Emerging roles for microRNA in the regulation of Drosophila circadian clock

Circadian rhythms with approximately 24-h periods are prevalent in all living organisms, from unicellular cyanobacteria to vertebrates. This endogenous timekeeping system endows organisms with the enormous advantage of anticipating and adjusting to environmental changes (e.g., light, food, temperature, etc.) [1]. Many physiological and behavioral processes, including hormone secretion, thermoregulation, metabolism, immune responses, and sleep–wake cycles, are modulated by the circadian clock [2–5]. Therefore, it is not surprising that disruption of the circadian clock is associated with pathogenesis, such as obesity, cardiovascular diseases, and even cancer [6–8]. Since the circadian clock plays a fundamental role in the maintenance of circadian-dependent processes, understanding the regulatory mechanism of the circadian clock is vitally important.

Drosophila has been instrumental in the study of circadian rhythms because of powerful genetic tools and their relatively simple circadian neuronal circuitry. After more than four decades of studies in Drosophila, we now have a deep understanding of the molecular components of the circadian clock that play a pivotal role in coordinating circadian readouts to environmental inputs. Thus, in this review we will focus on the circadian rhythms of Drosophila.

In flies, approximately 150 circadian neurons make up the clock network in the brain. This network can be categorized into seven subclasses based on anatomical location, cellular size, and neuropeptide expression: the small and large ventral lateral neurons (LNvs) expressing pigment dispersing factor (PDF), the fifth small LNv, the dorsal lateral neurons (LNds), and three groups of dorsal neurons (DN1, DN2, and DN3) [10]. The core proteins of the TTFL, encoded by clock (clk), cycle (cyc), period (per), and timeless (tim), are expressed in all clock neuron groups, but transcriptome studies of individual circadian neurons have identified that most cycling transcripts are expressed only in a specific subgroup [11], which is likely due to different regulatory mechanisms. Moreover, next-generation sequencing (NGS) results have demonstrated that approximately 50% of rhythmic primary transcripts do not give rise to rhythmic mRNA, and rhythmic proteins do not necessarily arise from rhythmic mRNA transcripts [12, 13]. For these reasons, we sought to explore when, where and how post-transcriptional modification is integrated into the circadian control system. This review will focus on miRNA-dependent post-transcriptional regulation, due to accumulating evidence demonstrating the involvement of miRNAs in the regulation of circadian timekeeping.

The endogenous circadian clock maintains a free-running (in the absence of rhythmic environmental stimulus) period of approximately 24 h by employing interconnected feedback loops that typically include a primary loop and an ancillary loop. In flies, the primary negative feedback loop is composed of transcriptional activators (CLK/CYC) and repressors (PER/TIM) in which heterodimeric transactivators bind to promotor regions to induce repressor expression. Following 6–8 h of cytosolic repressor accumulation and modification, heterodimerization occurs, and the repressors translocate to the nucleus, inhibiting their own transcriptional activity by interacting with activators. Consequently, the levels of repressor complexes drop, and a new cycle of transactivation starts. The ancillary feedback loop shares core transactivators with the primary loop but differ in that Vrille (VRI) and Par domain protein 1 (PDP1) repress and activate clk transcription, respectively (Fig. 1). The ancillary feedback loop is believed to stabilize the molecular clock by regulating the phase and amplitude of core pacemakers and rhythmic outputs [17, 18].

In addition to the transcriptional control of the molecular clock, post-translational regulation is built into the TTFL to regulate protein stability, activity, and subcellular localization. Among different post-translational regulations, phosphorylation is one of the most interesting, as phosphorylation levels of principal clock proteins oscillate in a temporal pattern. For example, PER protein phosphorylation peaks in the early morning/late evening, and dPER is hypophosphorylated in the early evening [19]. The kinases involved in the phosphorylation of the PER/TIM complex are glycogen synthase kinase-3 [GSK3; also known as Shaggy (SGG)], casein kinase II (CK2), and DOUBLETIME (DBT; Drosophila homolog of mammalian CK1δ/ε) [20–23]. There are also several important phosphatases that dephosphorylate PER and TIM, such as Protein Phosphatase 1 (PP1) and Protein Phosphatase 2A (PP2A) [24, 25] (Fig. 1).

The temporal delay between mRNA and protein oscillation of core clock genes indicates the existence of a layer of post-transcriptional regulation. When RNA sequencing and nascent RNA sequencing were applied to assess the whole genome circadian gene expression in fly heads, it was observed that ~ 77% strong, rhythmic, and mature mRNAs were derived from weak, non-oscillating nascent transcripts, suggesting that post-transcriptional modification has a major role in circadian mRNA expression¹². This builds upon prior research that showed that only approximately 22% of cycling mRNAs exhibit phases identical to those of their de novo transcripts, and certain circadian-related genes were even found to display constant mRNA levels over the course of a day [26]. Together, these data strongly suggest that post-transcriptional regulation is essential for circadian gene expression.

Although post-transcriptional mechanisms, including alternative splicing, mRNA modification, poly(A) tail length, and alternative polyadenylation [27–30], are known to regulate cyclic gene expression, mRNA stability has been the main focus in the study of post-transcriptional regulation of circadian rhythms. Not surprisingly, mRNAs with short half-lives are more likely to display significant oscillations in expression than mRNAs with long half-lives are. For example, a handful of core clock genes in the TTFL, namely, per, tim, and clk, show mRNA oscillations because of variable half-lives, which are likely to be mediated by post-transcriptional regulation. Many studies have demonstrated that mRNA stability depends upon environment and on developmental stage [31], but the most crucial regulatory factors include cis elements in mRNA structure and trans-acting factors (RNA-binding proteins, RBP), such as LARK, for example, which is an RBP that is a key regulator of translation of dbt [32].

A recent study showed that deletion of the 3′UTR of clk mRNA in Drosophila led to ectopic expression of CLK, generation of additional PDF-expressing neurons, and the induction of variable circadian behavior [33]. In addition, deletion of binding sites in clk for the miRNA bantam resulted in a similar phenotype characterized by ectopic expression of VRI and overgrowth of PDF-positive neurons, suggesting that miRNAs regulate the development of circadian neurons [33].

Microarray-based experiments were used to identify miRNAs that show circadian oscillations in the fly head. Several conserved miRNAs, such as miR-263a/b, for instance, show strong daily oscillations [34]. Further analysis demonstrated that, in clock mutants, miR-263a/b have attenuated phases but elevated levels [34].

Later, studies demonstrated correlations between miRNA biogenesis and circadian rhythms. To monitor circadian behavior, flies are usually entrained by several 12-h light/12-h dark (LD) cycles and are then released into constant darkness (DD) for a few days. In DD, circadian rhythms are only dependent on the endogenous clock, as there are no environmental perturbations, and the entrained behaviors persist under DD. Downregulation of GW182, a protein that interacts with Argonaute (Ago) protein to modulate miRNA-mediated translational repression or mRNA degradation, in circadian neurons results in arrhythmicity under DD and an advanced evening phase with absent morning anticipation under LD [35]. The underlying mechanism for GW182 regulation of circadian rhythms appears to be downstream of the PDF-receptor signaling pathway [35].

Dicer1, which encodes an important ribonuclease gene for miRNA biogenesis, has also been implicated in the modulation of the circadian clock. Flies with downregulated Dicer1 in circadian neurons exhibited a lower amplitude of rest to activity rhythms [36]. This relatively weak effect on circadian rhythms might be due to the increased activity of repressors under these conditions, which likely buffers the translation from miRNA deficiencies [36]. The reduced amplitude is likely due to the downregulation of some circadian-relevant miRNAs. Interestingly, another study showed that the period of Dicer-deficient mice was shortened by approximately 2 h due to faster translation of two core clock proteins, namely, PER1 and PER2 [37].

Overexpression of miR-279 in circadian neurons was found to cause arrhythmia in most flies [38]. Luo et al. [38] reported that miR-279 regulated circadian rhythm output pathways by targeting the key component, unpaired (upd), in the JAK/STAT pathway. Though UPD is absent in the main pacemaker neurons (sLNv), it is prevalent in PER-positive neurons, DNs and laterally located neurons (LLNs), all of which are in close proximity to PDF-containing dorsal projections [38]. This further indicated that miR-279 functions in circadian output. Notably, both overexpression and knockout of miR-279 caused a high degree of arrhythmicity, possibly due to the multiple target effects or its essential role in the output pathway. Additional data have demonstrated that miR-996 can rescue the behavioral phenotype in miR-279 mutants [40]. miR-996 is a circadian-relevant miRNA located downstream of miR-279, which shares a similar seed region. Overexpression of miR-996 in circadian neurons was found to result in a phenotype similar to that of miR-279 overexpression [40]. It seems that miR-996 is redundant compared to miR-279, based on the regulatory capacities and similar seed region, as well as abundance, but the presence of miR-996 may add to the robustness of miR-279 regulation of circadian rhythms or may participate in unique functions.

Another miRNA known to influence the circadian clock is miR-276a [39]. miR-276a is positively regulated by the transcription factor Chorion factor 2 (CF2) and has a cyclic expression pattern under LD. Interestingly, flies overexpressing CF2 have a phenotype similar to that seen in flies with an overexpression of miR-276a. miR-276a appears to target the central clock gene, tim [39]. Analyses of flies with spatial and temporal misregulation of miR-276a showed that this miRNA is necessary for circadian behavior robustness in PDF-positive neurons (LNvs). TIM and PER levels were reduced following miR-276a overexpression. Reduced PER levels in PDF-positive neurons was deemed to be a secondary effect of direct targeting of the tim 3′UTR by miR-276a. Deletions of miR-276a binding sites in the tim 3′UTR phenocopied the deletion of miR-276a at both molecular and behavioral levels [39]. Interestingly, TIM overexpression did not fully rescue the arrhythmic phenotype. Two possible explanations for this observation are that the circadian clock may be extremely sensitivity to TIM dosage or that miR-276a targets additional circadian-relevant mRNAs.

The free running period is one of the most important characteristics of circadian rhythms as it reflects the speed of the circadian clock. To date, two miRNAs, namely, bantam and let-7, have been found to be involved in the regulation of the free-running period. These two miRNAs target two central clock genes, clk and cwo, respectively [36, 41] (Fig. 2, Table 1).

It is worth noting that bantam was the first miRNA found to affect the circadian rhythms of flies. The miRNA, bantam, was first shown to play vital roles in germline maintenance, peripheral nervous system dendrite growth, and eye development [42–44]. Through tiling arrays, Kadener et al. [36] found that disruption of miRNA processing in the nucleus resulted in the accumulation of bantam within fly heads, and overexpression of bantam in the main pacemaker neurons lengthens the circadian period. The molecular mechanism behind the circadian-period regulatory function of bantam involves interactions with three conserved target sites of the clk 3′UTR. An interaction between clk mRNA and RISC has also been shown by AGO1 immunoprecipitation [36]. It appears that bantam has other functions, as restoration of clk levels by expression of clk with a 3′UTR with no bantam binding sites did not rescue the arrhythmicity phenotype of clk^AR.

Like bantam, let-7 also modulates the circadian period in Drosophila. Along with two other non-redundant miRNAs, miR-100 and miR-125, let-7 is a co-transcriptional product from a single polycistronic locus, known as the let-7 complex (let-7-C) [45]. Previous studies of let-7-C miRNA were focused on its developmental functions. These miRNAs are important for the remodeling of the abdominal neuromusculature, ovary morphogenesis, nervous system formation, and lifespan control [45–48]. Expression of let-7-C is localized in the α/β lobes of the mushroom bodies [47], brain structures known to be essential for learning and memory regulation, as well as for sleep and locomotor activity [49, 50]. A recent study showed that let-7 is also expressed in LNv neurons, further indicating the role of let-7 in circadian rhythm regulation [41]. Abnormal expression of let-7 in the main pacemaker neurons disrupts normal circadian rhythms: overexpression of let-7 lengthens the period, whereas knockout of let-7 eliminates morning anticipation [41]. Developmental defects were also excluded, as restricted overexpression in adulthood led to a long period of rest-activity rhythm [41]. Notably, cwo, which negatively regulates clk expression, was identified as the target of let-7 both in vitro and in vivo [41]. Other phenotypes resulting from let-7-C knockout, such as abnormal PDF expression in PDF neuron projections and faster PER accumulation in LNv [41], may be due to secondary effects.

Fruit flies exhibit a bimodal behavioral profile during each LD cycle, consisting of increased activity before dawn and dusk (termed morning anticipation and evening anticipation, respectively), both of which are controlled by dedicated circadian neurons [51]. Additionally, flies also display sharp and transient increases of activity during light–dark transitions, known as the “startle” response, which is a non-circadian response that persists even without a functional clock [51] or the main pacemakers [52]. miR-124 is conserved across the animal kingdom and is abundantly expressed in the central nervous system [53] (Fig. 2, Table 1). In the fly head, miR-124 expression is under circadian regulation, with trough and peak expression levels at mid-day and midnight, respectively [34]. Interestingly, miR-124-knockout flies show decreased morning anticipation under LD and an advanced evening phase under DD, but the circadian period is normal [54, 55]. In flies that do not express miR-124, circadian oscillators are functional with unaffected PER/TIM oscillation in all circadian neurons, and the photoreceptors show normal light response, suggesting that neither input or central clock functions are affected [54, 55]. Thus, miR-124 may function in circadian output. Several potential targets of miR-124 encode factors involved in the retrograde BMP signaling pathway, including Mef2, MMP1, and other positive components. However, expression of Mef2 and MMP1 and activation of BMP signaling regulate the endogenous period or rhythmicity under constant darkness [56–58], which is not consistent with the observed phenotype of miR-124 mutants. A double heterozygous mutation of sax and mad was found to partially correct the phase shift in the miR-124 mutant background [55], indicating the multifunction of miR-124 in regulating circadian rhythms.

Unlike miR-124-mediated regulation of the locomotion phase, the miR959-964 cluster mainly regulates the phase of feeding and immune function [59] (Table 1). The miR959-964 cluster, which includes six miRNAs, is encoded by two introns proximal to the protein-coding gene, CG31646 [59]. These six miRNAs show a robust circadian oscillation, which peaks at dusk (ZT12) [59]. The pri-miRNA cluster of miR959-964 displays an even stronger oscillation than individual mature miRNAs do. This oscillation is abolished, and levels accumulate in the arrhythmic per¹ mutant, and, oppositely, mature miR959-964 levels are reduced in the per¹ mutant. Surprisingly, expression pattern analysis did not localize expression of the miR959-964 to circadian-relevant neurons, despite the fact that elimination of four of six miRNAs (miR959-962) displayed mild effects on the circadian period (approximately 0.5 h shorter), and overexpression of the cluster in TIM-positive neurons extended the period for more than 1 h [59]. In fact, this cluster is expressed in the peri-cerebral fat body (adult head fat body), an organ involved in metabolism and immunity [60], a finding that is consistent with the predication that most mRNA targets of miR959-964 are implicated in metabolic and immune functions. The miR959 ~ 964 cluster negatively modulates immune response, as the knockdown line exhibited dramatically lowered survival of the pathogen Pseudomonas aeruginosa [59]. The circadian-modulated feeding behavior was positively regulated by these six miRNAs, and there is a bidirectional regulation between feeding and levels of this cluster [59]. The circadian clock controls feeding time, and miRNAs in this cluster are likely to downregulate factors involved in responses to stress and starvation that regulate multiple downstream physiological processes, such as metabolism, feeding, and foraging time, until the stress is offset.

One direct approach that aims to identify circadian-relevant miRNAs involves monitoring circadian behaviors after performing a forward genetic screen, which entails manipulation of miRNA expression via knockout, overexpression, or downregulation. Recently, a large collection of miRNA mutants generated by targeted homologous recombination has become available [61]. These stocks enable genetic screening for circadian-relevant miRNAs. In addition, the Gal4/UAS system is a powerful tool for manipulating gene expression due to the availability and diversity of transgenic flies with UAS-miRNA or UAS-miRNA-sponge constructs [62]. Thus, by using tissue- and cell-specific drivers, it is feasible to manipulate functions of specific miRNAs only in certain organs. In addition, targeted spatial and temporal knockout of specific genes has become readily available with the emergence of the CRISPR/Cas9 system [63].

One efficient way to identify circadian-relevant miRNAs is to do RNA-sequencing (RNA-seq) in either all circadian or specific circadian neurons [64]. miRNAs can be extracted from specific circadian neurons after FACS isolation or by manual sorting [64]. The comprehensive and high-throughput nature of RNA-seq enables identification of relatively short, non-coding transcripts. Moreover, this approach provides great insights into the oscillation of mature miRNAs and the abundance of pri-miRNAs through intron mapping.

Characterization of miRNA targets is essential for understanding the mechanisms of how miRNA functions. Bioinformatic tools, such as TargetScan [65] and PicTar [66], have been widely used for predicting potential miRNA targets. In the future, in vitro studies can be performed by co-transfecting the miRNA and its predicted target gene 3′UTR with a reporter. If the miRNA binds to its target, a reduced reporter signal will be observed. Additionally, manipulation of target gene expression should phenocopy the effect of the miRNA in vivo. Finally, if a mutant mRNA that does not have miRNA binding sites in the target resembles the miRNA knockout phenotype, this would provide strong evidence supporting miRNA-target interactions.

Emerging evidence has shown that miRNA-mediated post-transcriptional regulation is crucial to the generation and maintenance of a robust circadian clock in flies and mammals [9, 13, 26]. In this review, we discussed miRNAs known to modulate circadian rhythmicity and the free-running period, as well as locomotion and feeding phases in flies. Why is miRNA-mediated post-transcriptional regulation necessary in regulating the different attributes of circadian clock? One likely role is to contribute to the generation of the 24-h period. Although the established molecular mechanism of the circadian clock is based on the TTFL, it takes less than 24 h to close up the TTFL. Thus, additional regulatory levels, such as post-translational regulation and post-transcriptional regulation, for example, are required to generate the 24-h circadian period. A second likely role is to confer robustness to the running of the circadian clock. Multiple regulatory levels ensure that the circadian clock is less sensitive to changing environmental conditions. The third likely role is to enhance the mRNA oscillation of circadian-related genes as rhythmic mRNA expression is an important molecular output pathway of the circadian clock. Taken together, miRNA-mediated post-transcriptional regulation imposes benefits on the operation of the circadian clock.

As of yet, uncharacterized miRNAs may also play important roles in maintaining circadian rhythms. For example, miRNAs are found to be important for circadian photoresponses and seasonal adaptation in mammals [67], but no miRNA has been implicated in the circadian input pathways in flies yet. To date, 466 miRNAs have been identified in the fly genome (www.mirbase.org↗), but only a handful of miRNAs have been reported to be relevant to circadian clock function. Therefore, further characterization of miRNAs will be a vital step toward a more comprehensive understanding of the regulatory mechanisms underlying miRNA-mediated post-transcriptional control of the circadian clock.