A Role for the Locus Coeruleus in Hippocampal CA1 Place Cell Reorganization during Spatial Reward Learning

Abstract

During spatial learning, hippocampal (HPC) place maps reorganize to represent new goal locations, but little is known about the circuit mechanisms facilitating these changes. Here, we examined how neuromodulation via locus coeruleus (LC) projections to HPC area CA1 (LC-CA1) regulates the overrepresentation of CA1 place cells near rewarded locations. Using two-photon calcium imaging, we monitored the activity of LC-CA1 fibers in the mouse dorsal HPC. We find that the LC-CA1 projection signals the translocation of a reward, predicting behavioral performance on a goal-oriented spatial learning task. An optogenetic stimulation mimicking this LC-CA1 activity induces place cell reorganization around a familiar reward, while its inhibition decreases the degree of overrepresentation around a translocated reward. Our results show that LC acts in conjunction with other factors to induce goal-directed reorganization of HPC representations and provide a better understanding of the role of neuromodulatory actions on HPC place map plasticity.

Keywords: hippocampus; imaging; in vivo; locus coeruleus; noradrenaline; optogenetics; place cell; reward learning.

Figure 1.. Place cells are enriched at a translocated reward site during GOL.

A. The goal-oriented learning (GOL) task. Mice searched for an unmarked reward zone (RZ), and water rewards were delivered operantly within the fixed 10cm zone. The RZ was at the same location for 3 days, and then moved to a new location. B. Representative licking from one mouse. Histograms: fraction of total number of licks in each position bin (n=100). Blue shaded areas: RZs. C. Fraction of licks within the RZ aggregated by session and plotted by day (thick line and bars: mean±SEM, thin gray lines: individual animals). D. Left: time-averaged image from a representative recording session with 2p GCaMP6f imaging in CA1PCs. Right: relative GCaMP-calcium fluorescence (ΔF/F) traces from two example place cells. Scale bar: 50 μm. E. Example place cell tuning curves in the last session of RZ1 and RZ2. Rows: average tuning curves for individual cells along the linearized belt, normalized by peak activity. Blue shaded area: RZ location. F. Percentage of place cells (mean±SEM) in the peri-reward zone (pRZ) defined as the RZ +25 cm preceding the zone. The last session of day6 (RZ2) shows enrichment compared with day3 (RZ1). Dashed line: percentage of place cells expected from a uniform distribution along the belt (day3: 0.17±0.045, day6: 0.242±0.07, unpaired two-tailed t-test; t₍₁₈₎=−2.4, p=0.02. One-sample two-tailed t-test for expected value derived from the uniform distribution; day3: t₍₈₎=−0.36, p=0.72, day 6: t₍₉₎= 2.72, p=0.02). See also Figures S2–4.

Figure 2.. Locus coeruleus activity changes during GOL

A. Left: LC-CA1 axon labeling strategy. Cre-dependent virus [rAAV2/9:EF1a-(GCaMP6s)^Cre] was injected into the locus coeruleus (LC) of Th-IRES-Cre^+/− mice. LC axons in hippocampal (HPC) CA1 were imaged through a cannula. Middle and right: post hoc immunofluorescent staining with antibodies against tyrosine-hydroxylase (anti-Th) and GCaMP (anti-GFP) in LC and HPC to confirm labeling strategy. Scale bar: LC: 50 μm; inset: 20μm, HPC: 100 μm. B. Example of multi-day 2p imaging of LC-CA1 axons in CA1 stratum oriens (SO). Scale bar: 50 μm. C. A generalized linear model (GLM) trained with three covariates (position, velocity, and licking) to predict LC-CA1 activity. Example trace of LC-CA1 calcium activity (dashed line) and predicted by the GLM (solid line). D. Left: cross-validated R² calculated on the held out test data in RZ1 sessions and RZ2 sessions (mean±SEM, RZ1: 0.34±0.04; RZ2: 0.48±0.04; n=6 mice, two-tailed paired t-test, t₍₅₎=−5.97, p=0.004). Right: contribution of different variables estimated by the amount of information gained by including each variable during RZ1 and RZ2 (mean±SEM; position in RZ1: −0.01±0.013; position in RZ2: 0.14±0.03; velocity in RZ1: 0.15±0.03; velocity in RZ2: 0.12±0.03; licking in RZ1: 0.02±0.01; licking in RZ2: 0.03±0.01; n=6 mice, two-tailed paired t-tests between RZ1 and RZ2 for: velocity, t₍₅₎=−4.84, p=0.005; position, t₍₅₎=1.49, p=0.196; licking, t₍₅₎=−0.63, p=0.556). E. Example traces of LC-CA1 activity (dark magenta) and behavioral variables in RZ1 and RZ2 for 2 consecutive laps. In RZ1, axons are correlated with velocity, while in RZ2, there is a decorrelation between LC-CA1 activity and velocity (arrows). 1 arbitrary unit (AU) refers to 1 sigma of the Z-score trace for the fluorescence F (GCaMP6s), and speed in cm/s (velocity). F. Average peri-stimulus activity histogram of the LC-CA1 signal (dark magenta), velocity (lavender) and licking (black) in RZ1 and RZ2, centered around reward. The overshoot in LC-CA1 activity is only visible in RZ2 (arrow) (n=6 mice, mean: darker colors, SEM: lighter colors). It preceded the reward by an average of 24.33±2.02 cm. G. Slope of the linear fit between velocity and LC-CA1 signals. The signals are less correlated over the course of learning. Data are in mean±SEM for sessions in a given day collected from n=6 mice (Learning day1: 0.126±0.038; day2: 0.03±0.036; day3: 0.118±0.056; day4: 0.006±0.067; day5:−0.093±0.061; day6:−0.121±0.059. One-way mixed-effects model ANOVA F_{(5, 82)}=4.28, p=0.0017. Post-hoc Tukey test, day1 vs. day5, p=0.026. day1 vs. day6, p=0.008. day2 vs. day5, p=0.047. day2 vs. day6, p=0.016), and are correlated in RZ1 but decorrelated in RZ2 (Learning RZ1: 0.09±0.026; RZ2: −0.069±0.036. Two-tailed unpaired t-test, n₁=51 sessions, n₂=54 sessions, t₍₁₀₄₎=−3.57, p=0.0005). Two mice that did not learn the task showed signals correlated with velocity in both RZs (Non-learning RZ1: 0.175±0.037, RZ2: 0.194±0.03. Two-tailed unpaired t-test, n₁=15 sessions, n₂=16 sessions, t₍₃₀₎=0.399, p=0.693). H. The slope of the relationship between speed and fluorescence was correlated with behavioral performance, as measured by the fraction of licks in the RZ (Pearson’s R test, n=105 points, R=−0.346, p=0.0002). Each point is the average performance as a function of the average correlation coefficient for each session, and each mouse. See also Figures S1–4.

Figure 3.. Stimulating LC-CA1 axons induces CA1 place cell enrichment near a rewarded location during GOL.

A. Left: labelling strategy for optogenetic stimulation and imaging of LC-CA1 axons. The LC was injected with rAAV2/9:EF1a-(GCaMP6s)^Cre and rAAV2/9:EF1a-(bReaChes-tdTomato)^Cre. Right: time-averaged 2p images of GCaMP6s and bReaChes-tdTomato in CA1 in vivo, showing overlap between the two. Scale bar: 50 μm. B. ΔF/F traces of LC-CA1 axons expressing either GCaMP6s and bReaChes (opsin), or GCaMP6s only (control). Photostimulation: 1s pulse every 20s. C. Left: average ΔF/F traces in control and opsin mice, (n=3 mice per condition). Dotted line: SEM. Right: difference of ΔF/F between post and pre stimulation was higher for opsin mice (mean±SEM, control: 2×10⁻⁴±4×10⁻⁴; opsin: 0.0246±0.008. Two-tailed unpaired t-test, t₍₅₎=−2.9, p=0.044). D. Top left: labelling strategy for optogenetic stimulation of LC-CA1 axons and imaging of CA1PCs. rAAV2/9:EF1a-(bReaChes-tdTomato)^Cre was injected in the LC, and rAAV2/1:CaMKII-GCaMP6f was injected in CA1. Top right: averaged 2p Z-stack of CA1PCs and LC-CA1 axons. Scale bar: 50 μm. Bottom left: example LC injected with bReaChes-tdTomato and stained for tyrosine hydroxylase (anti-Th). Bottom right: example HPC stained for td-Tomato. LC-CA1 axons labelled with bReaChes-tdTomato can be seen in all layers (S.O.: stratum oriens, S.P.: str. pyramidale, S.R.: str. radiatum, S.L.M.: str. lacunosum-moleculare). Scale bars: LC and CA1,100 μm; LC inset, 20 μm. E. Stimulating LC-CA1 axons does not acutely affect CA1PC activity. Difference in ΔF/F during pre- vs. post-light stimulation. Left: each cross is one mouse, size of the cross is SEM. Right, average difference post- minus pre-light stimulation (mean±SEM, control: 0.0015±0.0047, opsin: 0.0097±0.0075. Two-tailed unpaired t-test, t₍₁₁₎=0.88, p=0.403; control, n=5 mice; opsin, n=6 mice). Red: Th-IRES-Cre^+/− mice injected with bReaChes-tdTomato in the LC (opsin), black: Th-IRES-Cre^−/− (control). Each dot is one mouse. F. Top: optogenetic stimulation of LC-CA1 axons in RZ1 of the GOL task. The LED stimulation began 10cm before reward and spanned the whole RZ. The peri-reward zone (pRZ) began 25cm before the LED zone. Bottom: opsin (Th-IRES-Cre^+/−, n=8 mice) and control (Th-IRES-Cre^−/−, n=4 mice) animals showed the same behavioral learning dynamics (mean±SEM, session1, control: 0.203±0.048; opsin: 0.332±0.083. session9, control: 0.514±0.182; opsin: 0.583±0.108. Mann-Whitney U test. session1, z₍₁₁₎=12.0, p=0.28, session 9, z₍₁₁₎=12.0, p=0.28). G. Heatmaps of place cells in the first and last sessions of control and opsin mice. Opsin mice show a large degree of enrichment before the light stimulation zone in the last session (session9). H. Percentage of place fields in the peri-reward zone (pRZ) in session1 (top) and session9 (bottom). Session9 of opsin mice shows enrichment compared with control (mean±SEM, session1, control: 0.216±0.019; opsin: 0.199±0.028. session9, control: 0.216±0.035; opsin: 0.402±0.059. Mann-Whitney U test. Session1, z₍₁₁₎=12.0, p=0.276, session9, z₍₁₁₎=3.0, p=0.017. Session1, control vs. expected distribution: t₍₃₎=1.3, p=0.28; opsin vs. expected distribution: t₍₇₎=1.21, 0.26. session9, control vs. expected t₍₃₎=1.95, p=0.14; opsin vs. expected: t₍₇₎=4.11, 0.004). See also Figure S4.

Figure 4.. Inhibition of LC-CA1 axons decreases CA1 place field enrichment

A. Left: labelling strategy for LC-CA1 optogenetic inhibition and imaging of CA1PCs. The LC was injected with rAAV2/9:EF1a-(ArchT-tdTomato)^Cre and rAAV2/1:CaMKII-GCaMP6f was injected in CA1. Middle: example LC injected with ArchT-tdTomato and stained for tyrosine hydroxylase (anti-Th). Right: example HPC section. LC-CA1 axons labelled with ArchT-tdTomato can be seen in all layers. Scale bars: LC and CA1: 50 μm; LC inset: 20μm. B. Optogenetic inhibition of LC-CA1 axons started on the session1 of RZ2 of the GOL task. The LED stimulation zone began 10cm before the reward and spanned the whole RZ. C. Left: opsin (Th-IRES-Cre^+/−, n=11 mice) and control (Th-IRES-Cre^−/−, n=9 mice) animals showed the same behavioral learning dynamics (mean±SEM, repeated measures ANOVA, opsin vs. control, within-subject factor of session, F_1,17 =1.89, p=0.17). Opsin and control mice showed the same behavioral performance in the last session of RZ2 (unpaired two-tailed t-test, t₍₁₉₎=0.35, p=0.72). D. Percentage of place cells in the peri-reward zone (pRZ) for control and opsin mice (mean±SEM). Dashed line: percentage of place cells expected from a uniform distribution (expected=0.176. control: 0.29±0.03, opsin: 0.18±0.02. Opsin vs. control, unpaired two-tailed t-test, t₍₁₉₎=3.02, p=0.007. One-sample two-tailed t-test for expected value, control: t₍₈₎=4.32, p=0.003, opsin: t₍₁₀₎=0.28, p=0.77).