pacemaker neurons (Lin et al., 2004; Peng et al., 2003). After entrainment to LD cycles, wild-type flies display robust 24 h rhythms of their locomotor activity in free-running constant dark (DD) conditions. Abnormalities in circadian rhythms are thus observed as an altered periodicity or rhythm amplitude of DD behaviors compared with that of wild-type flies.

Of the approximately 40 post-transcriptional regulatory genes, the RNAi-mediated depletion of several affected circadian locomotor behaviors (Fig. 1A). In particular, depletion of a NMD-relevant gene, Smg5, in either tim-expressing cells or PDF-expressing neurons dampened circadian locomotor rhythms, whereas the inhibition of tim-GAL4 activity by PDF-specific GAL80 (Stoleru et al., 2004) partially suppressed the Smg5 RNAi phenotype (Fig. 1B, Supplementary Table S1). We confirmed that tim-GAL4-driven expression of Smg5 RNAi resulted in Smg5 mRNA levels in adult heads that were ~60% of wild-type levels (data not shown). By contrast, transgenic flies displayed normal rhythmicity when Smg5 was depleted in other brain regions by repo–GAL4 (glial cells), dilp2 (Drosophila insulin-like peptide 2 )-GAL4 (pars intercerebralis), or cry13-GAL4 (cry -expressing cells).

Genetic manipulation of the NMD pathway in PDF-expressing neurons alters behavioral rhythms

We next examined if other NMD components similarly influence circadian behaviors. Indeed, transgenic depletion of each NMD factor (e.g., Upf1, Upf2, Upf3, Smg1, or Smg6 ) in tim-expressing cells caused behavioral arrhythmicity comparable to that observed with Smg5 depletion (Fig. 1C, Supplementary Table S2). Furthermore, the individual overexpression of these NMD factors (e.g., Upf1, Upf2, Smg5, or Smg6 ) in tim-expressing cells disrupted free-running locomotor rhythms (Supplementary Table S3), indicating that a specific dosage of NMD activity is required for clock function. RNA helicase UPF1 is a key component of the NMD machinery, whereas UPF2 and UPF3 act as adaptors/recruiters of the complex. Transgenic overexpression of Upf2.E801R, which disrupts UPF2–UPF3 interaction but rescues lethality in the Upf2 mutant (Avery et al., 2011), induced arrhythmicity in circadian behaviors comparable to the overexpression of wild-type Upf2. By contrast, transgenic overexpression of Upf2.FVRR, which disrupts UPF1–UPF2 interaction and cannot rescue lethality in the Upf2 mutant, negligibly affected circadian behaviors. The UPF1–UPF2 interaction might thus be limiting for clock function in Drosophila, as it is important for NMD and adult fly viability. Additionally, we observed that the overexpression of UPF1^D45B, a dominant-negative form of UPF1 (Micale et al., 2009), in PDF-expressing neurons was necessary and sufficient to lengthen the free-running periods in locomotor rhythms by >1 h (Fig. 1D, Supplementary Table S4). These results document the importance of the NMD pathway in PDF-expressing neurons for sustaining circadian behaviors.

NMD activity in PDF-expressing neurons does not affect neural development but is elevated by Clock mutation

To validate that PDF-expressing neurons express NMD-relevant genes and display NMD activity, we performed imaging analyses using confocal microscopy. We first employed an enhancer-trapping line for the Smg5 locus (PBac{IT.GAL4}Smg5^1409-G4) and visualized its enhancer activity in the adult fly brain by driving the expression of a fluorescent reporter protein. In this way, we indirectly detected Smg5 expression in PDF-expressing s-LNvs and l-LNvs as well as in cells over a broad region of the fly brain (Fig. 2A). To further assess NMD activity in PDF-expressing neurons, we utilized a transgenic RNA reporter harboring the SV40 3′ untranslated region (UTR)(UAS-nlsDsRed ::SV40-3′UTR) (Metzstein et al., 2006). The SV40 3′ UTR is a known target of NMD activity; therefore, the relative expression levels of the DsRed reporter protein are inversely proportional to endogenous NMD activity (Metzstein et al., 2006). We established transgenic flies expressing the NMD reporter in PDF-expressing neurons and compared their levels of DsRed expression among different genetic backgrounds. As expected, confocal imaging revealed that DsRed signals were higher in weak hypomorphic mutants of Upf2 (Upf2^2-8A) than in the wild type, validating the functionality and sensitivity of the NMD reporter in PDF-expressing neurons (Fig. 2C). Importantly, we found that the NMD reporter protein was expressed at very low levels in both s-LNvs and l-LNvs of arrhythmic Clk^Jrk mutants (Fig. 2D). These data suggest that NMD activity in PDF-expressing neurons is suppressed by wild-type Clk. Since the Clk^Jrk allele generates Clk transcripts harboring NMD-inducible PTCs, we cannot rule out the other possibility that the Clk^Jrk mutation potentiates the overall activity of the NMD pathway to promote the decay of its own transcripts.

Developmental defects in PDF-expressing neurons and their projections could be responsible for the abnormal circadian behaviors observed in some clock mutants. Although the Smg5 or Smg6 RNAi flies showed behavioral arrhythmicity, the cell bodies and axonal projections of their PDF-expressing neurons were intact (Fig. 2B). To further exclude developmental effects of NMD on circadian behaviors, we conditionally depleted Smg6 expression in PDF neurons by Pdf^GeneSwitch-Gal4 (Depetris-Chauvin et al., 2011). In this experimental condition, Smg6 RNAi expression was suppressed during development but oral administration of RU486 to adult flies induced GeneSwitch-Gal4 activity and the RNAi expression. The adult-specific depletion of Smg6 was sufficient to dampen the rhythmicity in circadian locomotor behaviors, consistent with the behavioral phenotypes caused by constitutive depletion (Fig. 2E, Supplementary Table S5). Collectively, these data suggest that the effects of the NMD pathway are not developmental but they maintain behavioral circadian rhythms in adult flies.

Hypomorphic mutations in Smg5 or Smg6 weaken morning anticipation and free-running locomotor rhythms

To elucidate the molecular mechanisms underlying NMD effects on circadian behaviors, we established mutant flies hypomorphic for either of the two NMD factors and analyzed their behavioral circadian phenotypes. The Smg5^e04233 mutant allele harbors a piggyBac insertion in the 5′ UTR of the Smg5 locus (Fig. 3A) and reduces Smg5 mRNA levels to 57% of the wild-type (data not shown). The Smg6^2a mutant allele possesses an ethyl methanesulfonate-induced missense mutation in the PilT N-terminal (PIN) domain (Fig. 3E) (Frizzell et al., 2012). This catalytic domain for nuclease activity is conserved between SMG5 and SMG6 proteins, implicating both factors in mRNA degradation at the last step of NMD pathway.

During the LD cycle, wild-type flies gradually increase their locomotor activity prior to the light transitions as a part of clock-dependent anticipatory activities to lights-on (i.e., morning anticipation) or lights-off (i.e., evening anticipation). By contrast, Smg5 mutants trans-heterozygous for Smg5^e04233 over the chromosomal deficiency of the entire Smg5 locus (i.e., Df(2L)BSC345 ) showed significantly weaker morning anticipation than either of the heterozygous controls (Figs. 3B and 3C). Smg5 mutants also exhibited less robust DD locomotor rhythms than their heterozygous controls (Figs. 3B and 3D, Supplementary Table S6). Dampened morning anticipation in LD cycles and low-amplitude rhythms in DD locomotor behaviors were similarly observed in mutants trans-heterozygous for Smg6^2a and chromosomal deficiency of the Smg6 locus (i.e., Df(3R)BSC494 ) (Figs. 3F–3H, Supplementary Table S7). These behavioral phenotypes in NMD mutants are thus consistent with their RNAi phenotypes.

Daily oscillations in levels of clock-relevant mRNAs are comparable between Smg6 mutants and heterozygous controls

PDF-expressing s-LNvs are a well-established neural locus important for morning anticipation and DD locomotor rhythms (Grima et al., 2004; Stoleru et al., 2004). However, we did not detect any gross defects in the cell bodies or axonal projections of these neurons in either Smg5 or Smg6 mutants (Fig. 4A). We reasoned that aberrant NMD of a specific mRNA substrate, but not neuronal defects, might underlie the observed circadian phenotypes. Such targets are expected to have elevated mRNA levels in NMD mutants due to the inefficient degradation of their transcripts, as has been observed for other NMD targets (Chapin et al., 2014; Johansson et al., 2007). Since it is technically difficult to quantify transcript levels in PDF-expressing neurons, we examined the mRNA levels of clock-relevant genes in Smg6 mutant heads. This analysis proved no significant differences between Smg6 mutants and heterozygous controls in the circadian expression of core clock transcripts (e.g., per, tim, vri, and Pdp1)(Fig. 4B). Smg6 mutation rather reduced Clk or cry expression only at a specific time of the day in LD cycles.

CrebB is a rate-limiting target of NMD to support robust behavioral rhythms

Mammalian ATF/CREB family genes are known targets of NMD (Hatano et al., 2013; Mendell et al., 2004). Moreover, a number of studies link ATF/CREB-dependent transcription with circadian clocks (Belvin et al., 1999; Ginty et al., 1993; Kim et al., 2016; Koyanagi et al., 2011; Lee et al., 2010; Lim et al., 2007; O’Neill et al., 2008; Obrietan et al., 1999; Scheving et al., 1998; Sun et al., 2015; Travnickova-Bendova et al., 2002). Accordingly, we examined the expression of the Drosophila ATF/CREB homolog, CrebB, in Smg6 mutants. The CrebB locus encodes several isoforms that are generated by alternative splicing and differentially include PTCs (Fig. 5A). A quantitative exon-specific analysis of these CrebB isoforms revealed that the levels of CrebB transcripts lacking exon 6 (CrebB-Δex6) were higher in Smg6 mutant heads than in those of the heterozygous controls. By contrast, the expression of CrebB transcripts containing exon 4 (CrebB-ex4) was lower in Smg6 mutants. The relative abundance of CrebB-Δex6 and CrebB-ex4 was similarly affected by Smg5 mutation (Fig. 5B), suggesting that specific isoforms of CrebB are bona fide targets of Drosophila NMD. In addition, modest but opposing effects of Smg5 and Smg6 on CrebB-RM levels may indicate their non-redundant roles in the post-transcriptional expression of this specific CrebB isoform.

To functionally validate that NMD-dependent regulation of CrebB is responsible for sustaining circadian behaviors, we transgenically overexpressed the CrebB-b isoform (Yu et al., 2006) in Smg5 RNAi or control flies and examined the effect on DD locomotor rhythms. Overexpression of the CrebB-b isoform rescued free-running locomotor rhythms in Smg5-depleted flies while it strongly dampened circadian behaviors in control flies (Fig. 5C, Supplementary Table S8). These results suggest that CrebB is rate-limiting for the NMD pathway to sustain high-amplitude behavioral rhythms.

NMD is one of the major RNA quality control mechanisms for degrading nonfunctional transcripts with PTCs and ensuring the translation of error-free mRNAs. In addition, NMD monitors the stability of normal mRNAs by targeting their selective cis-elements, such as upstream open-reading frames, long 3′ UTRs, introns in 3′ UTRs, and frame-shifting signals (Alonso, 2005; Chang et al., 2007; Hug et al., 2016; Nickless et al., 2017). It has been previously shown that NMD contributes to the circadian control of gene expression, though only a few clock genes have been identified as direct targets (Filichkin et al., 2014; Hatano et al., 2013; Kwon et al., 2014; Mendell et al., 2004; Shaul, 2015; Wu et al., 2017). In this report, we show that NMD is necessary to sustain circadian behaviors in Drosophila and localize this control to PDF-expressing pacemaker neurons. We further provide molecular and genetic evidence that NMD targets specific isoforms of CrebB to sustain circadian locomotor rhythms.

Daily rhythmic oscillations in cAMP levels are observed in a diversity of species (Eckel-Mahan et al., 2008; Fukuhara et al., 2004; O’Neill et al., 2008; Palacios-Munoz et al., 2018) and linked to the circadian clocks in part via the ATF/CREB family of transcriptional regulators (Ginty et al., 1993; Koyanagi et al., 2011; Obrietan et al., 1999; Scheving et al., 1998). In mammals, these transcription factors promote per transcription (Shimizu et al., 2007; Travnickova-Bendova et al., 2002), and the CLOCK-ARNTL (BMAL) heterodimer in mice negatively regulates ATF5 expression (Lemos et al., 2007). In Drosophila, CRE-dependent transcription exhibits both ultradian and circadian rhythms while these rhythms are disrupted by CrebB mutation (Belvin et al., 1999). As expected, transcriptional coactivators of the ATF/CREB family members, such as CREB-binding protein or CREB-regulated transcriptional coactivator support circadian rhythms in Drosophila and mammals via either CLK-CYC- or CREB-dependent transcription of circadian clock genes (Hosoda et al., 2009; Hung et al., 2007; Kim et al., 2016; Lee et al., 2010; Lim et al., 2007; Sakamoto et al., 2013).

Our genetic analyses revealed that the loss of NMD function potently dampened circadian locomotor rhythms without impairing the circadian expression of clock genes. Thus, the NMD-dependent clock function maps to the clock output pathway, which relays the timing information from molecular clocks (i.e., daily rhythmic gene expression) to behavioral rhythms. Given the isoform-specific effects of NMD on CrebB expression, we reason that NMD tunes the balance of CrebB isoforms with transcriptional activator or repressor function rather than silencing the overall activity of CrebB. There are two potential PTCs in CrebB transcripts: one in the intron after exon 2 and the other after exon 6. CrebB transcripts including exons 4 and 6 encode a transcriptional activator if their translation starts from the second start codon after the PTC following exon 2 (Fropf et al., 2013; Tubon et al., 2013; Yin et al., 1995a; Yin et al., 1995b). On the other hand, CrebB transcripts without exons 2, 4, or 6, such as CrebB-b, encode a transcriptional repressor (Davis et al., 1996; Fropf et al., 2013; Yin et al., 1994; Yin et al., 1995b) while the transcriptional activities of other splice variants have not been characterized (Fropf et al., 2013; Perazzona et al., 2004). Since our transcript analysis in LD cycles did not show any strong time-of-the-day-effects of Smg6 on the relative levels of CrebB-Δex6 and CrebB-ex4, it is less likely that NMD controls CrebB expression in a light- or circadian clock-dependent manner. As reported in a number of species (Barberan-Soler et al., 2008; Lewis et al., 2003; Tabrez et al., 2017; Traunmuller et al., 2014), we speculate that NMD is constitutively coupled to the alternative splicing of CrebB for the differential expression of CrebB isoforms, thereby sustaining robust behavioral rhythms in constant dark.

The transcriptional activity of two mammalian ATF/CREB homologs (i.e., ATF4 and ATF5) is implicated in circadian gene expression (Koyanagi et al., 2011; Lemos et al., 2007), and their transcript stability is regulated by NMD (Hatano et al., 2013; Mendell et al., 2004). Accordingly, we propose that NMD-dependent regulation of ATF/CREB expression is evolutionarily conserved among diverse species and represents an ancestral clock mechanism that sustains circadian rhythms. Future studies will elucidate the isoform-specific regulation of CrebB by NMD in Drosophila and whether there is feedback from the circadian clocks to control NMD activity.

We thank Bloomington Drosophila Stock Center, Kyoto Stock Center, Vienna Drosophila Resource Center, Dr. Mark Metzstein, Dr. Ronald Davis, Dr. Anand Singh, and Dr. Maria Fernanda Ceriani for providing the fly stocks and Dr. Young-Joon Kim for generating the UAS-Smg6 transformants. This work was supported by National Research Foundation of Korea (NRF) grants funded by the Korea government (MSIT) (no. 2012H1A2A1016929 to H.R., no. 2016R1A6A3A1193 2215 to J.L., no. 2018R1A5A1024261 to C.L., and no. 2016R1A2B4011111 to J.C.).