Tyramine induces dynamic RNP granule remodeling and translation activation in the Drosophila brain

Ribonucleoprotein (RNP) granules are dynamic condensates enriched in regulatory RNA binding proteins (RBPs) and RNAs under tight spatiotemporal control. Extensive recent work has investigated the molecular principles underlying RNP granule assembly, unraveling that they form through the self-association of RNP components into dynamic networks of interactions. How endogenous RNP granules respond to external stimuli to regulate RNA fate is still largely unknown. Here, we demonstrate through high-resolution imaging of intact Drosophila brains that Tyramine induces a reversible remodeling of somatic RNP granules characterized by the decondensation of granule-enriched RBPs (e.g. Imp/ZBP1/IGF2BP) and helicases (e.g. Me31B/DDX-6/Rck). Furthermore, our functional analysis reveals that Tyramine signals both through its receptor TyrR and through the calcium-activated kinase CamkII to trigger RNP component decondensation. Finally, we uncover that RNP granule remodeling is accompanied by the rapid and specific translational activation of associated mRNAs. Thus, this work sheds new light on the mechanisms controlling cue-induced rearrangement of physiological RNP condensates.


Introduction
Self-assembly of functionally related molecules into the so-called biological condensates has recently emerged as a prevalent process underlying subcellular compartmentalization (Alberti, 2017;Banani et al., 2017). Condensation of RNA molecules and associated regulatory proteins to form cytoplasmic ribonucleoprotein (RNP) granules, in particular, has been observed in virtually all cell types and species, ranging from bacteria to higher eukaryotes (Buchan, 2014;Cohan and Pappu, 2020). Different types of RNP granules have been defined based on their composition (e.g. S-foci), function (e.g. P-bodies), origin (e.g. Stress Granules), and/or the cell type they belong to (e.g. germ granules, neuronal granules) (Kiebler and Bassell, 2006;Anderson and Kedersha, 2009;Voronina et al., 2011;Buchan, 2014;De Graeve and Besse, 2018;Formicola et al., 2019). With the notable exception of Stress Granules, most of these RNP granules are found constitutively and have been implicated in the regulation of various aspects of RNA expression, from decay to subcellular RNA localization and translation (Besse and Ephrussi, 2008;Buchan, 2014;De Graeve and Besse, 2018;Formicola et al., 2019;Ivanov et al., 2019;Marnik and Updike, 2019;Trcek and Lehmann, 2019). Intriguingly, both repressor and activator functions have been assigned to RNA condensation: clustering of transcripts into translation factories, for example, appears to enhance translation (Pichon et al., 2016;Pizzinga et al., 2019;Chouaib et al., 2020;Dufourt et al., 2021), while recruitment of mRNAs to P-bodies, germ granules, or neuronal granules is rather associated

Tyramine induces a reversible remodeling of neuronal RNP granules in MB neurons
In resting brains of 10-15-day-old flies, 100-200 nm-sized cytoplasmic RNP granules are visible in the cell bodies of MB g neurons ( Figure 1A-A'' and C), a population of neurons known for its role in learning and memory (Keene and Waddell, 2007;Keleman et al., 2007;Akalal et al., 2010). These granules contain RNAs such as profilin, as well as regulatory proteins that dynamically shuttle between the granular and soluble pools . Among those are the RBP Imp and the DEAD box helicase Me31B, two conserved repressors of translation ( Figure 1A-A''; Minshall et al., 2001;Nakamura et al., 2001;Hüttelmaier et al., 2005;Hillebrand et al., 2010;Wang et al., 2017). To investigate the response of neuronal RNP granules to changes in neuronal state, we treated brain explants with different neurotransmitters and neuromodulators known to activate MBs and/or to be involved in learning and memory (Campusano et al., 2007;Martin et al., 2007;Majumdar et al., 2012;Silva et al., 2015;Iliadi et al., 2017;Cognigni et al., 2018;Sabandal et al., 2020). The number of Imp-positive RNP granules was scored after 30 min treatment ( Figure 1-figure supplement 1A,B). Tyramine, a bioamine found in trace amounts in both invertebrate and mammalian brains (Burchett and Hicks, 2006;Lange, 2009), triggered the strongest response, characterized by the decondensation of Imp molecules and a significant decrease in the number of Imp-containing granules ( Figure 1B,D and Figure 1-figure supplement 1C). Decondensation of Imp was accompanied by a significant, although less pronounced, relocalization of Me31B protein from the granular to the cytoplasmic pool ( Figure 1B'). This relocalization did not impact on the number of Me31B-positive granules ( Figure 1E), but translated into a decrease in the ratio between the granular and the cytoplasmic soluble pool of Me31B (partition coefficient; Figure 1F). Importantly, the re-localization of Imp and Me31B observed in the presence of Tyramine did not result from changes in protein levels, as similar levels of Imp and Me31B were observed with and without Tyramine treatment (Figure 1-figure supplement 2). These results thus indicate that the neuromodulator Tyramine triggers a remodeling of neuronal RNP granules characterized by the differential release of Imp and Me31B RNP components into the cytoplasm.
To test whether granule component decondensation was reversible, we transferred brain explants previously treated for 30 min with Tyramine to regular saline and fixed them after 60 min of recovery. While a significant decrease in the number of Imp-positive granules was observed after Tyramine treatment, bright Imp-positive granules were again visible after the recovery period and their number returned to baseline (Figure 1-figure supplement 1C). Similarly, the decreased partitioning of Me31B into granules was reverted after recovery ( Figure 1-figure supplement 1C). This thus suggests that Tyramine reversibly alters the phase behavior of RNP granule components.

Dynamics of Tyramine-induced RNP granule remodeling
Although the experiments described above highlighted that neuronal RNP granules dynamically reorganize, they did not provide detailed information about the temporal profile of RNP component decondensation. To monitor in real time the properties of RNP granules, we introduced via CRISPR/ Cas9 editing a GFP tag in the endogenous me31B locus and performed high-resolution real-time imaging of Me31B-GFP-expressing brain explants. This first revealed that granules exhibit a dynamic behavior characterized by successions of short movements and pauses, as well as both fusion and fission events (Figure 2A and Videos 1 and 2). To dynamically monitor the response of RNP granules to Tyramine, we then imaged Me31B-GFP-positive granules for 30 min after treatment ( Figure 2B-D and Videos 3 and 4) and quantitatively analyzed the partitioning of Me31B over time ( Figure 2B). This revealed that relocalization of Me31B from the granular to the cytoplasm pool is initiated within minutes after Tyramine treatment, but is a progressive rather than abrupt process. As shown in  Tyramine signals through the TyrR receptor to induce neuronal activation and RNP component decondensation Having discovered the impact of Tyramine on RNP granules, we next wondered how Tyramine signaling was mediated. A number of receptors responding to Tyramine have been identified in Drosophila (Ohta and Ozoe, 2014;El-Kholy et al., 2015), yet only one -the GPCR TyrR -has been shown to respond specifically to Tyramine, and not to other biogenic amines (Cazzamali et al., 2005;Huang et al., 2016). To thus test whether TyrR would mediate neuronal RNP granule remodeling, we treated TyrR Gal4 null mutant brain explants with Tyramine and analyzed the behavior of granule markers. As shown in Figure 3A-E, decondensation of both Imp and Me31B was significantly impaired upon TyrR inactivation, suggesting that TyrR is the main receptor involved. Both neuromodulatory and neurotransmitter functions have been assigned to Tyramine to date (Nagaya et al., 2002;Pirri et al., 2009;Huang et al., 2016;Jin et al., 2016;Schützler et al., 2019). To investigate whether Tyramine induced a calcium response in MB neurons, we monitored calcium transients upon exposure of brain explants to Tyramine, using a genetically encoded calcium indicator expressed in MBs (MB247-homer::GCamp3.0; Pech et al., 2015). Tyramine elicited calcium transients peaking 2-8 min after exposure ( Figure 3F and Videos 5 and 6), as well as a modest, but reproducible and dose-dependent, long-term increase in intracellular Ca 2+ ( Figure 3G). Importantly, inactivation of TyrR largely (although not completely) inhibited the main calcium peak induced by Tyramine ( Figure 3F), confirming the specificity of MB neuronal response. The slow response observed upon Tyramine exposure suggests that Tyramine may activate MB neurons indirectly through other neurons. Consistent with this idea, inhibiting the firing of MB neurons through conditional expression of the inward-rectifying potassium channel Kir 2.1 (Paradis et al., 2001) significantly suppressed Tyramine-induced Imp decondensation (Figure 1-figure supplement 1D), indicating the need for both evoked responses and GPCR-mediated signaling and suggesting the existence of TyrR-expressing neurons acting as a relay to transduce Tyramine signaling.

CamkII is required for Tyramine-induced RNP granule remodeling
To identify the proteins involved in the dynamic regulation of RNP granules, we expressed GFPtagged Imp proteins specifically in MB g neurons and immunoprecipitated GFP-fusions from adult head lysates (Figure 4-figure supplement 1A and Materials and methods). Co-precipitated proteins were identified by mass spectrometry, and heads expressing sole GFP were used as a specificity control. In total, 51 proteins were reproducibly identified, distributed into various functional categories (Figure 4-figure supplement 1B and Supplementary file 1). As expected, RBPs were strongly enriched in the bound fraction (22/51 proteins; p<0.001). Not all RBPs present in Imp RNP granules were however recovered , presumably because our immunoprecipitation approach mainly targeted soluble cytoplasmic complexes. Among the identified Imp highlighted. The morphology of a single MB g neuron is represented in red. The region imaged to analyze RNP granule behavior is boxed (turquoise dotted lines). (D, E) Normalized numbers of Imp-(D) or Me31B-(E) containing granules (per image field). Individual data points were color-coded based on the experimental replicate they belong to. Three (D) to four (E) replicates were performed and the mean value of each replicate is indicated as a symbol (triangle). At least 20 (D) or 12 (E) data points were collected for each replicate. ***, p<0.001 (t-test on individual data points). n.s. stands for not significant. (F) Distribution of Me31B partition coefficients. Partition coefficients were estimated by dividing the intensity of Me31B signal in individual RNP granules to the intensity of the cytoplasmic diffuse pool (see Materials and methods) and calculated for each granule detected in the imaged fields. The individual data points displayed on the graph were extracted from a single replicate. Three replicates were performed and the mean value of each is indicated as a symbol (triangle). Number of RNP granules: 622 granules distributed across 18 fields (control), 777 granules distributed across 18 fields (+ Tyramine). ***, p<0.001 (t-test on individual data points). Note that the p-value obtained when comparing the distributions of replicate means is 0.7 (Mann-Whitney test). For the list of values used to generate the graphs shown in D-F see Figure 1-source data 1. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Numerical data to support graphs in Figure 1D-F.     interactors, we focused our attention on Ca 2+ /calmodulin-dependent protein kinase II (CamkII), as it is a conserved kinase activated in response to calcium rises (Coultrap and Bayer, 2012). To validate the association between Imp and CamkII, we performed co-immunoprecipitation experiments in cultured S2R+ cells. As shown in Figure 4A, CamkII co-immunoprecipitated with GFP-Imp, but not with sole GFP. Furthermore, CamkII interacted with Imp both in the presence and in the absence of RNase, indicating that the Imp/CamkII interaction is RNA-independent. In vivo, both CamkII and phospho-CamkII (the active form of CamkII) were found diffusely localized in the cytoplasm of MB g LUT of ImageJ. Scale bar: 2 mm. (B) Me31B-GFP mean partition coefficients in function of time in brain explants treated (red) or not (black) with 10 mM Tyramine. Each data point represents the mean of the average partition coefficients measured for all granules present in fields imaged at a given time point. Tyramine was added at t = 2 min (orange arrow). (C, D) Image sequences extracted from movies recording the cell bodies of adult Mushroom Body (MB) g neurons endogenously expressing Me31B-GFP proteins. Brain explants were either maintained in saline (C), or supplemented with 10 mM Tyramine (D) at t = 2 min. Images were originally acquired every 30 s. Intensities are displayed using the 'Fire' LUT of ImageJ. Scale bar: 5 mm. Numbers of movies: 7 (ctrl) and 11 (+ Tyr). Note that in these experiments MB g neurons could not be unambiguously distinguished from other MB neuronal subpopulations. No difference could however be observed in the behavior of Me31B-GFP-positive granules within MB neurons. For the list of values used to generate the graphs shown in B see Figure 2-source data 1. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Numerical data to support graphs in Figure 2B.  To then investigate whether the function of CamkII was important for the remodeling of neuronal RNP granules in response to Tyramine, we expressed in MB neurons the ala peptide, a peptide derived from the CamkII autoinhibitory segment that binds to the catalytic site and was shown to selectively inhibit CamkII activity, both in vitro and in vivo (Griffith et al., 1993;Carrillo et al., 2010;Nesler et al., 2016;Newman et al., 2017). Conditional expression of the ala peptide significantly inhibited the decondensation of Imp molecules observed upon Tyramine treatment ( Figure 4B), demonstrating that CamkII is required cell-autonomously downstream of Tyramine to promote Imp decondensation. Ala-mediated inhibition of CamkII, however, did not inhibit the partial cytoplasmic relocalization of Me31B observed in response to Tyramine ( Figure 4C), indicating that CamkII specifically modulates the decondensation of Imp.

Tyramine induces the translational activation of granule-associated mRNAs
Neuronal RNP granules are thought to maintain associated mRNAs in a translationally silenced state (Krichevsky and Kosik, 2001;Fritzsche et al., 2013;El Fatimy et al., 2016;De Graeve and Besse, 2018). To investigate whether the observed release of granule components is accompanied by the translational derepression of granule-associated mRNAs, we monitored the translation of profilin, an mRNA known to be directly bound by Imp and present in MB RNP granules (Medioni et al., 2014;Vijayakumar et al., 2019). First, we analyzed the expression of a reporter in which the 3'UTR of profilin is fused to the coding sequence of EGFP. Constructs generated with the SV40 3'UTR were used as a negative control. As shown in Figure 5A, treating brains with Tyramine induced a significant increase in GFP signal intensity for the construct containing profilin 3'UTR, but not for that containing SV40 3'UTR. Furthermore, no significant increase in GFP expression was observed upon prior incubation of gfp-profilin 3'UTR-expressing brains with the translation inhibitor anisomycin ( Figure 5B), indicating that increased GFP levels result from increased protein synthesis. To extend our analysis to other mRNAs, we then monitored the response of two other reporters: one enriched in Imp and Me31B-positive granules in control conditions (gfp-cofilin 3'UTR), the other not (gfp-camk2 3'UTR) (K. Pushpalatha and F. Besse, unpublished). Remarkably, increased expression of gfp-cofilin 3'UTR, but not of gfp-camk2 3'UTR reporters, was observed upon Tyramine treatment ( Figure 5A), suggesting that translation activation is specific to granule-associated mRNAs. Further consistent with a model where RNP granule remodeling relieves associated mRNAs from translational repression, blocking Imp decondensation through inhibition of Video 1. Fusion between two Me31B-GFP-containing granules. Real-time imaging of Me31B-GFP-containing granules in the cell body of an intact adult brain explant. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s. Scale bar: 0.5 mm. https://elifesciences.org/articles/65742#video1 Video 2. Fission of a Me31B-GFP-containing granule.
Real-time imaging of Me31B-GFP-containing granules in the cell body of an intact adult brain explant. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s. Scale bar: 0.5 mm.
As GFP-based reporters reflect translation status indirectly, and with poor temporal dynamics, we then aimed at monitoring translation activation with high spatiotemporal resolution, using the Sun-Tag methodology (Pichon et al., 2016;Wang et al., 2016;Wu et al., 2016;Yan et al., 2016), recently deployed in Drosophila (Dufourt et al., 2021). SunTag-tagged profilin transcripts were coexpressed in MB g neurons together with scFv-GFP-NLS fusions to detect translation sites and brains were imaged in real-time. Strikingly, Tyramine induced within minutes the formation of bright GFPpositive cytoplasmic foci ( Figure 6A and Video 7). These foci were not observed in the absence of SunTag-tagged transcripts ( Figure 6A,B and Video 8). Furthermore, their formation was inhibited by puromycin ( Figure 6A,B and Video 9), indicating that they form through translation and likely correspond to translation foci. Remarkably, activation of profilin translation occurred as a burst, with kinetics very similar to that of Tyramine-induced calcium transients. As shown in Figure 6C and Video 7, indeed, the number of cells exhibiting SunTag foci peaked 2-8 min after Tyramine exposure, before progressively reverting to baseline values.
As translation peaked before RNP components completed their decondensation, we wondered if mRNAs would be released from granules rapidly after Tyramine treatment. smFISH experiments were thus performed 10 min after treatment to monitor the association of endogenous profilin transcripts ( Figure

Tyramine triggers RNP component decondensation and translation activation
Membrane-less RNP condensates enriched in transcripts under tight regulatory control, as well as regulators of RNA translation, transport or decay have been described in various cell types and organisms (Buchan, 2014). Neurons exhibit a particularly complex collection of RNP granules composed of both shared and distinct components, raising the question of how granule composition is established and dynamically regulated (Fritzsche et al., 2013;De Graeve and Besse, 2018;Formicola et al., 2019). Frameworks have been proposed to explain RNP granule compositional control, in which scaffold molecules (or nodes) establish a core network of multivalent interactions essential for both granule nucleation and further recruitment of more dynamically associated client Video 3. Behavior of Me31B-GFP-containing granules in brain explants treated with control saline. Real-time imaging of MB g cell bodies expressing endogenously expressed Me31B-GFP. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 30 min. Control HL3 buffer was added at t = 2 min. Scale bar: 3 mm. https://elifesciences.org/articles/65742#video3 Video 4. Dynamic response of Me31B-GFP-containing granules to Tyramine. Real-time imaging of MB g cell bodies expressing endogenously expressed Me31B-GFP. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 30 min, from intact adult brain explants. Tyramine was added at t = 2 min to reach a 10 mM final concentration. Scale bar: 3 mm. molecules (Banani et al., 2016;Ditlev et al., 2018;Sanders et al., 2020). In this model, modulation of scaffold or client partitioning properties can lead to the differential release of granule-enriched proteins and RNAs. Whether and how these frameworks apply to endogenous neuronal RNP granules has remained unclear. Our results uncovered that Tyramine stimulation differentially impacts on neuronal RNP components, as a nearly complete release of granule-associated Imp, but only a partial relocalization of granular Me31B was observed in response to Tyramine. These results, together with our observation that inactivation of me31B, but not of imp, prevents granule assembly (K. Pushpalatha and F. Besse, unpublished) are consistent with a model in which Me31B behaves as a scaffold, and Imp as a client whose partitioning into granules depends on Me31B, and is specifically modulated by Tyramine. Interestingly, the release of Imp from neuronal granules is associated with Figure 3 continued indicated as a symbol (triangle). At least 10 data points were collected for each replicate. ***, p<0.001 (Kruskal-Wallis test on individual data points with Dunn's post-test). n.s. stands for not significant. (E) Distribution of Me31B partition coefficients. Partition coefficients were estimated by dividing the intensity of Me31B signal in individual ribonucleoprotein (RNP) granules to the intensity of the cytoplasmic diffuse pool and calculated for each granule detected in the imaged fields. The individual data points displayed on the graph were extracted from a single replicate. Three replicates were performed and the mean value of each is indicated as a symbol (triangle). At least 10 fields were analyzed per condition. Number of RNP granules: 875 granules distributed across 18 fields (control), 558 granules distributed across 14 fields (+ Tyramine), and 475 granules distributed across 10 fields (+ Tyramine in TyrR Gal4 mutants). ***, p<0.001 (one-way ANOVA on individual data points with Dunnett's post-tests). n.s. stands for not significant. (F) Average fluorescence intensity (F) of GCamp3.0 signal in MB calyx upon exposure of control (green) or TyrR Gal4 mutants (orange) brain explants to 10 mM Tyramine. The MB247-homer::GCamp3.0 reporter was used to monitor Ca 2+ levels. Data are plotted as F(t)-F(t = 0)/F(t = 0) (DF/F(0); see Materials and methods). Tyramine was added at t = 2 min (red arrow). Error bars correspond to S.E.M. Number of brains analyzed: 18 (control) and 13 (TyrR Gal4 mutants). (G) Dose-dependent long-term increase in Ca 2+ levels upon exposure to Tyramine. Intensities of GCamp3.0 signal are plotted as F(t = 30 min) -F(t = 0). Numbers of brains analyzed: 5 (-), 5 (1 nM), 6 (300 nM), 6 (30 mM), and 8 (10 mM). For the list of values used to generate the graphs shown in D-G see Figure 3-source data 1.
The online version of this article includes the following source data for figure 3: Source data 1. Numerical data to support graphs in Figure 3D-G.
Video 5. Tyramine induces a rise in the intracellular Ca 2+ concentration of control MB neurons. Real-time imaging of MB neurons (calyx region) expressing a Homer::GCamp3.0 construct under the control of the MB247 promoter. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 15 min, from intact adult brain explants.
Tyramine was added at t = 2 min. Scale bar: 10 mm. https://elifesciences.org/articles/65742#video5 Video 6. Tyramine-induced calcium response is inhibited in TyrR-/-mutant context. Real-time imaging of a TyrR Gal4 brain in which MB neurons expressed a Homer::GCamp3.0 construct under the control of the MB247 promoter. MB calyx region is shown. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 15 min, from intact adult brain explants. Tyramine was added at t = 2 min. Scale bar: 10 mm.  . CamkII interacts with Imp and is required for Tyramine-induced Imp decondensation. (A) CamkII co-immunoprecipitates with Imp. FLAG-CamkII constructs were co-transfected with either GFP-Imp or GFP (negative control) in S2R+ cells. GFP proteins were immunoprecipitated and the bound fractions (right) used for western blot. Input fractions (left) were used as a control of expression. Anti-FLAG and anti-GFP antibodies were used to detect respectively CamkII (MW » 55-60 kDa) and Imp (MW » 90 kDa) fusion proteins. Cell lysates were treated (+) or not (À) with RNase prior to immunoprecipitation. (B) Normalized numbers of Imp-containing granules in brain explants treated (+ Tyramine) or not (ctrl) with 10 mM Tyramine. + ala refers to the condition where the CamkII inhibitory peptide ala was expressed specifically in MB neurons, using the tub-Gal80ts;;OK107-Gal4 driver. Individual data points were color-coded based on the experimental replicate they belong to. Three replicates were performed for each condition and the mean value of each replicate is indicated as a symbol (triangle). At least 12 data points were collected for each replicate. (C) Distribution of Me31B partition coefficients. Me31B partition coefficients were estimated by dividing the intensity of Me31B signal in individual ribonucleoprotein (RNP) granules to the intensity of the cytoplasmic diffuse pool and calculated for all the granules detected in imaged fields. The individual data points displayed on the graph were extracted from a single replicate. Three replicates were performed and the mean value of each is indicated as a symbol (triangle). Number of RNP granules: 622 granules distributed across 18 fields (control), 777 granules distributed across 18 fields (+ Tyramine) and 663 granules distributed across 16 fields (+ Tyramine +ala). *, p<0.05; ***, p<0.001 (one-way ANOVA on individual data points with Dunnett's post-tests). n.s. stands for not significant. For the list of values used to generate the graphs shown in B, C see Source data 1. Numerical data to support graphs in Figure 4B, C. the translational activation of its target RNA profilin, suggesting that differential release of client RBPs, rather than complete granule disassembly, might represent a means to modulate the expression of specific sets of client-associated RNAs in response to distinct stimuli. As further revealed by our real-time imaging experiments, translation activation and granule component decondensation, although both starting within minutes after Tyramine exposure, exhibit distinct temporal profiles. While translation activation occurs as a burst, decondensation of RNP components is a continuous and progressive process. This may result from the gradual depletion of the pool of translationally repressed mRNAs normally dynamically recruited to RNP granules.

CamkII activity is required downstream of Tyramine to inhibit Imp partitioning
Phosphorylation is a rapid and reversible post-translational modification that dramatically impacts on both intra-and inter-molecular interactions. Not surprisingly then, phosphorylation was shown in vitro and in cells to regulate the partitioning of RNP components into condensates, either positively or negatively depending on the context (Hofweber and Dormann, 2019;Kim et al., 2019). In living systems, studies have so far pointed to a role for granule-enriched kinases in the phosphorylation of scaffold proteins, leading to granule disassembly. This process is for example required for clearance of Stress Granule upon recovery (Wippich et al., 2013;Krisenko et al., 2015;Shattuck et al., 2019), or for dissolution of anteriorly localized P-granules in C. elegans zygotes (Wang et al., 2014).
If and how phosphorylation modulates the recruitment of client molecules to regulate RNP composition rather than assembly/disassembly has so far remained poorly explored. Here, we showed that the calcium-activated CamkII kinase interacts with Imp, and is required downstream of Tyramine to trigger Imp decondensation. As Tyramine induces a calcium response in MB neurons, our results thus suggest a model in which Tyramine-induced activation of CamkII may promote its association with Imp, thus inhibiting the partitioning of Imp into somatic RNP granules ( Figure 6-figure supplement 1). Whether CamkII directly phosphorylates Imp remains to be investigated. Although Imp contains four CamkII consensus phosphorylation sites (RXXS/T), mutating these four sites into RXXA by CRISPR/Cas9-mediated engineering did not prevent the decondensation of phosphomutant Imp proteins in response to Tyramine (Figure 4-figure supplement 1D). CamkII may thus either phosphorylate Imp through non-canonical sites, as described for other targets (Kennelly and Krebs, 1991;Sun et al., 1994;Huang et al., 2011;Herren et al., 2015), or phosphorylate partner(s) of Imp essential for its association with neuronal RNP granules. Interestingly, our results suggested that CamkII is not enriched within RNP granules, raising the hypothesis that kinases may modulate RNP granule composition by targeting the soluble pool of client molecules, rather than the granule-associated pool.

Tyramine signals through the TyrR receptor to activate MB neurons and trigger RNP granule remodeling
Tyramine is a biogenic amine produced from Tyrosine in neurons expressing the Tyrosine decarboxylase enzyme (Lange, 2009). Although it has for long exclusively been considered as a precursor of Octopamine, recent work has revealed Tyramine-dependent but Octopamine-independent function in Drosophila, identifying Tyramine as a neuroactive chemical modulating different aspects of animal physiology and behavior (Huang et al., 2016;Schützler et al., 2019). A number of GPCRs responsive to Tyramine have been cloned and pharmacologically tested, and three have been defined as Tyramine receptors (Oct-Tyr, TyrR, and TyrRII). Only one was shown to bind Tyramine with high affinity and specificity: TyrR (also termed CG7431 or TAR2) (Cazzamali et al., 2005;Ohta and Ozoe, 2014). As indicated by single-cell transcriptomic analyses, neither TyrR nor any of the other Tyramine receptors is significantly expressed in MB neurons (Davie et al., 2018). Our results, however, have shown that both the calcium response and the remodeling of neuronal RNP granules triggered in . Consistent with this idea, the calcium response we observed in MBs in response to Tyramine is slow, peaking 2-8 min after addition of Tyramine. Furthermore, Tyramine-dependent remodeling of neuronal RNP granules was significantly blocked upon Kir2.1-mediated hyperpolarization of MB neurons, a process that should inhibit evoked responses but not GPCR-mediated signaling. The identity of the TyrR-expressing neurons relaying Tyramine signal remains to be investigated. Previous experiments based on a knock-in line in which the TyrR locus has been deleted and replaced by Gal4 showed that this receptor is expressed in a restricted number of brain neurons, some of them (e.g. TyrIPS neurons) projecting in the vicinity of MBs and thus representing potential candidates (Huang et al., 2016). In the future, it will also be interesting to determine in which physiological or behavioral contexts the neuronal circuit activated by Tyramine regulates the translation of RNP granule-associated RNAs, and whether this regulation might be local. Although the physiological roles of Tyramine have so far been poorly characterized, this conserved bioamine has been involved in the regulation of both metabolic and behavior traits (Lange, 2009;Ohta and Ozoe, 2014;Ma et al., 2016;Schützler et al., 2019;Roeder, 2020). Potentially relevant to the known functions of MB neurons, Tyramine was in particular shown to regulate the avoidance behavior of flies exposed to repulsive olfactory cues (Kutsukake et al., 2000) and to dampen male courtship drive (Huang et al., 2016). Whether these responses involve post-transcriptional regulatory mechanisms such as those described in the context of long-term olfactory habituation in flies (Bakthavachalu et al., 2018) however remains to be investigated. Thus, testing whether Tyramine, similar to its derivative Octopamine (Burke et al., 2012;Wu et al., 2013), modulates learning and memory processes known to depend on RNA regulation represents another interesting line of future research.

Materials and methods
Key resources Continued on next page Figure 5 continued three outlier data were omitted from the graph (although they were considered to calculate the mean of the corresponding replicate and to perform statistical tests). (C) Normalized GFP signal intensities produced by the GFP-profilin 3'UTR reporter. GFP-profilin 3'UTR was expressed solely (ctrl), or together with ala (+ ala), and brain explants were treated (+) or not (À) with 10 mM Tyramine for 30 min. The ala inhibitory peptide was expressed conditionally in adult MB neurons using tub-Gal80ts;;OK107-Gal4. (D) Proportion of gfp-profilin 3'UTR RNA molecules contained in Me31B-mTomatopositive granules. Co-localization was measured using the JACoP plugin of ImageJ (see Materials and methods and Figure 4-figure supplement 1C) and values normalized to controls. Individual data points in A-D were color-coded based on the experimental replicate they belong to. Three replicates were performed for each condition and the mean value of each replicate is indicated as a symbol (triangle). At least 10 data points were collected for each replicate. ***, p<0.001 (t-tests on individual data points for A, D and Kruskal-Wallis test on individual data points with Dunn's post-tests for B, C). n.s. stands for not significant. For the list of values used to generate the graphs shown in A-D see Figure 5-source data 1.
The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Numerical data to support graphs in Figure 5.
The UASp_scFvGFP_NLS plasmid was generated through Gibson assembly, by inserting the 10Â UAS and p-transposase promoter sequences from pVALIUM22 into Not1/Xho1-digested pNo-sPE_scFvGFP_NLS plasmid (Dufourt et al., 2021) using NEBuilder HiFi DNA Assembly Master Mix. Fragments were amplified using the following primers: 965scfvgfpF (5'-ggccagatccaggtcg-cagcggccgcGCGGCCGCATAACTTCGTATAATG-3'); 966scfv (5'-cggggcccatCTCGAGTGA Video 7. Tyramine induces the assembly of SunTagprofilin foci. Real-time imaging of MB g cell bodies coexpressing SunTag-profilin RNAs and ScFv-GFP-NLS fusions under the control of VT44966-Gal4. Note that VT44966-Gal4 is expressed at high level only in a subset of MB g neurons. Some cells initially contained a big cluster of ScFv-GFP fusions; these cells were excluded from the analysis. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 30 min, from intact adult brain explants. Tyramine was added at t = 2 min to reach a 10 mM final concentration. Scale bar: 3 mm. The line expressing GFP-Imp proteins with mutated CamkII consensus sites from the endogenous locus was generated through CRISPR/Cas9 gene editing, following a two-step procedure. First, an imp-RMCE line in which the upstream GFP exon of the G080 line was preserved, but most of the imp locus deleted and replaced by a cassette containing the 3xP3-RFP selection marker flanked by inverted attP sites was generated by homology-dependent repair, using two gRNAs (upstream: ACA TTGCATTGCAGCTGAGTTGG and downstream: GCGAGCTCACAACAGTAAGGAGG) and a dsDNA donor plasmid with 1 kb long-homology arms (chrX:1078009-10799070 and 10797081-10798308; Wellgenetics). Second, a donor pBS-KS-attB1-2 plasmid in which the four RXXS/T consensus sites of Imp were mutated into RXXA was generated through Gibson assembly and integrated through cassette exchange (RMCE) in the imp-RMCE line. Individuals with a correctly oriented cassette were selected by PCR and their genomic imp locus sequenced.
Video 8. ScFv-GFP-positive foci are not observed in the absence of SunTag-profilin transcripts. Real-time imaging of MB g cell bodies expressing ScFv-GFP-NLS fusions under the control of VT44966-Gal4. Note that VT44966-Gal4 is expressed at high level only in a subset of MB g neurons. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 12 min, from intact adult brain explants. Tyramine was added at t = 2 min to reach a 10 mM final concentration. Scale bar: 3 mm. https://elifesciences.org/articles/65742#video8 Video 9. Puromycin disrupts the assembly of SunTagprofilin foci. Real-time imaging of MB g cell bodies coexpressing SunTag-profilin RNAs and ScFv-GFP-NLS fusions under the control of VT44966-Gal4. Note that VT44966-Gal4 is expressed at high level only in a subset of MB g neurons. Signal intensities are displayed using the 'Fire' LUT of ImageJ. Images were acquired every 30 s for 12 min, from intact adult brain explants. Brain explants were incubated with 250 mM Puromycin for 15 min before addition of Tyramine. Tyramine was added 2 min after the movie starts, to reach a 10 mM final concentration. Scale bar: 3 mm.
smFISH Drosophila brains were dissected in cold RNase-free HL3 and treated for 10 min with 10 mM Tyramine. Samples were then fixed for 1 hr in 4% formaldehyde in PBS and dehydrated overnight in ethanol 70%. Brains were then briefly rinsed in wash buffer (10% formamide in 2Â SSC) before overnight incubation at 45˚C in Hybridization Buffer (100 mg/mL dextran sulfate, 10% formamide in 2Â SSC) supplemented with Quasar 570-labeled Stellaris Probes at a final concentration of 0.25 mM.

Image acquisition
Fixed samples were imaged using a Plan Apo 63X NA 1.4 oil objective, on a Zeiss LSM880 inverted confocal microscope equipped with an Airy Scan module. Images were acquired with a pixel size of 0.043 mm and were processed with the automatic Airy Scan processing mode (strength 6.0). Note that MB g neurons were located based on their position within MBs (estimated by DAPI) and the differential expression of Imp in MB neuron sub-types (Medioni et al., 2014).

Live imaging
Brains of 10-15-day-old flies were dissected in HL3 at room temperature, and then transferred and mounted on a four chambered 35 mm dish with 20 mm bottom well (IBL, #D35C4-20-0-N) poly-lysinated before use (3 hr incubation at RT or overnight at 4˚C). Once correctly oriented, brains were stabilized on the plate using low melting agarose (NuSieve GTG Agarose #50080, 0.07%, dissolved in HL3). Live imaging of RNP granule remodeling and time-course calcium imaging was performed using a 40Â NA 1.1 water objective, on a Zeiss LSM880 inverted confocal microscope equipped with an Airy Scan module. Images were acquired every 30 s for 32 min. Tyramine (dissolved in HL3) or HL3 were added as a single drop 2 min after the start of imaging. To inhibit translation, puromycin (sigma, #P8833) was applied at a final concentration of 250 mM and incubated for 15 min prior to addition of Tyramine. For Tyramine dose-response experiments, MB247-homer::GCamp3.0 brains were mounted on 35 Â 10 mm petri dishes (Nunc, #153066) coated with poly-lysine and covered with HL3. MB calyx regions were then imaged with an upright Leica DM6000 TCS SP5 confocal microscope equipped with a HCX APOL 40X water (0.8 NA) objective. Two images of MB calyces were taken for each brain: one at t = 0 (before addition of Tyramine) and one at t = 30 min.

Image analysis RNP granule and RNA detection
For RNP quantifications, 114.9 mm 2 ROIs containing six to seven cells were cropped from single z slices (two independent ROIs per brain), treated with a Gaussian Blur filter, resized by a factor two using the Image Pyramid plugin and converted from 32-bit to 16-bit images, all in ImageJ. Granules were detected and quantified using the Small Particle Detection (SPaDe) algorithm available under the following link :https://hal.inria.fr/hal-01867805/document. Minimal granule size was set to four pixels and threshold defined so as to optimize measured F1 scores (as described in De Graeve et al., 2019). F1 scores were calculated by comparing the spatial coordinates of manually annotated RNP granules with the binary masks generated by SPaDe with different thresholds. Threshold used for detection of Imp and Me31B-positive granules was hence set to 0.6234. In experiments where Imp antibody staining was used, RNP granules of a size smaller than 13 pixels were excluded from analysis. A similar procedure was used for detection of smFISH RNA spots, using 0.3434 as SPaDe threshold.

Me31B partition coefficient measurements
Me31B partition coefficients were defined as the ratio between the maximal intensities of individual RNP granules and the average diffuse cytoplasmic signal. Maximal RNP granule intensities were measured with ImageJ on the original raw images, using the masks generated by SPaDe. For cytoplasm measurements, masks were obtained using the following procedure: first, a Gaussian Blur filter (s:8) was applied on images in which granule-containing pixels were previously blanked; second, the mean of the intensity signal was measured and used as lower limit for a thresholding interval (the upper limit being the maximum possible intensity value, 65000). Mean intensities were then measured on raw images using these masks. To monitor RNP remodeling over time, two regions of 102.6 mm 2 , each containing six to seven cells, were cropped per movie. For each time point, Me31B partition coefficients were calculated for each granule of the cropped regions as described previously, and mean partition coefficients calculated.

Me31B and Imp protein levels
Protein signals in the whole cytoplasm were measured in living samples at t = 0 and t = 32 using two regions per movie of respectively 102.6 mm 2 for Me31B and 71.2 mm 2 for Imp. Individual t = 32 values were normalized to their corresponding t = 0 measurements.

Intracellular calcium quantification
Real-time calcium imaging was quantified as follows: a 155 mm 2 region was selected in the MB calyx of each imaged hemisphere and average intensity measured for each time point. Data were plotted as F(t)-F(0)/F(0), where F(0) is the mean of the four intensity values obtained before Tyramine was added (t0 to t3), and F(t) represents the intensity value measured at time t.
For the dose-dependent c response to Tyramine, GCamp3 mean fluorescence intensity was measured for each hemisphere in ImageJ, on the t = 0 and t = 30 images using a 17.5 mm 2 ROI. Measures were normalized such that the value of each t0 was set to 1.

ScFv-GFP puncta analysis
The number of ScFv-GFP-containing spots was counted manually from movies. Cells containing at t = 0 a big cluster of ScFv-GFP fusions were excluded from the analysis.

Translational reporter expression
For the 3'UTR reporter experiments, average GFP intensity was measured with ImageJ on maximum intensity projections, within cell body regions of 173.6 mm 2 . Two ROIs were considered per brain.

Colocalization analysis
To quantify the number of profilin RNA contained in RNP granules, binary masks were first generated by SPaDe from the RNP granule channel (GFP-Imp or Me31B-mTomato) and from the smFISH channel (GFP-profilin or profilin). Co-localization was then assessed using the ImageJ JACoP plugin (https://imagej.nih.gov/ij/plugins/track/jacop.html) and the object-based method (Bolte and Cordelières, 2006), where RNA spots were defined by their center of mass and RNP granules considered as particles, respectively. The fraction of RNP spots contained in granules was calculated for each condition.
Immunoprecipitation-mass spectrometry 201Y-Gal4, UAS-GFP-Imp and 201-Gal4, UAS-GFP flies were amplified at 25˚C in bottles. 3-5-dayold flies were collected and immediately frozen. Heads were collected at 4˚C using two prechilled sieves of different mesh sizes (630 mm on top and 400 mm at the bottom) and crushed into powder using prechilled mortar and pestle. The head powder (3 g for each experiment) was then transferred to a prechilled 15 mL glass Dounce Tissue Grinder and homogenized in 10 mL DXB buffer (25 mM HEPES), pH 6.8, 50 mM KCl, 1 mM MgCl 2 , 1 mM dithiothreitol (DTT), 250 mM sucrose, 1/100 Halt Protease, and Phosphatase Inhibitor Cocktail (# Thermo Scientific). The homogenate was cleared by two consecutive centrifugations (10,000 rpm for 10 min at 4˚C). 250 mL of GFP-Trap_A beads (Chro-moTek, Germany) were then added to the cleared lysate in a 15 mL falcon tube and incubated on a rotator for 1.5 hr at 4˚C. Beads were pelleted by mild centrifugation (2000 rpm for 2 min at 4˚C), washed twice with DXB buffer and once with DXB buffer + 0.1% NP 40. Proteins bound to the beads were eluted by addition of 1.6 mL of 0.2M glycine, pH 2.5 and incubation for 10 min on a rotator. Eluates were then collected and neutralized with 400 mL of 1M Tris-HCl, pH 8, before being concentrated on an Amicon Ultra-2 mL centrifugal filter, MWCO 10 kDa (Merck Millipore, USA) to obtain a final volume of 50 mL that was loaded on a polyacrylamide gel. Coomassie-stained gels were cut, trypsin-digested, and further processed through LC-MS/MS by the EMBL proteomic core facility. Peptides were searched using Mascot (search engine, v.2.2.07) against the Uniprot_Drosophila melanogaster database, and then uploaded and analyzed using the Scaffold three software. Proteins fulfilling the following requirements were considered as Imp interactors: (i) represented by at least 10 peptides, (ii) represented with at least twice more peptides in the GFP-Imp pull-downs than in the GFP control IPs, and (iii) meeting the previous requirements for the two independent replicates performed.

Statistical analysis
Statistical details for each experiment can be found in the corresponding figure legends. SuperPlots were used to show individual data points with a color code referring to the original experimental set. Average values of each biological replicate are shown using the same color-code (Lord et al., 2020).
All statistical analyses were performed using GraphPad Prism 8. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.