Abstract
Social recognition memory (SRM) is a key determinant of social interactions. While the cerebellum emerges as an important region for social behavior, how cerebellar activity affects social functions remains unclear. We selectively increased the excitability of molecular layer interneurons (MLIs) to suppress Purkinje cell firing in the mouse cerebellar vermis. Chemogenetic perturbation of MLIs impaired SRM without affecting sociability, anxiety levels, motor coordination or object recognition. Optogenetic interference of MLIs during distinct phases of a social recognition test revealed the cerebellar engagement in the retrieval, but not encoding, of social information. c-Fos mapping after the social recognition test showed that cerebellar manipulation decreased brain-wide interregional correlations and altered network structure from medial prefrontal cortex and hippocampus-centered to amygdala-centered modules. Anatomical tracing demonstrated hierarchical projections from the central cerebellum to the social brain network integrating amygdalar connections. Our findings suggest that the cerebellum organizes the neural matrix necessary for SRM.
Similar content being viewed by others
Introduction
Social behavior, defined as interactions among conspecifics, is evolutionarily conserved and critical for survival of both individuals and species. Social behavior involves processes that detect, store, and respond to social stimuli, which requires concerted actions of multiple brain areas including the medial prefrontal cortex (mPFC), anterior cingulate cortex (ACC), hippocampus, and amygdala1,2,3. Damage to any of these regions can affect social behavior, making it susceptible to pathological conditions, e.g., Alzheimer’s disease, schizophrenia, and autism spectrum disorder (ASD)4,5,6. Despite its significance, our understanding of the neural basis for social behavior is incomplete.
The cerebellum, which is not typically considered to be part of the social brain1,2,3, has emerged as a key node in an integrative network that links diverse functions of sense, motor, emotion, cognition and working memory7,8,9. In the cerebellar cortex, Purkinje cells (PCs) receive excitatory and inhibitory inputs and carry the sole output. While synaptic excitation facilitates PC firing, the PC firing activity is strongly controlled by feedforward inhibition from molecular layer interneurons (MLIs)10. Nearly all PCs send GABAergic axons to inhibit the cerebellar nuclei (CN)11. Neurons in the CN directly project to subcortical regions (first-order targets) such as the ventral thalamus (vTH). Via the first-order neurons, the CN connects to second-order targets virtually in the entire neocortex12. The neocortex, in turn, sends input back to the cerebellar cortex via the pontine nuclei. Through these feedback loops, the cerebellum is viewed as an estimator of the internal state for the operation of movement and cognition13. The diverse functionality of the cerebellum is thought to come from the topographical organization of the cerebello-cortical circuits. Functional magnetic resonance imaging (fMRI) of the human brain has revealed that lobules I-V and VIII are involved in sensorimotor tasks via interactions with the somatosensory and motor cortices; lobules VI-VII in social cognition, working memory, and language via interactions with the prefrontal, temporal and parietal cortices; lobules IX and X (flocculus) in eye movement and balance via interactions with the vestibular system14,15. However, these functions are generalized because motor and non-motor dichotomy of certain lobules can occur16. The anterior and posterior lobes are further divided into a central vermis and lateral hemispheres. The vermis and its projections to the fastigial nuclei (FN) in the medial CN receive input from the spinocerebellar tract and send output to the brainstem, thalamus, and cerebral cortex. In addition to motor coordination and postural control, a characteristic function of the vermis is to regulate emotion via interactions with the limbic circuits17. A pioneer study demonstrated that stimulating the vermis, but not the hemispheres, generated responses in the limbic regions in animals18. Work with neuronal tracing further shows the FN directly projects to the thalamus and hypothalamus19,20,21,22, while its links to the hippocampus and amygdala may be indirect21. From the thalamus, the FN connects to the limbic cortex including the mPFC and ACC21,22. This provides a structural base for the vermis-dependent aggression, spatial memory, and social association23,24,25.
Social interactions rely on intentions, emotions, and memories of social encounters, covering distinct yet interactive processes of social cognition and social memory1,2,3. Social cognition is the ability to imitate the actions of others (mirroring) and to understand the mental states of oneself and others (mentalizing)1,2,3. Social memory (or social recognition memory, SRM) is the ability to distinguish familiar from novel conspecifics by recalling previous encounters; it is supported by the integrity of the mPFC, ACC, hippocampus, and amygdala26,27,28,29,30. fMRI studies suggest the human cerebellum partakes in the neural networks for social cognition31,32. While there is no direct evidence for its involvement in SRM, the cerebellum is known to encode procedural memories, most notably eyeblink conditioning33. Early reports indicate MLIs contribute to the consolidation of learned eyeblink response by suppressing PC activity and releasing the CN from tonic PC inhibition to enable the eyeblink reflex34,35,36,37. Blocking MLI-PC synaptic transmission or excitation of MLIs in the floccular lobe alters the adaptation of the vestibulo-ocular reflex38,39. Inhibiting MLIs in lobule VI affects reverse learning and in lobule VII affects novelty-seeking behavior in juvenile mice, signifying a role of MLIs in cognitive and social development beyond the motor domain37. Interestingly, postmortem analysis of ASD patients shows an increase of GAD67 mRNA in MLIs, implying dysregulated feedforward inhibition to PCs40. This is confirmed in various ASD mouse models, where low PC activity, at least partially caused by MLI over-inhibition, leads to social impairments41,42,43,44. However, it remains to be determined how the cerebellum transforms such abnormal local activity into social behavior deficits.
Using chemogenetic and optogenetic techniques, we mimicked the dysregulated MLI-PC inhibition observed in ASD patients40 and mouse models42,43,44 by selectively increasing the excitability of MLIs to reduce PC firing in the vermis. Targeting MLIs also helped minimize potential changes in movement25 or aggression23 that could confound the analysis of social interactions45, if PCs were directly silenced. We evaluated the effects of this manipulation on mouse behaviors. As it specifically affected SRM, we conducted functional mapping of the cerebello-cortical axis by measuring the expression of an immediate early gene c-Fos after the SRM test46, and traced the inter-nucleus connections from the FN to the cerebral cortex. Our findings indicate that MLIs fine-tune the cerebellar intrinsic circuits to contextualize the extrinsic networks for SRM, thereby enabling normal social behavior.
Results
Chemogenetic excitation of MLIs in the cerebellar vermis disrupts SRM
We used an adeno-associated virus (AAV)-mediated chemogenetic approach to express an excitatory Designer Receptors Exclusively Activated by Designer Drugs (DREADD, AAV8-hSyn-DIO-hM3Dq-mCherry) in major vermal lobules of the anterior (vermis IV/V) or posterior (vermis VI/VII) cerebellum of adult c-kitIRES-Cre mice (Fig. 1a). Via Cre-LoxP recombination, the c-kit promoter allows selective transduction of MLIs postnatally47. The vermis was targeted because of its association with the limbic system and neuropsychiatric disorders17. To examine transduction efficiency, we imaged cerebellar sections several weeks after viral infusions and behavioral tests. Despite individual variability, robust yet localized expression of the hM3Dq receptor (marked with mCherry) was observed in lobule IV/V or VI/VII (Fig. 1b, Supplementary Fig. 1). Within the individual cerebellum, the hM3Dq was mostly present in midsagittal sections and diminished in sections >500 µm away from the midline (Supplementary Fig. 1). By normalizing the area containing mCherry to the total area of lobule IV/V or VI/VII in the midsagittal sections, we found 29.2 ± 2.7% of lobule IV/V and 35.8 ± 3.0% of lobule VI/VII (n = 6 mice/group) were affected. Higher-resolution imaging revealed that hM3Dq expression was predominantly in MLIs (Fig. 1b) although limited Cre activity could appear in Golgi and glial cells47. Among MLIs, stellate cells (SCs) innervate the distal dendrites of PCs and do not directly affect PC firing, whereas basket cells (BCs) execute powerful inhibition of PC activity by forming GABAergic synapses on the soma and ephaptic connections on the axon initial segment of PCs10. Thus, we focused on the BC-PC microcircuits by co-labeling PCs with an anti-Calbindin antibody. As shown before10, a BC innervated several adjacent PC somas (Supplementary Fig. 1). We validated our chemogenetic approach with ex vivo patch-clamp recordings. Activation of the hM3Dq with clozapine-N-oxide (CNO; 10 µM) significantly increased the frequency of action potentials (APs) in BCs expressing the excitatory DREADD but had no effect on non-transduced BCs (Fig. 1c, d; statistical details in Supplementary Table 1). As a result of synaptic inhibition10,43, CNO reduced the firing rate of PCs that were connected to hM3Dq-expressing BCs and increased the coefficient of variation (CV) of inter-AP intervals in these PCs. CNO did not affect PCs innervated by non-transduced BCs (Fig. 1e, f).
Next, we evaluated the impact of the chemogenetic excitation of MLIs on mouse behavior by intraperitoneal (i.p.) injection of CNO (1 mg/kg) 30–40 min before each assay25,43. Compared to their littermate controls that lacked Cre recombinase, c-kitIRES-Cre mice transduced with the hM3Dq in lobule IV/V or VI/VII showed no differences in open-arm avoidance (measured with the elevated plus maze test), locomotion (measured with the open field test), motor coordination (measured with the rotarod test), or social play (measured with the reciprocal social interaction test) (Supplementary Fig. 2), in line with early reports that interfering vermal activity in adult mice did not affect their general performance25,37,43. To further scrutinize their social behavior, we conducted the three-chamber social test that consisted of a sociability trial immediately followed by a social novelty trial (Fig. 1g). In the sociability trial, all groups explored the stranger mouse more than the empty cup and rendered positive sociability indices (Fig. 1g, top). In the social novelty trial, both control mice and those expressing the DREADD in vermis IV/V explored the new stranger more than the old one, while mice expressing the DREADD in vermis VI/VII did not show such a preference, resulting in a social novelty index near 0 (Fig. 1g, bottom).
Social novelty preference relies on remembering previous encounters and is driven by seeking the unknown48. To determine whether the lack of social novelty preference was caused by memory loss, we performed the social recognition test with a prolonged interval (45 min) between a learning and a testing trial (Fig. 1h). The time delay was introduced to assess memory retention. In the learning trial, all animals displayed normal social tendency toward the stranger mouse and had positive social learning scores (Fig. 1h, top), consistent with findings from the three-chamber social test (Fig. 1g). In the testing trial, the control group again showed intact social novelty preference. However, the vermis-perturbed groups explored the new and the old strangers indiscriminately, accompanied with low social recognition scores (Fig. 1h, bottom). To determine whether the chemogenetic perturbation reduced memory capacity or motivation to seek novelty in general, we did the object recognition test, which was identical to the social recognition test except that the social stimuli were replaced with inanimate objects (Fig. 1i). Surprisingly, all groups spent more time exploring the novel object than the old object with positive object recognition scores (Fig. 1i), indicating well-preserved object-based memory and novelty-seeking behavior. The total exploration time among groups was comparable in all tests (Supplementary Table 1). Together, these results suggest that the vermis is important for SRM without affecting anxiety levels, locomotion, exploratory motivation, social approach, or object recognition memory in animals.
Optogenetic stimulation of MLIs in the vermis interferes with retrieval, but not encoding, of SRM
SRM is a type of declarative memory that consists of encoding, storage/consolidation, and retrieval of social information49. To specify at which of these stages the cerebellum is engaged, we employed an optogenetic approach that permits precise temporal control of neuronal firing with a bacterial artificial chromosome (BAC) transgenic mouse line that uses the nNOS promoter to express channelrhodopsin-2 (ChR2) in MLIs (nNOS-ChR2)50. ChR2 (fused to YFP) was found in most MLIs in the cerebellum (Fig. 2a). Co-labeling PCs with an anti-Calbindin antibody confirmed that ChR2 was exclusively present in the somas, dendrites, and axon terminals of MLIs. To measure neuronal responses to photostimulation, we recorded APs from MLIs (BCs) and PCs while delivering an 8 Hz train of light pulses (470 nm, 25 ms duration) (Fig. 2b). This low frequency ensures reliable generation of APs in response to light flashes;50 and it falls within the range of theta oscillations (4–10 Hz), which play a role in learning and memory51. Specifically, theta-burst stimulation of the human cerebellum affects episodic memory52. As shown in Fig. 2b–d, the photostimulation rapidly increased AP frequency in MLIs, while reducing it in postsynaptic PCs. Upon cessation, MLI and PC activity returned to baseline levels. Additionally, the photostimulation decreased regularity of AP firing in PCs, indicated by a dramatic increase in the CV of inter-AP intervals (Fig. 2e). When all inter-AP intervals were plotted on a histogram and fit with a Gaussian function, we found the photostimulation created two distinct peaks centering around 14.5 ms and 54.2 ms, in contrast to the unimodal pattern with a peak around 17.5 ms without stimulation (Fig. 2f). Based on the mean values of the Gaussian distribution, the 1st peak likely represented spontaneous PC activity whereas the 2nd peak correlated to the stimulus pattern.
Applying the same light pulses through optic fibers implanted in lobule IV/V or VI/VII, we studied the behavior consequences of photostimulating MLIs. In the open field test, nNOS-ChR2 mice and their littermate controls (not expressing ChR2) traveled a similar distance, and the photostimulation did not change their locomotor activity (Supplementary Table 1). To know how the vermis was involved in the process of SRM, we photostimulated in different sessions of the social recognition test. When the light was presented during the learning trial (Fig. 2g), all mice explored the stranger more than the empty cup with positive social learning scores (Fig. 2g, top). They also explored the novel stranger more than the familiar one with positive social recognition scores (Fig. 2g, bottom). Perturbing lobule IV/V yielded less exploration in total as compared to the other groups (Supplementary Table 1), which may reflect deficits in motivation, attention and/or sensorimotor integration associated with interference of the anterior cerebellum25. Overall, perturbation of the cerebellum at the encoding stage did not impair sociability, social novelty preference or SRM.
In contrast, when the light was presented during the testing trial (Fig. 2h), all mice spent more time exploring the stranger than the empty cup with positive indices of sociability (Fig. 2h, top). Yet, unlike the control group that explored the novel stranger more than the old one, the vermis-perturbed groups did not prefer the novel stranger, leading to poor indices of social recognition (Fig. 2h, bottom). As there was no group difference in total exploration in this case (Supplementary Table 1), the impaired social recognition was likely accounted for by an inability to recall previous encounters rather than by confounding factors such as a lack of inquisitiveness. When the light was presented in the testing trial of the object recognition test (Fig. 2i), all animals preferred the new object to the old one with positive indices of object recognition (Fig. 2i), implying no effect of this manipulation on non-social recognition memory. These results suggest that the vermis plays a specific role in the retrieval (or consolidation), but not encoding, of social information.
Perturbing the vermis decreases functional connectivity and disorganizes network structure of brain regions essential for SRM
To understand how the cerebellum mediates SRM, we monitored neuronal activity via a molecular marker, c-Fos, in 24 brain regions following the social recognition test on c-kitIRES-Cre mice and their littermate controls (Fig. 3a). As shown in Fig. 1h, activation of vermal MLIs by the excitatory DREADD disrupted SRM. 90 min after the test, we collected brain tissue for c-Fos immunostaining. Basal levels of c-Fos are normally low but can be elevated by external stimuli, making it a useful tool to identify activated neurons46. The 90 min window was chosen because c-Fos expression peaks 1-2 h after stimulation53. We examined c-Fos in brain regions with well-established roles in SRM26,27,28,29,30. By counting the number of c-Fos-positive cells in each region and normalizing these values to the mean number of c-Fos-positive cells in the control group, we found significant effects on c-Fos levels for the mPFC, parahippocampal cortices, amygdala, hypothalamus, and ventral tegmental area (VTA) (Supplementary Table 1). Perturbing the vermis, irrespective of the lobules, elevated c-Fos expression in the temporal association cortex (TeA), ectorhinal cortex (Ect), and perirhinal cortex (PRh) compared to the control group (Supplementary Fig. 3). These parahippocampal cortices interact with the hippocampus and cerebral cortex to support high-order cognitive processes such as contextual associations and emotional inferences54. In other regions, changes in c-Fos depended upon which lobule was targeted. For instance, perturbing lobule IV/V mostly reduced c-Fos levels in subdivisions of the mPFC (anterior cingulate cortex (rostral), Acg: p = 0.024; prelimbic cortex, PL: p = 0.044; but not infralimbic cortex, IL: p > 0.05; Fig. 3b). In contrast, perturbing lobule VI/VII increased c-Fos levels in subdivisions of the amygdala (central nucleus of amygdala, CeA: p = 0.003; basolateral amygdala, BLA: p = 0.007; but not basomedial amygdala, BMP: p > 0.05; Fig. 3c). The differences are likely related to distinct connections of the anterior and posterior cerebellum to the cortical and subcortical regions14,15.
To identify subtypes of the c-Fos-positive cells, we co-labeled glutamatergic and GABAergic neurons with anti-CaMKII and anti-GAD67 antibodies, respectively. As illustrated in Supplementary Fig. 4, c-Fos-positive cells in these regions were primarily glutamatergic neurons (~60–80%), with small proportions of GABAergic neurons (~10–20%) and other cell types (~10–20%). This result suggests that while various cells are activated, excitatory neurons that typically represent the output from each region play a major role in mediating the cellular response to SRM.
To gain insight into functional connectivity, we calculated Pearson correlation coefficients (r) of each pair of brain regions across the subjects in each group based on the c-Fos measurements. We then created interregional correlation matrices with the values of Pearson’s r from 1 to −1 indicated by a gradient spectrum from red to indigo, respectively (Fig. 4a). As the matrices exhibited discrete patterns for each group, we further analyzed the network connectivity by only including the most positive (r > 0.6) and the most negative (r < −0.6) correlations in the network plots (Fig. 4b). The plots comprised the brain regions (nodes) and their connections (edges; red lines: r > 0.6, indigo lines: r < −0.6). A chi-square test revealed that the control group had more edges, and proportionally more positive correlations, than the vermis-perturbed groups (vermis IV/V: χ2 = 66.03, p < 0.001, vermis VI/VII: χ2 = 27.42, p < 0.001; Fig. 4c). Perturbing the vermis reduced the mean r of all brain regions compared to the control group (Fig. 4d), implying a decrease in concerted brain-wide activity. Figure 4e–h highlighted four key regions for SRM. Both vermis groups had lower r than the control group in the mPFC, ACC, and hippocampus, except that the difference for the vermis VI/VII group in the ACC was not significant (p = 0.097). Conversely, the vermis groups had higher r than the control group in the amygdala though the increase of the vermis IV/V group was mild (p = 0.079); this may reflect the lobule-specific changes of c-Fos expression in the amygdala (Fig. 3c).
To define network properties, we did graph theoretical analysis by focusing on the positive correlations in the matrix of each group. First, we computed degree (the number of edges that a node has) and betweenness (the number of shortest paths that pass through a node) to assess the centrality of a network node. A high value for “degree” indicates a brain region that is connected to many others, and a high “betweenness” indicates a brain region that has close control over other regions55,56. For degree, the Ect, hippocampus (CA1 and dentate gyrus [DG]), and primary motor cortex (M1) were ranked high (>0.8) in the control group, while the amygdala (BLA and BMP), parietal association cortex (PtA), lateral entorhinal cortex (lEnt), and Acg were ranked high in the vermis groups (Fig. 5a–c, left). For betweenness, the Ect and TeA were ranked high (>0.8) in the control group, while the PtA and Acg were ranked high in the vermis groups (Fig. 5a–c, middle). Next, we quantified within-community Z-scores (the extent of a node connected in its own community) and participation coefficients (the extent of a node connected to other communities) to outline the modularity of the network. A brain region with high rankings in both Z-score and participation coefficient is classified as a connector hub that regulates the interactions within and between communities (modules)55,56. In the control group, the Ect, TeA, hippocampus (CA1, CA2/3 and DG), mPFC (Acg and IL), M1, and ACC had high ranks for Z-scores and participation coefficients, whereas in the vermis groups, the amygdala (BLA and BMP), PtA, lEnt, Acg, and VTA had high ranks for Z-scores, but their participation coefficients were zero (Fig. 5a–c, right). Lastly, we simulated the spatial distribution of the degree and betweenness centralities in a force atlas format, where distinct modules were color-coded. The control group contained two interactive modules (Fig. 5d). One module (pink) was centered on the hippocampus and mPFC, which critically regulate mnemonic processes48. The other module (green) included the parahippocampal cortices and CA2/3, which integrate sensory information for memory establishment26,54. Perturbation of the vermis destroyed these modules and produced solitary, amygdala-centered networks (Fig. 5e, f). Given that dynamic interactions among the amygdala, hippocampus, parahippocampus and mPFC are crucial for emotion-based memory57,58, we propose that the cerebellum supports SRM by promoting and coordinating functional connectivity of the corticolimbic system.
Anatomical tracing of the fastigial nucleo-cortical connections
The vermis influences the corticolimbic system through projections to the FN17. To track the FN output connections, we injected AAV1-hSyn-Cre unilaterally into the FN of floxed tdTomato (Ai9) mice to label FN neurons via Cre-LoxP recombination59 (Fig. 6a). Since AAV1-hSyn-Cre has an anterograde transsynaptic property60, this experiment allowed us to trace FN axons to their first- and second-order targets in the forebrain. We transduced neurons mainly in the FN while avoiding the interposed and dentate nuclei (Fig. 6b). As most fastigial axons decussate to innervate downstream nuclei12,17, we followed their projections contralateral to the injection site in the 24 brain areas that were examined for c-Fos expression (Fig. 4).
In line with recent anatomical studies in mice21,22, we found prominent labeling of somas and processes of neurons in the vTH (ventrolateral and ventromedial thalamus) across individual mice (Fig. 6c, Supplementary Fig. 5). Labeled neurons were also found in the anteromedial (AM), centrolateral (CL), centromedian (CM), mediodorsal (MD), and paraventricular (PV) thalamus (Fig. 6c, top and middle). This reinforces that the thalamus is a major efferent pathway for the FN to the neocortex21,22. Moreover, we observed neurons expressing tdTomato sparsely distributed in the dorsomedial and lateral hypothalamus (Fig. 6c, bottom). The cerebellar-hypothalamic contacts may be important for somatic and visceral integration19,20. In the midbrain, labeled axons (no somas) were present in the VTA, consistent with the previous report on a lack of monosynaptic connection between the FN and VTA21. Labeled neurons and their processes were found in the substantia nigra pars compacta (SNc) and red nucleus (RN), indicating direct FN projections to the basal ganglia and rubrospinal tract involved in motor and cognitive functions7,8,9.
Among the second-order targets, axonal (not somatic) labeling was evident in subcortical regions, e.g., the dorsal striatum (dSTR) and nucleus accumbens (NAc) (Fig. 7b, Supplementary Fig. 5). But the labeling was scarce in the hippocampus and amygdala (Fig. 7d), confirming no mono- or disynaptic innervation from the FN to these areas21. The engagement of hippocampus and amygdala in cerebellum mediated SRM (Fig. 5) is likely through their intense crosstalk with other brain regions, such as the cerebral cortex61,62. In fact, we observed a high density of labeled axons in somatosensory (S1) and motor (M1) cortices, as well as in areas involved in decision-making, emotion and attention, e.g., orbitofrontal cortex (OFC), ACC, TeA, and PtA (Fig. 7a–c). There was moderate labeling in the mPFC (Acg and PL), retrosplenial regions (RSD and RSG), ectorhinal (Ect), perirhinal (PRh), and entorhinal (lEnt) cortices (Fig. 7a, c, e), which are crucial for episodic memory54. The IL was sparsely labeled here (Fig. 7a) but not in another study21, possibly due to differences in injection sites or floxed mice (Ai9 vs Ai14). The abundance of labeled neurons and their processes was summarized for the 24 regions (Supplementary Fig. 5). Collectively, these data provide an anatomical basis for cerebellar regulation of SRM. However, the functional connectivity for each region derived from the c-Fos mapping (Figs. 3–5) does not simply correlate to its structural connection with the medial cerebellum (Figs. 6, 7), suggesting the complexity of the inter-nucleus circuits subserving SRM.
Amygdalar integration in the fastigial nucleo-cortical circuits
To understand how the amygdala, specifically BLA, becomes a central hub in the SRM network after vermal manipulation (Fig. 5), we combined anterograde tracing of the FN output with retrograde tracing of the BLA input by infusing AAV1-hSyn-Cre in the FN and AAVrg-hSyn-EGFP in the contralateral BLA of Ai9 mice (Fig. 8a). As shown in Figs. 6 and 7, AAV1-hSyn-Cre59,60 permitted transduction of tdTomato in the FN (Fig. 8b) and its downstream neurons. Being a retrograde virus, AAVrg-hSyn-EGFP63 labeled the BLA (Fig. 8c) and its upstream neurons with EGFP. Using the two fluorescent signals, we probed the anatomical relationship between the FN and BLA. In line with the literature64, the BLA was connected with the cortex including auditory, visual, somatosensory, motor (M1), prefrontal (PL and IL), temporal (TeA), ectorhinal (Ect), perirhinal (PRh), and entorhinal (lEnt) cortices, as well as the subcortex including basal ganglia (dSTR) and thalamus (Supplementary Fig. 6). In the thalamus, we found a great number of EGFP-expressing neurons in the PV, parafascicular (PF), MD, and medial geniculate (MG) nuclei but rarely in the vTH (Fig. 8d, e), contrasting to the preferential target of the FN in the ventral part (Fig. 6c). The thalamic input to the BLA is known for conveying sensory information for fear learning and memory64. In the PV nucleus where both tdTomato- and EGFP-expressing somas were present, they did not overlap, implying that the FN projects to different neurons from those sending input to the BLA. In the midbrain, EGFP-labeled somas were found in the VTA while EGFP-labeled axons (not somas) were in the NAc (Fig. 8f, g). This represents a well-known VTA-to-BLA-to-NAc circuit involved in reward seeking behavior65. In the VTA and NAc, very few EGFP-labeled somas or axons contained tdTomato, the indicator for the FN projections. In the cortices that were identified as the second-order FN targets (Fig. 7), e.g., M1, OFC, PtA, Ect and PRh, we noticed EGFP-expressing somas, but they rarely made close contact with axons expressing tdTomato (Fig. 8h, i). The cerebral cortex-BLA interactions are critical for coordinating emotional and cognitive responses57,58,64. The separate signals in all brain regions suggest that the FN does not connect with the BLA through hierarchical neurotransmission. However, the coexistence of the FN output and BLA input within the same regions signifies a role of the BLA in integrating sensory and emotional information in the FN-neocortex circuits (Supplementary Fig. 7).
Discussion
By manipulating cerebellar activity with spatiotemporal control, we have uncovered a new role for the cerebellum in incorporation of working memory into social behavior. Specifically, we found that activating MLIs in the vermis during memory recall impaired social, but not non-social, recognition memory. Neuroanatomical tracing and c-Fos functional mapping revealed the circuit basis for the cerebellum to co-activate and coordinate the limbic structures essential for emotional responses and cognitive functions in support of SRM. This work provides evidence for the internal forward model, in which the cerebellum conducts mental activity by orchestrating the cerebello-cortical network13.
Although the cerebellum is known for motor coordination, our manipulations did not affect animals’ motor performance on the accelerating rotarod or exploratory activity in less challenging conditions, e.g., in the open field (Supplementary Fig. 2). This can be for several reasons. First, our interference was restricted to a single lobule, as indicated by the localized hM3Dq expression (Fig. 1). Activation of MLIs in a lobule might not be sufficient to disturb general movement, even though it was sufficient to produce a SRM deficit. The sensitivity of the non-motor function to the cerebellar interference is congruent with other studies using similar approaches23,24,37. Second, we manipulated MLIs instead of PCs to attenuate, rather than silence, the output of the cerebellar cortex (Figs. 1, 2). This tactic limited potential alterations in motor skills, anxiety levels, social preference, and territorial aggression that might confound the assessment of SRM45. Lastly, the subregions that we perturbed are less involved in online motor control. While the anterior vermis may be part of the sensorimotor representation14,15 and targeting lobule IV/V with photostimulation occasionally impeded exploratory behavior (Fig. 2g), the posterior vermis typically represents affective and cognitive functions14,15 and interfering with lobule VI/VII had no effect on total exploration in the object and social recognition tests (Figs. 1, 2). Though these two tests had the same structure, vermis-perturbed animals were able to discriminate between objects by spending more time on the new object than the old one yet failed to show such a preference in identifying conspecifics. This demonstrates an explicit contribution of the cerebellum to SRM.
PCs fire APs spontaneously to encode output information of the cerebellar cortex to the neocortex. Reduced PC activity leads to motor and non-motor dysfunctions in neurodegenerative (e.g., ataxia) and neurodevelopmental (e.g., ASD) disorders66. It is difficult to pinpoint whether the extent of PC firing deficits, the lobule-specific targeting of genetic mutations, or the conditions of other brain regions result in certain behavior phenotypes. In the case of ASD, cerebellar abnormalities including PC loss and impaired PC activity and plasticity underlie a spectrum of behavioral deviations in humans and animal models67,68. However, it is noteworthy that restoring PC activity is to date the most effective avenue to rescue social deficits in ASD mouse lines41,42,43,44. APs elicited from PCs are driven by PC intrinsic excitability and regulated by MLI inhibition, both of which can alter PC firing rate and regularity10. Further analysis is needed to conclude whether global changes in speed or specific patterns of PC activity are crucial for processing social information.
By spatially and temporally tuning PCs, MLIs actively participate in learning and memory to define motor and non-motor functions34,35,36,37,38,39. For example, chemogenetic inhibition of MLIs in paravermal lobule VI or crus 1 causes deficient eyeblink conditioning responses37. The same manipulation in lobule VI affects reverse learning and in lobule VII affects novelty-seeking behavior. However, most of the effects are present in juvenile mice and diminish when animals reach maturity37. Our manipulation of MLIs in the adult mice also had little effect on learning, emotion, or sociability (Supplementary Fig. 2) except for SRM (Figs. 1, 2). This unexpected finding may be due to our experimental design. Opposite to suppressing MLIs37, we increased their activity based on a common phenotype among ASD mouse models42,43,44. Activation of the hM3Dq in the c-kitIRES-Cre mice might have involved a small number of Golgi and glial cells in addition to MLIs47 (Fig. 1). Yet, photostimulation of the nNOS-ChR2 mice solely excited MLIs50 and produced the same SRM deficit as the chemogenetic manipulation (Fig. 2), arguing for the significant contribution of MLIs to SRM. Future investigation is required to determine how MLIs encode SRM with in vivo recordings of APs or Ca2+ signals from MLIs during social recognition tasks.
SRM exploits working memory while judging a recently encountered conspecific26,27,28,29,30. The cerebellar involvement in working memory is well documented12. Schmahmann and Sherman first reported patients with damage to the posterior vermis displayed working memory loss along with other cognitive and affective symptoms69. Reduced volume or activity of the vermis is often present in neuropsychiatric disorders17. Animal work shows that interfering with PC activity in the vermis does not affect object recognition memory per se, like our data in Fig. 1 & 2, but compromises location- or context-based memory24,25. Also, the cerebellum participates in decision-making and consolidation of reward-related go/no-go memory tasks70,71,72. As social interactions are rewarding, the cerebellum may promote SRM via its connections to the reward pathway73, which significantly overlaps with the social brain network1,2,3.
We constructed the c-Fos-derived neural network underlying SRM and identified two main modules centering on the mPFC-hippocampus and Ect-TeA (Figs. 3–5). The mPFC-hippocampus module may be essential for processing episodic information of “what”, “where” and “when”48; and the Ect-TeA module may be critical for integrating sensory information into the mPFC-hippocampus memory system54. Perturbing the cerebellum broke down the highly interactive clusters and shifted the brain network to an amygdala-centered module (Fig. 5). The emergence of amygdala as a hub reinforces our results that the vermis is of particular importance for emotion-related memory (Figs. 1, 2). As described before17,19,20,21,22, the vermis and its projections to the FN sent output to diverse brain areas (Figs. 6, 7). The monosynaptic links to the limbic system (thalamus and hypothalamus) and brainstem, as well as disynaptic links with the prefrontal cortex, temporal cortex, parietal cortex, and striatum may specifically mediate the cognitive and affective processes in SRM7,8,9. However, we did not find mono- or disynaptic connections from the FN to the hippocampus and amygdala (Figs. 6, 7). Their functional connectivity with the vermis may be through the cerebral cortex61,62. A study proposes that the cerebellum regulates coherence of gamma oscillations between mPFC and hippocampus for spatial memory74. It is tempting to speculate that the cerebellum may impact hippocampal activity in a similar way to enable SRM. Our analysis of the amygdalar integration revealed coexistence yet nonoverlap of FN outputs and BLA inputs in the SRM network (Fig. 8, Supplementary Fig. 7). Likely via neuronal interactions within these regions, the vermis has influenced amygdalar activity and connectivity (Figs. 3–5). In general, direct synaptic contacts convey more information than indirect ones. However, the indirect FN-BLA connections may be crucial for processing complex information to modulate the effects of emotion on memory57,58. The fact that cerebellar perturbation can alter these distal regions supports our view that the cerebellum is an integral part of the cerebello-cortical network for SRM.
Our data suggest that the cerebellum is necessary for retrieval, but not encoding, of SRM (Fig. 2). Memory encoding is the act of storing external features and internal states in a subset of neurons (engram) across various brain regions. Memory retrieval requires concurrent reactivation of the engram cells49. Here, we showed the SRM task primarily activated excitatory neurons in the local circuits that were responsible for sending information to other brain regions (Supplementary Fig. 4), providing a mechanism for the interregional coactivation of engram cells. The neuronal activity in the encoding and retrieval processes may differ. For instance, the hippocampal CA2/3 and DG are only active during memory formation while the subiculum (and CA1 to a less degree) is active during memory recall75,76. This selective activation in the initial learning may explain why c-Fos expression in the hippocampus was insensitive to the changes in later performance of memory retrieval (Supplementary Fig. 3). In addition to recollection of contextual cues, recognition memory entails familiarity for acquaintance-based discrimination77. While recollection involves the medial temporal lobe including parahippocampus and amygdala, familiarity involves the PRh. Remarkably, c-Fos levels in all the regions were elevated by the vermal perturbation that led to the SRM deficit (Fig. 3, Supplementary Fig. 3). This strengthens the specificity of the identified neural network for recognition memory. Our finding that interfering with the vermis decreased brain-wide functional connectivity (Fig. 4) indicates a unique position of the cerebellum in reinstatement of memory traces by coactivating the engrams.
The results of this study must be interpreted with caution due to several technical issues. First, the readouts of c-Fos expression have a modest temporal resolution. Despite being a validated tool46, c-Fos expression is a slow transcriptional process and does not allow real-time monitoring of neural activity or resolve the differences between events occurring close in time. Thus, the c-Fos-based neural network may reflect interwound responses to encoding and recalling of SRM. Further inquiry using in vivo electrophysiology and Ca2+ imaging is needed to address how the engram cells are recruited and reactivated in a task-related manner with temporal accuracy. Second, the conceptualized network model of SRM inevitably embeds other information, such as the sensory and motor processing during the social recognition task as well as environmental factors, e.g., fear or stress from experimental handlings. While the cerebellum was taken for examining the distribution of fluorescent proteins and not used for c-Fos staining, c-Fos in the vermis can be reactive to external stimuli including odorants78. To minimize the variables, all groups were tested under the same conditions and c-Fos counts were normalized to the average value of the control group for each brain region (Figs. 3–5). Thereby, the network disruption may largely result from the cerebellar interference. Lastly, it is worth noting that the chemo- and optogenetic manipulations are artificial, from which we can infer what the cerebellum can do but not exactly what it does physiologically.
Notwithstanding these considerations, our study presents the first evidence for a causal link of the cerebellum to social memory, which advances our understanding of the neural substrates for social behavior. As social impairments are commonly associated with neuropsychiatric disorders4,5,6, our results will help develop diagnostic and therapeutic strategies by targeting cerebellar modulation of the higher-order functions with cell-type- and region-specific precision.
Methods
Subjects
C-kitIRES-Cre mice and nNOS-ChR2 BAC mice were provided by Hiroki Taniguchi (Ohio State University, USA)47 and George J Augustine (Nanyang Technological University, Singapore)50, respectively. Ai9 (RCL-tdT) mice (strain # 007909) were purchased from Jackson Laboratory (Bar Harbor, ME)59. Both sexes (8-12 weeks old) were included, and no sex differences were observed. Mice were kept under a 12-hour light-dark cycle (light on from 07:00 to 19:00) and reared 3-5 per cage with food and water ad libitum. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) and the Institutional Biosafety Committee (IBC) of University of Minnesota, in accordance with the National Institutes of Health guidelines. Heterozygous c-kitIRES-Cre were bred to obtain mice with or without (control) Cre recombinase, genotyped with 2 primers: 5′-CGGTCGATGCAACGAGTGATG-3′ and 5′-AGCCTGTTTTGCACGTTCACC-3′. Male nNOS-ChR2 were crossed with female C57BL/6 J to obtain wild-type (control) or nNOS-ChR2 mice, genotyped with 2 primers: 5′-AGTAGCTCAGGTTCCTGTGGG-3′ and 5′-GCAAGGTAGAGCATAGAGGG-3′.
Stereotactic surgery
c-kitIRES-Cre, nNOS-ChR2 and their littermate controls were randomly assigned to vermis IV/V and VI/VII surgical groups. Animals were anesthetized with a cocktail of ketamine and xylazine (100 and 10 mg/kg, respectively, i.p.). Carprofen (5 mg/kg, s.c.) was administered as an analgesic. Mice were fixed on a stereotactic frame (Stoelting, Wood Dale, IL) and an incision was made to expose the skull. For viral infusion, a hole was drilled above the cerebellum of c-kitIRES-Cre mice and their wild-type littermates according to the coordinates: AP -6.0 or -7.0 mm (for vermis IV/V or VI/VII, respectively), ML ± 0.0 mm, relative to bregma79. AAV8-hSyn-DIO-hM3Dq-mCherry (titer>3.0 × 1013 vg/ml, 0.1–0.15 µl, University of Minnesota Viral Vector and Cloning Core) was infused into the cerebellum (DV −1.0 mm) with a 10 µl Hamilton syringe controlled by a microinjection pump (QSI 53311, Stoelting) at a flow rate of 0.09 µl/min. After infusion, the needle was left in place for >7 min and then slowly retracted. The wound was sutured and disinfected with 70% ethanol and iodine. For optic-fiber implantation, a hole was drilled above the cerebellum of nNOS-ChR2 mice and their wild-type littermates according to the coordinates: AP −6.4 or −7.0 mm (for vermis IV/V or VI/VII, respectively), ML ± 0.0 mm, relative to bregma79. An optic fiber (∅200 µm) attached with a ceramic ferrule (FT200EMT and CFLC230-10, Thorlabs, Newton, NJ) was implanted in the hole 0.0–0.2 mm below the cerebellum surface. Two screws and dental cement were applied to secure the optic fiber on the skull. For neuronal tracing, AAV1-hSyn-Cre (plasmid #105553, titer >1 × 1013 vg/ml, Addgene, Watertown, MA) was unilaterally infused into the FN of Ai9 (RCL-tdT) mice according to the coordinates: AP −6.24 mm, ML 0.7 mm, DV −3.3 mm, relative to bregma79. AAVrg-hSyn-EGFP (plasmid #50465, titer ≥7 × 1012 vg/ml, Addgene) was infused into the BLA according to the coordinates: AP −1.34 mm, ML 3.0 mm, DV −4.6 mm, relative to bregma79. After operations, mice were placed on a heating pad (37 °C) to maintain body temperature until awake. Electrophysiology and behavioral tests were conducted 3 weeks after surgeries. Imaging of labeled neurons was done 4–5 weeks after surgeries. Mice (n = 1–2/group) were excluded when injections were outside the intended areas.
Electrophysiology
c-kitIRES-Cre mice (n = 6) and their wild-type littermates (n = 7) injected with AAV8-hSyn-DIO-hM3Dq-mCherry and nNOS-ChR2 mice (n = 3) were used for recordings. Following decapitation, the brain was immediately dissected, and sagittal cerebellar slices were sectioned at a thickness of 300 µm using a vibratome (VT1200S, Leica Biosystems, Buffalo Grove, IL) in ice-cold modified artificial cerebrospinal fluid (ACSF). It contained (in mM): sucrose (217.6), KCl (3), glucose (10), NaH2PO4 (2.5), NaHCO3 (26), MgCl2 (2), and CaCl2 (1), continuously bubbled in 95% O2 and 5% CO2 (pH 7.4). Slices were then incubated in oxygenated standard ACSF including (in mM): NaCl (125), KCl (2.5), glucose (10), NaH2PO4 (1.25), sodium pyruvate (2), myo-inositol (3), ascorbic acid (0.5), NaHCO3 (26), MgCl2 (1), and CaCl2 (2) (pH 7.4) at 37 °C for 30 min prior to experimentation. Slices were transferred to a recording chamber mounted on a BX51WIF Olympus microscope with a 60x water immersion objective. They were perfused with the standard ACSF at a rate of ~1 ml/min. PCs and MLIs were identified by their size and location. Patch electrodes had resistances of 2.5-3 and 4.5-6 MΩ for PCs and MLIs, respectively. APs at the soma were recorded in the cell-attached mode with GΩ seal at −60 mV for PCs and −70 mV for MLIs. Intracellular solution contained (in mM): K-gluconate (97.5), KCl (32.5), EGTA (0.1), HEPES (40), MgCl2 (1), ATP (2), GTP (0.5) (pH 7.3). All recordings were acquired on-line at 35 °C, filtered at 4 kHz, digitized at 50 kHz with a dual-channel amplifier (MultiClamp 700B, Molecular Devices, San Jose, CA) and digitizer (Digidata 1550B, Molecular Devices). Data were analyzed off-line with MiniAnalysis 6.0.7 (Synaptosoft, Fort Lee, NJ), Clampfit 10 (Molecular Devices) and Excel 2016 (Microsoft, Redmond, WA). To activate hM3Dq receptor, CNO (BML-NS105, Enzo Life Sciences, Farmingdale, NY) was applied to slices43. To excite ChR2, 470 nm LED light (LEDD1B, Thorlabs) was delivered through the 60x objective coupled with a GFP filter cube (Olympus)50. Light pulses were controlled using pCLAMP 10 (Molecular Devices). Light intensity (maximum = 8.2 mW/mm2 at the tissue) was measured with a PM100D power meter and a S121C photodiode sensor (Thorlabs).
Behavioral testing
Either in chemogenetic (using c-kitIRES-Cre) or in optogenetic (using nNOS-ChR2) experiments, three groups (n = 6–10 mice with 3–5 females/group) were included: control, vermis IV/V, and vermis VI/VII. For chemogenetic experiments, CNO was administered in all groups (1 mg/kg, i.p.) 30–40 min prior to each test. CNO was stored in aliquots at −20 °C. It was diluted in saline, kept on ice, and protected from light while in use. For optogenetic experiments, only open field, object recognition and social recognition tests were conducted. Light pulses were generated with a Polygon DMD pattern illuminator (Mightex, Toronto, Canada) powered by 470 nm LED through a cable connected to the optic fiber implanted on the animal’s head. Light intensity (maximum=5.9 mW/mm2 at the tissue) was measured with a PM100D power meter and a S121C photodiode sensor (Thorlabs).
Apparatus
An elevated plus maze consisted of a central platform (5 × 5 cm), two walled arms (25 × 5 × 25 cm) and two arms without walls (25 × 5 cm) in the shape of a cross. It was elevated 30 cm high, and the two types of arms were situated opposite to each other. A digital rotarod (Rotamex, Columbus Instruments, Columbus, OH) was used to test motor coordination. A box composed of 3 chambers (20 × 40 x 30 cm each) with passages (5 × 5 cm) within the walls that divided the chambers was used for three-chamber social tests. A box (40 × 40 × 30 cm) made of polyvinyl chloride was used for open field, object recognition, social recognition, and reciprocal social interaction tests. Different sets of objects with variable textures (smooth or rough), sizes (diameter 7–9 cm, height 14–17 cm) and shapes (column, irregular) were used for object recognition. Objects were weighed enough to prevent being moved by animals. Assignment of objects was counterbalanced to minimize object and/or place preference. Age-, sex-, and strain-matched wild-type mice served as strangers for social tests. Metal grid cups (diameter 10 cm, height 12 cm) were used to constrain strangers. Placement of strangers was counterbalanced between subjects. All tests were done in a quiet room (< 40 dB) under dim light (~ 50 lx) during 10:00-16:00. Test designs were based on previous studies using similar experimental protocols25,43,44. Apparatus was cleaned with 70% ethanol between subjects. A camera was connected to a computer to record and analyze behavior via tracking software (ANY-maze, Stoelting). Object and social exploration were manually counted by experimenters who were blind to the design. Other measurements were automatically made with the ANY-maze software.
Elevated plus maze
This test was used to assess anxiety-like behavior. A subject was placed onto the central platform (facing an open arm) and allowed to freely explore for 5 min. Entries to the center, open and closed arms, time spent in these areas, and time spent on head-dip and body-extension were analyzed.
Open field
This test was used to assess locomotor and exploratory activity. A subject was put into the center of the open field and monitored for 15 min. Traveling distance and duration, rearing, grooming, thigmotaxis, and center stay (virtual central square 13.3 × 13.3 cm) were analyzed in 5 min time bins.
Rotarod
This test was used to assess motor coordination and motor learning. A subject was placed on a rotarod rotating at a speed of 4 rpm until it could stay on for >1 min. Then, the speed accelerated from 4 to 40 rpm in 5 min, which was taken as one trial. Three trials were conducted with an inter-trial interval of 15 min. Rotating speed and time staying on the rotarod before falling were recorded.
Reciprocal social interaction
This test was used to assess pro-social behavior in a native situation without physical barriers. A subject was placed into the open field along with a stranger for 10 min. An interaction was defined as the animal’ head being in physical contact with any body part of the other’s, except for the tail. Duration of physical contacts was recorded.
Three-chamber social test
This test was used to assess social and social novelty preferences. It consisted of three sessions: habituation, sociability, and social novelty (9 min each). In the habituation trial, a subject was placed into the middle chamber and allowed to freely explore the whole apparatus. In the sociability trial, a stranger that had never been contacted by the subject was put underneath the cup in one of the side chambers. An identical empty cup was placed on the other side. In the social novelty trial, another stranger was put underneath the previously empty cup. Physical contact around the cups by the subject’s nose, head and forelimbs was defined as explorative behavior. Sociable index = [time for exploring the stranger mouse − time for exploring the empty cup] / total exploration time. Social novelty index = [time for exploring the novel stranger − time for exploring the old stranger] / total exploration time. Positive values indicate intact social and social novelty preferences. Animals that showed a bias for either chamber (stayed >7 min) in the habituation trial were excluded (n = 1–2/group).
Social recognition
This test was used to assess social recognition memory. It included a learning trial (7 min) and a testing trial (7 min) with an inter-trial interval of 45 min. In the learning trial, a stranger was put underneath a cup on one side of the open field and an identical empty cup was placed on the other side. In the testing trial, another stranger was put underneath the previously empty cup. Physical contact around the cups by the subject’s nose, head and forelimbs was defined as explorative behavior. Sociable index = [time for exploring the stranger mouse − time for exploring the empty cup] / total exploration time. Social novelty index = [time for exploring the novel stranger − time for exploring the old stranger] / total exploration time. Positive values indicate intact sociability and social memory29. Animals that did not explore all presented stimuli in either trial were excluded (n = 1–2/group).
Object recognition
This test was used to assess object-based recognition memory. It included a learning trial and a testing trial with an inter-trial interval of 45 min. In the learning trial, two distinct objects were placed in the open field and a subject freely explored them for 7 min. In the testing trial, one of the objects was replaced with a novel object and the subject explored them for 7 min. Object exploration was defined as physical contact with the objects by the head, nose, or forelimbs, but not other behaviors, such as climbing the objects or sitting next to them. To minimize individual differences, the index = [time for exploring the novel object − time for exploring the old object]/total exploration time was calculated. A positive value indicates intact object memory. Animals that did not explore all presented stimuli in either trial were excluded (n = 1-2/group).
Histology, Immunohistochemistry, and Imaging
Mice were deeply anesthetized and perfused with ice-cold phosphate-buffered saline (PBS) followed by 4% paraformaldehyde. The brain was removed, immersed in 4% paraformaldehyde at 4 °C overnight, transferred into 30% sucrose, and stored at 4 °C until processed. The brain was sectioned into 40–50 µm-thick slices with a microtome (SM2010R, Leica Biosystems). The cerebellum was taken for examining the distribution of fluorescent proteins and not used for c-Fos immunostaining.
c-Fos
Same batch of c-kitIRES-Cre and their wild-type littermates that underwent behavioral testing were used for c-Fos staining (n = 6 mice/group). 90 min after the social recognition test, mice were sacrificed, and their brains were fixed. Brain sections were pretreated with 0.3% H2O2 for 10 min and incubated in a blocking solution (2% goat serum and 0.2% Triton X-100 in PBS) at 37 °C for 30 min. Sections were then incubated with a rabbit anti-c-Fos antibody (1:4000, ab190289, Abcam, Cambridge, MA) at 4 °C for 24–48 h. Sections were washed with 0.2% Triton X-100 and incubated with biotinylated anti-rabbit IgG (1:200, BA-1000, Vector Labs, Burlingame, CA) at room temperature for 2 h. After another rinse with 0.2% Triton X-100, sections were incubated in ABC Kit (Vector Labs) for 1 h. 3,3’ diaminobenzidine chromogen was used for visualization. Sections were mounted on slides, cover-slipped with a mounting medium (H-5000, Vector Labs) and imaged with a Leica DMi8 microscope under a 10x lens. 24 brain regions were identified based on the mouse brain atlas79. The number of c-Fos-positive cells in each region was counted bilaterally across 2-4 serial sections with a 200 µm interval. Region of interest (ROI) was adjusted to the size of each brain region and the same ROI was applied to the same region between groups. The number of c-Fos-positive cells in each group was divided by the average value of the control group and presented as fold change. Experimenters who were blind to the design counted labeled cells with semi-automated analysis by ImageJ (NIH).
Calbindin, CaMKII, and GAD67
To mark PCs, brain sections were incubated with a rabbit anti-calbindin D28K antibody (1:1000, PA1-931, Invitrogen, Waltham, MA) at 4 °C for 24–48 h. After being washed in PBS, sections were incubated with a goat anti-rabbit Cy5 (1:500, A10523, Invitrogen) at room temperature for 3 h. After another rinse with PBS, sections were mounted on slides and cover-slipped with a mounting medium (H-1500, Vector Labs). To identify subtypes of c-Fos positive neurons following the social recognition test, sections were incubated with primary antibodies of guinea pig anti-c-Fos (1:1000, 226308, Synaptic Systems, Gottingen, Germany), rabbit anti-CaMKII (1:200, PA5-99558, Invitrogen) and mouse anti-GAD67 (1:500, MAB5406, Millipore, Burlington, MA) at 4 °C for 24–48 h. After being washed in PBS, sections were incubated with secondary antibodies of goat anti-guinea pig Alexa 488 (1:1000, A11073, Invitrogen), goat anti-rabbit Alexa 555 (1:1000, A21428, Invitrogen) and goat anti-mouse Alexa 647 (1:1000, A21236, Invitrogen) at room temperature for 3 h. After another rinse with PBS, sections were mounted on slides and cover-slipped with a mounting medium (H-1500, Vector Labs). Images were taken with a Zeiss LSM 710 confocal microscope and Zen 2.0 software (Oberkochen, Germany).
Neuronal tracing
Ai9 (RCL-tdT) mice (n = 3) infused with AAV1-hSyn-Cre in the FN and AAVrg-hSyn-EGFP in the BLA were used for anatomical tracing. Via Cre-LoxP recombination, neurons in the FN were labeled with tdTomato59. As AAV1-hSyn-Cre has a transsynaptic property60, it further transduced FN downstream neurons. Example images from such a downstream region (PL) was shown in Supplementary Fig. 8. AAVrg-hSyn-EGFP is a retrograde virus63. It targeted the BLA and its upstream neurons. 4-5 weeks after viral infusion, brain sections were collected in series with a 200 µm interval, mounted on slides and cover-slipped with a mounting medium (H-1500, Vector Labs). Images were taken with a Zeiss LSM 710 confocal microscope and Zen 2.0 software. Location of somas, dendrites and axons of tdTomato- and EGFP-expressing neurons was identified based on the mouse brain atlas79.
Interregional c-Fos correlations and graph theoretical analysis
Matrices of pairwise correlations were created by calculating Pearson coefficients (r) from the numbers of c-Fos-positive cells inter-regionally within each group (n = 6 mice). r was Fisher Z transformed. Statistics was conducted to compare average r between groups. Positive and negative r were indicated in red and indigo respectively in a gradient spectrum. Each node represented a brain region, and each connecting line represented Pearson’s r between regions. Graph theoretical analysis was used to illustrate network properties for social memory. The network was constructed with unweighted adjacency matrices for each group based on significant (uncorrected p < 0.05) and positive (r > 0.6) interregional correlations of c-Fos counts. Clusters were identified through enumeration of all potential community structures and optimization of the modularity. Degree (the number of edges that a node has) and betweenness (the number of shortest paths that pass through a node) were calculated to reveal the centrality of a node in the network. Within-community connectivity (Z-score, the extent of a node connected in its own community) and between-community connectivity (participation coefficient, the extent of a node connected to other communities) were calculated to reveal the modularity of the network. Graph theoretical analysis was performed in MATLAB (R2015b) with Brain Connectivity Toolbox (https://sites.google.com/site/bctnet/). Calculation details were described previously55,56.
Statistics
One-way or mixed two-way ANOVA with a between factor (“group”) and a within factor (“object”, “interval” or “trial”) was applied to analyze data. Paired or unpaired t-tests were used when main effects were found. One sample t-test was used to compare the indices to 0 (chance level). Fisher’s LSD was applied to post hoc tests when appropriate. Sample sizes were based on previous studies using similar experimental protocols25,43,44. For behavioral tests, n denoted the number of mice. For other analyses, n denoted the number of cells or samples from >3 mice/group. All measurements were taken from distinct samples. n values were given in the Source Data file and figure legends. All statistics were two-tailed tests with significance set as p < 0.05 using IBM SPSS statistics 25 (IBM Corp, Armonk, NY) and GraphPad Prism 9.3.1 (GraphPad Software, San Diego, CA). Data were presented as the mean ± standard error of the mean (SEM). Statistical details including p values and degrees of freedom were summarized in Supplementary Table 1.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All data are available in the main text and supplementary information. Source data are provided with this paper and deposited to the Figshare (https://doi.org/10.6084/m9.figshare.22647700). The mouse brain atlas used in this study is publicly available79.
References
Adolphs, R. The neurobiology of social cognition. Curr. Opin. Neurobiol. 11, 231–239 (2001).
Insel, T. R. & Fernald, R. D. How the brain processes social information: searching for the social brain. Annu Rev. Neurosci. 27, 697–722 (2004).
Frith, C. D. & Frith, U. Mechanisms of social cognition. Annu Rev. Psychol. 63, 287–313 (2012).
Pelphrey, K., Adolphs, R. & Morris, J. P. Neuroanatomical substrates of social cognition dysfunction in autism. Ment. Retard Dev. Disabil. Res Rev. 10, 259–271 (2004).
Couture, S. M., Penn, D. L. & Roberts, D. L. The functional significance of social cognition in schizophrenia: a review. Schizophr. Bull. 32, S44–S63 (2006).
Elamin, M., Pender, N., Hardiman, O. & Abrahams, S. Social cognition in neurodegenerative disorders: a systematic review. J. Neurol. Neurosurg. Psych. 83, 1071–1079 (2012).
Cacciola, A. et al. The Known and Missing Links Between the Cerebellum, Basal Ganglia, and Cerebral Cortex. Cerebellum 16, 753–755 (2017).
Bostan, A. C. & Strick, P. L. The basal ganglia and the cerebellum: nodes in an integrated network. Nat. Rev. Neurosci. 19, 338–350 (2018).
Stoodley, C. J. & Tsai, P. T. Adaptive Prediction for Social Contexts: The Cerebellar Contribution to Typical and Atypical Social Behaviors. Annu Rev. Neurosci. 44, 475–493 (2021).
Kim, J. & Augustine, G. J. Molecular Layer Interneurons: Key Elements of Cerebellar Network Computation and Behavior. Neuroscience 462, 22–35 (2021).
De Zeeuw, C. I. & Berrebi, A. S. Postsynaptic targets of Purkinje cell terminals in the cerebellar and vestibular nuclei of the rat. Eur. J. Neurosci. 7, 2322–2333 (1995).
Strick, P. L., Dum, R. P. & Fiez, J. A. Cerebellum and nonmotor function. Annu Rev. Neurosci. 32, 413–434 (2009).
Sokolov, A. A., Miall, R. C. & Ivry, R. B. The Cerebellum: Adaptive Prediction for Movement and Cognition. Trends Cogn. Sci. 21, 313–332 (2017).
Krienen, F. M. & Buckner, R. L. Segregated fronto-cerebellar circuits revealed by intrinsic functional connectivity. Cereb. Cortex 19, 2485–2497 (2009).
Stoodley, C. J. & Schmahmann, J. D. Functional topography in the human cerebellum: a meta-analysis of neuroimaging studies. Neuroimage 44, 489–501 (2009).
Stoodley, C. J. & Schmahmann, J. D. Evidence for topographic organization in the cerebellum of motor control versus cognitive and affective processing. Cortex 46, 831–844 (2010).
Schmahmann, J. D., Oblak, A. L. & Blatt, G. J. in Handbook of the Cerebellum and Cerebellar Disorders (eds Mario U. Manto et al.) 605–624 (Springer International Publishing, 2022).
Heath, R. G., Dempesy, C. W., Fontana, C. J. & Myers, W. A. Cerebellar stimulation: effects on septal region, hippocampus, and amygdala of cats and rats. Biol. Psych 13, 501–529 (1978).
Dietrichs, E. & Haines, D. E. Interconnections between hypothalamus and cerebellum. Anat. Embryol. (Berl.) 179, 207–220 (1989).
Zhu, J. N., Yung, W. H., Kwok-Chong Chow, B., Chan, Y. S. & Wang, J. J. The cerebellar-hypothalamic circuits: potential pathways underlying cerebellar involvement in somatic-visceral integration. Brain Res Rev. 52, 93–106 (2006).
Fujita, H., Kodama, T. & du Lac, S. Modular output circuits of the fastigial nucleus for diverse motor and nonmotor functions of the cerebellar vermis. Elife 9, e58613 (2020).
Pisano, T. J. et al. Homologous organization of cerebellar pathways to sensory, motor, and associative forebrain. Cell Rep. 36, 109721 (2021).
Jackman, S. L. et al. Cerebellar Purkinje cell activity modulates aggressive behavior. Elife 9, e53229 (2020).
Zeidler, Z., Hoffmann, K. & Krook-Magnuson, E. HippoBellum: Acute Cerebellar Modulation Alters Hippocampal Dynamics and Function. J. Neurosci. 40, 6910–6926 (2020).
Chao, O. Y., Zhang, H., Pathak, S. S., Huston, J. P. & Yang, Y. M. Functional Convergence of Motor and Social Processes in Lobule IV/V of the Mouse Cerebellum. Cerebellum 20, 836–852 (2021).
Hitti, F. L. & Siegelbaum, S. A. The hippocampal CA2 region is essential for social memory. Nature 508, 88–92 (2014).
Okuyama, T., Kitamura, T., Roy, D. S., Itohara, S. & Tonegawa, S. Ventral CA1 neurons store social memory. Science 353, 1536–1541 (2016).
Zinn, C. G. et al. Major neurotransmitter systems in dorsal hippocampus and basolateral amygdala control social recognition memory. P Natl Acad. Sci. USA 113, E4914–E4919 (2016).
Tanimizu, T. et al. Functional Connectivity of Multiple Brain Regions Required for the Consolidation of Social Recognition Memory. J. Neurosci. 37, 4103–4116 (2017).
Phillips, M. L., Robinson, H. A. & Pozzo-Miller, L. Ventral hippocampal projections to the medial prefrontal cortex regulate social memory. Elife 8, e44182 (2019).
Molenberghs, P., Cunnington, R. & Mattingley, J. B. Brain regions with mirror properties: a meta-analysis of 125 human fMRI studies. Neurosci. Biobehav Rev. 36, 341–349 (2012).
Van Overwalle, F., Ma, Q. & Heleven, E. The posterior crus II cerebellum is specialized for social mentalizing and emotional self-experiences: a meta-analysis. Soc. Cogn. Affect Neurosci. 15, 905–928 (2020).
Gerwig, M., Kolb, F. P. & Timmann, D. The involvement of the human cerebellum in eyeblink conditioning. Cerebellum 6, 38–57 (2007).
Heiney, S. A., Kim, J., Augustine, G. J. & Medina, J. F. Precise control of movement kinematics by optogenetic inhibition of Purkinje cell activity. J. Neurosci. 34, 2321–2330 (2014).
ten Brinke, M. M. et al. Evolving Models of Pavlovian Conditioning: Cerebellar Cortical Dynamics in Awake Behaving Mice. Cell Rep. 13, 1977–1988 (2015).
Boele, H. J. et al. Impact of parallel fiber to Purkinje cell long-term depression is unmasked in absence of inhibitory input. Sci. Adv. 4, eaas9426 (2018).
Badura, A. et al. Normal cognitive and social development require posterior cerebellar activity. Elife 7, e36401 (2018).
Rowan, M. J. M. et al. Graded Control of Climbing-Fiber-Mediated Plasticity and Learning by Inhibition in the Cerebellum. Neuron 99, 999–1015.e1016 (2018).
Wulff, P. et al. Synaptic inhibition of Purkinje cells mediates consolidation of vestibulo-cerebellar motor learning. Nat. Neurosci. 12, 1042–1049 (2009).
Yip, J., Soghomonian, J. J. & Blatt, G. J. Increased GAD67 mRNA expression in cerebellar interneurons in autism: implications for Purkinje cell dysfunction. J. Neurosci. Res. 86, 525–530 (2008).
Tsai, P. T. et al. Autistic-like behaviour and cerebellar dysfunction in Purkinje cell Tsc1 mutant mice. Nature 488, 647–651 (2012).
Cupolillo, D. et al. Autistic-Like Traits and Cerebellar Dysfunction in Purkinje Cell PTEN Knock-Out Mice. Neuropsychopharmacology 41, 1457–1466 (2016).
Chao, O. Y. et al. Targeting inhibitory cerebellar circuitry to alleviate behavioral deficits in a mouse model for studying idiopathic autism. Neuropsychopharmacology 45, 1159–1170 (2020).
Yang, Y. M. et al. Identification of a molecular locus for normalizing dysregulated GABA release from interneurons in the Fragile X brain. Mol. Psych. 25, 2017–2035 (2020).
Simmons, D. H., Titley, H. K., Hansel, C. & Mason, P. Behavioral Tests for Mouse Models of Autism: An Argument for the Inclusion of Cerebellum-Controlled Motor Behaviors. Neuroscience 462, 303–319 (2021).
Bullitt, E. Expression of c-fos-like protein as a marker for neuronal activity following noxious stimulation in the rat. J. Comp. Neurol. 296, 517–530 (1990).
Amat, S. B. et al. Using c-kit to genetically target cerebellar molecular layer interneurons in adult mice. PLoS One 12, e0179347 (2017).
Chao, O. Y., de Souza Silva, M. A., Yang, Y. M. & Huston, J. P. The medial prefrontal cortex— - hippocampus circuit that integrates information of object, place and time to construct episodic memory in rodents: Behavioral, anatomical and neurochemical properties. Neurosci. Biobehav Rev. 113, 373–407 (2020).
Josselyn, S. A. & Tonegawa, S. Memory engrams: Recalling the past and imagining the future. Science 367, eaaw4325 (2020).
Kim, J. et al. Optogenetic mapping of cerebellar inhibitory circuitry reveals spatially biased coordination of interneurons via electrical synapses. Cell Rep. 7, 1601–1613 (2014).
Herweg, N. A., Solomon, E. A. & Kahana, M. J. Theta Oscillations in Human Memory. Trends Cogn. Sci. 24, 208–227 (2020).
Dave, S., VanHaerents, S. & Voss, J. L. Cerebellar Theta and Beta Noninvasive Stimulation Rhythms Differentially Influence Episodic Memory versus Semantic Prediction. J. Neurosci. 40, 7300–7310 (2020).
Chaudhuri, A., Zangenehpour, S., Rahbar-Dehgan, F. & Ye, F. Molecular maps of neural activity and quiescence. Acta Neurobiol. Exp. (Wars.) 60, 403–410 (2000).
Aminoff, E. M., Kveraga, K. & Bar, M. The role of the parahippocampal cortex in cognition. Trends Cogn. Sci. 17, 379–390 (2013).
Newman, M. E. & Girvan, M. Finding and evaluating community structure in networks. Phys. Rev. E Stat. Nonlin Soft Matter Phys. 69, 026113 (2004).
Guimera, R. & Amaral, L. A. Cartography of complex networks: modules and universal roles. J. Stat. Mech. 2005, nihpa35573 (2005).
Hermans, E. J. et al. How the amygdala affects emotional memory by altering brain network properties. Neurobiol. Learn Mem. 112, 2–16 (2014).
McEwen, B. S., Nasca, C. & Gray, J. D. Stress Effects on Neuronal Structure: Hippocampus, Amygdala, and Prefrontal Cortex. Neuropsychopharmacology 41, 3–23 (2016).
Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13, 133–140 (2010).
Zingg, B. et al. AAV-mediated anterograde transsynaptic tagging: Mapping corticocollicular input-defined neural pathways for defense behaviors. Neuron 93, 33–47 (2017).
Sigurdsson, T. & Duvarci, S. Hippocampal-Prefrontal Interactions in Cognition, Behavior and Psychiatric Disease. Front Syst. Neurosci. 9, 190 (2015).
Janak, P. H. & Tye, K. M. From circuits to behaviour in the amygdala. Nature 517, 284–292 (2015).
Tervo, D. G. et al. A Designer AAV Variant Permits Efficient Retrograde Access to Projection Neurons. Neuron 92, 372–382 (2016).
McDonald, A. J. Functional neuroanatomy of the basolateral amygdala: Neurons, neurotransmitters, and circuits. Handb. Behav. Neurosci. 26, 1–38 (2020).
Sun, L. et al. Reactivating a positive feedback loop VTA-BLA-NAc circuit associated with positive experience ameliorates the attenuated reward sensitivity induced by chronic stress. Neurobiol. Stress 15, 100370 (2021).
Cook, A. A., Fields, E. & Watt, A. J. Losing the Beat: Contribution of Purkinje Cell Firing Dysfunction to Disease, and Its Reversal. Neuroscience 462, 247–261 (2021).
Wang, S. S., Kloth, A. D. & Badura, A. The cerebellum, sensitive periods, and autism. Neuron 83, 518–532 (2014).
Hampson, D. R. & Blatt, G. J. Autism spectrum disorders and neuropathology of the cerebellum. Front Neurosci. 9, 420 (2015).
Schmahmann, J. D. & Sherman, J. C. The cerebellar cognitive affective syndrome. Brain 121, 561–579 (1998).
Deverett, B., Koay, S. A., Oostland, M. & Wang, S. S. Cerebellar involvement in an evidence-accumulation decision-making task. Elife 7, e36781 (2018).
Deverett, B., Kislin, M., Tank, D. W. & Wang, S. S. Cerebellar disruption impairs working memory during evidence accumulation. Nat. Commun. 10, 3128 (2019).
Ma, M. et al. Molecular layer interneurons in the cerebellum encode for valence in associative learning. Nat. Commun. 11, 4217 (2020).
Carta, I., Chen, C. H., Schott, A. L., Dorizan, S. & Khodakhah, K. Cerebellar modulation of the reward circuitry and social behavior. Science 363, eaav0581 (2019).
Liu, Y. et al. Causal Evidence for a Role of Cerebellar Lobulus Simplex in Prefrontal-Hippocampal Interaction in Spatial Working Memory Decision-Making. Cerebellum 21, 762–775 (2022).
Zeineh, M. M., Engel, S. A., Thompson, P. M. & Bookheimer, S. Y. Dynamics of the hippocampus during encoding and retrieval of face-name pairs. Science 299, 577–580 (2003).
Eldridge, L. L., Engel, S. A., Zeineh, M. M., Bookheimer, S. Y. & Knowlton, B. J. A dissociation of encoding and retrieval processes in the human hippocampus. J. Neurosci. 25, 3280–3286 (2005).
Eichenbaum, H., Yonelinas, A. P. & Ranganath, C. The medial temporal lobe and recognition memory. Annu Rev. Neurosci. 30, 123–152 (2007).
Hernandez-Briones, Z. S. et al. Olfactory stimulation induces cerebellar vermis activation during sexual learning in male rats. Neurobiol. Learn Mem. 146, 31–36 (2017).
Franklin, K. B. J. & Paxinos, G. The mouse brain in stereotaxic coordinates, compact third edition. (Academic Press; 3rd edition, 2008).
Acknowledgements
This study was supported by the National Institutes of Health (NIH) grants R15 NS112964 and R01 MH129300 to Y.M.Y. and R01 NS112289 to J.M.C., the Singapore Ministry of Education grant MOE2017-T3-1-002 to G.J.A., the Brain & Behavior Research Foundation Young Investigator grant to O.Y.C., and the Winston and Maxine Wallin Neuroscience Discovery Fund to YMY. AAV8-hSyn-DIO-hM3Dq-mCherry was generated by Dr. Ezequiel Marron Fernandez de Velasco in the University of Minnesota Viral Vector and Cloning Core. AAV1-hSyn-Cre and AAVrg-hSyn-EGFP were purchased from Addgene (gifts from Drs. James M Wilson and Bryan Roth). We thank the Structural Circuits Core (University of Minnesota) for providing technical support and Dr. Pei Wern Chin (Nanyang Technological University) for critical input.
Author information
Authors and Affiliations
Contributions
O.Y.C. and Y.M.Y. designed the project, performed the experiments, and analyzed the data with assistance from S.S.P. and HZ. G.J.A., J.M.C., C.K. and H.T. provided the mouse lines and technical input. O.Y.C. and Y.M.Y. wrote an early version of the manuscript and all other authors contributed to the final edition.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Diego Restrepo, Roy Sillitoe and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Chao, O.Y., Pathak, S.S., Zhang, H. et al. Social memory deficit caused by dysregulation of the cerebellar vermis. Nat Commun 14, 6007 (2023). https://doi.org/10.1038/s41467-023-41744-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-023-41744-2
This article is cited by
-
Cerebellar vermis joins the brain’s social network
Communications Biology (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.