PKCδ is required in MB neurons for LTM formation.

(A) Schematic representation of Drosophila MB. The MB includes ∼2,000 intrinsic neurons per brain hemisphere. Their cell bodies are located in the dorsal posterior part of the brain. MB neurons each send a single neurite into the neuropil, which first traverses the calyx, a dendritic region where MB neurons receive olfactory input from projections neurons, and then extends into a long axonal branch. The bundled axons of MB neurons form a fascicle called the peduncle, which traverses the brain to the anterior part, where axons branch to form horizontal and vertical lobes according to three major branching patterns – α/β, α’/β’ and γ (Aso et al., 2014) – that define as many neuronal categories. The MB lobes receive input from dopaminergic neurons (the DANs), which signal stimuli of positive and negative valences in a region-specific manner (Aso et al., 2014). During associative learning, dopamine release on coincidentally odorant-activated MB output synapses modulates the synaptic drive to the network of MB output neurons, which bias subsequent odor-driven behavior (Heisenberg, 2003; Hige, 2018). Aversive LTM induced by spaced training is more specifically encoded within the α/β neurons (Pascual and Préat, 2001; Séjourné et al., 2011; Yu et al., 2006), and we previously showed that LTM retrieval involves the depression of an attraction-mediated pathway efferent from the MB vertical lobes (Aso et al., 2014; Bouzaiane et al., 2015; Dolan et al., 2018; Séjourné et al., 2011). However, according to another recent study, LTM retrieval mobilizes in parallel another MB output circuit efferent from the medial lobes (Jacob and Waddell, 2020). (B) Expression of PKCδ RNAi in adult MB neurons induced a significant reduction in the mRNA level of PKCδ measured by RT-qPCR in fly heads. Relative Quantification (RQ) was performed, indicating the foldchange of mRNA levels relative to the control genotype (n=11, t20=2,826, p=0.0104). (C) PKCδ knockdown in adult MB neurons impaired memory after 5x spaced conditioning (n=17-25, F2,58=12.59, p<0.0001). Without the induction of PKCδ RNAi expression, memory formed after 5x spaced conditioning was normal (n=15-17, F2,45=0.4100, p=0.6661). Memory formed after 5x massed training (n=13-14, F2,37=0.6511, p=0.5273) and 1x training (n=18, F2,51=0.8135, p=0.4490) was normal in flies knocked down for PKCδ in adult MB neurons. (D) The cyto-δCKAR sensor was expressed in adult MB neurons and visualized in the CFP and YFP channels. Cytosolic PKCδ activity levels are recorded within the vertical lobes of the MBs (indicated with dashed line). Scale bar=50 µm. In naïve flies, application of 250 µM of PDBu (black arrow), a pharmacological activator of PKCδ, resulted in the increase of the cyto-δCKAR response, reaching a plateau, as compared to the DMSO control (n=6, t10=5.658, p=0.0002). Quantification of the mean cyto-δCKAR response was performed 280 s after PDBu application on a time window of 560 s (black line). (E) In naïve flies, application of 250 µM of PDBu (black arrow) resulted in an increase in the cyto-δCKAR response that is abolished when PKCδ is knocked down in adult MB neurons (n=5-6, t9=4.182, p=0.0024). Quantification of the mean cyto-δCKAR response was performed 280 s after PDBu application on a time window of 560 s (black line). (F) To compare post-conditioning cytosolic PKCδ activities (between 30 min and 1 h 30 min post-conditioning, in yellow on the imaging time frame), cyto-δCKAR traces were normalized to the plateau value reached after addition of PDBu (saturation of the sensor), thus the activity level of cytosolic PKCδ is estimated as the cyto-δCKAR signal value before PDBu application. Cytosolic PKCδ activity is increased in the vertical lobes after 5x spaced associative paired conditioning as compared to a non-associative spaced conditioning (unpaired) protocol (n=9-10, t17=3.178, p=0.0055). Quantification of the mean post-training PKCδ activity was performed on a time window of 120 s before PDBu application (black line). (G) After 5x massed paired conditioning, cytosolic PKCδ activity was not changed as compared to 5x massed unpaired conditioning (n=8, t14=0.3282, p=0.7476). (H) Similarly, after 1x paired conditioning, cytosolic PKCδ activity was not changed as compared to 1x unpaired conditioning (n=8, t14=0.004121, p=0.9968). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by one-way ANOVA with post hoc testing by the Tukey pairwise comparisons test (C) or by unpaired two-sided t-test (B, D-H). Asterisks refer to the least significant P-value of post hoc comparison between the genotype of interest and the genotypic controls (C), or to the P-value of the unpaired t-test comparison (B, D-H) using the following nomenclature: *p<0.05, **p<0.01, ***p<0.001, ns: not significant, p>0.05. See also Figure 1 – supplement 1 and Supplementary Table 1.

PKCδ is required in α/β neurons for LTM formation.

(A) When PKCδ was knocked down specifically in the α/β neurons of MBs, memory after 5x spaced conditioning was impaired (n=18, F2,51=9.521, p=0.0003). Without PKCδ RNAi induction, memory after 5x spaced conditioning was normal (n=14, F2,39=0.4673, p=0.6301). Memory was also normal after 5x massed training (n=18, F2,51=1.771, p=0.1805) and 1x training (n=12, F2,33=0.5069, p=0.6070) in flies expressing PKCδ RNAi in adult α/β MB neurons. (B) Using the same approach as detailed in Figure 1F, the post-training activity of cytosolic PKCδ was measured specifically in the α/β neurons, between 30 min and 1 h 30 min post-conditioning (in yellow on the imaging time frame). After 5x spaced paired conditioning, cytosolic PKCδ activity was increased in the α lobes as compared to unpaired conditioning (n=9-11, t18=2.723, p=0.0140). (C) After 1x paired conditioning, cytosolic PKCδ activity was not changed as compared to unpaired conditioning (n=12, t22=1.159, p=0.2590). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by one-way ANOVA with post hoc testing by the Tukey pairwise comparisons test (A) or by unpaired two-sided t-test (B-C). Asterisks refer to the least significant P-value of a post hoc comparison between the genotype of interest and the genotypic controls (A) or to the P-value of the unpaired t-test comparison (B-C) using the following nomenclature: *p<0.05, **p<0.01, ns: not significant, p>0.05. See also Figure 2 – figure supplement 1 and Supplementary Table 1.

PKCδ regulates mitochondrial pyruvate metabolism upon LTM formation.

(A) The pyruvate sensor Pyronic was expressed in adult MB neurons and the pyruvate FRET signal was quantified in the vertical lobes. In naïve control flies (tub-Gal80ts; VT30559, UAS-Pyronic), PDBu pretreatment (250 µM, 3 min) elicited a faster pyruvate accumulation following sodium azide application (5 mM, black arrows) as compared to flies pretreated with the DMSO solvent alone (left panel, slope measurement n=10-11, t19=2.311, p=0.0322). This PDBu pretreatment induced increase in pyruvate accumulation is abolished when PKCδ is knocked down in MB neurons as compared to DMSO pretreatment (right panel, slope measurement n=9-14, t21=0.6734, p=0.5080). Notably, the plateaus of the pyruvate accumulation kinetics are lower when PDBu is injected onto the brains as compared to the DMSO controls. Indeed, PDBu injection is followed by a slight increase in the measured Pyronic ratio that is not sensitive to PKCδ knockdown, indicating a PDBu indirect effect independent of PKCδ on pyruvate neuronal levels (Figure 3 – figure supplement 1B). (B) In control flies (tub-Gal80ts; VT30559, UAS-Pyronic), spaced training elicited a faster pyruvate accumulation in axons of MB neurons after sodium azide application (5 mM; black arrow) as compared to non-associative unpaired training (left panel, slope measurement n=10, t18=6.751, p<0.0001). PKCδ knockdown in adult MB neurons (tub-Gal80ts;VT30559, UAS-Pyronic, UAS-PKCδ RNAi) impaired the spaced training induced increase in pyruvate accumulation in axons of MB neurons following sodium azide application (right panel, slope measurement n=10, t18=0.3794, p=0.7089). As for Figures 1 and 2, imaging was performed between 30 min and 1 h 30 min post-conditioning, represented in yellow on the imaging time frame. (C) Left panel: schema of the regulation of the pyruvate dehydrogenase (PDH) complex. In mitochondria, PDHC catalyzes the conversion of pyruvate to acetyl-CoA, which enters the tricarboxylic acid cycle (TCA). PDH can be inactivated through phosphorylation by PDK. In contrast, PDP can activate a phosphorylated PDH. In our model, PKCδ regulates PDH activity via PDK inhibition. Right panel: after 5x spaced conditioning, flies coexpressing the PDK and PKCδ RNAi in the α/β MB neurons at the adult stage (tub-Gal80ts;c739>UAS-PKCδ RNAi, UAS-PDK RNAi) exhibited normal memory formation as compared to genotypic controls (tub-Gal80ts;c739/+ and +/UAS-PKCδ RNAi, UAS-PDK RNAi). Flies solely expressing the PKCδ RNAi in α/β neurons at the adult stage (tub-Gal80ts; c739>UAS-PKCδ RNAi) exhibited the reported LTM defect, whereas flies that were only knocked down for PDK in adult α/β neurons (tub-Gal80ts;c739>UAS-PDK RNAi) formed normal memory (n=12, F4,55=4.746, p=0.0023). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by unpaired two-sided t-test (A-B) or by one-way ANOVA with post hoc testing by the Tukey pairwise comparisons test (C). Asterisks refer to the P-value of the unpaired t-test comparison or to the least significant P-value of post hoc comparison between the genotype of interest and the genotypic controls using the following nomenclature: *p<0.05, ****p<0.0001, ns: not significant, p>0.05. See also Figure 3 – figure supplement 1.

MP1 neurons control PKCδ activity in the MB neurons via DAMB signaling.

(A) Top panel: Schema of the MB structure and its afferent MP1 neuron. The dopaminergic neuron MP1 (depicted in green) in the PPL1 cluster projects onto the peduncle area of the MBs. Lower panel: The mito-δCKAR sensor was expressed in adult MB neurons and visualized in the CFP and YFP channels. The region of recording for panel B is at the level of the MB peduncles (dashed line). (B) Naïve flies expressing the mito-δCKAR sensor in MB neurons together with the dTrpA1 (a heat-sensitive cation channel) in MP1 neurons (30E11-LexA driver) were subjected to a thermal treatment consisting of three 2-min periods at 30°C separated by 2 min (red vertical lines); control flies expressed the mito-δCKAR in MB neurons but not the dTrpA1 channel (no 30E11-LexA driver). Mitochondrial PKCδ activity was recorded before (baseline) and immediately after the activation periods. Quantification of the mean mito-δCKAR response was performed 120 s after the last cycle of thermal activation on a time window of 480 s (black line). In naïve flies expressing the dTrpA1 channel in MP1 neurons (30E11-LexA;VT30559 > UAS-mito-δCKAR, LexAop-dTrpA1), activation of MP1 increased mitochondrial PKCδ activity as compared to control flies (VT30559 > UAS-mito-δCKAR, LexAop-dTrpA1) (n=9-10, t17=3.830, p=0.0013). When DAMB was knocked down in MB neurons (30E11-LexA;VT30559>UAS-mito-δCKAR, UAS-DAMB RNAi, LexAop-dTrpA1), MP1 activation failed to increase mitochondrial PKCδ activity as compared to control flies (VT30559>UAS-mito-δCKAR, UAS-DAMB RNAi, LexAop-dTrpA1) (n=8-9, t15=0.3148, p=0.7572). (C) The mito-δCKAR sensor was expressed in adult MB neurons and mitochondrial PKCδ activity was recorded in the vertical lobes (dashed line). As in the peduncle region, in naïve flies expressing the dTrpA1 channel in MP1 neurons (30E11-LexA;VT30559>UAS-mito-δCKAR, LexAop-dTrpA1), activation of MP1 increased mitochondrial PKCδ activity in the vertical lobes as compared to control flies (VT30559 > UAS-mito-δCKAR, LexAop-dTrpA1) (n=7, t12=2.622, p=0.0223). Data are expressed as mean ± SEM with dots as individual values and were analyzed by unpaired two-sided t-test. Asterisks refer to the P-value of the unpaired t-test comparison using the following nomenclature: *p<0.05, **p<0.01, ns: not significant, p>0.05. Scale bar=30 µm. See also Figure 4 – figure supplement 1.

PKCδ translocates to the mitochondria of α/β neurons upon LTM formation.

The mito-δCKAR sensor was expressed in adult α/β neurons and the post-training activity of mito-PKCδ was measured in the peduncle (A) or vertical lobes (B), between 30 min and 1 h 30 min post-conditioning (in yellow on the imaging time frame). The same approach was used as in Figure 1F-H with PDBu application to reach saturation level of the sensor and quantification of the mean post-training PKCδ activity performed on a time-window of 120 s before PDBu application (black line). (A) After 5x spaced paired conditioning, mitochondrial PKCδ activity in the peduncle was increased as compared to unpaired conditioning (left panel, n=7-9, t14=3.555, p=0.0032). When DAMB was knocked down in adult α/β neurons, mitochondrial PKCδ activity was not changed after 5x spaced paired conditioning as compared to unpaired conditioning (right panel, n=8-9, t15=1.089, p=0.2934). (B) At the level of the vertical lobes, after 5x spaced paired conditioning, mitochondrial PKCδ activity was increased as compared to unpaired conditioning (left panel n=7-9, t14=3.932, p=0.0015). When DAMB was knocked down in adult α/β neurons, mitochondrial PKCδ activity was not changed after 5x spaced paired conditioning, as compared to unpaired conditioning (right panel n=8-9, t15=0.5247, p=0.6075). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by unpaired two-sided t-test. Asterisks refer to the P-value of the unpaired t-test comparison using the following nomenclature: **p<0.01, ns: not significant, p>0.05.

The DAMB/PKCδ axis is specific to LTM formation.

The pyruvate sensor Pyronic was expressed in adult MB neurons and pyruvate accumulation was measured after conditioning in the vertical lobes. For each panel, the period of imaging after conditioning is indicated in yellow on the imaging time frame. (A) 30 min to 1 h 30 min after 1x paired conditioning, an increased pyruvate flux was measured in control flies (tub-Gal80ts;VT30559, UAS-Pyronic) as compared to non-associative unpaired 1x training (slope measurement n=17-20, t35=4.075, p=0.0003). Here, we pooled the data obtained from control flies acquired in parallel to DAMB and PKCδ knockdown flies (panels C and D). (B) 3 h to 4 h 30 min after 1x paired conditioning of control flies (tub-Gal80ts;VT30559, UAS-Pyronic), the rate of pyruvate accumulation was similar as compared to unpaired conditioning (slope measurement n=8, t14=0.8022, p=0.4359). (C) 30 min to 1 h 30 min after 1x paired conditioning, the pyruvate flux was still increased when DAMB was knocked down in the MBs at the adult stage (tub-Gal80ts;VT30559, UAS-Pyronic, UAS-DAMB RNAi) (slope measurement n=9-12, t19=3.528, p=0.0022). (D) Similarly, PKCδ knockdown in adult MB neurons (tub-Gal80ts; VT30559, UAS-Pyronic, UAS-PKCδ RNAi) did not impair the 1x conditioning induced increase in pyruvate accumulation in MB neuron axons (slope measurement n=8-10, t16=4.568, p=0.0003). (E) 3 h to 4 h 30 min after the last cycle of 5x spaced conditioning of control flies (tub-Gal80ts;VT30559, UAS-Pyronic), the pyruvate flux was increased as compared to 5x spaced unpaired conditioning (slope measurement n=9-7, t14=3.889, p=0.0016). (F) 8 h to 9 h 30 min after the last cycle of 5x spaced conditioning of control flies (tub-Gal80ts;VT30559, UAS-Pyronic), the pyruvate flux was still increased as compared to 5x spaced unpaired conditioning (slope measurement n=8, t14=3.628, p=0.0027). (G) 24 h after the last cycle of 5x spaced conditioning of control flies (tub-Gal80ts;VT30559, UAS-Pyronic), the pyruvate flux was similar as compared to 5x spaced unpaired conditioning (slope measurement n=9, t16=0.3256, p=0.7489). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by unpaired two-sided t-test. Asterisks refer to the P-value of the unpaired t-test comparison using the following nomenclature: **p<0.01, ***p<0.001, ns: not significant, p>0.05.

Schema of the energy-based gating of LTM, illustrating the role of PKCδ downstream of the DAMB signaling cascade.

After 1x training, the pyruvate flux to the mitochondria in the vertical lobes of MB neurons is increased (Rabah et al., 2022). This increased energy state is not maintained but is progressively turned off, and after 3 h, the MBs are back to their basal energy state. Upon spaced training, in addition to the mechanism occurring after 1x training, the enhanced oscillatory activity of MP1 neurons activates DAMB signaling in the efferent α/β MB neurons. The DAMB receptor preferentially couples with Gq and the produced DAG activates PKCδ, which results in its translocation to mitochondria. There, PKCδ can activate the PDH, by releasing the inhibition exerted by PDK. The enhanced PDH activity promotes an increase in mitochondrial pyruvate flux, and this enhanced energy state gates memory consolidation, thereby enabling LTM formation. After 3 and even 8 hours, the lasting effect of PKCδ on mitochondrial metabolism maintains the high energy state of the MBs needed to support LTM formation.

Control experiments for behavior analysis and δCKAR imaging experiments.

(A) Behavior analysis with a second PKCδ RNAi: 24-h memory after 5x spaced conditioning was also specifically impaired after induction of the PKCδ RNAi expression in adult MB (n=14-18, F2,45=6.764, p=0.027), and it was not impacted without induction (n=11-15, F2,37=0.6226, p=0.5421). No memory defect was found after 5x massed training (n=16, F2,45=0.8968, p=0.4151) or 1x training (n=17-18, F2,48=0.6592, p=0.5219) in flies knocked down for PKCδ in adult MB. (B) Pan-neuronal expression of either PKCδ RNAi at the adult stage significantly reduced PKCδ mRNA levels. Relative Quantification (RQ) was performed, indicating the foldchange of mRNA levels relative to the control genotype (n=5-6, F2,13=11.85, p=0.0012). (C) In naïve flies, application of 5 µM of Bis IV (black arrow), a PKC inhibitor, resulted in a decrease in the cyto-δCKAR response, and this response was abolished when PKCδ was knocked down (n=5-6, t9=3.007, p=0.0148). Quantification of the mean cyto-δCKAR response was performed 20 s after Bis IV application on a time window of 250 s (black line). (D) After 5x spaced training, Bis IV injection also decreased the cyto-δCKAR response as compared to the DMSO control (n=6, t10=4.825, p=0.0007). Data are expressed as mean ± SEM with dots as individual values, and were analyzed either by one-way ANOVA with post hoc testing by the Tukey pairwise comparisons test (A-B), or by an unpaired two-sided t-test (C-D). Asterisks refer to the least significant P-value of post hoc comparison between the genotype of interest and the genotypic controls (A-B) or to the P-value of the unpaired t-test comparison (C-D) using the following nomenclature: *p<0.05, **p<0.01, ***p<0.001, ns: not significant, p>0.05.

Control analysis of 24-h memory formed after massed training upon PKCδ knockdown in α/β neurons.

(A) 24-h memory after 5x massed training was normal when PKCδ RNAi was expressed in the α/β neurons at the adult stage as compared to its genotypic controls (n=18, F2,51=1.771, p=0.1805). Data are expressed as mean ± SEM with dots as individual values, and were analyzed by one-way ANOVA with post hoc testing by the Tukey pairwise comparisons test. ns: not significant, p>0.05.

Additional information on pyruvate flux measurement using the Pyronic sensor.

(A) Schema representing the measure of the mitochondrial pyruvate flux using the pyruvate FRET sensor Pyronic. Here, mitochondrial respiration is blocked by sodium azide at the beginning of the recording, thereby stopping pyruvate mitochondrial uptake. The expected kinetic of pyruvate accumulation following the azide treatment is represented in grey, showing that pyruvate accumulates from the arrest of its mitochondrial uptake until saturation of the sensor. The rate of pyruvate accumulation (i.e. the slope of the measured kinetic, measured between 10 and 70% of the plateau) reflects the rate at which pyruvate was consumed by mitochondria for energy production before their blockade. (B) Complete recording of pyruvate imaging data shown in Figure 3A. In control flies (tub-Gal80ts;VT30559, UAS-Pyronic), PDBu injection is followed by a progressive increase in the measured Pyronic ratio (black line) as compared to DMSO injection (n=11, t20=5.437, p<0.0001). This effect is not sensitive to PKCδ knockdown (tub-Gal80ts;VT30559, UAS-Pyronic, UAS-PKCδ RNAi flies, n=9-14, t21=3.542, p=0.0019). As PDBu is known to activate other PKCs (Wu-zhang et al., 2012), this observation likely reflects a PDBu off-target effect independent of PKCδ on pyruvate neuronal levels. This effect explains why, when normalized just before azide injection (as in Figure 3A), the pyruvate accumulation plateaus appear lower when PDBu is injected onto the brains as compared to the DMSO controls. Data are expressed as mean ± SEM with dots as individual values, and were analyzed by unpaired two-sided t-test. Asterisks refer to the P-value of the unpaired t-test comparison using the following nomenclature: **p<0.01, ****p<0.0001.

Subcellular addressing of the cyto-δCKAR and mito-δCKAR probes.

(A) Comparison of fluorescence sub localization in the soma of MB neurons in flies expressing either cyto-δCKAR, mito-δCKAR or mito-DsRed (DsRed is addressed to mitochondria using the COX8 targeting sequence (Lutas et al., 2012)) under the control of the VT30559 driver. Mito-δCKAR punctate signal, similar to what is observed for mito-DsRed signal, indicates mitochondrial localization of the probe, in contrast to the diffuse signal observed with cyto-δCKAR. This comparison was done in the soma area because it is the only MB region where the density of mitochondria allows their discrimination by conventional microscopy. Images were acquired by 2-photon microscopy, cyto-and mito-δCKAR sensors were visualized through the YFP channel, and the mito-DsRed signal was acquired by 750 nm excitation and 580-610 nm detection. Scale bar=20µm.

Sensory acuity controls, related to Figure 1.