PAM neurons form functional synapses with DPM neurons.

(A) The schematic and expression patterns of PAM and DPM neurons. PAM neurons labeled by R58E02-LexA are visualized by GFP (green), and DPM neurons labeled by eyeless-GAL80;MB-GAL80;c316-GAL4 are visualized by RFP (magenta). (B) Fragment of spGFP1-10 and nSyb-GFP1-10 is expressed in R58E02-GAL4 and LexA+ neurons, and fragment of spGFP11 is expressed L0111-LexA and eyeless-GAL80;MB-GAL80;c316-GAL4+ neurons, respectively. Reconstituted GFP signals are shown as puncta (green). The mushroom body is outlined with dashed lines. (C) Newly generated VT064246-LexA (II) and VT064246-LexA (III) recapitulate expression pattern of DPM neurons labeled with VT064246-GAL4. (D) Schematic of restricted trans-Tango for direct synaptic contacts between PAM-GAL4 (R58E02) labeled upstream neurons and DPM-LexA (VT064246) labeled targeted postsynaptic to PAM neurons. Anti-GFP-labeled neurons are the downstream of PAM neurons. Anti-RFP-labeled neurons are non-downstream neurons. Right lower panels, the control of restricted trans-Tango. Signals of downstream target of DPM neurons was not detected without PAM driver R58E02-GAL4. Scale bar: 20 μm.

PAM neurons exert direct inhibitory control over DPM neurons.

(A) Ex vivo functional connection between PAM neurons and DPM neurons. Averaged GCaMP traces and quantification of DPM neurons in response to activation of P2X2 expressing PAM neurons by application of 2.5 mM ATP. n = 12-23. Groups that are statistically no differences using the same letter, while distinct letters denote statistically significant differences between groups. (B) In vivo calcium imaging for functional connection between PAM neurons and DPM neurons. Averaged GCaMP traces and quantification of DPM neurons in response to activation of CsChrimson expressing PAM neurons by red light. The red box indicates the 30-s, 2-Hz, 627-nm light pulse. n = 14-16. **p<0.01; ****p<0.0001. Curves and bar graph are represented as mean ± SEM with individual values. See also Table 3.

Activation of PAM neurons during consolidation impairs long-term memory, possibly via inhibiting DPM neurons.

(A) The paradigm of a single session training with 2 min association of sucrose (Green) and an odor, and the paradigm of 24 h appetitive memory with blocking the output of neurons for 2.5 h before and during training at 32 ℃ (left). Inactivation of PAM neurons during training impaired 24 h memory (Right). n = 8-10. (B) Schematic of innervations of DPM neurons on all lobes of MB (left). Inactivation of DPM neurons for 2.5 h after training disrupts 24 h memory (right). (C) Inactivation of DPM neurons labeled by VT064246-GAL4 after training for 2.5 h impaired 24 h memory. n = 28-30. (D) Inactivation of DPM neurons during training did not impair 24 h memory. n = 6-9. (E) Schematic of innervations of R58E02-GAL4 labeling (PAM) neurons on the horizontal lobes of MB, including subdomains of γ1-5, β’2, β1-2 and α1 (left). Different gray scales were for each compartment’s visibility, not projection intensity. Activation of PAM neurons for 1 h after training disrupts 24 h memory (right). (F) Flies form 24 h appetitive memory without activation of R58E02-GAL4+ neurons. n = 14-18. (G) Inactivation of PAM neurons after training for 2.5 h did not impair 24 h memory. n = 14-15. Red line: 30 ℃; blue line: 32 ℃. n.s.: not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Bar graph are represented as mean ± SEM with individual values. See also Table 1.

Increased or decreased activity of PAM/DPM neurons results reduced and fragmented sleep.

(A) Timeline for sleep recording with activation of PAM neurons for 1 h before, during and after activation till supposed 24 h memory test. Total sleep (upper panel), number of sleep episodes (middle panel), and P(wake) (probability of transition from a sleep to an awake status) (lower panel) in 1 h bin for activation of PAM neurons. n = 40-43. (B) Inactivation of PAM neurons labeled by R58E02-GAL4 under starvation resulted in a significant decrease of sleep at early night, with increased number of sleep episodes and P(wake). n = 33-39. (C) Timeline for sleep recording with inactivation of DPM neurons labeled by c316-GAL4 for 2.5 h before, during and after activation till supposed 24 h memory test. Total sleep, number of sleep episodes, and P(wake) in 1 h bin for inactivation of DPM neurons. n = 33-43. (D) Inactivation of DPM neurons labeled by VT064246-GAL4 with starvation resulted in a decreased and fragmented sleep after activation at night, but almost no effect on P(wake). n = 11-37. Red line: 30 ℃; blue line: 32 ℃; LP1: light period on day 1; red box: activation of neurons for 1 h; DP1 (grey box): dark period on day 1; LP2: light period on day 2; ZT: Zeitgeber time. See also Table 4.

PAM-DPM interactions in sleep regulation are likely to be context-dependent.

(A) Activation of PAM neurons without starvation resulted a reduction of sleep at early night, as well as a slight decrease in number of sleep episodes, but an increase in P(wake). n = 39-42. (B) Inactivation of PAM neurons had no effect on sleep without starvation. n = 22-27. (C) Inactivation of DPM neurons labeled by eyeless-GAL80;MB-GAL80;c316-GAL4 without starvation had barely no effects on sleep. n = 35-47. (D) Inactivation of DPM neurons labeled by VT064246-GAL4 had mild effect in sleep at early phase after inactivation without starvation. n = 27-36. Sleep data were presented in 1 h bin. White box: light period; red box: activation period; grey box, dark period; solid fill: starvation; patterned fill: no starvation. See also Table 4.

The PAM-α1 subset of PAM neurons contributes to memory consolidation.

(A, C) The expression pattern of LTM-PAM and STM-PAM neurons labeled by R15A04-GAL4 and R48B04-GAL4, respectively. Upper panels, cell body; middle panels, projections on the MB; lower panels, schematic of the projections on the horizontal lobes. Scale bar: 20 μm. (B, D) The paradigm of 24 h memory by activation of LTM/STM-PAM neurons for 1 h after training. Activation of R15A04-GAL4+ neurons impairs 24 h memory (B). Activation of R48B04-GAL4+ neurons has no effect on 24 h memory (D). n = 15-21. (E) The expression patterns of specific subtypes of LTM-PAM neurons labeled by split-GAL4 lines: MB299B and MB043B (α1), MB213B (β1/β2), MB025B (β’1), MB032B (β’2), MB315C (γ5). Scale bar: 20 μm. (F) Left panel, activation and no-activation of subtypes of LTM-PAM neurons. Right panel, activation of MB299B and MB043B which project to α1 subdomain impairs LTM compared to the no-activation group. n = 5-12. (G) The differences in cell number/identity between the two drivers of MB299B and MB043B. “+” with greyscales indicated the levels of projections of a driver in MB subdomains; “-” indicated no projections. See also Tables 1 and 2.

The PAM-α1 subset of PAM neurons contributes to linking memory consolidation and sleep.

(A) Flies with activation of MB299B-GAL4+ neurons after training for 1 h exhibited impaired 24 h memory compared to their genetic controls. n = 14-17. (B) Activation of PAM-α1 neurons labeled by MB043B after training for 1 h impaired 24 h memory. n = 16-20. (C) Flies with blockade the output of MB299B-GAL4+ neurons after training for 2.5 h exhibited an impairment of 24 h memory compared to genetic controls. n = 11-14. (D) Inactivation of PAM-α1 neurons labeled by MB043B after training for 2.5 h did not affect 24 h memory. n = 16-25. (E) 1 h activation of PAM-α1 neurons labeled by MB299B prior to and during the test impaired 24 h memory. n = 11-17. (F) Activation of PAM-α1 neurons labeled by MB299B during the test did not affect 24 h memory. n = 18-23. (G) Inhibition of PAM-α1 neurons labeled by MB299B prior to and during the test for 30 min did not affect 24 h memory. n = 9-10. (H) Upper panel: the paradigm for sleep recording under starvation. Lower panel: total sleep, number of sleep episodes, and P(wake) in 1 h bin for activation of PAM-α1 neurons labeled by MB299B. n = 22-30. (I) Inactivation of PAM-α1 neurons labeled by MB299B under starvation had no effects on sleep. n = 33-39. (J) Activation of PAM-α1 neurons labeled by MB043B for 1 h reduced total sleep, increased the number of sleep episodes and P(wake) especially during the dark period. n = 23-30. (K) Inactivation of PAM-α1 neurons labeled by MB043B under starvation significantly reduced total sleep, and increased both the number of sleep episodes and P(wake), especially during the dark period. n = 30-39. See also Tables 1 and 4.

A specific PAM-α1-DPM inhibitory microcircuit maintains dynamic changes in activity post-learning.

(A) Characterization of anatomical connections of PAM-α1-DPM microcircuit. Left panel: 3D view of expression patterns of PAM-α1 neurons (colored) and DPM neurons (grey) from NeuPrint. Right panel: synaptic connections between 14 PAM-α1 neurons and 2 DPM neurons with their identity number based on the Hemibrain connectome dataset. (B) Activation of P2X2 expressing PAM-α1 neurons labeled by MB299B by application of ATP reduced GCaMP levels in DPM neurons labeled by VT064246-LexA. n = 10-18. (C-F) Paradigm for appetitive olfactory training and vehicle un-trained control, and representative neural activity profiles reflected by Tric-LUC expression in drivers labeling PAM-α1 and DPM neurons. Data were continuously recorded for three hours post-training in freely moving flies. (D-E) Dynamic activity changes of PAM-α1 (blue) and DPM (magenta) neurons under un-trained/trained conditions. Dotted lines represent un-trained groups, and solid lines represent trained groups. n = 20-24. (F) Quantification of the averaged changes in PAM-α1 and DPM neuronal activity at each hour post-training. *: significant difference between the DPM neuron groups under un-trained versus trained conditions; #: significant difference between PAM-α1 and DPM neuron groups. See also Table 3.

The PAM-α1-DPM inhibitory microcircuit binds memory consolidation and sleep in an internal-state dependent manner.

(A) Simultaneous silence of PAM-α1-DPM microcircuit resulted in sleep reduction and fragmentation, increased possibility of waking. n = 36-40. (B) Simultaneous silence of PAM-α1-DPM microcircuit without starvation had little effect on sleep and the number of sleep episodes and P (wake). n = 29-31. (C-D) Sleep profiles of before, during and after activation of PAM-α1 neurons till supposed 24 h memory test with/without 0.1 mM gaboxadol (THIP) treatment. Application of THIP rescued sleep reduction induced by activation of PAM-α1 neurons. n = 38-42 (C), n = 37-44 (D). (E) The paradigm of 24 h memory test with activation of PAM-α1 neurons for 1 h after training. (F-G) THIP rescued 24 h memory deficit induced by 1 h activation of PAM-α1 neurons during consolidation compared to genetic controls (F, n = 15-17) or compared to no THIP group (G, n = 17). See also Tables 1, 2 and 4.

Dop1R1 expressed in DPM neurons mediates sleep-memory integration.

(A) Colocalization of DPM neurons with distinct dopamine receptor subtypes. DPM neurons labeled by VT064246-LexA were visualized by GFP (green), and dopamine receptors labeled by Dop1R1-GAL4, Dop1R2-GAL4, DopEcR-GAL4 and Dop2R-GAL4 were visualized by RFP (magenta). The yellow dashed box outlines the cell body of one DPM neuron, with the upper and lower panels on the right showing the split channels for DPM neurons and the corresponding dopamine receptor subtype (white arrow), respectively. Scale bar: 20 μm. (B) Averaged cAMP traces (middle panel) and quantification (right panel) of DPM neurons in response to DA with the presence of 2.0 mM TTX with or without Dop1R1 or Dop1R2 knockdown. n = 11 for all groups. (C) Knock down Dop1R1 on DPM neurons decreased the total amount of daytime sleep and increased P(wake). n = 40-44. (D-E) Knock down Dop1R1 on DPM neurons had no effect on 2min, but impaired 24 memory. n = 14-19 (D), n = 20-22 (E). (F) Knock down Dop1R2 on DPM neurons, had no effect on sleep. n = 31-32. (G-H) Knock down Dop1R2 on DPM neurons had no effect on 2min, but impaired 24 memory. n = 17-20 (G), n = 15-16 (H). See also Tables 1 and 3.

MBON-α1 neurons regulate sleep irrespective of starvation state, and their activity disruption during memory consolidation impairs LTM.

(A, C) Activation of MBON-α1 neurons labeled by MB310C under starvation (A, n = 26-32) or no starvation (C, n = 31 for all groups) condition had no effects on sleep. (B, D) Inactivation of MBON-α1 neurons for 2.5 h significantly reduced total sleep, increased the number of sleep episodes and P(wake) especially during the dark period under starvation (B, n = 37-44) and no starvation (D, n = 33-43) condition. (E) Flies with activation of MBON-α1 neurons after training for 1 h exhibited impaired 24 h memory compared to their genetic controls. n = 17-20. (F) Flies with blockade the output of MBON-α1 neurons after training for 2.5 h exhibited an impairment of 24 h memory compared to their genetic controls. n = 17-18. See also Tables 1 and 4.

Schematic of activity-dependent parallel and specific regulatory circuits for sleep-memory regulation.

Increased activity of PAM-α1 neurons strongly inhibits DPM, leading to disrupted sleep and impaired LTM. MBON-α1 likely acts as a parallel circuit relative to PAM-α1-DPM pathway, interfering with the coupling between sleep and memory. DA signaling through Dop1R1 receptors triggers an increase in cAMP and a decrease in Ca²⁺, disrupting both sleep and LTM. By contrast, signaling via Dop1R2 receptors only reduces Ca²⁺ levels, specifically impairing LTM. However, inactivated PAM-α1 may target different downstream targets to separately regulate sleep or memory.

Activation of DPM neurons increases sleep but has no effect on memory consolidation.

(A) Activation of DPM neurons labeled by eyeless-GAL80;MB-GAL80;c316-GAL4 under starvation increased total sleep during the activation period, with little to no effect afterwards. n = 35-42. (B) Activation of DPM neurons for 1 h immediately following training did not affect 24 h memory. n = 13-16. See also Tables 1 and 4.

The PAM-α1 subset linking memory consolidation and sleep depends on internal starvation status.

(A-B) Sleep profiles of total sleep, the number of sleep episodes, and P(wake) with 1 h activation of MB299B (A) or MB043B (B) before, during and after activation. Sleep was not affected without starvation. n = 28-31. (C) Inactivation of MB299B for 2.5 h without starvation had no effects on sleep. n = 30-37. (D) Inactivation of MB043B for 2.5 h without starvation resulted a significant reduction in total sleep, a mild increase of number of sleep episodes during the dark period, and no effect on P(wake). n = 30-39. See also Table 4.

Verification of Dop1R1 and Dop1R2 knockdown efficiency and their effects on neuronal activity.

(A-B) Relative mRNA levels of Dop1R1 (A) and Dop1R2 (B) in adult fly heads. Knockdown of either Dop1R1 or Dop1R2 in elav-GAL4 labeling cells resulted in a significant reduction in transcript levels compared to controls. Data are presented as mean ± SEM (n = 5-8). **p < 0.01. (C-E) Averaged GCaMP traces (upper panels) and quantification (lower panels) of DPM neurons in response to DA with the presence of 2.0 mM TTX with or without Dop1R1 or Dop1R2 knocked-down. (C, n = 12-13; D, n = 12 for all groups; E, n = 15-27). See also Tables 1 and 3.