Plasticity in Prefrontal Cortex Induced by Coordinated Synaptic Transmission Arising from Reuniens/Rhomboid Nuclei and Hippocampus

Abstract

The nucleus reuniens and rhomboid nuclei of the thalamus (ReRh) are reciprocally connected to a range of higher order cortices including hippocampus (HPC) and medial prefrontal cortex (mPFC). The physiological function of ReRh is well predicted by requirement for interactions between mPFC and HPC, including associative recognition memory, spatial navigation, and working memory. Although anatomical and electrophysiological evidence suggests ReRh makes excitatory synapses in mPFC there is little data on the physiological properties of these projections, or whether ReRh and HPC target overlapping cell populations and, if so, how they interact. We demonstrate in ex vivo mPFC slices that ReRh and HPC afferent inputs converge onto more than two-thirds of layer 5 pyramidal neurons, show that ReRh, but not HPC, undergoes marked short-term plasticity during theta frequency transmission, and that HPC, but not ReRh, afferents are subject to neuromodulation by acetylcholine acting via muscarinic receptor M2. Finally, we demonstrate that pairing HPC followed by ReRh (but not pairing ReRh followed by HPC) at theta frequency induces associative, NMDA receptor dependent synaptic plasticity in both inputs to mPFC. These data provide vital physiological phenotypes of the synapses of this circuit and provide a novel mechanism for HPC-ReRh-mPFC encoding.

Keywords: associative plasticity; neuromodulation; prefrontal cortex; short-term plasticity; thalamocortical.

Figure 1

Electrophysiological characterization of L5 pyramidal cells receiving optogenetically activated nucleus reuniens/rhomboid synapses. (A) Representative widefield-fluorescence image showing neuronal transduction of nucleus reuniens (Re) and rhomboid nucleus (Rh) following injection of AAV9:CaMKii:hChR2 (E123T/T159C):mCherry (red) and DAPI (blue). VRe, ventral reuniens; Xi, xiphoid; PaXi, paraxiphoid; CM, central medial; AM, anteromedial; VM, ventromedial; MD, mediodorsal; Sub, submedius thalamic nuclei; mt, mammillothalamic tract. (B) Monochrome image of mCherry positive fibers in PFC following AAV injection into ReRh. Dotted lines denote the boundaries of prelimbic cortex. mCherry signal is amplified with anti-mCherry antibody. Cg1, cingulate cortex; IL, infralimbic cortex; PrL, prelimbic cortex. Scale bar = 500 μm. (C) Schematic of acute mPFC slice with whole-cell recording from layer 5 pyramidal neuron in PrL, light activation of soma and proximal dendrites via microscope objective (blue) and stimulation of hippocampal fiber bundle using conventional stimulating electrode. (D) Representative ReRh (blue) and HPC (black) EPSPs. Blue arrow denotes light activation. (E) Proportion of cells receiving different permutations of ReRh and HPC inputs, 187 cells from 65 animals. Passive membrane properties measured from −100 pA current injection split by synaptic input. RMP plotted as mean ± standard deviation, one-way ANOVA F_(3,183) = 1.2, P = 0.32. Other parameters one or more column failed Shapiro–Wilk test for normality, box plots show median and interquartile range, whiskers max and min data points. Kruskal–Wallis test P values: Tau = 0.074, Rinput = 0.031, Sag % = 0.0036, sag + rebound = 0.84. ^*/^** = P < 0.05/0.01 Dunn’s multiple comparisons post-hoc.

Figure 2

ReRh and HPC synapses are indistinguishable. (A) Average waveforms of ReRh and HPC EPSPs, stimulation denoted by triangle, electrical stimulation artifacts removed for clarity. Traces show mean ± SEM waveform, scale bars = 2 mV/50 ms. (B) Latency from stimulation to EPSP peak, box plot shows median, 25th and 75th percentiles, whiskers maxima and minima. Individual values shown as open circles. (Mann–Whitney test, U = 407, P = 0.077, n = 33 cells from 28 animals). (C) EPSP peak amplitude versus area of ReRh and HPC EPSPs. Linear regression slopes were not significantly different (F_(1,62) = 1.5, P = 0.23). (D) EPSP peak amplitude versus decay slope from 90% to 15% of EPSP peak. Linear regression slopes were not significantly different (F_(1,62) = 0.3, P = 0.57). (E) Representative −70-mV AMPAR-mediated (negative-going) and +40-mV NMDAR-mediated (positive-going) EPSCs resulting from ReRh (blue) and HPC (black) stimulation. (F) NMDAR:AMPAR ratios of ReRh and HPC EPSCs were not significantly different (Mann–Whitney U = 99, P = 0.60, n = 15 cells from 10 animals). (G) EPSC_AMPA versus EPSC_NMDA for ReRh and HPC inputs. Linear regression slopes were not significantly different (F_(1,26) = 1.05, P = 0.32). (H) EPSC_NMDA inhibition by bath application of GluN2B selective antagonist Ro25–6981 (3 μM) as indicated by gray shaded region. Mann–Whitney U = 12, P = 0.4. N = 6 cells from 5 animals.

Figure 3

Reuniens/rhomboid inputs to prelimbic cortex depress at theta frequency. (A) ReRh inputs to L5 pyramidal neurons undergo strong short-term depression, and show different plasticity pattern to HPC inputs at 5 Hz (repeated measures 2-way ANOVA: main effect of pathway F_(1,11) = 8.5, P = 0.014; main effect of response number F_(1.9,20.6) = 5.1; P = 0.018, interaction F_{(3.1, 34.1)} = 4.4, P = 0.0095) and 10 Hz (pathway F_(1,11) = 5.0, P = 0.048; response number F_(3.2,35.4) = 27.9, P = 1.1 × 10⁻⁹; interaction F_(4.3,47.0) = 8.5, P = 0.00002; Greenhouse–Geisser correction applied, both frequencies). N = 12 cells from 11 animals. Scale bars = 0.3 mV/200 ms. (B) ReRh inputs to L2/3 pyramidal cells show equal degree of short-term depression as inputs to L5 at 5 Hz (Repeated-measures 2-way ANOVA: main effect of layer F_(1,19) = 0.71, P = 0.41; main effect of response number F_(8,61.0) = 15.3, P = 2.6 × 10⁻¹⁶; interaction F_(3.2,61) = 0.45, P = 0.73) and 10 Hz (main effect of layer: F_(1,19) = 0.12, P = 0.73; main effect of response number F_(8,62.7) = 30.3, P = 6.8 × 10⁻²⁸; interaction F_(3.3,62.7) = 0.9, P = 0.44; Greenhouse–Geisser correction applied, both frequencies. L2/3 n = 9 cells from 5 animals; L5 data repeated from Fig 3A). (C) Following injection of AAV9-CaMKii-hChETA_TC-mCherry into intermediate/ventral HPC, acute mPFC slices were made and HPC-mPFC transmission evoked by electrical and optogenetic stimulation were compared. EPSPs evoked by ChETA_TC and electrical stimulation were of similar amplitude (paired t-test, t₍₁₂₎ = 0.78, P = 0.45). No difference in short-term plasticity was observed at 5 (main effect of stimulation method 5 Hz: F_(1,12) = 0.07, P = 0.79; main effect of response number F_(2.1,30.0) = 0.45, P = 0.65; interaction F_(3.1,36.9) = 2.2, P = 0.10) or 10 Hz stimulation frequency (stimulation method F_(1,12) = 0.38, P = 0.55; response number F_(2.1,24.7) = 4.6, P = 0.02, interaction F_(2.6,31.7) = 1.1, P = 0.38). Greenhouse–Geisser corrections applied. N = 13 cells from 5 animals.

Figure 4

ReRh and HPC are inhibited by group II mGluR activation, but high probability of release underlies short-term depression of ReRh inputs. (A) Activation of group II mGluRs with LY354740 (500 nM) reveals no difference in acute depression of ReRh and HPC inputs measured during the final 10 min of drug application (2-way repeated measures ANOVA: main effect of timepoint F_(2,20) = 13.3, P = 0.0002; pathway F_(1,10) = 0.02, P = 0.90; interaction F_(2,20) = 0.81, P = 0.46; Sidak’s post-hoc comparisons shows difference vs. baseline, ReRh P = 0.024, HPC P = 0.0009). N = 7 cells from 7 animals. (B) LY341495 (100 μM) did not affect short-term plasticity of ReRh input at 5 or 10 Hz (RM ANOVA; 5 Hz: main effect of drug F_(1,8) = 4.8, P = 0.059; main effect of response number F_(4.0,32.3) = 23.6, P = 3 × 10⁻⁹; interaction F_(4.4,35.2) = 0.85, P = 0.51; 10 Hz: main effect of drug F_(1,8) = 0.47, P = 0.51; response number F_(2.6,20.7) = 28.8, P = 2.9 × 10⁻⁷; interaction F_(2.5,20.2) = 2.6, P = 0.089). Greenhouse–Geisser corrections applied. N = 9 cells from 6 animals. (C) Activity-dependent block of isolated NMDA EPSCs by MK-801 (40 μM). Example blockade of EPSC_NMDA measured at +40 mV in ReRh and HPC pathway, traces colored by stimulus number in presence of MK-801, normalized to amplitude of first response, scale bars: 0.25 of normalized peak/100 ms. Plot of ReRh versus HPC amplitudes shows data lie above the identity line. Decay of ReRh is significantly faster than decay of HPC (single exponential curve constrained to Y0 = 1, ReRh τ = 5.4, HPC τ = 7.9, extra sum of squares F-test, F_(2,532) = 19.0, P < 0.0001). N = 10 cells from 8 animals.

Figure 5

Cholinergic modulation of hippocampal, but not nucleus reuniens/rhomboid inputs to prelimbic cortex via M2 muscarinic receptors. (A) Pooled data showing 10-min bath application of cholinergic agonist carbachol (shaded region) at different concentrations has no effect upon ReRh (left) input to PrL, but reversibly depresses HPC inputs (right) in a concentration-dependent manner. Example EPSPs for each pathway at baseline (−10 to −1 min), acute (10–19 min) and washout (40–49 min) shown from a representative 10 μM experiment (scale bars = 1 mV, 100 ms; n = cells/animals: 100 nM = 9/8; 1 μM = 6/4; 10 μM = 8/7). (B) Selective M1 muscarinic antagonist pirenzipine does not block depression of HPC by 10-μM CCh. Pirenzipine (1 μM) pre-applied during light shaded region and co-applied with CCh during dark shaded region. EPSPs from a representative cell are shown above pooled data, scale bars = 1 mV/100 ms. N = 7 cells from 6 animals. (C) Selective M2 muscarinic antagonist AF-DX 116 blocks depression of HPC by 10-μM CCh. AF-DX 116 (2 μM) pre-applied during light shaded region and co-applied with CCh during dark shaded region. EPSPs from a representative cell are shown above pooled data, scale bars = 1 mV/100 ms for ReRh and 2 mV/100 ms HPC. N = 5 cells from 5 animals. For plots of individual experiments and statistics please refer to Supplementary Fig. S5.

Figure 6

Dopaminergic modulation of basal ReRh and HPC inputs to mPFC. (A) D1R-like dopamine agonist SKF81297 bath applied at 0.5 μM caused a modest, reversible increase in transmission which was not pathway specific (2-way ANOVA, main effect of timepoint: F_(2,48) = 5.4, P = 0.0078; pathway F_(1,24) = 0.4, P = 0.54; interaction F_(2,48) = 1.8, P = 0.17). Values = mean ± SEM. Representative EPSPs at baseline, final 10 mins of drug application and final 10 min of recording, scale bars = 3 mV/100 ms. N = 13 cells from 13 animals. (B) SKF81297 at 10 μM did not result in a significant alteration of ReRh or HPC EPSPs (2-way ANOVA, main effect of timepoint F_(2,64) = 1.1, P = 0.34; main effect of pathway F_(1,32) = 0.07, P = 0.8, interaction F_(2,64) = 0.34, P = 0.71). Values = mean ± SEM. Representative EPSPs shown, scale bars = 3 mV/100 ms. N = 18 cells from 16 animals. (C) D2R-like dopamine agonist quinpirole (10 μM) does not affect basal ReRh or HPC transmission (2-way ANOVA, main effect of timepoint F_(2,20) = 0.05, P = 0.31, interaction F_(2,20) = 1.2, P = 0.31). Data shown are mean ± SEM, representative EPSPs (scale bars = 2 mV/100 ms), raw EPSP amplitudes for individual experiments. N = 6 cells from 6 animals.

Figure 7

Pairing of HPC and ReRh inputs repeated at theta frequency induces NMDA receptor-dependent, associative LTD. (A) Schematic diagram showing hypothesized tripartite circuit dynamics. Information arising in HPC (green pathway) may project directly to mPFC (solid arrow) and feed forward disynaptically via ReRh (dashed arrows), HPC EPSPs would therefore precede ReRh in mPFC resulting in negative lag. Conversely, a signal originating in ReRh may reach mPFC directly and disynaptically via HPC (purple) resulting in the opposite temporal activation profile. (B) Theta-frequency pairing of HPC and ReRh inputs at −70 mV (HPC stimulus preceding ReRh by 10 ms, 100 pairs delivered at 5 Hz, Vm = −70 mV) does not induce synaptic plasticity in either pathway (Follow up: plasticity measured at 30–40 min after pairing; paired t-test: ReRh t₍₆₎ = 0.6, P = 0.56; HPC t₍₆₎ = 0.8, P = 0.45). Traces show example averaged EPSPs at baseline (ReRh; blue [above]/HPC; black [below]) and 30–40 mins (follow-up; red) and the first 15 pairings (gray; middle traces) with HPC stimulation denoted by black and ReRh by blue triangles, respectively. Scale bars EPSPs: 5 mV, 100 ms, pairing: 5 mV, 50 ms. Right: EPSP amplitudes for individual experiments. N = 7 cells from 7 animals. (C) Depolarization to −50 mV during 5 Hz pairing (−10 ms delay) induces LTD of ReRh and HPC inputs to PFC. Traces as in B except scale = 2 mV in non-pairing traces. Example experiment traces show baseline and 30–40 min EPSPs. Note incidence of spiking during pairing protocol. Individual experiment EPSP sizes at baseline and final 10 min (Wilcoxon signed ranks: ReRh Z = −2.7, P = 0.008; HPC Z = −2.7, P = 0.008). Normalized final EPSP amplitudes for −10 ms delay pairing performed at −70 and −50 mV (2-way ANOVA, effect of membrane potential: F_(1,28) = 11.3, P = 0.0023, main effect of pathway: F_(1,28) = 0.005, P = 0.95, interaction: F_(1,28) = 0.14, P = 0.7). N = 9 cells from 9 animals. (D) Pairing with +10 ms lag (ReRh precedes HPC by 10 ms) with depolarization does not induce plasticity in either pathway (Wilcoxon signed ranks ReRh: Z = −0.14, P = 0.89, HPC: Z = −0.42, P = 0.67). Traces as in (B). Normalized final EPSP amplitudes for +10 ms and − 10 ms at −50-mV membrane potential (2-way ANOVA, main effect of pairing order: F_(1,30) = 5.2, P = 0.030, main effect of pathway: F_(1,30) = 0.12, P = 0.73, interaction: F_(1,30) = 0.0002, P = 0.99). N = 8 cells from 7 animals. (E) Bath application of NMDA receptor antagonist D-AP5 (50 μM), as indicated by gray shading, blocks induction of LTD by −10 ms pairing. Traces as in (B), except EPSP scale bars = 1 mV. (Wilcoxon signed ranks ReRh: Z = −0.84, P = 0.4; HPC Z = −1.3, P = 0.21). Normalized amplitudes for −50 mV pairing in absence or presence of D-AP5 (2-way ANOVA, main effect of drug: F_(1,.30) = 21.0, P < 0.0001, main effect of pathway: F_(1,30) = 0.9, P = 0.35, interaction: F_(1,30) = 2.6, P = 0.12). N = 8 cells from 8 animals.