PKA and cAMP/CNG Channels Independently Regulate the Cholinergic Ca2+-Response of Drosophila Mushroom Body Neurons1,2,3

The mushroom bodies (MBs) are the most prominent structures in adult Drosophila brain. They have been involved in several crucial functions, such as learning and memory, sleep, locomotor activity, and decision making.


Introduction
Several behavioral and genetic studies performed in different invertebrate organisms, such as honeybees, locusts, and the fruitfly, Drosophila melanogaster, demonstrated the critical role of the mushroom bodies (MBs) in olfactory learning and memory (L&M) (Heisenberg, 1998;Menzel, 2012), as well as in other functions such as sleep (Joiner et al., 2006), locomotor activity (Martin et al., 1998;Besson and Martin, 2005), and decision making (Tang and Guo, 2001). In Drosophila, a combination of genetic and behavioral studies, based on an extensive library of mutants and transgenic animals, have identified a number of genes and signaling cascades that contribute to memory formation McGuire et al., 2005;Keene and Waddell, 2007;Busto et al., 2010). Genes such as dunce (dnc) (Dudai et al., 1976), encoding a phosphodiesterase (PDE) that degrades cAMP, and rutabaga (rut) (Duerr and Quinn, 1982), encoding an adenylate cyclase (AC) that synthesizes cAMP, are known to regulate cAMP levels. Although these genes were identified more than 30 years ago, their precise roles and physiological consequence of disrupting cAMP levels is largely unknown. Furthermore, the Ca 2ϩ -response that contributes to several cellular processes and even gene expression, yielding in vivo memory formation in the MBs, is still not well characterized. This lack of information is mainly due to the limited electrophysiological access to the neurons in the brain of adult Drosophila, though in bigger invertebrates such as honeybee, some electrophysiological studies have been performed (Schäfer et al., 1994). Indeed, the majority of studies, notably on dnc and rut mutants, have been performed on other neurons, such as motoneurons (Ueda and Wu, 2009) or on dissected brains, at different developmental stages (Lee and O'Dowd, 2000).
More recently, physiological approaches using fluorescent markers to image the fly brain have begun to explore MB physiology. Except in the few studies that have investigated both the calyx and the lobes (Tomchik and ), up to now the majority of calcium imaging studies were focused on single MB regions, either on a part of the calyx/cells bodies 2004;Honegger et al., 2011) or on the lobes (Yu et al., 2006;Akalal et al., 2010). Thus, in adult flies, except for protein kinase A (PKA) quantification (Gervasi et al., 2010), the direct in vivo effect of disturbing the cAMP signaling pathway on the Ca 2ϩ -response and on the overall cellular physiology of the Kenyon cells (KCs) still remains poorly characterized.
In this work, we took advantage of the in vivo bioluminescence imaging technique recently developed (Martin et al., 2007) to simultaneously monitor neuronal Ca 2ϩactivity of the whole MB structure, including the calyx/ cell-bodies (CCB) and the lobes, continuously, over a long time period. We recorded the nicotine (cholinergic)induced Ca 2ϩ -response, employing both genetics (using mutants and/or targeted RNAi) and pharmacological approaches to manipulate different components of the cAMP signaling pathway. We show that the downregulation or upregulation of cAMP levels results in a proportional change of the Ca 2ϩ -response, while acute increase in the cAMP levels is sufficient to trigger a Ca 2ϩ -response. Finally, genetic manipulation of PKA, a cAMP effector, suggests that cAMP also has a PKA-independent effect, via the cyclic nucleotide-gated Ca 2ϩ -channel (CNG).

Flies
Flies were maintained on standard medium at room temperature (24°C). P[UAS-GFP-aequorin] (GA) transgenic flies (Martin et al., 2007) were used in conjunction with the P[GAL4]OK107 line to target GA to the MBs. P[GAL4]OK107 (Bloomington Stock Center) is expressed in a large population (approximately 90%) of KCs (Aso et al., 2010). Imaging experiments were performed on progeny of flies containing both the

Pharmacology
To stimulate the flies, we used either acetylcholine or nicotine. To investigate the roles of cAMP pathway, forskolin and IBMX were used. Nicotine (Sigma) was prepared as a 10 mM stock solution in H 2 O and diluted to 25 M in Drosophila Ringers just prior experiment. Forskolin (Sigma) was prepared as a 13 mM stock in ethanol and then dissolved in Drosophila Ringers to 13 M. IBMX (Sigma) was daily dissolved at 40 mM in 100% ethanol and diluted further in Drosophila Ringers at 200 M final concentration. 8Br-cAMP (Sigma) was dissolved at 20 mM in 100% ethanol and diluted further in Drosophila Ringers at 200 M final concentration. All drugs were applied using the previously described perfusion system.

Determination of the nicotine-induced Ca 2؉ -response
The duration and the total photons (TP) were determined using an automated statistical analysis of the signal script developed at the laboratory (a routine programmed in Microsoft Visual Basic/Excel, available on request). Briefly, a sliding window of 20 data points (5 s) was compared using t test with a control window of 240 data points (30 s) corresponding to the recorded resting phase before nicotine application. The response of the KCs was considered (quantified) between the time we obtained more than six of 10 consecutive p values above 0.025 (starting response) and more than six of 10 consecutive p values below 0.025 (end of response).

Quantitative and statistical analysis
We used the Photon Viewer (2.1) software (Science Wares) written in LabView 7.1 (National Instruments) to analyze the imaging data. Nicotine-induced bioluminescence signals are presented as photons/s (within the ROI). Image recordings were obtained from five to 15 flies for each genotype. All statistics were done using InVivoStat (2.1) software (Clark et al., 2012), a biostatistics front-end for the open-source statistic package based on the R project (http://www.r-project.org/). The dataset was analyzed via one-way ANOVA followed by a planned comparison on the predicted means to compare the level of the selected effect using the Benjamini-Hochberg's with a rank transformation (Benjamini and Hochberg, 1995;Benjamini et al., 2001).

Results
In Drosophila, as in mammals, the olfactory integration network is composed of at least two successive integration nodes (synapses) linked by nerve bundles. The odor, transduced by the olfactory receptors neurons (ORNs), is first integrated in the antennal lobe glomeruli (Wilson and Mainen, 2006;Wilson, 2013). This integration occurs through a complex network of local interneurons and dendrodendritic connections with the projection neurons (PNs) (Ng et al., 2002;for review, see Wilson, 2013). The PNs send their axons to two distinct structures: the calyx of the MBs (which represent the dendritic arborisation of KCs) and the lateral horn. Synaptic connections between PNs and the KCs are the second critical site of integration of the olfactory input (Murthy et al., 2008;Turner et al., 2008;Cassenaer and Laurent, 2012). These excitatory PNϪKC synapses are cholinergic, with KCs expressing ionotropic nicotinic acethylcholine receptor (nAchR) (Fayyazuddin et al., 2006). In this study, we focused on intrinsic KC physiology, a part of the network described above. In order to stimulate nAChRs on the MBs and trigger Ca 2ϩ -responses, we first applied the endogenous agonist acetylcholine (Ach), which triggers a Ca 2ϩresponse when applied to dissected brains, as reported previously (Yu et al., 2003). However, Ach is unstable and can potentially affect other receptor types, like the muscarinic Ach receptor, located in other parts of the olfactory integration network, such as the antennal lobes (Blake et al., 1993). Hence, we use nicotine, which is more stable and allows better and more reliable stimulus control. As previously reported in pupae MB cultures (Campusano et al., 2007), nicotine application evokes a similar response pattern as Ach application, but is more reproducible.

Nicotine induced a characteristic Ca 2؉ -response in mushroom bodies
We used a 20ϫ objective, which allowed visualisation of the entire MB at once, and recorded responses from the CCB and various MB lobes (Fig. 1AϪC), which could be subdivided into the vertical lobe, comprised of the ␣/␣' lobes, and the medial lobes, comprised of the ␤/␤' and ␥ lobes. In the absence of any stimulus, we observed neither basal nor oscillatory Ca 2ϩ -activity in the KCs (the constitutive neurons of the MBs). A 1 min application of nicotine (25 M, at 2 ml/min) evoked a typical response pattern in the MBs. The response started in the CCB and propagated into the axonal projections at the level of the MB lobes (Fig. 1DϪF). A typical nicotine-evoked Ca 2ϩresponse was composed of two distinct phases in the CCB, and only one phase in the MB lobes. The CCB response first showed a rapid exponential activity increase (0 s corresponds to the beginning of the response), and peaked at approximately 9 s (Fig. 1E). This first phase reached ϳ2200 photons/s (ph/s), the signal then decreased slightly for ϳ2 s, and rose again to give a second lower peak of ϳ1800 ph/s, ϳ15 s after the first response started. The responses finally decreased slowly, and terminated after ϳ80 s. To simplify, the response can be summarized into two components, which are defined by the first and the second peak. In addition, the use of different angles of view to observe the MB permitted the identification of substructures associated with both response components. The first component corresponds to the response in the calyx (Fig. 1H, green ROI), while the second component, which occurs slightly after, corresponds the response in the cell bodies of the KCs (Fig.  1H, orange ROI). Indeed, a refinement of the two ROIs, which was possible on few flies according to their precise angles of view, allowed spatiotemporal separation of these two components of the response (Fig. 1G). However, as the two components partly overlap in the majority of the flies imaged, it made it difficult to precisely and systematically separate the two components and to define their individual durations. Consequently, only the overall response of the CCB and duration were taken into account in this study. Similarly, the spatial resolution obtained at the level of the MB lobes did not allow us to precisely discriminate different subneuronal populations from each other. Therefore, the ␣/␣' lobes are considered altogether as the vertical lobe (VL), while the ␤/␤' and ␥ lobes are considered altogether as the medial lobes (ML) in this study. Moreover, due to the position of the fly's head and the recording angle, the VLs partly overlapped with the peduncles of the MBs. Thus, in order to avoid any bias in subsequent analysis, we only quantified the response in the CCB and ML. In summary, the first component of the response corresponds to the calyx (dendritic branches), whereas the second component corresponds to the cell bodies (Fig. 1E). The response in the ML (Fig. 1E, red curve) was delayed compared to the CCB response, and was composed of a single peak of approximately 1100 ph/s, which occurred roughly 10 s after response initiation. The ML response lasted for about 55 s in total. We also quantified the total number of emitted photons for the response in the CCB and ML. The TP average was ϳ39000 photons from the CCB and ϳ13000 from the ML (see Fig. 3). Finally, to confirm that this robust Ca 2ϩresponse does not significantly vary with genetic background, results were obtained with additional control lines (VDRC-GD-60000 and VDRC-KK-61000), which were recorded and then shown to share the same characteristics as the CS trans-heterozygotes flies (Fig. 2IϪJ,LϪO, blue bars).

Decreasing cAMP decreases the Ca 2؉ -response
In L&M, the role played by the cAMP pathway within the MB has been extensively studied using genetic and behavioral approaches. However, the involvement of this pathway in modulation of the MB Ca 2ϩ -response is still only partly documented. Thus, in order to determine the cAMP pathway's role in the MB Ca 2ϩ -response modulation, we disrupted it using two different complementary strategies: mutations and targeted MB-specific RNAi (simultaneously under the control of the same P[Gal4] GFPaequorin driver line: OK107). We first tested two different mutants of the rut gene encoding AC: the loss of function rut 1 (Feany, 1990) and rut 2080 , a P-element insertion (Levin et al., 1992). rut 1 showed a global decrease in activity ( Fig.  2A), despite a similar pattern of activity to control flies, with two components in the CCB and one in the ML. The first component had a mean value of ϳ880 ph/s, while the second had a mean of ϳ600 ph/s. These values correspond to ϳ40% of the average control response intensity. The response amplitude in the ML was about half (51%) as strong as that of control flies. The total duration was slightly diminished in the CCB (59% of the control response), but not significantly affected in the ML (Fig.  2LϪM). The TP was the most affected parameter, with decreases of 32% in the CCB and 36% in the ML, compared to control rut 1 flies (Fig. 2L, M). The second mutant rut 2080 (Fig. 2B) had a 61% decreased response amplitude in the CCB compared to the control, while the response amplitude in the ML was not significantly reduced.
The TP was significantly reduced only in the CCB (Fig.  2L,M). Finally, response duration was not significantly reduced in rut 2080 (Fig. 2N,O).
The cAMP pathway is a ubiquitous signaling pathway involved in several other critical processes, such as apo-ptosis (Zhang et al., 2006) and cellular fate (Bilodeau et al., 2000). Hence, in order to overcome unspecific effects of the pathway, potentially induced by expression outside the desired structure, we disrupted it locally only in the MBs using targeted RNAi, under the control of the Gal4/  Because of the recording angle, the response in the calyx, unavoidably, partially overlaps with the response in the cell bodies. H, Accumulated (10 s) bioluminescence image of the nicotine-evoked response corresponding to each ROI, separately. Because it is not possible to perfectly separate the response from the two ROIs, we use a single ROI comprised of both of them: the CCB complex (ROI circled in blue). For the medial lobes (red circle in C), again here, since we privileged the overall view of the MBs, this approach did not permit us to separate the response of the various sublobes, such as ␤, ␤', or ␥.   UAS system (Brand and Perrimon, 1993). In addition, since each RNAi could have higher or lower efficiency, we used two independent RNAi constructs. In general, the two RNAis gave similar results (Fig. 2C,D). In CCB, they reduced the nicotine-evoked response by ϳ73% compared to controls, without disturbing either the kinetic continued (n ϭ 7), rut-RNAi(1) (n ϭ 8), and rut-RNAi(2) (n ϭ 6). E؊H, Bioluminescent Ca 2ϩ -activity evoked by nicotine with upregulated cAMP production in UAS-rut (n ϭ 8), UAS-G ␣s ‫ء‬ (n ϭ 7), flies incubated 10 min with forskolin (13 M; n ϭ 6), and flies incubated 10 min with 8Br-cAMP ( Values are mean Ϯ SEM. Statistics: A؊D, One-way ANOVA was followed by a planned comparison of the predicted means to compare the levels of the selected effect using the Benjamini-Hochberg's test with rank transformation: ‫ء‬p Ͻ 0.05; ‫‪p‬ءء‬ Ͻ 0.01; ‫‪p‬ءءء‬ Ͻ 0.001 (for complete statistics, see Tables 1 and 2). properties or the total response duration. Surprisingly, in ML, the first RNAi (RNAi-1) did not change the Ca 2ϩresponse, whereas with the second RNAi (RNAi-2), the response was reduced to ϳ44% compared to the control. Altogether, these data show that a defect in rut-AC activity, resulting in decreased cAMP, leads to an overall Ca 2ϩresponse decrease in the MBs, without modifying its general kinetic proprieties.

Increasing cAMP increases the Ca 2؉ -response
Next, we looked at the inverse effect: an increase of cAMP production. First, we overexpressed the cAMP-producing enzyme rutabaga using the Gal4-UAS system (Brand and Perrimon, 1993). UAS-rut has been commonly used in order to rescue rut mutations (Zars et al., 2000). With this approach, the flies show a significant increase of 152% of the CCB response both in amplitude and the TP, while the other parameters remained unmodified (Fig. 2E,LϪO). We then targeted UAS-G ␣s ‫ء‬ expression to the MBs. G␣ s ‫ء‬ is a mutated, constitutively active form of G␣ s protein, which results in rut upregulation. G␣ s ‫ء‬ has been successfully used to disrupt olfactory L&M in MBs (Connolly et al., 1996). Flies expressing the UAS-G␣ s ‫ء‬ showed a significant increase in the Ca 2ϩ -response (Fig. 2F). The first component in the CCB and the response in the ML showed an increase of 162% and 156%, respectively, compared to control amplitude. The second component was less affected in the CCB, showing ϳ125% increase. Both TP and duration were significantly increased in the CCB and ML (Fig. 2LϪO). AC stimulation in flies expressing UAS-rut and UAS-G␣ s ‫ء‬ is constitutive (chronic) and independent of any physiological regulation. To assess the effect of acute AC stimulation, we pharmacologically induced cAMP production using forskolin, an AC stimulator (de Souza et al., 1983). Forskolin (13 M) dissolved in ethanol was applied 10 min prior nicotine application. The results obtained by application of the vehicle alone (ethanol) on the control lines present the same characteristics than the normal CS trans-heterozygotes flies and showed no spontaneous activity (Fig. 2J,LϪO). Among the eight flies recorded under these conditions, two directly responded to forskolin application. These responses were synchronous in all MB parts, while the CCB response was made up of a single component (data not shown). The flies responding directly to forskolin, prior to nicotine application, were not taken into account for subsequent quantification. The remaining flies that were not responsive to forskolin application had double the amplitude and TP compared to controls, both in CCB and ML (Fig. 2G), following nicotine application after 10 min. However, although the response duration increased in the CCB (120%) was not significant, it was significantly increased in the ML (236%) (Fig. 2N,O). In addition, we stimulated cAMP effectors with the membrane-permeable PDE-resistant cAMP analog 8Br-cAMP ( Fig. 2H) (Delgado et al., 1991). Flies were incubated in 200 M 8Br-cAMP for 10 min prior to nicotine application. The results resemble those observed with forskolin, but the response kinetics was different. The peak was delayed in both the CCB and the ML, suggesting that response kinetics might be particularly sensitive to cAMP degradation.
Previous experiments resulted in chronic or acute disruption of cAMP synthesis regulation. Conversely, in order to increase the quantity of cAMP without impairing rut AC, we used a different set of strategies to decrease its deg-  (2) radation through PDE activity. dnc was the first L&M mutant described (Byers et al., 1981). The dnc gene encodes the only known PDE catalyzing the degradation of cAMP into 5'AMP in the MBs. In order to impair dnc in the MBs, we used three different strategies. First, we used the historical hypomorphic mutation dnc 1 . Surprisingly, in this mutant, we observed a significant decrease of amplitude and TP (in CCB: amplitude 35%, TP 42%; while in ML: 39% amplitude, TP 45%, compared to control levels) (Fig.  3A). The kinetics of the response was similar to controls. Similar to rut, dnc is involved in many cellular mechanisms in the adult as well as throughout development (Balling et al., 1987). Next, we looked at the local effect of PDE disruption using RNAi against dnc, targeted to the MBs. In these MB-dnc-deficient flies, we observed increased Ca 2ϩ -response in all parts of the MBs (Fig. 3B), which is in agreement with the previous experiments of cAMP upregulation (G␣ s ‫,ء‬ forskolin). Importantly, in the CCB, the amplitude of the two components was significantly increased by 244%, while the amplitude increase was even greater in the ML (300% increase). TP increased by 240% in the CCB and 282% in the ML (Fig. 3JϪM). Duration was slightly increased (129%) in the CCB, but more affected in the ML (196% increased).
Since the above RNAi results contradicted those obtained from dnc 1 , we used two additional independent approaches. First, we undertook a second RNAi (RNAi-2) experiment using a different construct (from NIG Japan). This experiment resulted in a very similar Ca 2ϩ -response pattern (Fig. 3C) to that observed using the first RNAi (Fig.  3B) construct. Both components of the CCB were increased by 172%, while the ML increased by 182%. TP was increased by 214% in the CCB and 192% in the ML (Fig. 3JϪM). The modification of the duration was strikingly similar to the effect observed using the first RNAi. Secondly, we used a pharmacological agent, IBMX (Beavo et al., 1970;Gervasi et al., 2010), to block PDE activity. Flies were preincubated with IBMX (200 M) for 10 min prior to nicotine application. Similar to forskolin, we observed spontaneous activity in four of 10 flies. These responses had slight variability in amplitude and TP, but were similar in kinetics (Fig. 3D, Movie 1). In the four flies showing IBMX-induced activity, Ca 2ϩ -response pattern was rather different from the nicotine-evoked response profile. It consisted of one fast-synchronous component in the ML and CCB without clear propagation or rebound. The six remaining flies that were nonresponsive to IBMX showed a significantly increased nicotine response (Fig. 3E). This increase was again very comparable to the effect of the RNAi. In the CCB and the ML, the amplitude was increased by 248%. TP increased by 240% in the CCB and 275% in the ML (Fig. 3JϪM). Durations were not significantly affected in the CCB and the ML. Except for dnc 1 , all results show a direct link between Ca 2ϩ -response modulation and cAMP levels for both acute and chronic upregulation. Although this increase appears to be a general phenotype, careful observation shows that acute and chronic modification of cAMP levels leads to slightly different effects. Indeed, acute activity, using pharmacological agents (IBMX, fors-kolin), did not increase response duration in the CCB, whereas chronic activity, using genetic approaches such as G ␣s ‫ء‬ and dnc-RNAi, increased this duration. These results suggest the existence of different molecular mechanisms in Ca 2ϩ modulation in acute versus long-term cAMP increase. In order to further investigate the effect of a dnc disturbance, we overexpressed the gene using the UAS-dnc construct (cDNA), which is commonly used to rescue dnc mutations (Cheung et al., 1999). In these flies, specifically in the MBs, we observed a radically reduced response (Fig. 3F) (even lower than observed with rut mutation) with a TP, which reduces to 15% of the control response in the CCB and 16% in the ML (Fig. 3JϪM). However, this result seems to be in agreement with the results obtained with the rut, but in this case (UAS-dnc), the duration was also severely reduced. This result is similar to the two independent RNAis and the pharmacology, while completely opposite to what was observed in dnc 1 mutation.
Finally, we generated two different rut and dnc doubledeficient lines by combining the mutation rut 2080 with dnc-RNAi(1), and rut-RNAi(1) with dnc-RNAi(2) (Fig.  3G,H). In both lines, TP and the duration in the CCB were restored (Fig. 3J,K), while the amplitude was partly restored (75% and 62%, respectively, compared to control). The kinetics of the responses remained disrupted (Fig.  3G,H). The response in the ML was delayed in both cases, but was differently disrupted in the two double mutant lines. In rut 2080 ;dnc-RNAi, response amplitude was restored to control levels, and in dnc-RNAi;rut-RNAi, the amplitude decreased by 56%, while TP reduced by 55%, compared to control. Although their kinetics are quite complex, these results suggest that another AC could potentially play a role in Ca 2ϩ -response modulation, while the Ca 2ϩ -sensitive rut AC is required to obtain a normal response pattern.
Calmodulin affects the Ca 2؉ -response in the CCB, but not in the lobes cAMP is at the center of the theoretical model of the coincidence detector. This model places MBs at the intersection of two kinds of stimuli: the unconditioned stimuli (negative or positive reinforcement) and the conditioned stimulus (generally the odors) (Heisenberg, 2003;. Theoretically, this integration relies on the dual regulation of rut AC by Ca 2ϩ -calmodulin (CaM) (supposedly in association with the nAchR odor response) present in the CCB, and the G-protein (G s ) coupled to a metabotropic receptor (associated with neuromodulators released by reward or nociceptive pathways) present in the MB lobes (Heisenberg, 2003;McGuire et al., 2005;Keene and Waddell, 2007;Waddell, 2010). Despite a number of studies supporting this model, the importance of CaM on Ca 2ϩ -response modulation has not been directly demonstrated in vivo so far, although a rut CaMindependent deficiency is associated with defective memory (Livingstone et al., 1984). Hence, to explore the role of CaM, we silenced its expression using two independent RNAi constructs. The two RNAi constructs have a rather similar effect on the Ca 2ϩ -response (Fig. 4A,B).

UAS-dnc
Duration ( The response in CCB was delayed and lowered to 75% of control by both RNAi constructs. The response in the ML seemed to be slightly increased with the first RNAi construct, but remained unaffected with the second. TP was decreased (by 50%) in the CCB, but unchanged in the ML (Fig. 4G,H). The duration did not vary significantly in any part of the MBs with either construct (Fig. 4I,J). These results indicate that CaM has an effect in the CCB, but not in the MB lobes, and thus suggests a regionalized effect of CaM on the Ca 2ϩ -response.
Beyond its effect on rut AC, CaM is known to interact with several Ca 2ϩ -regulated proteins through notably the Ca 2ϩ /calmodulin-dependent protein kinase II (CaMKII) (Yao and Wu, 2001;Lisman et al., 2002;Trudeau and Zagotta, 2003) and caki (renamed recently CASK) (Martin and Ollo, 1996;Hodge et al., 2006;Gillespie and Hodge, 2013;Malik et al., 2013), as well as several other targets, both in Drosophila and other model organisms. In order to characterize further the putative link between CaM and cAMP, we used two different complementary approaches. First, we applied forskolin on CaMdeficient flies to stimulate directly cAMP production and circumvent the Ca 2ϩ /CaM effect on rut. In this combination, we observed a restoration of the response in the CCB both in amplitude and TP (Fig. 4C). However, the kinetics of the response remains disturbed, being characterized by a longer rising time that led to an overall longer duration in the CCB. In addition, surprisingly, in the ML we observed a striking diminution of the TP to 59% (Fig. 4H) and the amplitude to 44% of controls (Fig. 4C).
Second, we used dnc-RNAi or IBMX in a CaM-deficient context, thus combining the disruption of the degradation of cAMP with a decreased endogenous production of cAMP due to the deficit of Ca 2ϩ /CaM stimulation on rut. In both cases, these combinations failed to restore the control response level in the CCB (Fig. 4D,E). The response of ML were even weaker than previously observed with forskolin (IBMXϩCaM-RNAi ϭ 25%; dnc-RNAiϩCaM-RNAi ϭ 36%) (Fig. 4GϪJ). However, the outcome effects on the Ca 2ϩ -response of these three different experimental approaches are different: the first case (CaM-RNAi alone; Fig. 4A,B) yields a decrease in the CCB without affecting the ML; the second case (CaM-RNAi ϩ forskolin; Fig. 4C) yields a restoration of the Ca 2ϩ -response in the CCB but a striking decrease in the ML; and the third case (CaM-RNAi ϩ IBMX or dnc-RNAi; Fig. 4D,E) yields no restoration of the Ca 2ϩ -response in the CCB, but rather a striking decrease in the ML. In summary, altogether these three approaches share a differential effect between the CCB and the lobes, confirming that CaM effects are dissociated between these two compartments and therefore might be regionalized in the MB.

PKA is a major modulator of the Ca 2؉ -response
The cAMP-dependent PKA plays several roles in many species, particularly neural plasticity in mammals (Nguyen and Kandel, 1996;Kandel, 2012). The impairment of PKA activity in Drosophila has been related with strong L&M phenotypes (Skoulakis et al., 1993;Yamazaki et al., 2010). cAMP was also shown to locally regulate PKA in MBs (Gervasi et al., 2010). PKA is a multimeric holoenzyme composed of two regulatory and two catalytic subunits. Following activation by cAMP, PKA plays various roles such as K ϩ -channel phosphorylation (Drain et al., 1994;Esguerra et al., 1994) and transcriptional regulation through cAMP response element-binding protein (CREB) (Yin et al., 1995b). In order to assess PKA's role in Ca 2ϩresponse regulation, we first used UAS-mC‫,ء‬ a constitutively active catalytic subunit of murine PKA (Li et al., 1995). This constitutively active subunit was previously shown to impair sleep when specifically expressed in the continued by nicotine application (n ϭ 6, D) and IBMX spontaneous Ca 2ϩ -response (n ϭ 4, E). F, Bioluminescent Ca 2ϩ -activity with upregulated degradation of cAMP through an overexpression of dnc-PDE using UAS-dnc transgenic construct (n ϭ 8). G, H, Bioluminescent nicotine-evoked Ca 2ϩ -activity with downregulated cAMP production, combined with downregulated cAMP degradation in rut 2080 ; dnc-RNAi(1) (n ϭ 7, G), and in rut-RNAi(2);dnc-RNAi(2) (n ϭ 7, H). I, Bioluminescent image (accumulation time: 120 s) of the nicotinic Ca 2ϩ -response of a typical fly for each genotype. J, K, Total number of photons during the nicotine response in the CCB (J) and in the ML (K). L, M, Total duration of the response in the CCB (L) and in the medial lobe (M). Values are mean Ϯ SEM. Statistics: same as for Figure 2.
Movie 1 In vivo bioluminescence imaging of Ca 2ϩ -responses in the KCs of the MBs induced by nicotine application. On the left, we observe a wild-type control-Canton-S fly; the right corresponds to the so-called spontaneous activity induced by the IBMX application. Each frame represents 1 s of light accumulation and is shifted by 250 ms, seen at 25 frames/s. The light emission is coded in pseudocolors (2-6 photons/pixel) (MP4 ϭ 2.89 Mb).
MBs (Joiner et al., 2006;Pitman et al., 2006). Flies expressing the mC‫ء‬ in the MBs showed a significant increase in the Ca 2ϩ -response (Fig. 5A, Movie 2). The first exponential phase in the CCB culminated with an intensity of 287% compared to controls. Consequently, the second component was almost invisible, since it merged with the decreasing phase of the first exponential. The response in ML was increased by 351%. The responses were also significantly prolonged in the calyx, albeit to a weaker level (100 ph/s), with an average duration of 444 s in the CCB (450% increase) and 193 s in the ML (350% increase), following response initiation (Fig. 5F). Conversely, TP had a significantly greater increase (532%) in the ML compared to the CCB (381% increase) (Fig. 5E). Next, to assess if blocking PKA yields an opposite effect on Ca 2ϩresponse, we expressed a mutated regulatory subunit of PKA, UAS-R‫,ء‬ which constitutively blocks the catalytic subunit by competing with the endogenous regulatory subunit (Li et al., 1995). These flies not only had reduced Ca 2ϩ -responses (Fig. 5B), but these responses were delayed and their intensity halved for both components in the CCB. In contrast, the ML responses did not appear to be delayed, although their intensity decreased. TP also halved in CCB and ML, but its duration was diminished only in the CCB (Fig. 5E,F).
Given similar results for AC and PKA regulation, we next studied the effect of blocking PKA under high cAMP levels, the rationale being that this would potentially reveal a direct putative role of cAMP on Ca 2ϩ -response, independent of its effect via PKA. In order to do this, we coexpressed G ␣s ‫ء‬ and R‫ء‬ in KCs. Interestingly, the double-transgenic flies displayed an intermediate pheno- type (Fig. 5C). The response was delayed in the CCB, similar to the R‫ء‬ flies, but its intensity reached a level roughly similar to controls in both CCB and ML. The results obtained with PKA impairment reveal a strong positive influence of this effector on Ca 2ϩ -response globally. However, the results with the double-transgenic flies led us to question whether an additional cAMP-regulated PKA-independent mechanism may regulate the Ca 2ϩresponse.

The Ca 2؉ -response is modulated through CNGs
In order to explore the cAMP-dependent, PKAindependent effect in MB response modulation, we looked for direct potential targets of cAMP. Amongst them, the CNGs are a class of channels that can be opened by cAMP or cGMP (Baumann et al., 1994). These channels are mainly permeable to Ca 2ϩ , although they are also permeable to most other cations. Four genes in Drosophila are predicted to encode CNGs: cngc, cng-b, cngl, and cg42260 (Kaupp and Seifert, 2002). For our experiments, we used RNAi against two different CNGs expressed in the adult brain: cngc and cngl (Baumann et al., 1994;Miyazu et al., 2000). CNGC has mostly been studied for its role in the response to hypoxia (Vermehren-Schmaedick et al., 2010). It is expressed in the adult brain, and was shown to be very responsive to cGMP, and to a lesser extent to cAMP (tenfold less). Calcium flux, through this channel, is also blocked in a voltage-dependent manner (Baumann et al., 1994). We used two different RNAi constructs against cngc, which led to concordant results for TP and duration, but with different kinetic phenotypes (Fig. 6A,C). Both RNAi showed a decreased TP in the CCB (38% and 47%) and the ML (41% and 45%), but duration was only affected in the ML (45%) of cngc-RNAi(2). In cngc-RNAi(1), the TP was decreased, but the overall response followed similar kinetics as controls (two components in the CCB and a delayed response in the ML) (Movie 3). cngc-RNAi(2) had a disrupted waveform: the ML response peaked before the CCB and was followed by a low-activity tail. In addition, it was narrower than the control response. The CCB response in the second RNAi was only made of one component, which had a longer rising time. These different effects between the two RNAi constructs may potentially be due to different expression levels.
In order to directly determine the role of cAMP on CNGC, we increased cAMP in a cngc knocked-down background using three independent methods. First, we coexpressed G␣s‫ء‬ with RNAi(1) (Fig. 6E), which led to increased amplitude, similar to control levels, but not to G␣s‫ء‬ levels (compared to Fig. 2F). Furthermore, values for TP and duration were in between control flies and those that had cngc-RNAi alone. We then applied the pharmacological agents forskolin and 8Br-cAMP to the cngc-RNAi flies. Forskolin (Fig. 6G) led to a very similar phenotype as cngc-RNAi(1);G␣s‫ء‬ flies, except for the response in the ML, which was decreased compared to controls, but similar to the cngc-RNAi(1) response. Next, we applied 8Br-cAMP to cngc-RNAi(2) (Fig. 6I) flies, which resulted in a response that shared all the characteristics of the second RNAi. In summary, the combination of increased cAMP and cngc-RNAi leads to an effect that is in between controls and cngc-RNAi on its own. In other words, it never restores the enhanced cAMP phenotype to the CCB or ML, suggesting that the CNGC channel plays an early and crucial role in the Ca 2ϩ -response modulated by cAMP.
Finally, we investigated the physiological effects of the combination of the depletion of CNGC and the PKA transgenic construct: UAS-mC‫ء‬ and UAS-R‫.ء‬ With the coexpression of UAS-mC‫ء‬ (the constitutively activated PKA subunit) and cngc-RNAi(1), the nicotinic responses are very close to those of the mC‫ء‬ flies, both in duration and TP. However, the double transgenes (cngc-RNAi(1);UAS-mC‫)ء‬ show some slight modification of their kinetics (Fig.  6K), which are globally slower, and also exhibit a significant diminution of the TP in the medial lobe (Fig. 7B). We then suppressed the PKA activity using the combination of UAS-R‫ء‬ and cngc-RNAi. While double-transgenic flies show a kinetic phenotype very close to what was observed with cngc-RNAi(1) alone (Fig. 6, A vs M), their TP and response duration remained at the same value as observed in both single-transgenic flies (which were themselves very similar to each other) (Fig. 7). The results obtained with the manipulation of PKA suggest that cngc is not necessary for the amplification of the response observed in UAS-mC‫ء‬ flies while it remains critical for the kinetic properties. UAS-R‫ء‬ experiments seem to indicate that we already obtained the maximal effect (the minimal Ca 2ϩ -response) using cngc knockdown.

Movie 2
In vivo bioluminescence imaging of Ca 2ϩ -responses in the KCs induced by nicotine application on a fly expressing the UAS-mC‫.ء‬ Left, Control-CS (same fly as in Movie 1). Right, A fly expressing the UAS-mC‫,ء‬ a constitutively activated PKA. We remark that the level and the duration of activity is importantly increased in mC‫ء‬ fly. However, for the sake of the visualization of the movie, notice that the duration of this movie does not exactly correspond to the calculated duration reported in Figure 5F, because the accumulation time of the signal is settled and displayed differently. Each frame represents 1 s of light accumulation and is shifted by 250 ms, seen at 25 frames/s. The light emission is coded in pseudocolors (2-6 photons/pixel) (MP4 ϭ 2.89 Mb). CNG-like (CNGL), encoded by cngl, is another cAMP/ cGMP-sensitive channel that is expressed at high levels in the central brain of Drosophila (Miyazu et al., 2000). It shares most structural characteristics of CNGC, except for the aspartic acid residue found in other CNG channels that is necessary for cGMP selectivity (Varnum et al., 1995). This residue is substituted by a valine in CNGL channels (Miyazu et al., 2000). No data on CNGL physiology or ion selectivity are available yet, but the structural features suggest that, similar to CNGC, CNGL contains a voltage-sensitive region that is permeable to most cations, and probably mainly to Ca 2ϩ (Miyazu et al., 2000). We used two independent RNAi constructs directed against cngl that gave roughly similar results (Fig. 6B,D). The amplitude and TP in both RNAi were dramatically reduced in the CCB (26% and 31%, respectively) and the ML (35% and 33%, respectively). The kinetic properties were slightly different between the two RNAi lines: they both followed a two-component response, but the rising phase and overall duration were longer (a duration comparable to control) in RNAi(2). Next, we combined CNGL impairment with cAMP stimulation using the same tools as was previously done with CNGC (G␣s‫;ء‬cngl-RNAi(2) (Fig. 6F). Similarly, treating cngl-RNAi(2) flies with either forskolin or 8Br-cAMP had a strikingly similar effect on the responses (Fig. 6, H and J vs F). The CCB amplitude was restored, but the rising phase was longer than controls, increasing the duration of the response. The response in the ML was significantly decreased and delayed compared to controls. TP in the CCB was between 398% and 509% higher than in flies containing cngl-RNAi alone (Fig.  7), and in between TP values of the control and cAMPenhanced flies. In contrast, the response in the ML ranged from 24% to 38% of that of the cAMP-enhanced flies and was between cngl-RNAi and control levels. The results with CNGL knocked-down suggest that this channel has a major role in response generation (result with RNAi alone), but can be supplemented in the CCB by enhancement of another class of channels sensitive to cAMP. However, CNGL seems to be essential for propagating the response to the ML. A noticeable feature of both CNGs is that their phenotype is more severe than that observed in the rut-deficient files, suggesting that either other ACs are involved in CNG stimulation or they play a similar role as the voltage-gated cation channels through their voltage-sensitive characteristics.
Finally, as previously performed with cngc, we combined the cngl-RNAi with UAS-mC‫ء‬ and UAS-R‫.ء‬ As observed with cngc-RNAi(1);UAS-mC‫ء‬ flies, the response of the cngl-RNAi(2);UAS-mC‫ء‬ flies were very similar to UAS-mC‫ء‬ flies, with few modifications of the kinetic property of the response; the response amplitude seemed to be lower than in mC‫ء‬ (Fig. 6L) flies, but this diminution is compensated by an increased duration of response tail (Fig. 7C,D), giving a final TP at the same level as observed in mC‫ء‬ transgenic flies (Fig. 7A,B). The effect of the combination of UAS-R‫ء‬ and cngl-RNAi was also very similar to the phenotype observed in cngl knockdown flies both in the CCB and ML (Figs. 6B,N, 7). These results altogether suggest that the overactivation of the PKA mediate the amplification of the response through a partner, which might play an early role in the response and seems to be independent from cngl, which only disturbs the kinetic of the response. As previously seen with cngc, the UAS-R‫ء‬ experiments seem to indicate that we already obtained the maximal effect (the minimal Ca 2ϩ -response) using cngl knockdown.

Movie 3
In vivo bioluminescence imaging of Ca 2ϩ -responses in the KCs induced by nicotine application on a fly expressing the cngc-RNAi(1). Left, Control-CS fly. Right, A fly expressing the cngc-RNAi(1). We remark that the level and the duration of activity are importantly reduced in this fly. Each frame represents 1 s of light accumulation and is shifted by 250 ms, seen at 25 frames/s. The light emission is coded in pseudocolors (2-6 photons/pixel) (MP4 ϭ 2.89 Mb). measure functional Ca 2ϩ -response continuously, over long time periods, with high sensitivity. This probe was previously used to study odor-induced Ca 2ϩ -response in ORNs (Murmu et al., 2010; and the MBs, and to detect spontaneous Ca 2ϩ -activity in neurons and glial cells (Minocci et al., 2013). In this study, we used this approach to simultaneously record and temporally correlate the responses of the CCB and lobes, and vizualized Ca 2ϩ -activity propagation in the axonal projections within the MB lobes.
responses due to activation of afferent circuits. Nonetheless, the RNAi knockdown approach that targets only the MB neurons supports our conclusion that the physiological Ca 2ϩ -effects described here are indeed due to the disturbance of the given targeted pathway or channels within the MBs.
The level of the Ca 2؉ -response is proportional to the level of cAMP As mentioned before, although the dnc and rut mutants have been identified for more than 30 years, their in vivo physiological effects on the Ca 2ϩ -response in the KCs of the adult fly still remain largely unknown. We used several independent strategies (genetic and pharmacology) to demonstrate that the nicotine-induced Ca 2ϩ -response is proportional to the level (higher or lower) of cAMP. The downregulation through the two alleles of rut, or via two different rut-RNAis, leads to an ϳ50% Ca 2ϩ -response decrease in both the CCB and lobes. Similarly, the overexpression of dnc leads to even a stronger phenotype, suggesting the PDE activity in the response modulation is of critical importance. Conversely, the upregulation using two independent dnc-RNAis or by targeted overexpression of rut and a constitutively active G-protein subunit (G ␣s ‫)ء‬ significantly increases the Ca 2ϩ -response. Unexpectedly, however, the response is decreased with dnc 1 , a contrasting result compared to the other strategies used to increase cAMP. These contradictory results are likely due to defects accumulated during fly development. Indeed, it was reported that dnc, which is expressed at various developmental stages, has diverse roles in cells and notably affects the survival of KCs throughout development, leading to smaller CCB in adult flies (Balling et al., 1987). We also use a pharmacological strategy to upregulate the level of cAMP. Again, either increasing its synthesis by directly stimulating adenylyl cyclase using forskolin or diminishing its degradation using IBMX yields similar results: a huge increase of the Ca 2ϩ -response. Thus, the similar results obtained by these independent approaches demonstrate that cAMP levels determine the level of the Ca 2ϩ -response. Based on literature describing L&M defects caused by disruption of the cAMP pathway, it seems that fine regulation of the level of cAMP is the crucial parameter. This is because both its decrease or its increase disrupts L&M, as well as other MB functions such as centrophobism (Besson and Martin, 2005;Lebreton andMartin, 2009) or sleep (Joiner et al., 2006). Research to date has only investigated the L&M function of rut; however, as already discussed (Tomchik and Gervasi et al., 2010), a number of other putative ACs have been reported in the Drosophila genome (DAC39E, DAC78C, DAC76E, CG32158, CG32301, CG32305). Thus, it is possible that these additional noncharacterized ACs could be active in the MB lobes or the CCB.

Acute increase of cAMP triggers a transient Ca 2؉ -response in the KCs
In this study, we used two fundamentally different and complementary methods to disrupt cAMP pathway: on one hand, we used the genetic approach (mutants, tar-geted overexpression, and targeted RNAi), which induce chronic modifications of the pathway; and on the other hand, we used the pharmacological approach, which corresponds to an acute effect (stimulation or blockade). Interestingly, an acute increase of cAMP synthesis, or reducing its degradation by pharmacological approaches (forskolin or IBMX), induced Ca 2ϩ -activity in the MBs. In contrast, chronic dysregulation (mutant, targeted RNAi) did not induce any spontaneous Ca 2ϩ -activity. These results suggest that chronic misregulation of cAMP can be (at least partially) compensated by other mechanisms (e.g., signaling pathway partners such as PKA, or upregulation or downregulation of various channels) in neurons, while an acute modification is not compensated, and therefore is sufficient to trigger a response in the KCs. We also found that a chronic upregulation of cAMP in the calyx led to a prolonged response (dnc-RNAi, G␣ stransgene), possibly through the effect of PKA, while acute upregulation (by pharmacology: IBMX and forskolin) does not affect the duration. This suggests that the effect might be, at least partially, through the CNGs. Interestingly, it was reported that KCs cultured from late stage pupae showed spontaneous Ca 2ϩ -transients in a cell autonomous fashion (Jiang et al., 2005). In addition, functionally behavioral genetic approaches demonstrated that synaptic transmission between KCs and their downstream partners is important in memory retrieval, but not necessary for memory acquisition or storage (Dubnau et al., 2001). Consequently, we hypothesize that this spontaneous triggering following an acute increase in cAMP, either pharmacologically evoked or naturally occurring, could represent a molecular and cellular mechanism for reminiscence or retrieval, since the KCs seem to be able to activate themselves spontaneously (cell autonomously) without any afferent stimuli.

PKA-dependent and -independent effect of cAMP
PKA is the best known effector of cAMP. The constitutive activation of the catalytic subunit strikingly increases the Ca 2ϩ -response both in CCB and lobes. Moreover, it significantly prolongs the response duration by up to 444 s in the CCB (more than fourfold). In contrast, blocking the regulatory subunit using a dominant negative form (UAS-R‫)ء‬ decreases the Ca 2ϩ -response. However, the latter also delays the response in the CCB. PKA likely acts by phosphorylating the K ϩ -channels (Delgado et al., 1992;Brüggemann et al., 1993;Zhou et al., 2002) and/or regulating CREB (Yin et al., 1995a;1995b). The chronic inactivation of PKA may have phosphorylated some K ϩchannels, resulting in a modified resting membrane potential. This, in turn, may have led to less excitable cells, which consequently delayed the Ca 2ϩ -response. Additionally, the residual response following PKA blockade (UAS-R‫)ء‬ revealed that cAMP can act by itself, and thus represents its PKA-independent effect. To corroborate these results, we increased the production of cAMP in PKA-blocked flies, which indeed increased the Ca 2ϩresponse (UAS-R‫;ء‬UAS-G␣ s ). Furthermore, the knockdown of the different CNGs, which decreased the Ca 2ϩresponse, suggests that these channels play a crucial role in KC responsiveness. Moreover, cAMP supplementation is never sufficient to restore the kinetic properties of the response in a CNG-deficient context. The different response patterns displayed after cAMP enhancement in a CNG knockdown context suggest that the two different CNGs are playing segregated and sequential roles in cAMP-dependent regulation of MBs responsiveness: CNGC seems to play an early role on the overall response, while CNGL seems to play its role mainly in the medial lobe (for example, response level only partly rescued in the medial lobes). The overactivation of PKA in CNGdeficient context [cng-RNAi(1) or cng-RNAi(2);UAS-mc‫]ء‬ does not restore the kinetic proprieties of the responses but completely restores the quantitative parameters (TP and duration) observed with an overactivation of the PKA alone, suggesting that CNGs are not required for PKA's modulation of the response but are critical for its kinetics. Inversely, the blockade of the PKA activity in CNGdeficient flies [cng-RNAi(1) or cng-RNAi(2);UAS-R‫]ء‬ yields to a CNG phenotype, confirming that the CNG-dependent modulation of Ca 2ϩ -response is independent of the PKA.

The differential effect of calmodulin between the CCB and the lobes
The rut-AC is a Ca 2ϩ /calmodulin-dependent enzyme (Livingstone et al., 1984;Levin et al., 1992). However, a function of CaM, per se, within the KCs has not yet been directly described. Here, we functionally demonstrated that CaM knockdown results in a segregated and regionalized effect. While it significantly decreases the Ca 2ϩresponse in the CCB, it does not affect the responses in the lobes. This effect suggests that either its modulation is necessary for Ca 2ϩ -responsiveness in the CCB, but not in the MB lobes, or a calmodulin-independent AC activation occurs in the lobes, putatively through another AC. Rut protein is present in the ␣,␤,␥ lobe branches (Han et al., 1992), while functional subdivisions in cAMP synthesis within the MBs has been reported (Gervasi et al., 2010). Since Rut-AC can be activated either by Ca 2ϩ /calmodulin or via G-protein stimulation, a first hypothesis could be that in the CCB, Rut-AC activation occurs through calmodulin, while in the lobes it could be activated through G-proteins.
However, the results obtained from the CaM knockdown (CaM-RNAi) combined to either the forskolin or the IBMX or dnc-RNAi (Fig. 4CϪE), which all yield a clear dissociated effect between the CCB and the lobes, could suggest a second alternative-for instance, the implication of intermediate partners. Indeed, the effect of the CaM knockdown in the CCB, which seems to be compensated solely by the direct stimulation of the rut-AC (forskolin), is consistent with the canonical model involving CaM directly in the rut stimulation. However, the unexpected striking decreased Ca 2ϩ -response in the ML due to the increase of the cAMP (IBMX or dnc-RNAi) combined with CaM knockdown, suggesting that another partner, hypothetically coregulated by cAMP (directly or through PKA) and/or CaM (directly or through CaMKII and/or CASK), could play a role in the ML regulation by inhibiting the Ca 2ϩ -response when cAMP is increased in In the CCB: the conditional stimulus (e.g., an olfactory stimulus) triggers nicotinic inputs, which activate the nAchR located on the KCs of the MBs, allowing the Ca 2ϩ -entry. Calcium binds to CaM and subsequently activates the production of cAMP by RUT. In agreement with the coincidence detector model, in certain conditions, e.g., when the conditional stimulus is simultaneously applied with an unconditional stimulus (e.g., a nociceptive electric shock), the dopaminergic receptors could costimulate the rut-AC to increase further the cAMP level through G ␣s . The resulting increase in cAMP stimulates the PKA as well as the CNGs (CNGC and CNGL), which both participate in amplifying the Ca 2ϩ -entry. At the same time, the PKA allows the Ca 2ϩ -entry and/or the persistence of the Ca 2ϩ -entry, likely by affecting the repolarisation of the cells, possibly through the K ϩ -channels. In parallel and simultaneously, the Ca 2ϩ -entry modifies the voltage of the cells that allows the voltage-gated calcium channels (VGCC) to also participate to the Ca 2ϩ -entry. Altogether, these activities trigger actions potentials (APs) that propagate to the lobes. In the medial lobes: at the axon terminals, the APs open the VGCC, allowing Ca 2ϩentry. This Ca 2ϩ stimulates the CaM. In parallel or simultaneously (as for instance in certain environmental conditions), a neuromodulator (e.g., dopamine, octopamine) activates a metabotropic receptor, which stimulates a G-protein, and then stimulates the RUT to increase the cAMP. Then, the cAMP stimulates the CNGs (and more likely the CNGL; as suggested by our results in Fig. 6F,H,J). Moreover, according to our results (knocking-down the CaM and handling the cAMP level; Fig. 4), we hypothesize that the CaM might act as an inhibitor of the cAMP-stimulation of the CNGL (red line). This could be achieved absence of the CaM. Interestingly, the Ca 2ϩ -response of the forskolin in the CaM-RNAi (Fig. 4C) resembles the effect of the forskolin, G ␣s , and 8-Br-cAMP in the cngl-RNAi context (Fig. 6F,H,J), suggesting that CNGL could be a putative target. Ca 2ϩ /calmodulin modulation of different CNGs has also been already reported in other systems as olfactory and visual systems (Trudeau and Zagotta, 2003). Therefore, CASK and/or CaMKII are good candidates for these putative intermediate partners (see Fig. 8 for a schematic model) since both of them have been reported to be involved in learning and memory formation (Malik et al., 2013), as well as in calcium signaling in Drosophila larvae . The differential Ca 2ϩ -response between the CCB and the lobes makes calmodulin a new and interesting candidate for alternative AC activation and so potentially in relation to the coincidence detector hypothesis. However, the L&M effect of the knockdown of calmodulin, specifically within the MBs, would first have to be determined. T1 Crossref missing T2 Crossref missing continued either through the CaMKII or CASK, since both of them have been implicated in learning and memory, or directly by the CaM on the CNGL (since in some organisms, certain CNGs have been reported to be sensitive to CaM; Kaupp and Seifert, 2002). These successive events lead to the fine tuning of the Ca 2ϩ -level that mobilize the synaptic vesicles and the output. This hypothetical concomitant inhibition by the CaM and the cAMP on the CNGL could represent a coincidence detector.