Inhibitory Neurons Marked by the Connectivity Molecule Kirrel3 Regulate Memory Precision

Abstract

The homophilic adhesion molecule Kirrel3 drives synapse formation between dentate granule (DG) neurons and GABA neurons, and Kirrel3 gene variants are associated with neurodevelopmental disorders in humans. However, the circuit function and behavioral relevance of Kirrel3-expressing neurons are unknown. Using intersectional genetics, we identified a population of Kirrel3-expressing GABA neurons that regulate memory discrimination in male and female mice. Using chemogenetics with in vivo electrophysiology and behavioral assays, we discovered that activating Kirrel3-expressing GABA neurons, but not parvalbumin neurons, potently inhibits CA3 neuron activity and impairs contextual memory discrimination during recall, revealing a critical role for these neurons in the retrieval of precise memories. Light and electron microscopy of Kirrel3-expressing GABA neurons suggests that they receive direct excitation from DG neurons and project onto CA3 dendrites. Together, this multiscale approach demonstrates how cell type-specific expression of adhesion molecules mark subsets of neurons that control key features guiding memory and behavior.

Keywords: CA3; Kirrel3; inhibitory neuron; memory discrimination; memory generalization; mossy fiber synapse.

Extended Data Figure 2-1.

Summary of statistics for all experimental results

Extended Data Figure 3-1.

A) Examples of cues used for different contexts. B-D) Baseline metrics for K3-Flp;Gad2-Cre mice treated with either saline (gray) or DCZ (purple). n = 19 saline, 17 DCZ Each data point represents a mouse with males a triangle, females a circle. B) Fear acquisition curve for K3-Flp;Gad2-Cre mice shows the percent time spent freezing for 30 sec prior to each foot shock. Two-way ANOVA mixed effect test indicates a significant difference for shock number (indicating all mice conditioned normally) but no significant difference between treatment (saline vs DCZ). C) Baseline freezing equals the percent time freezing for each K3-Flp;Gad2-Cre mouse in the 3 minutes prior to the first foot shock. D) Baseline Activity Index shows overall motion of each K3-Flp;Gad2-Cre mouse in the 3 minutes prior to the first foot shock. E) Data from brain sections of Kirrel3-Flp; Gad2-Cre mice transfected with CV-hM3D virus. Plot indicates the percent area of hM3D-mCherry signal in CA3 versus outside CA3 as a measure of proper CA3 DREADD virus targeting and expression for each K3-Flp;Gad2-Cre mouse used in behavior experiments. Each row indicates a mouse, and each dot represents a brain section from that mouse. Error bars represent SEM. F-I) Data for WT WT mice treated with either saline (gray) or DCZ (cyan). n = 12 saline, 9 DCZ. Each data point represents a mouse with males a triangle, females a circle. F) shows % time freezing in each context, G) shows fear acquisition curves (two-way ANOVA mixed effect test indicates a significant difference for shock number indicating all mice conditioned normally but no significant difference between saline vs DCZ), H) shows the percent time spent freezing for 30 sec prior to each foot shock and I) shows the activity index for WT mice. See Extended Data Figure 2-1 for detailed statistics.

Extended Data Figure 4-1.

A) Confocal images of adult triple heterozygous mice (K3-Flp;Gad2-Cre;RC-FLTG) stained for GFP to label K3-GABA neurons (green) and the indicated GABA neuron marker (magenta). B) Table showing the percent of K3-GABA neurons that express the indicated marker as assessed by immunostaining triple transgenic mice. n = 3 female adult mice. The total number of neurons counted for each marker is indicated. C) Representative image of a hippocampal section from an adult PV-Cre mouse expressing Cre-dependent hM3D-mCherry (magenta) and stained with anti-PV antibodies (green) indicates infected neurons express PV as expected. Arrowheads (yellow) point to overlap (white). D) Representative image of a PV-Cre mouse with a CA3 injection of the Cre-dependent DIO-hM3D-mCherry AAV. E) Fear acquisition curve shows freezing percent (%) for 30 sec prior to each foot shock. Two-way ANOVA mixed effect test indicates a significant difference for shock number indicating all mice conditioned normally but no significant difference between saline vs DCZ. F) Baseline freezing and G) activity index prior to foot shock for PV-Cre mice. For E-G, n = 19 saline, 16 DCZ. Error bars represent SEM. Each data point represents a mouse with males a triangle, females a circle. See Extended Data Figure 2-1 for detailed statistics.

Extended Data Figure 5-1.

A) Fear acquisition curve shows freezing percent (%) for 30 sec prior to each foot shock for K3-GABA mice treated with saline or DCZ prior to the conditioning context only. Two-way ANOVA mixed effect test indicates a significant difference for shock number indicating all mice conditioned normally but no significant difference between saline vs DCZ. B) Baseline freezing and C) activity index prior to foot shock for animals in A. n = 15 saline, 10 DCZ. Error bars represent SEM, no significant differences as indicated. D) Fear acquisition curve shows freezing percent (%) for 30 sec prior to each foot shock for K3-GABA mice treated with saline or DCZ prior to Contexts A and C only on day 2. Two-way ANOVA mixed effect test indicates a significant difference for shock number indicating all mice conditioned normally but no significant difference between saline vs DCZ. E) Baseline freezing and F) activity index prior to foot shock for animals in D. n = 11 saline, 14 DCZ. Error bars represent SEM, no significant differences as indicated. In all cases, each data point represents a mouse with males a triangle, females a circle. See Extended Data Figure 2-1 for detailed statistics.

Figure 1.. Generation and validation of a Kirrel3-Flp driver mouse line.

A) Schematic of the Flp-2A insertion site in the mouse Kirrel3 gene. B) Western blot of brain lysates from 2 months old mice show that the Flp transgenic line expresses normal levels of Kirrel3 protein. Lysates from Kirrel3 knockout (KO) and wildtype (WT) are used as controls. Coomassie stained membrane indicates equal protein loading. C) A hippocampal section from an adult Kirrel3-Flp;Gad2-Cre;RC-FLTG triple transgenic mouse (heterozygous for all transgenes indicated) stained with antibodies against TdTomato (red) and GFP (green). D) A magnified image of a hippocampal section labeled by fluorescent in situ hybridization (FISH). Arrows indicate K3-GABA neurons co-expressing Kirrel3, GFP, and GAD1 mRNA. E) Quantification of FISH as indicated. Each column represents a mouse and each square represents one hippocampal section from that mouse. Average percentages from all data points are shown with a dotted line. n = 7 adult mice (3 male and 4 female) Error bars represent SEM.

Figure 2:. Intersectional genetics enables specific targeting and activation of K3-GABA neurons.

A) Schematic of the ConVERGD (CV)-hM3D-mCherry AAV construct. B) Representative hippocampal images from an adult mice (genotypes indicated) infected with CV-hM3D-mCherry targeted to area CA3. mCherry is only present when both Flp and Cre are expressed (left). C) Magnified image from a hippocampal section processed for dual FISH/immunohistochemistry from a Kirrel3-Flp;Gad2-Cre mouse infected with CV-hM3D-mCherry. D) Schematic of the experimental design for E and F. E) Representative images showing cFos (green) in K3-GABA neurons expressing hM3D-mCherry (magenta) specifically after DCZ injection (right). Yellow arrowheads indicate K3-GABA neurons that express cFos (merge is white). F) Quantification of the % of DREADD-expressing K3-GABA neurons that express cFos for saline and DCZ treated mice after foot shock. n = 4 males each. Each column represents one mouse, up to 3 sections and 6 hippocampi counted per mouse. Average from all data points is shown with a dotted line. Error bars represent SEM, nested t-test. See Extended Data Figure 2-1 for detailed statistics.

Figure 3.. K3-GABA activation impairs memory discrimination.

A) Schematic of behavioral paradigm and experimental design for K3-Flp;Gad2-Cre mice. B) Time spent freezing (%) when placed in indicated contexts after saline (gray) or DCZ (blue) injection. All mice were adult K3-Flp;Gad2-Cre heterozygotes injected with CV-hM3D-mCherry. n = 19 saline, 17 DCZ mice. C) Context discrimination ratios for K3-GABA mice on day 1 (A vs B). mean +/− SEM, unpaired t-test. D) Context discrimination ratios for K3-GABA mice on day 2 (A vs C). mean +/− SEM, unpaired t-test. E) Experimental design for wildtype (WT) no DREADD control mice. F, G) Discrimination ratios from WT mice for contexts A vs B (F) or contexts A vs C (G). n = 12 saline; 9 DCZ. mean +/− SEM, unpaired t-test. In all graphs, males are represented with a triangle, females a circle. See Extended Data Figure 3-1 for AAV targeting analysis and additional behavior data including fear acquisition curves, baseline freezing, and baseline activity metrics. See Extended Data Figure 2-1 for detailed statistics.

Figure 4.. Activating parvalbumin-expressing neurons does not impair memory discrimination.

A) Example sections from adult PV-Cre mice expressing the Cre-dependent DIO-hM3D-mCherry AAV and stained for cFos (green). B) % of hM3D-expressing PV neurons that are cFos positive after saline (gray) and DCZ injection (red) following foot shock. n= 6 mice each condition, global average indicated by dotted line, nested t-test. C) Experimental design for PV-Cre mice. D) Time spent freezing (%) when placed in indicated contexts after saline (gray) or DCZ (red) injection. All mice were adult PV-Cre heterozygous mice injected with Cre-dependent hM3D-mCherry. n= 19 saline, 16 DCZ. E,F) Data from D plotted as discrimination ratios for contexts A vs B (E) or contexts A vs C (F). Error bars represent SEM, t-tests. Each data point represents a mouse with males a triangle, females a circle. See Extended Data Figure 4-1 for additional data related to the lack of overlap between Kirrel3 and PV neurons and additional acquisition and baseline behavioral data. See Extended Data Figure 2-1 for detailed statistics.

Figure 5.. K3-GABA activation during recall impairs memory discrimination.

A) Experimental design showing saline or DCZ injection only prior to foot shock conditioning in context A in K3-Flp;Gad2-Cre heterozygotes mice. B) Mouse behavior plotted as percent time spent freezing when placed in indicated contexts. Here, DCZ refers to the mice that received DCZ at conditioning. They were not treated with DCZ prior to each context. n= 15 saline, 10 DCZ mice. C,D) Data from B plotted as discrimination ratios for contexts A vs B (C) or contexts A vs C (D). Error bars indicate SEM, t-tests. E) Experimental design showing saline or DCZ injection only prior to contexts A and C on day 2 in K3-Flp;Gad2-Cre heterozygotes mice. F) Mouse behavior plotted as percent time spent freezing for when placed in indicated contexts. Here, DCZ refers to the mice that received DCZ on day 2 only. They were not treated with DCZ on other days. n= 11 saline, 14 DCZ mice. G,H) Data from F plotted as discrimination ratios for contexts A vs B (G) or contexts A vs C (H). Error bars indicate SEM, t-tests. In all graphs, each data point represents a mouse with males a triangle, females a circle. See Extended Data Figure 5-1 for additional acquisition and baseline behavioral data. See Extended Data Figure 2-1 for detailed statistics.

Figure 6.. K3-GABA neuron activity strongly inhibits CA3 neuron spike rate.

A) Experimental design for in vivo recordings. B) Raster plots of representative single unit CA3 recordings from mice before and after saline or DCZ injection. 10 units each graph. C) Mean firing rate of CA3 single units for each recording session for hM3D-mCherry infected K3-Flp;Gad2-Cre mice treated with saline or DCZ. n=5 saline; 8 DCZ, paired t-test. D) Mean running speed for hM3D-mCherry infected K3-Flp;Gad2-Cre mice treated with saline or DCZ. n=5 saline; 8 DCZ, paired t-test. E) Mean firing rate of CA3 single units for each recording session for hM3D-mCherry infected PV-Cre mice treated with saline or DCZ. n=8 saline; 11 DCZ, paired t-test. F) Mean running speed for hM3D-mCherry infected PV-Cre mice treated with saline or DCZ. n=8 saline; 11 DCZ, paired t-test. G) Firing rates were normalized as a percent change from saline for each genotype. Graph summarizing and comparing data from C and E. ANOVA with post-tests. H) Average number of DREADD expressing neurons per section from each mouse used for in vivo recording experiments in B-G. nested t-test. See Extended Data Figure 2-1 for detailed statistics.

Figure 7.. Activating K3-GABA neurons specifically decreases cFos expression in area CA3.

A) Schematic of DG-to-CA3 connectivity and the potential connectivity of K3-GABA neurons to mediate either direct inhibition to CA3 or indirect inhibition via DG. B) Experimental design for C-I. C) Percent of DREADD-expressing K3-GABA neurons that are cFos positive. n=6 saline, 7 DCZ mice. nested t-test. D-G) Relative density (normalized to saline controls) of cFos-positive cells in CA3, DG, CA1, and hilus in hM3D-mCherry infected K3-Flp;Gad2-Cre mice treated with saline or DCZ. n=6 saline, 7 DCZ mice. nested t-test. H) Percent of DREADD-expressing PV neurons that are cFos positive. n=4 saline, 4 DCZ mice. nested t-test. I) Relative density (normalized to saline controls) of cFos-positive cells in CA3 in hM3D-mCherry infected PV-Cre mice treated with saline or DCZ. n=4 saline, 4 DCZ mice. nested t-test. For all graphs, each point represents a hippocampus section, each column represents a mouse, males a triangle, females a circle, and the average for all data points in a condition is shown as a dotted line. In all graphs, blue data points indicate analyses done in mice with K3-GABA neurons activated and red data points indicate mice with PV neurons activated. Error bars show SEM. See Extended Data Figure 2-1 for detailed statistics.

Figure 8.. K3-GABA neurons receive direct input from DG mossy fiber boutons.

A,B) Hippocampal section from an 84 days old Kirrel3-Flp;Gad2-Cre mouse injected with an AAV that mediates expression of a Flp and Cre-dependent GFP. Kirrel3-GABA neurons are green, the DRAQ5 nuclear stain is magenta. The arrow points to a Kirrel3-GABA neuron shown magnified in B. C) The same Kirrel3-GABA neuron identified in B is outlined in green in an EM view. D) A 3D reconstruction of the identified K3-GABA dendrite (green). Gray planes note the top and bottom of the SL layer. DG axons with mossy fiber boutons that synapse onto the Kirrel3-GABA dendrite are orange, DG axons with en passant synapses are blue and other axons are gray. E,F) A magnified view of the Kirrel3-GABA dendrite and innervating axons. Colors are the same as in D. The arrow points to a CA3 thorny excrescence spine (pink) G) EM image of the Kirrel3-GABA dendrite (green) receiving multiple inputs from two mossy fiber boutons (orange). A nearby CA3 neuron and its associated multi-headed thorny excrescence spine are shown (pink). Arrowheads indicate synaptic densities from mossy fiber boutons to the Kirrel3-GABA dendrite. H) A less magnified EM image showing axons connected via large mossy boutons (orange) and en passant synapses (blue). Large bundles of DG axons can be seen (asterisk). I) Hippocampal section from an adult Kirrel3-Flp;Gad2-Cre mouse infected with a Flp and Cre-dependent GFP AAV. K3-GABA neurons express GFP (green), Hoechst nuclear stain (blue). J) Higher magnification image of a Kirrel3-Flp;Gad2-Cre mouse sparsely infected with a Flp and Cre-dependent GFP AAV. Note Kirrel3-GABA axon labeling primarily in the SR and SO layers. Hoechst nuclear stain (blue). K) Hippocampal section from a PV-Cre mouse infected with a Cre-dependent AAV driving expression of mCherry. PV neurons express mCherry (magenta) and Hoechst nuclear stain (blue). SO; stratum oriens, SP; stratum pyramidale, SL; stratum lucidum, SR; stratum radiatum