MicroRNA-92a is a circadian modulator of neuronal excitability in Drosophila

To identify PDF cell cycling miRNAs, green fluorescent protein (GFP)-labelled PDF cells were manually sorted for RNA extraction and miRNA libraries constructed with an optimized protocol. Although enough miRNA reads were obtained, sample variation was substantial and precluded using the RNA-sequencing data to identify cycling expression patterns; they usually have modest amplitudes (two- to threefold). Therefore, RT–qPCR was used, which produced more consistent results. On the basis of hints (potential cyclers) from the sequencing results, 16 miRNAs were tested, of which 10 showed reproducible daily expression profiles between two biological replicates; 6 of the 10 are rhythmically expressed (and). As only a few miRNAs were found cycling in heads, the results here suggest more cycling miRNAs in PDF cells than the whole heads. ²⁸ Fig. 1a Supplementary Fig. 1 ^{22 23 24 25 26 27}

We focus in this manuscript on mir-92a. It was the only PDF cell cycling miRNA with lower levels during most of the daytime and higher levels during the nighttime (). In the first day of constant darkness (DD1), miRNA cycling persists but with a 4–8 h phase advance compared to LD, suggesting that light also affects mir-92a expression (and). Fig. 1a Fig. 1b Supplementary Fig. 2

To test further whether the cycling expression is under the control of the core molecular clock, the same LD assay was done inflies (;. They are completely arrhythmic because of a nonsense mutation in, a core clock gene. There is no indication of residual mir-92a cycling in this background, indicating that it is indeed downstream of the core molecular clock (). per per PDF-GAL4;UAS-mCD8::GFP) per 0 0 Fig. 1c

To address the functions of mir-92a, we first tested whether manipulating mir-92a levels affects the well-characterized circadian morphological changes in the PDF cell termini. They undergo daily fasciculation–defasciculation cycles under circadian control. To this end, mir-92a was overexpressed, or knocked down using a miRNA sponge (SP), in PDF cells with co-expression of mCD8::GFP (mir-92aOE ormir-92aSP, respectively). The morphological cycles of the PDF cell termini were quantified with Sholl analysis, that is, assaying the intersections between axon branches and concentric circles (see Methods). ^{13 29} PDF-GAL4;UAS-mCD8::GFP;UAS- UAS-

Compared with wild-type (WT) flies (w1118,in agenetic background), overexpression of mir-92a (mir-92aOE,mir-92aOE) specifically in PDF cells maintains projections in the fasciculated state at both ZT2 and ZT14 as indicated by non-cycling and low numbers of axonal crosses (). Specifically, overexpression of mir-92a shows an ∼37% decrease in maximal axonal crosses at ZT2 and no significant differences in axonal crosses at ZT14 (and). In contrast, the knockdown results in the opposite, namely the defasciculated state, with no significant differences between the control (scramble,scramble) and knockdown (mir-92aSP,mir-92aSP) at ZT2 but an ∼24% increase in maximal axonal crosses in the mir-92a knockdown at ZT14 (and). No differences of axonal length were observed among genotypes. To address possible developmental effects from use of, the same experiments were performed using the inducible geneswitch driverand the transgenes were only activated in adults. The same results were obtained (), indicating that adult-specific manipulation of mir-92a levels changes the fasciculation–defasciculation state of the PDF termini. PDF-GAL4;UAS-mCD8::GFP/+ w1118 PDF-GAL4;UAS-mCD8::GFP;UAS- PDF-GAL4;UAS-mCD8::GFP;UAS- PDF-GAL4;UAS-mCD8::GFP;UAS- PDF-GAL4 PDF-GSG, Fig. 2a Fig. 2a Supplementary Fig. 3 Fig. 2a Supplementary Fig. 3 ³⁰ Supplementary Fig. 4

Defasciculation can reflect higher neuronal excitability and fasciculation lower excitability. Rat mir-92a is also implicated in synaptic scaling, suggesting that fly mir-92a suppresses neuronal excitability. To address this possibility more directly, we tested how mir-92a levels affect PDF cell depolarization with high concentration of KCl, by monitoring changes in fluorescence levels of the voltage sensor ArcLight (mir-92aOE ormir-92aSP); they decrease when neurons are depolarized. Brains were attached to the bottom of a chamber with adult haemolymph-like saline (AHL), and baseline fluorescence recorded with a microscope. After 60 s of baseline recording, KCl was perfused into the chamber, which caused an immediate and drastic decrease in fluorescence levels (and). Interestingly, mir-92aOE significantly decreased the response, and mir-92aSP consistently but insignificantly increased the response (). ^{30 31 32 11} Fig. 2b Supplementary Movie 1 Fig. 2b PDF-GAL4;UAS-ArcLight;UAS- UAS-

To further test the effect of mir-92a, nicotine was used to stimulate PDF cells andwas used to focus on adult-specific effects. Nicotine is a more physiological agonist and has been shown to fire PDF cells via nicotinic receptors. To this end, the Casensor GCaMP6 was co-expressed with eithermir-92a ormir-92aSP under the control ofAlthough robust increases of fluorescence levels were induced by 3 × 10M nicotine, we obtained negative results with the mir-92a manipulations. Changes among genotypes were not statistically significant, perhaps because of relatively large variations among brains as well as the weakerdriver compared to(see below). This was despite a similar trend as the KCl stimulation, namely decreased responses with mir-92a overexpression and increases with mir-92a knockdown (). PDF-GSG UAS- UAS- PDF-GSG. PDF-GSG PDF-GAL4 ³³ Supplementary Fig. 5 2+ −6

An independent approach to address mir-92a function was to monitor Calevels in PDF cells using animaging system. CaLexA is an artificial transcription factor usually located in the cytoplasm when Calevels are low. Higher Calevels cause CaLexA to translocate to the nucleus, where it can bind to the upstream LexAop element and activate luciferase expression of a transgene. Luciferase levels therefore positively correlate with Calevels in this system. To test whether mir-92a levels change Calevels as another proxy of neuronal excitability, luciferase levels were measured every hour for three consecutive days in living WT flies and in flies with manipulated levels of mir-92a (mir-92aOE ormir-92aSP). 2+ 2+ 2+ 2+ 2+ in vivo PDF-GAL4;UAS-CaLexA;LexAop-luciferase;UAS- UAS- ³⁴

Calevels in WT flies show a rhythmic pattern with peaks in the morning, consistent with previously reported electrophysiological assays and GCaMP6 imaging of PDF cells (). Compared to WT, overexpressing mir-92a significantly lowers the luciferase levels, especially during the light period (LP) when mir-92a levels are low (). On the contrary, knocking down mir-92a increases luciferase levels, especially during the dark period when mir-92a levels are high (). 2+ Fig. 2c ^{10 12} Fig. 2c Fig. 2c

GCaMP6 was also used to monitor Calevels in PDF cells. Flies that expressedtogether with eithermir-92aOE ormir-92aSP undercontrol were dissected, and the fluorescence levels of their PDF termini measured and quantified. Whereas PDF termini show apparent spontaneous activity, we were not able to accurately estimate the spiking rates and amplitudes from the Casignals because of noise and heterogeneity within individual neurons. We could however quantify the differences between baseline fluorescence levels among genotypes. Consistent with theCaLexA results, knocking down mir-92a levels resulted in an ∼224% increase in baseline fluorescence levels at ZT18–22, and overexpression resulted in an ∼40.6% decrease at ZT6–10 (). Since higher Calevels are associated with higher neuronal excitability, these GCaMP6 data as well as the CaLexA data support the hypothesis that mir-92a suppresses neuronal excitability. 2+ 2+ 2+ UAS-GCaMP6f UAS- UAS- PDF-GAL4 in vivo Supplementary Fig. 6

To test whether the effect of mir-92a on neuronal excitability is restricted to PDF cells, mir-92a levels were either up- or downregulated in neurons regulating fly sleep. Dopaminergic neurons are well-characterized wake-promoting neurons in mammals and in flies. Overexpression of mir-92a in these neurons (;mir-92aOE) might lower their excitability and therefore increase sleep. Indeed, sleep duration is significantly increased in the overexpression flies, and sleep duration is significantly reduced when mir-92a is knocked down (). Similar changes in sleep duration were observed when mir-92a levels were manipulated in another set of wake-promoting neurons driven by(). As a negative control, sleep duration ofmir-92aOE/+ ormir-92aSP/+ flies without GAL4 drivers was measured and was indistinguishable from WT flies (). ³⁵ Fig. 3a Supplementary Fig. 7A ⁸ Supplementary Fig. 7B TH-GAL4 UAS- Dvpdf-GAL4 UAS- UAS-

A subset of dorsal clock neurons contain 8–10 cells per brain and are sleep-promoting; the driver isfrom Janelia Research Campus. Overexpression of mir-92a in these cells led to a significant decrease in sleep duration, consistent with the neuronal excitability hypothesis (). Although mir-92a knockdown in these cells had no effect (), there are many possible reasons for a negative result, for example, thedriver is not sufficiently strong, or endogenous mir-92a is not well expressed in these neurons. PDFR-GAL4 GAL4 ³⁴ Fig. 3b Fig. 3b

We also assayed sleep duration in mir-92a null (mir-92aKO) flies. It is affected and in the ‘correct' direction: the mutant flies show ∼17% less total sleep and ∼44% sleep loss during the LP (). These sleep results taken together further confirm that mir-92a suppresses neuronal excitability and also indicate that this regulation is not restricted to PDF cells. Fig. 3c

Since mir-92a suppresses neuronal excitability, we wanted to test whether neuronal excitability also affects mir-92a levels. This possibility might also be relevant to the circadian cycling of mir-92a levels. As light pulses during the nighttime fire PDF cells and phase shift the circadian clock, an effect of light/excitability on mir-92a levels could be via an effect on the core clock. We therefore subjected entrained flies to a 10-min light pulse at either ZT15 or ZT21 and assayed PDF cell mir-92a levels after a subsequent 50 min in the dark. (The same 10 min protocol is used for traditional phase shift assays, with maximum phase delays observed at ZT15 and maximum phase advances at ZT21.) ⁸

mir-92a levels increased by approximately twofold in PDF cells after the ZT15 light pulse, and they decreased by ∼2.5-fold after the ZT21 light pulse (). The results appear specific for PDF cells: the light pulses had no effects on mir-92a levels in dopaminergic neurons, which are not activated by light (). Fig. 4a Supplementary Fig. 8

To test whether manipulating mir-92a levels in PDF cells has an effect on behaviour, circadian rhythms, sleep and phase shift responses of the mir-92aOE or SP flies (mir-92aOE ormir-92aSP) were assayed. There was no effect on rhythmic strength or circadian period (), reflecting perhaps the more modest effect on excitability compared to previous experiments. However, sleep and phase shift responses were altered. Knockdown of mir-92a in PDF cells resulted in decreases of sleep duration in both the LP and dark period, whereas overexpression decreases and broadens evening peak activity (). As the role(s) of PDF cells in regulating sleep and circadian rhythms is currently enigmatic, it is unclear how to interpret these changes. However, PDF cells have a more straightforward relationship to phase-shifting. Here dimmer light (0.69 mW cm) was used to ensure that the system was not saturated by strong light. Surprisingly perhaps, mir-92aOE leads to bigger shifts both at ZT15 and ZT21, and mir-92aSP causes smaller shifts (). This observation is consistent with an effect of mir-92a levels on neuronal excitability, with the direction of the effects possibly due to a more labile or sensitive circadian clock when the neuronal excitability of PDF cells is decreased by mir-92aOE and the opposite by the mir-92aSP. PDF-GAL4;UAS- UAS- Supplementary Fig. 9A ^{36 37} Supplementary Fig. 9B ^{8 38} Fig. 4b −2

To identify a mir-92a target responsible for the observed phenotypes, translating ribosome affinity purification (TRAP) was performed on flies with mir-92a either up- or downregulated throughout the circadian system (mir-92aOE compared to/+ (control), andmir-92aSP compared toscramble (control)). Potential targets should show increased mRNA levels (indicated by the Input) and/or translating mRNA levels (indicated by the immunoprecipitation (IP)/Input) with mir-92a downregulation, and decreased mRNA levels with mir-92a upregulation. In addition, these targets should have predicted mir-92a-binding sites in their 3′UTRs with TargetScan (). There were 26 genes from our TRAP data that met these criteria (). We then tested these potential candidates with RNA interference (RNAi) in the CaLexA system, the sleep assay and the PDF projection morphology to focus on candidates with the same phenotype as mir-92aOE flies. Tim-GAL4;UAS-RiboTag;UAS- Tim-GAL4;UAS-RiboTag Tim-GAL4;UAS-RiboTag;UAS- Tim-GAL4;UAS-RiboTag/UAS- http://www.targetscan.org/↗ Supplementary Table 1

a homologue of mammalianandwas the only candidate that met all these criteria (see below). It is an NAD-dependent deacetylase of the Sirtuin family. There is one predicted mir-92a-binding site in the3′UTR. The site is conserved amongspecies but does not exist in mammals according to TargetScan. The TRAP results indicate no detectable changes inmRNA input levels between WT and mir-92a OE or knockdown flies, whereas the IP/Input levels anticorrelate with mir-92a levels. The data therefore suggest that mir-92a suppressesexpression by inhibiting its translation without markedly affecting mRNA stability (). Unfortunately, no good antibody against fly SIRT2 is available for western blots or immunostaining. sirt2, sir2 sirt3 sirt2 Drosophila sirt2 sirt2 Fig. 5a

To confirm that mir-92a suppressesexpression by binding to the predicted site in the 3′UTR, a luciferase reporter assay was performed in S2 cells. Either a WT or a binding site-mutated3′UTR was inserted into the psiCHECK2 vector downstream of thereporter gene. An internalreporter was expressed separately from the vector as a transfection control (). sirt2 sirt2 renilla firefly luciferase Supplementary Fig. 10

Co-transfection ofand-mir-92a together with the reporter plasmid carrying WT 3′UTR results in a significantly lower renilla/firefly luciferase bioluminescence ratio compared to controls in which the irrelevant geneor the irrelevant miRNA mir-184 was co-transfected instead of mir-92a (). Moreover, a mutation of the mir-92a-binding site in the 3′UTR eliminates the mir-92a suppression (). The results confirm that mir-92a suppressesexpression bothand. Ub-GAL4 UAS dsRed sirt2 in vitro in vivo Fig. 5b Fig. 5b

If mir-92a suppresses neuronal excitability by downregulatingexpression,RNAi should phenocopy mir-92aOE. Three RNAi lines are available from the Transgenic RNAi Project. One shows high percentage lethality at the pupal stage at 25 °C (#36868 from Bloomington Stock Center), but the other two (#32482, RNAi-1 and #31613, RNAi-2) show ∼85% knockdown efficiency: both lower endogenousin heads to ∼15% when driven by(). sirt2 sirt2 sirt2 Tubulin-GAL4 Supplementary Fig. 11

was knocked down in PDF cells (RNAi) with both RNAi lines, and PDF projections were maintained in the fasciculated state during the day as well as the night (). At ZT2, knockdown ofdecreases maximal axonal crosses by ∼31% and ∼38%, respectively, with no significant differences observed at ZT14 (). This is similar to the mir-92aOE phenotype shown above (). In addition,RNAi abolished the effect of the mir-92aSP in maintaining the PDF projections defasciculated in both the day and the night, indicating thatis epistatic to (downstream of) mir-92a (). AsRNAi driven byis also sufficient to maintain projections in the fasciculated state, the phenotype is probably not due to developmental effects (). sirt2 PDF-GAL4;UAS-mCD8::GFP;UAS-sirt2 sirt2 sirt2 sirt2 sirt2 PDF-GSG Fig. 6a Supplementary Fig. 12 Fig. 2a Supplementary Fig. 12 Supplementary Fig. 13

In theCaimaging assay, decreased Calevels were also observed with aknockdown (RNAi;), similar to mir-92aOE (). In addition, adulthood-specific (GSG) sirt2 RNAi decreased PDF neuron responsiveness to nicotine stimulation (), comparable to the effect of mir-92a overexpression (). in vivo sirt2 PDF-GAL4;UAS-CaLexA;LexAop-luciferase;UAS-sirt2 PDF- 2+ 2+ Fig. 6b Fig. 2c Supplementary Fig. 14 Supplementary Fig. 5

In dopaminergic neurons,RNAi (RNAi) caused increased sleep, which phenocopies mir-92aOE (;RNAi-2 increased sleep duration only when combined withto increase the knockdown efficiency;). Knockdown ofin flies co-expressing the mir-92aSP in dopaminergic neurons increased the sleep duration back to normal levels, indistinguishable from WT flies, indicating once again thatactivity is epistatic to mir-92a ().RNAi also causes a bigger phase shift response at both ZT15 and ZT21 like mir-92aOE (). sirt2 TH-GAL4;UAS-sirt2 sirt2 UAS-dicer2 sirt2 sir2 sirt2 Fig. 6c Supplementary Fig. 15 Fig. 6d Fig. 6e

An amorphic strain of(#8839 from Bloomington Stock Center) was also assayed. We first checked whether this phenotypically null strain shows any PDF cell fasciculation phenotype. However, the projections also show substantially increased branching (), which complicates the fasciculation assay. To better quantify and distinguish defasciculation versus branch overgrowth, we assayed the defasciculation index (DI); it is the percentage of axonal intersections across concentric rings outside of a 15° cone. High DI indicates defasciculation, and low suggests fasciculation. A low DI at ZT2 as well as ZT14 indicates that PDF cell projections maintain a fasciculated state in theamorphic strain, and increases of axonal crosses indicate an increased branching at both ZT2 and ZT14 (). This indication of increased branching could be due to non-cell autonomous effects and/or a complete lack of functional SIRT2 in the animals, for example, a developmental effect. sirt2 sirt2 Supplementary Fig. 16A ³¹ Supplementary Fig. 16

We also assayed theamorphic flies behaviourally. They show ∼23% increase in total sleep duration and almost 100% increase of sleep during LP (). This is similar toRNAi in dopaminergic neurons and opposite to mir-92a null flies (and). Strikingly,amorphic flies are also highly arrhythmic, showing significantly lower rhythmicity index compared to w1118 WT flies (). This may be due to the disruption of cycling neuronal excitability in circadian neurons. sirt2 sirt2 sirt2 Supplementary Fig. 17A Fig. 3c 6c Supplementary Fig. 17B

In summary,RNAi phenocopies mir-92aOE in PDF cells and can reverse the phenotypes caused by mir-92aSP in the morphological and behavioural assays. Not surprisingly, althoughamorphic flies present a more complicated picture, the data taken together indicate that mir-92a suppresses neuronal excitability via the downregulation ofexpression. sirt2 sirt2 sirt2

Flies were reared on standard cornmeal/agar medium with yeast under 12:12 h LD cycles at 25 °C.;was described in ref..mir-92aOE,andRNAi flies were from the Bloomington stock centre.mir-92aOE was backcrossed six times to w1118 flies.scramble andmir-92aSP were kind gifts from Davie Van Vactor lab.flies were described in ref..was described in ref..CaLexAwas a recombined stable line as described in ref..was described in ref..was described in ref..was described in ref..was a kind gift from Dr JH Park.was a gift from the Zipursky lab. PDF-GAL4 UAS-mCD8::GFP UAS- PDFR-GAL4 UAS-sirt2 UAS- UAS- UAS- UAS-ArcLight UAS-GCaMP6f PDF-GAL;UAS- ;LexAop-luciferase PDF-GSG TH-GAL4 PDF-GAL4 Dvpdf-GAL4 UAS-RiboTag ^{53 11 54 34 30 8 9}

psiCHECK2 plasmid is commercially available from Promega. To insert the3′UTR, the vector was first digested with XhoI and NotI and the PCR product amplified from gDNA with the3′UTR forward and reverse primers () and then incorporated into the vector with Gibson Assembly (NEB). To mutate the mir-92a-binding site in the3′UTR, the Agilent Quickchange Kit was used with the following primers: sirt2 mut 3′UTR forward and reverse (). sirt2 sirt2 sirt2 Supplementary Table 2 Supplementary Table 2

Trikinietics Acitivity Monitors (Waltham, MA) were used to measure the locomotor activity of individual flies around 7-day old. For sleep assays, female flies were used, and the data analysed with sleep analysis scripts developed by the Griffith lab at Brandeis University using MATLAB (MathWorks, Natick, MA). For phase shift assays, male flies were entrained and subjected to a 10-min 0.69 mW cmlight pulse at a given time point and left in DD for 7 days. Circadian rhythm behaviour including rhythmicity, period and phase shift was analysed as described. −2 ⁸

Immunostaining was done as described. Briefly, fly heads were fixed in PBS with 4% paraformaldehyde supplemented with 0.008% Triton X-100 for 1 h at 4 °C before dissection. A mouse anti-GFP antibody (1:1,000, Sigma-Aldrich, G6539), a mouse anti-PDF antibody (1:10, Developmental Studies Hybridoma Bank, University of Iowa, Iowa city, IA, described in ref.) and Alexa Fluor 488 (1:500, Invitrogen Cat#: A-11001) were used as primary and secondary antibodies. Brains were imaged at × 20 on a Leica SP5 confocal microscope. Images are maximum projections of Z sections. Sholl analysis in FIJI was used for axonal crosses, and quantification was according to software instructions. DI was calculated with modified Sholl analysis as described in ref.. ^{38 8 31}

Fly brains were dissected in AHL consisting 108 mM NaCl, 5 mM KCl, 2 mM CaCl, 8.2 mM MgCl, 4 mM NaHCO, 1 mM NaHPO, 5 mM trehalose, 10 mM sucrose and 5 mM HEPES, and mobilized to the bottom of a perfusion Sylard-bottom (Dow Corning, Midland, MI) chamber filled with AHL using a pin anchored to the Sylgard. Depolarization buffer (high KCl for ArcLight experiments) containing 28 mM NaCl, 85 mM KCl, 2 mM CaCl, 8.2 mM MgCl, 4 mM NaHCO, 1 mM NaHPO, 5 mM trehalose, 10 mM sucrose and 5 mM HEPES, or 3 × 10M Nicotine diluted in AHL (for GCaMP6f experiments, Sigma-Aldrich N3876, concentration modified according to ref.) was perfused into the chamber using a gravity-fed ValveLink perfusion system (Automate Scientific, Berkeley, CA). Imaging was done with an Olympus BX51WI fluorescence microscope (Olympus, Center Valley, PA) under an Olympus × 60 (0.90 W, LUMPlanFI) water-immersion objective and was captured using a charge-coupled device camera (Hamamatsu ORCA C472-80-12AG). The following filter sets were used for excitation and emission (Chroma Technology, Bellows Falls, VT): excitation, HQ470/ × 40; dichroic, Q495LP; emission, HQ525/50 m. Frames were captured with μManager with 2 Hz with 4 × 4 binning with 500 ms exposure time and 50 ms intervals. Fluorescence levels were quantified with FIJI. 2 2 3 2 4 2 2 3 2 4 ^{55 56 33 56 57 58} −6

The recording was done as described in ref.. Basically, food containing 1% agar and 5% sucrose was heated to melt and supplemented with 20 mM of D-luciferin potassium salt (GOLDBIO), 250 μl of which was then distributed into every other well in a 96-well plate. Plates were allowed to cool down completely before use. Individual flies were loaded into each well and the plate was then sealed with a transparent adhesive (TopSeal-A PLUS, Perkin Elmer). Every well was punctured with two to three small holes to allow air circulation, and recording was with a TopCount NXT luminescence counter (Perkin Elmer) in an incubator under LD cycles at 25 °C. Data were then analysed with MATLAB. ³⁴

For cell-extracted RNA, ∼100 GFP-labelled PDF cells were manually sorted and stored in 100 μl TRIzol reagent (Invitrogen). A detailed procedure for cell sorting is described in ref.. RNA was then extracted following the supplier's protocol. To quantify individual miRNAs, total extracted RNA was ligated with a 3′ adaptor and then a 5′ adaptor for RT–PCR. The PCR product was then diluted (1:20) and quantified by qPCR using a universal reverse primer and a miRNA-specific forward primer (). 2S rRNA was amplified along with miRNAs and served as a normalization RNA. The strategy is adapted from a miRNA deep-sequencing protocol and allows easier and more higher-throughput screening. Stem–loop qPCR confirmed the results for mir-92a (). ²⁸ Supplementary Table 2 Supplementary Fig. 18

Flies expressing(FLAG tag) in addition to mir-92a manipulation (mir-92aOE (w1118 as control) ormir-92aSP (scramble as control)) were collected on dry ice and decapitated. Fly heads were homogenized (lysate kept as input) and immunoprecipitated with Sigma M2 anti-FLAG magnetic beads. RNA was then extracted from the beads with TRIzol reagents (IP). RT–qPCR and high-throughput sequencing were performed with both the Input and IP RNA for quantification. For high-throughput sequencing, data were mapped to dm3 genome using Tophatand expression levels were quantified using Cufflinks. Tim-GAL4;UAS-RiboTag UAS- UAS- UAS- ^{59 60}

S2 cells plated in 96-well plates (Costar, 3610) were co-transfected with 12.5 ng of each plasmid mixed with 2 μl of Cellfectin II Reagent (Thermo Fisher). Plasmids includedmir-92a,,mir-184 and psiCHECK2. The Promega Dual-Luciferase Reporter Assay System was used to measure luciferase levels 3 days post transfection.is the reporter for 3′UTR activity, andis the transfection efficiency control (). Ub-GAL4, UAS- UAS-dsRed UAS- Renilla luciferase Supplementary Fig. 10

The authors declare that all data supporting the findings of this study are available within the article and its, or from the corresponding author on request. Supplementary Information