The hippocampal CA2 region is essential for social memory

Abstract

The hippocampus is critical for encoding declarative memory, our repository of knowledge of who, what, where and when. Mnemonic information is processed in the hippocampus through several parallel routes involving distinct subregions. In the classic trisynaptic pathway, information proceeds from entorhinal cortex (EC) to dentate gyrus to CA3 and then to CA1, the main hippocampal output. Genetic lesions of EC (ref. 3) and hippocampal dentate gyrus (ref. 4), CA3 (ref. 5) and CA1 (ref. 6) regions have revealed their distinct functions in learning and memory. In contrast, little is known about the role of CA2, a relatively small area interposed between CA3 and CA1 that forms the nexus of a powerful disynaptic circuit linking EC input with CA1 output. Here we report a novel transgenic mouse line that enabled us to selectively examine the synaptic connections and behavioural role of the CA2 region in adult mice. Genetically targeted inactivation of CA2 pyramidal neurons caused a pronounced loss of social memory--the ability of an animal to remember a conspecific--with no change in sociability or several other hippocampus-dependent behaviours, including spatial and contextual memory. These behavioural and anatomical results thus reveal CA2 as a critical hub of sociocognitive memory processing.

Figure 1. Genetic targeting of the CA2 subfield using the Amigo2-Cre mouse line

a, Bilateral hippocampal injection (n = 64) of Cre-dependent YFP AAV in Amigo2-Cre mice resulted in specific expression of YFP (green) in CA2 PNs. b, Extent of transduction. Left, adapted reference atlas images. Center, YFP expression. Right, mm from bregma along rostrocaudal axis. c-g, Magnified images of boxed area in (b). c, YFP (green). d, RGS14 staining (red, n = 4). e, Merge of (c) and (d) showing YFP and RGS14 overlap. f, GABA staining (red, n = 3). g, merge of (c) and (f) showing no GABA and YFP overlap. Panels show coronal sections with Nissl counterstain (blue). Scale bars, 1000 μm, 400 μm, 200 μm in (a), (b), (c-g), respectively.

Figure 2. Genetically targeted tracing of the CA2 circuit

a-g, Monosynaptic inputs to CA2 revealed with pseudotyped rabies virus (n = 8). Cells labeled with rabies (magenta); Nissl (green). Sagittal sections (a-d) and coronal sections (e-g). a, b, Labeled neurons in CA2 and CA3 ipsilateral (a) and contralateral (b) to hemisphere of rabies virus injection. Rabies labeling shows monosynaptic inputs from lateral EC (c), medial EC (d), medial septum (MS), nucleus of the diagonal band (NDB) (e), median raphe (MR) (f), and lateral supramammillary nucleus (SUMl) (g). Fluorescent processes in (c,d) may represent dendritic or axonal labeling. h, i, Output of CA2 revealed by axonal YFP signal (green, n = 6). Nissl stain (magenta). i, Magnification of boxed area in (h). Note strong labeling of CA2 projections to SO and SR of CA1. Scale bars, 200 μm. slm, stratum lacunosum-moleculare.

Figure 3. Electrophysiological verification of CA2 inactivation with tetanus toxin

a, Experimental setup for photostimulation of CA2 PNs. b, c, Action currents recorded from CA2 PNs expressing YFP and ChR2 (n = 6) (b) or TeNT and ChR2 (n = 4) (c) in response to five 2-ms blue (470 nm) light pulses (blue bars). d, Experimental setup for current-clamp recordings of photostimulated PSPs in CA1 PNs. e, PSPs recorded when YFP (n = 14, green) or TeNT (n = 14, magenta) was co-expressed with ChR2 in CA2 PNs. f, Mean input-output curve of PSP as function of light intensity when YFP or TeNT was co-expressed with ChR2 in CA2 PNs. Data show mean ± s.e.m.

Figure 4. Inactivation of CA2 pyramidal neurons abolishes social memory

a, Left, sociability test. Middle, Both YFP (n = 11) and TeNT (n = 13) groups preferred the chamber with a littermate (YFP, P = 0.0083; TeNT, P = 0.0055; multiplicity adjusted P values) and did not differ significantly (two-way ANOVA: Treatment × Chamber F(1,44) = 0.013, P = 0.91; Treatment F(1,44) = 1.566, P = 0.22; Chamber F(1,44) = 17.49, P = 0.0001). Right, The groups had similar interaction time difference scores (littermate – empty) (P = 0.9154, two-tailed t-test). b, Left, social novelty test. Middle, The YFP, but not the TeNT, group preferred the novel animal (YFP, P = 0.0012; TeNT, P = 0.3593; multiplicity adjusted P values); the groups differed significantly (two-way ANOVA: Treatment × Chamber F(1,44) = 11.25, P = 0.0016). Right, TeNT group showed a significantly lower difference score (novel – littermate) than the YFP group (P = 0.0109, two-tailed t-test). c, d, Left, direct interaction test using the same (c) or different (d) stimulus animals in two trials. c, Middle, The YFP, but not the TeNT, mice displayed decreased investigation of a familiar stimulus mouse during trial 2 (YFP, n = 15, P < 0.0001; TeNT, n = 16, P = 0.1499; multiplicity adjusted P values); the two groups differed significantly (two-way RM ANOVA: Treatment × Trial F(1,29) = 24.23, P < 0.0001). Right, difference score (trial 1 – trial 2) of TeNT group was less than that of YFP group (P < 0.0001; two-tailed t-test). d, Middle, The two groups explored the two different stimulus animals for similar amounts of time (two-way RM ANOVA: Treatment × Trial F(1,29) = 0.0068, P = 0.93; Treatment F(1,29) = 2.405, P = 0.13; Trial F(1,29) = 3.278, P = 0.0806), with similar difference scores (Right, P = 0.93, two-tailed t-test). e, 5-trial social memory assay. The YFP group (n = 15), but not the TeNT group (n = 14), habituated to repeated presentation of the same stimulus mouse (trials 1-4) and dishabituated to the novel mouse (trial 5). Two-way RM ANOVA confirmed a significant difference between the groups (Treatment × Trial F(4,108) = 7.26, P < 0.0001; Treatment F(1,27) = 7.86, P = 0.009; Trial F(4,108) = 15.41, P < 0.0001). Data show mean ± s.e.m.

Extended Data Figure 1. Generation of Amigo2-Cre mouse line

λ Red-mediated homologous recombination with galK positive and negative selection was used to make seamless changes to the bacterial artificial chromosome (BAC). PCR cassettes shown in orange, and Amigo2 locus shown in blue. The PCR cassette contained two homology arms (H1, 58nt; H2, 62nt) that flanked the galactose kinase (galK) cassette. The homology arms flanked the Amigo2 start codon. Recombination followed by positive selection was used to obtain the galK integrate. Recombination of the modified BAC with a PCR cassette containing the Cre open reading frame (ORF) and polyA (PA) flanked by the same homology arms yielded the final BAC used to generate the transgenic line.

Extended Data Figure 2. Amigo2-Cre mice express Cre in a genetically defined population of CA2 PNs

Coronal sections of hippocampus from Amigo2-Cre mice injected in dorsal hippocampus with a Cre-dependent AAV to express YFP (shown in green) in CA2. a, Coronal section of ventral hippocampus (~2.8 mm caudal to bregma, see Figure 54 of Franklin & Paxinos for reference image) showing CA2 axons (green) from dorsal CA2. Note absence of YFP in ventral CA2 neurons (RGS14 stain in red). b, 97.22 ± 0.46% of YFP⁺ cells (n = 4 mice, 2948 cells) express the CA2 marker PCP4 (red). c, 98.45 ± 0.33% of YFP⁺ cells (n = 4 mice, 2870 cells) express the CA2 marker STEP (red). d, Nearly no YFP⁺ cells (0.17 ± 0.13%; n = 4 mice, 2870 cells) express the CA1 marker WFS1 (red). e-f, Magnification of boxed area in (b) showing YFP signal (e) PCP4 staining (f) and a merge of the two (g). h-j, Magnification of boxed area in (c) showing YFP signal (h) STEP staining (i) and a merge of the two (j). k-m, Magnification of boxed area in (d) showing YFP signal (k) WFS1 staining (l) and a merge of the two (m). Nissl stain shown in blue. Scale bars, 400 μm (a-d) and 100 μm (e-m).

Extended Data Figure 3. Amigo2-Cre mice express Cre in RGS14⁺ CA2 PNs but not in GABAergic inhibitory neurons

Cre⁺ neurons expressing YFP (shown in green) co-label with RGS14 staining (shown in red), but do not co-label with GABA staining (shown in red in separate images). a, Reproduction of section −1.06 mm shown in Fig. 1b. b, e, Magnification of area boxed in (a). c, RGS14 staining of section shown in (b). d, Merge of (b, c) demonstrating YFP and RGS14 co-labeling. f, GABA staining of section shown in (e). g, Merge of (e, f) showing no overlap of GABA and YFP. h, Reproduction of section −1.46 mm shown in Fig. 1b. i, l, Magnification of area boxed in (h). j, RGS14 staining of section shown in (i). k, Merge of (i, j) demonstrating YFP and RGS14 co-labeling. m, GABA staining of section shown in (l). n, Merge of (l, m) showing no overlap of GABA and YFP. o, Reproduction of section −2.18mm shown in Fig. 1b. p, s, Magnification of area boxed in (o). q, RGS14 staining of section shown in (p). r, Merge of (p, q) demonstrating YFP and RGS14 co-labeling. t, GABA staining of section shown in (s). u, Merge of (s, t) showing no overlap of GABA and YFP. Scale bars, 200 μm. Nissl stain shown in blue.

Extended Data Figure 4. Specificity of the pseudotyped rabies virus

a, b, No labeled cells were observed (n = 3 mice) following injection of the (EnvA)SAD-ΔG-mCherry virus when TVA was not expressed in CA2. b, Magnification of boxed area in (a). Rabies labeling would have appeared in magenta; Nissl stain shown in green. Scale bars, 200 μm.

Extended Data Figure 5. Inactivation of CA2 does not alter locomotor activity or anxiety-like behavior

a, There was no significant difference (P = 0.31, two-tailed unpaired t-test) between CA2-YFP and CA2-TeNT groups in the distance traveled in the open field (OF) test (YFP, 53.14 ± 4.62m, n = 8; TeNT, 47.04 ± 3.70m, n = 10). b, There was also no significant difference (P = 0.55, two-tailed unpaired t-test) between the groups in the number of rearing events recorded during the OF session (YFP, 378.0 ± 17.36, n = 8; TeNT, 354.7 ± 30.99, n = 10). c, d, Inactivation of CA2 did not alter anxiety-like behavior measured in the elevated plus maze (EPM). The number of open arm entries was not significantly different (P > 0.99, two- tailed unpaired t-test) between the groups (YFP, 14.00 ± 1.46, n = 8; TeNT, 14.00 ± 1.54, n = 10). Additionally, the time spent in the open arms (YFP, 163.7 ± 10.43s n = 8; TeNT, 155.1 ± 16.38s n = 10) did not differ significantly (P = 0.68, two-tailed unpaired t-test) between the groups. Results are presented as mean ± s.e.m.

Extended Data Figure 6. Spatial learning and memory assayed with the Morris water maze (MWM) task is unaltered by CA2 inactivation

a, Schema of the experimental design. On days 1 and 2, mice were trained to find a platform with a visible flag. On days 3-7, mice were trained to find a hidden platform located in the SW quadrant of the water maze. Spatial memory was assayed on day 8 with the platform removed. Reversal training was conducted on days 9-13 with the platform now hidden in the NW quadrant. Spatial memory of the novel location was tested on day 14. b, Path length to the platform was not altered significantly by CA2 inactivation (two-way repeated measures ANOVA: Treatment × Time F(11,770) = 0.67, P = 0.77; Time F(11,770) = 21.87, P < 0.0001; Treatment F(1,70) = 2.85, P = 0.10). c, Latency to find the platform did not differ significantly between the two groups (two-way repeated measures ANOVA: Treatment × Time F(11,770) = 0.78, P = 0.66; Time F(11,770) = 25.23, P < 0.0001; Treatment F(1,70) = 2.84, P = 0.10). YFP, n = 8; TeNT, n = 10. d, Spatial memory during the probe trial was unaffected by CA2 inactivation. The percent of time spent in the target quadrant (YFP, 33.00 ± 2.66%; TeNT, 38.6 ± 4.79%) was not significantly different between the two groups (P = 0.36, two-tailed unpaired t-test). e, Spatial memory following reversal training was unaffected by CA2 inactivation. There was no significant difference between the groups in percent time spent in the target quadrant during the probe trial following reversal training (YFP, 36.38 ± 5.75%; TeNT, 36.40 ± 2.92%; P > 0.99, two-tailed unpaired t-test). Results are presented as mean ± s.e.m.

Extended Data Figure 7. Contextual fear conditioning memory and auditory fear conditioning memory are unaffected by inactivation of CA2

a, Schema of the experimental design. Delay fear conditioning was employed to test hippocampal-dependent contextual fear memory and amygdala-dependent auditory fear memory. b, There was no significant difference in percent freezing between the groups (two-way repeated measures ANOVA: Treatment × Day F(4,68) = 0.31, P = 0.87; Treatment F(1,17) = 0.13, P = 0.73; Day F(4,68) = 100.8, P < 0.0001; YFP, n = 11; TeNT, n = 8). Prior to training on day 1, neither group exhibited a fear response to context A (YFP, 2.45 ± 1.06%; TeNT, 0.75 ± 0.49%) or to the tone (YFP, 3.09 ± 1.31%; TeNT, 1.63 ± 0.84%). On day 2 after training, robust fear responses to context A were measured in both groups (YFP, 24.09 ± 2.88%; TeNT, 26.00 ± 4.10%). Both groups exhibited low levels of freezing on day 3 in novel context B (YFP, 6.55 ± 1.52%; TeNT, 4.00 ± 0.87%) demonstrating context specificity of the fear memory and a lack of fear generalization. Both groups exhibited robust freezing to the tone on day 3 (YFP, 35.82 ± 4.93%; TeNT, 34.63 ± 3.96%), demonstrating intact auditory fear memory. c, Freezing data plotted in 30s bins. Shaded areas represent tone presentation. Red line represents shock delivery. Left, two-way repeated measures ANOVA revealed no significant difference between groups in freezing on day 1 (Treatment × Time F(6,102) = 1.135, P = 0.3474; Treatment F(1,17) = 1.116, P = 0.3056; Time F(6,102) = 6.348, P < 0.0001). Middle, two-way repeated measures ANOVA revealed no significant difference between groups in freezing on day 2 (Treatment × Time F(9,153) = 0.9741, P = 0.4637; Treatment F(1,17) = 0.1326, P = 0.7203; Time F(9,153) = 6.335, P < 0.0001). Right, two-way repeated measures ANOVA revealed no significant difference between groups in freezing on day 3 (Treatment × Time F(7,119) = 0.2490, P = 0.9716; Treatment F(1,17) = 0.6517, P = 0.4307; Time F(7,119) = 50.87, P < 0.0001). Results are presented as mean ± s.e.m.

Extended Data Figure 8. Object recognition memory and preference for novelty is preserved in CA2-TeNT animals

a, Schema of the experimental design for the novel object recognition task. b, The groups did not differ significantly in exploration of object 1 (YFP, 16.75 ± 1.57s; TeNT, 19.60 ± 2.24s) or object 2 (YFP, 16.50 ± 1.97s; TeNT, 15.90 ± 1.66s) averaged over the course of the first 4 trials (two-way ANOVA: Treatment × Object F(1,32) = 0.80, P = 0.38; Object F(1,32) = 1.05, P = 0.31; Treatment F(1,32) = 0.34, P = 0.56; YFP, n = 8; TeNT, n = 10). c, Both groups explored the novel object (YFP, 21.23 ± 2.37s; TeNT, 24.37 ± 2.81s) more than the familiar object (YFP, 7.41 ± 0.92s; TeNT, 8.57 ± 1.48s). Statistical analysis revealed a significant effect of object, but not CA2 inactivation or interaction of the two (two-way ANOVA: Treatment × Object F(1,28) = 0.22, P = 0.64; Object F(1,28) = 48.46, P < 0.0001; Treatment F(1,28) = 1.02, P = 0.32). Multiple comparison testing revealed a significant difference between exploration of the novel object compared to exploration of the old object for both the YFP group (P = 0.0002) and the TeNT group (P < 0.0001). d, Schema of the experimental design for another variation of the novel object recognition task. e, The groups did not differ significantly in time spent exploring object 1 (YFP, 21.50 ± 2.31s; TeNT, 22.18 ± 3.57s) or object 2 (YFP, 22.02 ± 2.23s; TeNT, 22.36 ± 2.81s) during trial 1 of day 4 (two-way ANOVA: Treatment × Object F(1,44) = 0.004, P = 0.95; Object F(1,44) = 0.02, P = 0.90; Treatment F(1,44) = 0.03, P = 0.85; YFP, n = 12; TeNT, n = 12). f, Both groups explored the novel object (YFP, 21.49 ± 1.91s; TeNT, 22.73 ± 1.82s) more than the familiar object (YFP, 13.74 ± 1.83s; TeNT, 16.53 ± 1.64s). Statistical analysis revealed a significant effect of object, but not CA2 inactivation or interaction of the two (two-way ANOVA: Treatment × Object F(1,44) = 0.18, P = 0.67; Object F(1,44) = 15.02, P = 0.0004; Treatment F(1,44) = 1.25, P = 0.27). Multiple comparison testing revealed a significant difference between exploration of the novel object compared to exploration of the old object for both the YFP group (P = 0.008) and the TeNT group (P = 0.02). Results are presented as mean ± s.e.m.

Extended Data Figure 9. Olfaction is unaffected by CA2 inactivation

a, There was no significant difference between the groups in latency to find a buried food pellet (YFP, 63.93 ± 8.22s, n = 15; TeNT, 67.06 ± 9.42s n = 16; P = 0.81, two-tailed unpaired t-test). b, There was no significant difference between the groups (YFP, n = 15; TeNT, n = 14) in performance on the olfactory habituation/dishabituation task (two-way repeated measures ANOVA: Treatment x Trial F(11,297) = 0.933, P = 0.51; Treatment F(1,27) = 0.08, P = 0.78; Trial F(11,297) = 60.21, P < 0.0001). Results are presented as mean ± s.e.m.