Dorsolateral septum GLP-1R neurons regulate feeding via lateral hypothalamic projections

Abstract

Objective: Although glucagon-like peptide 1 (GLP-1) is known to regulate feeding, the central mechanisms contributing to this function remain enigmatic. Here, we aim to test the role of neurons expressing GLP-1 receptors (GLP-1R) in the dorsolateral septum (dLS; dLS^GLP-1R) that project to the lateral hypothalamic area (LHA) on food intake and determine the relationship with feeding regulation.

Methods: Using chemogenetic manipulations, we assessed how activation or inhibition of dLS^GLP-1R neurons affected food intake in Glp1r-ires-Cre mice. Then, we used channelrhodopsin-assisted circuit mapping, chemogenetics, and electrophysiological recordings to identify and assess the role of the pathway from dLS^GLP-1R →LHA projections in regulating food intake.

Results: Chemogenetic inhibition of dLS^GLP-1R neurons increases food intake. LHA is a major downstream target of dLS^GLP-1R neurons. The dLS^GLP-1R→LHA projections are GABAergic, and chemogenetic inhibition of this pathway also promotes food intake. While chemogenetic activation of dLS^GLP-1R→LHA projections modestly decreases food intake, optogenetic stimulation of the dLS^GLP-1R→LHA projection terminals in the LHA rapidly suppresses feeding behavior. Finally, we demonstrate that the GLP-1R agonist, Exendin 4 enhances dLS^GLP-1R →LHA GABA release.

Conclusions: Together, these results demonstrate that dLS-GLP-1R neurons and the inhibitory pathway to LHA can regulate feeding behavior, which might serve as a potential therapeutic target for the treatment of eating disorders or obesity.

Keywords: Feeding; GLP-1; Hypothalamus; Obesity; Septum.

Figure 1

Inhibition of dLS^GLP−1Rneurons acutely increases food intake. (A) Brain schematic of viral injection for chemogenetic dLS^GLP−1R neurons inhibition. (B) Transduction of dLS^GLP−1R neurons with hM4Di-mCherry. (C) CNO application hyperpolarized the resting membrane potential and reduced the firing rate of dLS^GLP−1R neurons (paired t-test, t(12) = 6.281, p < 0.0001, n = 12 cells from 3 mice). (D–F) Chemogenetic inhibition of dLS^GLP−1R neurons increased food intake (D) during the dark cycle when fed (two-way ANOVA, main effect of Group: F(1,120) = 14.44,p = 0.0002; main effect of time: F(4,120) = 71.72, p < 0.0001, no interaction between Group and Time: F (4,120) = 1.691, p = 0.1566), (E) during the light cycle when fed (two-way ANOVA, main effect of Group: F(1,120) = 53.89, p < 0.0001; main effect of time: F(4,120) = 58.53, p <0.0001, interaction between Group and Time: F(4,120) = 4.284, p = 0.0028) and (F) refeeding following an overnight fast (two-way ANOVA, main effect of Group: F(1,120) = 28.55, p < 0.0001; main effect of Time: F(4,120) = 68.25, p < 0.0001, interaction between Group and Time: F(4,120) = 2.463, p = 0.0488). n = 13 mice per group. Data are presented as mean ± SEM. Sidak's multiple comparisons test: ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Figure 2

Activation of dLS^GLP−1Rneurons has no significant effect on food intake. (A) Brain schematic of viral injection for dLS^GLP−1R neuron activation. (B) Transduction of dLS^GLP−1R neurons with hM3Dq-mCherry. (C) CNO application has no significant effect on the resting membrane potential and the firing rate of dLS^GLP−1R neurons (paired t-test, t(17), p = 0.2523, n = 17 cells from 3 mice), after CNO application, 35% of neurons exhibited resting membrane potential hyperpolarization, and 29% of neurons exhibited resting membrane potential depolarization. (D–F) Chemogenetic activation of dLS^GLP−1R neurons has no significant effect on food intake (D) during the dark cycle when fed (two-way ANOVA, no main effect of Group: F(1,105) = 3.402, p = 0.0679; main effect of Time: F(4,105) = 36.10, p < 0.0001, no interaction between Group and Time: F(4,105) = 0.3067, p = 0.8730.), (E) during the light cycle when fed (two-way ANOVA, no main effect of Group: F(1,105) = 0.06226, p = 0.8034; main effect of Time: F(4,105) = 18.67, p < 0.0001, no interaction between Group and Time: F(4,105) = 0.01724, p = 0.9994), and (F) refeeding following an overnight fast (two-way ANOVA, no main effect of Group: F(1,105) = 0.1432, p = 0.7059; main effect of Time: F(4,105) = 84.5, p < 0.0001, no interaction between Group and Time: F (4,105) = 0.04575, p = 0.9960). n = 13 control and 10 hM3Dq mice. (G–I) Local inhibitory connections in dLS. Schematic showing the experiment used to record postsynaptic currents in local dLS neurons induced by optogenetic stimulation of dLS^GLP−1R neurons (G). In brain slice recordings, blue light stimulation evoked robust IPSCs, which were blocked by picrotoxin but not CNQX (H–I) (one-way ANOVA, F (4,25) = 98.81, p < 0.0001, ∗∗∗∗Sidak's multiple comparisons test p < 0.0001 vs. ACSF). n = 6 cells from 3 mice. Data are presented as mean ± SEM.

Figure 3

dLS^GLP−1Rneurons project to LHA. (A) Glp1r-ires-Cre mice were injected in dLS with pAAV-hSyn-DIO-EYFP bilaterally. (B and C) Viral-mediated expression of EYFP in dLS^GLP−1R neurons soma (B) and axonal projections in the LHA (C) in Glp1r-ires-Cre mice. (D) Glp1r-ires-Cre:Ai14 mice were injected in LHA with retro-pAAV-Ef1a-DIO-EYFP bilaterally. (E and F) Images of coronal brain sections containing the dLS. White arrows indicate GLP-1R+/EYFP+ neurons. (G) Schematic of experiment used to record postsynaptic currents in LHA neurons induced by optogenetic stimulation of dLS^GLP−1R neurons. Glp1r-ires-Cre mice were injected in dLS with pAAV-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA bilaterally and LHA neurons were patched. (H) Representative trace of photostimulation (470 nm LED)-evoked IPSC in LHA neurons, which can be blocked by picrotoxin but not CNQX. (I) Normalized IPSC amplitude before and after CNQX, TTX, 4AP and PTX application (one-way ANOVA, F(4,20) = 10.7, p < 0.0001, ∗∗Sidak's multiple comparisons test p < 0.01 vs. ACSF). n = 5 cells from 3 mice. Data are presented as mean ± SEM.

Figure 4

dLS^GLP−1R → LHA inhibition increases feeding. (A) Brain schematic of viral injection for dLS^GLP−1R → LHA neuron inhibition. (B) Transduction of dLS^GLP−1R → LHA neurons with hM4Di-mCherry. (C) CNO application hyperpolarized the resting membrane potential and reduced the firing rate of dLS^GLP−1R → LHA projection neurons (paired t-test, t(6) = 2.897, p = 0.0339, n = 6 cells from 3 mice). (D–F) Chemogenetic inhibition of dLS^GLP−1R → LHA projection neurons increased food intake (D) during the dark cycle when fed (two-way ANOVA, main effect of Group: F(1,40) = 26.01, p < 0.0001; main effect of Time: F(4,40) = 36.65, p < 0.0001, no interaction between Group and Time: F(4,40) = 1.993, p = 0.1142), (E) during the light cycle when fed (two-way ANOVA, main effect of Group: F(1,60) = 27.88, p < 0.0001; main effect of Time: F(4,60) = 11.68, p < 0.0001, no interaction between Group and Time: F(4,60) = 2.344, p = 0.0649), and (F) refeeding following an overnight fast (two-way ANOVA, main effect of Group: F(1,55) = 44.88, p < 0.0001; main effect of Time: F(4,55) = 104, p < 0.0001, interaction between Group and Time: F(4,55) = 3.795, p = 0.00085). n = 7 control an d6 hM4Di mice. Sidak's multiple comparisons test: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Data are presented as mean ± SEM.

Figure 5

Activation of dLS^GLP−1R projection neurons has no significant effect on food intake. (A) Brain schematic of viral injection for dLS^GLP−1R → LHA neuron activation. (B) Transduction of dLS^GLP−1R → LHA neurons with hM3Dq-mCherry. (C) CNO application depolarized the resting membrane potential and increased the firing rate of dLS^GLP−1R→LHA projection neurons (paired t-test, t(8) = 2.702, p = 0.0306, n = 8 cells from 3 mice). (D–F) Chemogenetic activation of dLS^GLP−1R → LHA projection neurons has no significant effect on food intake (D) during the dark cycle when fed (two-way ANOVA, no main effect of Group: F(1,45) = 0.07858, p = 0.7805; main effect of Time: F(4,45) = 23.01, p < 0.0001, no interaction between Group and Time: F(4,45) = 0.1826, p = 0.9463), (E) during the light cycle when fed (two-way ANOVA, a marginally significant main effect of Group: F(1,50) = 3.272, p = 0.0765; main effect of Time: F(4,50) = 6.024, p = 0.0005, no interaction between Group and Time: F(4,50) = 0.4554, p = 0.7680). (F) Chemogenetic activation suppressed food intake during refeeding following an overnight fast (two-way ANOVA, main effect of Group: F(1,50) = 11.01, p = 0.0017; main effect of Time: F(4,50) = 64.4, p < 0.0001, no interaction between Group and Time: F(4,50) = 1.013, p = 0.4096). n = 7 control and 5 hM3Dq mice. Data are presented as mean ± SEM.

Figure 6

dLS^GLP−1R→LHA activation suppresses feeding. (A and B) ChR2-EYFP expression in the dLS (A) and axonal projections in the LHA (B) in Glp1r-ires-Cre mice. (C) Photostimulation of dLS^GLP−1R → LHA projections significantly suppressed food intake (two-way ANOVA, main effect of Group: F(1,36) = 3.382, p = 0.0742; main effect of Stimulation: F(2,36) = 3.816, p = 0.0314, no interaction between Group and Stimulation: F(2,36) = 2.576, p = 0.0900. ∗Sidak's multiple comparisons test p < 0.05). Data are presented as mean + SEM. n = 6 controls and 8 ChR2 mice. (D) Experimental paradigm for recording postsynaptic currents in LHA slices induced by optogenetic stimulation of dLS^GLP−1R neurons. (E) Representative evoked IPSCs after paired-pulse optogenetic stimulation. (F) Quantification of evoked IPSCs (paired t-test, t(10) = 2.312, p = 0.0461) and PPR (paired t-test, t(10) = 3.135, p = 0.0120) before and after Exn4 application. n = 10 cells from 3 mice. ∗p < 0.05.