Decoding neural circuits that control compulsive sucrose seeking

Abstract

The lateral hypothalamic (LH) projection to the ventral tegmental area (VTA) has been linked to reward processing, but the computations within the LH-VTA loop that give rise to specific aspects of behavior have been difficult to isolate. We show that LH-VTA neurons encode the learned action of seeking a reward, independent of reward availability. In contrast, LH neurons downstream of VTA encode reward-predictive cues and unexpected reward omission. We show that inhibiting the LH-VTA pathway reduces "compulsive" sucrose seeking but not food consumption in hungry mice. We reveal that the LH sends excitatory and inhibitory input onto VTA dopamine (DA) and GABA neurons, and that the GABAergic projection drives feeding-related behavior. Our study overlays information about the type, function, and connectivity of LH neurons and identifies a neural circuit that selectively controls compulsive sugar consumption, without preventing feeding necessary for survival, providing a potential target for therapeutic interventions for compulsive-overeating disorder.

Figure 1. Phototagging LH-VTA projections reveals two populations of neurons with different response latencies to photostimulation

(A) Wild-type mice (n=12) were injected with AAV₅-DIO-ChR2-eYFP into the lateral hypothalamus (LH) and HSV-EF1α-IRES-Cre-mCherry into the ventral tegmental area (VTA). (B) Horizontal brain slices containing the LH were prepared for whole-cell patch-clamp recordings in ChR2-expressing and non-expressing LH neurons. (C) Individual traces recorded in current-clamp mode showing the response of ChR2-expressing (green, n=10) and non-expressing (grey, n=14) cells to a 5 ms pulse of 473 nm light. The box and whisker plot shows the average response latency for each ChR2-expressing cell ex vivo. (D) Photoresponse latencies in vivo were calculated by measuring the time from stimulation to 4 standard deviations (SD) above the baseline firing rate. (E) A bimodal distribution of excitatory photoresponse latencies was identified in recorded units (n=198) and divided into Type 1 (green; n=19) and Type 2 units (blue; n=34). (F) Type 1 units responded to photostimulation with fast excitation (3-8 ms latency). Inset shows the overlaid average traces for spontaneous spiking (black) and light-evoked spiking (blue) from a representative unit. (G) Type 2 units responded to photostimulation with delayed excitation (80-120 ms latency). (H) Scatterplot depicting the peak-trough duration of the waveform plotted against the average firing rate for each unit. (I) Normalized histogram showing the distribution of peak-trough durations for Type 1 units and (J) Type 2 units. (K and L) Diagrams illustrating two possible circuit models. (K) Type 1 units project directly from the LH to the VTA, while Type 2 units represent a population in the LH that is receiving feedback from the VTA, or (L) Type 2 units represent a population in the LH that is receiving input from collaterals of Type 1 units. Dotted lines indicate the presence of either a monosynaptic or polysynaptic connection. Scale bar: y-axis, 0.2 mV; x-axis, 500 μs. See also Figure S1.

Figure 2. Inhibition of the VTA selectively attenuates the photoresponse of Type 2, but not Type 1, units

(A) Mice expressing ChR2 in LH-VTA projections received an additional injection of AAV₅-CaMKIIα-eNpHR3.0-eYFP into the VTA to allow for transient inhibition of VTA neurons by yellow light. Three epochs of phototagging were conducted (LH photoactivation: ON-ON-ON, VTA photoinhibition: OFF-ON-OFF). (B) Type 1 (n=6/121 units, n=6 animals) photoresponse properties were unaffected (0%; n=0/6 attenuated or abolished) by VTA inhibition. Inset circles represent the number of units photoresponsive during each epoch. Inset shows the overlaid average traces for spontaneous spiking (black) and light-evoked spiking (blue) from a representative unit. (C) Type 2 (n=15/121 units, n=6 animals) photoresponse properties were abolished (67%; n=10/15) or attenuated (87%; n=13/15) during NpHR-mediated VTA inhibition. (D) No significant difference in max z-score was detected between epochs with and without inhibition of the VTA for Type 1 units (two-tailed, Paired Student's t-test, p=0.71). The max z-score was significantly lower in the ON (LH blue light illumination + VTA photoinhibition) epoch relative to the first OFF epoch (LH blue light illumination only) for Type 2 units (two-tailed, Paired Student's t-test, **p=0.0015). (E) There was a significant difference in max z-score (normalized to the OFF epoch) during photoinhibition of the VTA between Type 1 units compared to Type 2 units (two-tailed, Unpaired Student's t-test, *p=0.014). Error bars indicate +SEM. Scale bar: y-axis, 0.2 mV; x-axis, 500 μs. See also Figure S3.

Figure 3. Type 1 units predominantly respond to the port entry, while Type 2 units respond to both the conditioned stimulus and the port entry

(A) Mice with optrodes implanted in the LH and expressing ChR2 in LH-VTA projections were trained on a task where 50% of nosepokes (NP) were followed by a cue (conditioned stimulus; CS) that predicts the delivery of sucrose (unconditioned stimulus; US) at the delivery port. In vivo electrophysiological recordings were performed during the behavioral task followed by phototagging in the same recording session to identify units by projection target. (B) Perievent raster histograms for a representative Type 1 unit that responded to port entry, but not to the reward-predictive cue. Inset shows overlaid average traces for spontaneous spiking (black) and light-evoked spiking (blue) from a representative unit. (C) Population z-score plots showing the average response of all Type 1 units (n=19/198 units, n=12 animals). (D) Perievent raster histograms for a representative Type 2 unit that responded to the reward-predictive cue, but not to port entry. (E) Population z-score plots show the average response of all Type 2 units (n=34/198 units, n=12 animals). (F) Heat map representation of the individual z-scores of all units. (G) Of all Type 1 units, 63% responded exclusively to the port entry (n=12/19), while 11% responded to both the port entry and the reward-predictive cue (n=2/19). Within the Type 1 units that responded to the port entry, 64% (n=9/14) were excited (red) upon port entry while 36% (n=5/14) were inhibited (blue), and within the units that responded to the reward-predictive cue, 100% (n=2/2) were inhibited by the cue. (H) Of all Type 2 units, 35% (n=12/34) responded exclusively to the reward-predictive cue, 26% (n=9/34) responded exclusively to the port entry, and 12% (n=4/34) responded to both. Within the Type 2 units that responded to the cue, 100% (n=16/16) were excited by the cue while none were inhibited, and within the units that responded to port entry, 77% (n=10/13) were inhibited upon port entry while 23% (n=3/13) were excited. (I) Graphical representation of z-scores during the experimental windows for cue, no cue, and port entry, for Type 1, Type 2, and “no photoresponse” units. (J) Diagram of recorded units demonstrating whether they responded to the cue or port entry (PE) and whether that response was with excitation (+) or inhibition (−). Error bars indicate +SEM. Scale bar: y-axis, 0.2 mV; x-axis, 500 μs. See also Figure S2.

Figure 4. LH-VTA neurons encode the conditioned response of sucrose-seeking

(A) The original partial reinforcement sucrose self-administration task was modified so that in 30% of trials during which the reward-predictive cue was present, the expected sucrose delivery was omitted (15% of all trials). (B) Perievent raster histograms of a Type 1 unit that showed no difference in response to port entry with reward omission. Inset shows overlaid average traces for spontaneous spiking (black) and light-evoked spiking (blue) from a representative unit. (C) Perievent raster histograms of a Type 2 unit that showed a significantly different response to port entry upon omission of the expected reward. (D) Of all Type 1 units recorded (n=17/122 units, n=6 animals), only 12% (n=2/17) showed a significant difference in their response when the expected reward was omitted. In contrast, of all Type 2 units recorded (n=18/122 units, n=6 animals), 67% (n=12/18) showed a significant difference in their response when the expected reward was omitted (Chi-square=10.9804, ***p=0.0009). (E) Unexpected sucrose delivery occurred in the absence of predictive cues. Perievent raster histogram of a Type 1 unit that did not respond to port entry following unpredicted reward delivery. (F) Population z-score plot showing the average response of all Type 1 units to the port entry following unpredicted reward delivery. (G) Perievent raster histogram of a Type 2 unit that showed an increase in firing rate to port entry following unpredicted reward delivery. (H) Population z-score plot Type 2 unit responses to port entry following unpredicted reward delivery separated into those that showed a significant response and those that showed no significant response. (I) Of all Type 1 units recorded (n=8/105 units, n=6 animals), 0% (n=0/8) showed a significant response to the port entry following unpredicted reward delivery. In contrast, of all Type 2 units recorded (n=16/105 units, n=6 animals), 50% (n=8/16) showed a significant response to the port entry following unpredicted reward delivery (Chi-square=6, *p=0.0143). (J) Schematic of LH-VTA loop and the components of reward processing encoded by Type 1 and 2 cells. CR=conditioned response; CS=conditioned stimulus; US=unconditioned stimulus. Scale bar: y-axis, 0.2 mV; x-axis, 500 μs.

Figure 5. Excitation of LH-VTA projections promotes, while inhibition attenuates, compulsive sucrose-seeking

(A) Mice received injections of AAV₅-CaMKIIα-ChR2-eYFP (n=8), AAV₅-CaMKIIα-eNpHR3.0-eYFP (n=14), or AAV₅-CaMKIIα-eYFP (n=6 controls for ChR2, n=8 controls for NpHR) into the LH and an optic fiber was implanted above the VTA. (B) Mice were trained on a Pavlovian conditioned approach task wherein a cue predicted sucrose delivery to a port located across a shock grid. On test day, mice were presented with 20 cues during a baseline period without shock, 20 cues when the shock grid was on, and 20 cues during which 10 Hz blue or constant yellow light was delivered while the shock floor remained on. (C) Mice in the ChR2 group showed a significant increase in the number of port entries per cue during the ‘Shock+Light’ epoch relative to eYFP controls (n=8 ChR2, n=6 eYFP; 2-way ANOVA revealed a group × epoch interaction, F_2,24=20.47, p<0.0001; Bonferroni post hoc analysis, *p < 0.05). The difference between the number of port entries per cue during the ‘Shock+Light’ epoch and ‘Shock’ epoch was also significantly different between the ChR2 and eYFP control groups (two-tailed, Unpaired Student's t-test, **p=0.0090). (D) Mice in the NpHR group showed a significant decrease in the number of port entries per cue during the ‘Shock+Light’ epoch relative to eYFP controls (n=13 NpHR, n=8 eYFP; 2-way ANOVA revealed a group × epoch interaction, F_2,38=116.63, p<0.0001; Bonferroni post hoc analysis, *p < 0.05). The difference score was also significantly different between the NpHR-expressing and eYFP control mice (two-tailed, Unpaired Student's t-test, **p=0.0062). (E) Mice were placed into an open chamber with two cups, one containing food and the other without, and behavior in three experimental epochs was recorded (light OFF-ON-OFF). ChR2-expressing mice showed a significant increase in feeding (measured by time spent consuming food) compared with eYFP controls during the epoch paired with blue light stimulation (n=8 ChR2, n=6 eYFP; 2-way ANOVA revealed a group × epoch interaction, F_2,24=4.23, p=0.0268; Bonferroni post hoc analysis, **p < 0.01). (F) NpHR-expressing mice showed no significant differences from eYFP control mice in time spent feeding in any of the epochs (n=9 NpHR, n=7 eYFP). (G) To examine the effect of light stimulation on analgesia, mice had their tails placed into a heated water bath and the latency to tail withdrawal was measured during two counterbalanced epochs (light ON/OFF). ChR2-expressing mice showed no significant difference in tail withdrawal latency (normalized to OFF epoch) during blue light stimulation compared to eYFP controls (n=8 ChR2, n=6 eYFP), (H) nor did NpHR-expressing mice during yellow light stimulation (n=5 NpHR, n=8 eYFP). Error bars indicate +SEM. See also Figure S4.

Figure 6. The LH sends a mixture of excitatory and inhibitory projections to both dopamine (DA) and GABA neurons in the VTA

(A) AAV₅-CaMKIIα-ChR2-eYFP was injected into the LH and at least 6 weeks later 300 μm thick horizontal brain slices were prepared containing the VTA. Whole-cell patch-clamp recordings were made in VTA neurons, and ChR2-expressing LH terminals were activated by illumination with 473 nm light via an optic fiber resting on the brain slice. (B) Neurons were filled with biocytin during recording, and DA neurons were identified by immunohistochemistry for tyrosine hydroxylase (TH) (n=27). (C) The net effect of optical stimulation of LH terminals was assessed in current-clamp mode, which revealed that 55% of DA neurons (n=15/27) showed a net excitatory response, while 30% (n=8/27) responded with net inhibition, and 15% (n=4/27) showed no response. An example of an excitatory postsynaptic potential (EPSP, red trace), an inhibitory postsynaptic potential (IPSP, blue trace), and a non-responsive cell (grey trace) are shown below each bar. (D) The distribution of all recorded TH+ neurons plotted on horizontal midbrain slices with colors indicating the response to LH terminal photostimulation. (E) VTA DA neurons received only AMPAR-mediated input (67%, n=6/9), only GABA_AR-mediated input (11%, n=1/9) or both of these currents (22%, n=2/9). (F) VTA GABA neurons were identified by the presence of mCherry (n=24), achieved by injection of Cre-dependent AAV₅-EF1α-DIO-mCherry into the VTA of VGAT::Cre mice. (G) Optical stimulation of LH terminals in current-clamp mode showed that GABA neurons respond with either net excitation (46%, n=11/24) or net inhibition (54%, n=13/24) to LH input. (H) The distribution of each recorded GABA neuron plotted on horizontal midbrain slices with colors indicating the response to LH terminal stimulation. (I) The monosynaptic input to VTA GABA cells recorded in the presence of TTX and 4AP revealed that GABA neurons receive a mixture of AMPAR-mediated and GABA_AR-mediated input from the LH (AMPA only: 18%, n=2/11; AMPA & GABA_A: 73%, n=8/11; GABA_A: 9%, n=1/11). MT=medial terminal nucleus of the accessory optic tract. See also Figure S5 and S6.

Figure 7. Photoactivation of the GABAergic, but not the glutamatergic, component of the LH-VTA projection increased feeding behaviors

(A) In order to selectively activate glutamatergic or GABAergic LH-VTA projections, VGLUT2::Cre and VGAT::Cre mice received an injection of AAV₅-DIO-ChR2-eYFP or AAV₅-DIO-eYFP into the LH and had an optic fiber implanted over the VTA. In the sucrose-seeking task, there were no significant differences in the number of port entries per cue in any epoch for LH^glut-VTA:ChR2 mice (n=7) compared to LH^glut-VTA:eYFP control mice (n=6) nor (B) in LH^GABA-VTA:ChR2 mice (n=6) compared to LH^GABA-VTA:eYFP mice (n=8). (C) There was no significant difference between LH^glut-VTA:ChR2 mice and eYFP controls in feeding behavior. (D) However, LH^GABA-VTA:ChR2 mice showed a significant increase in time spent feeding during light stimulation compared to LH^GABA-VTA:eYFP controls (2-way ANOVA revealed a group × epoch interaction, F_2,24=4.78, p=0.0178; Bonferroni post hoc analysis, **p < 0.01). (E) Neither LH^glut-VTA:ChR2 mice, (F) nor LH^GABA-VTA:ChR2 mice showed a difference in tail withdrawal latency compared to their respective controls. (G) \LH-VTA:ChR2 mice showed a significant increase in time spent gnawing during the light ON epoch compared to eYFP controls (2-way ANOVA revealed a group × epoch interaction, F_2,24=4.78, p=0.0179; Bonferroni post hoc analysis, ***p < 0.001). (H) There was no significant difference between LH^glut-VTA:ChR2 and LH^glut-VTA:eYFP controls in gnawing behavior. (I) However, LH^GABA-VTA:ChR2 animals also showed a significant increase in time spent gnawing during the light ON epoch compared to LH^GABA-VTA:eYFP controls (2-way ANOVA revealed a group × epoch interaction, F_2,24=18.91, p<0.0001; Bonferroni post hoc analysis, ****p < 0.0001). (J) The difference score for gnawing behavior between the ON and OFF epochs was significantly greater in LH^GABA-VTA:ChR2 animals in comparison with either wild-type LHVTA:ChR2 or LH^glut-VTA:ChR2 animals (1-way ANOVA, F_2,18=16.76, p<0.0001; Bonferroni post hoc analysis, ***p <0.001). (K) Frequency-response curve showing the effect of different blue light stimulation frequencies (OFF, 5 Hz, 10 Hz) on behavior in LH^GABA-VTA:ChR2 animals. See also Figure S7.