Empathy is crucial for our emotional experience and social interactions, and its abnormalities manifest in various psychiatric disorders. Observational fear is a useful behavioral paradigm for assessing affective empathy in rodents. However, specific genes that regulate observational fear remain unknown. Here we showed that 129S1/SvImJ mice carrying a unique missense variant in neurexin 3 (Nrxn3) exhibited a profound and selective enhancement in observational fear. Using the CRISPR/Cas9 system, the arginine-to-tryptophan (R498W) change in Nrxn3 was confirmed to be the causative variant. Selective deletion of Nrxn3 in somatostatin-expressing (SST+) interneurons in the anterior cingulate cortex (ACC) markedly increased observational fear and impaired inhibitory synaptic transmission from SST+ neurons. Concordantly, optogenetic manipulation revealed that SST+ neurons in the ACC bidirectionally controlled the degree of socially transmitted fear. Together, these results provide insights into the genetic basis of behavioral variability and the neurophysiological mechanism controlling empathy in mammalian brains.

Large image of Figure 1.


Empathy is the ability to recognize and share the feelings of others. The neurocognitive processing of this affective and cognitive information is fundamental for our emotional and social lives (Bernhardt and Singer, 2012de Waal, 2008). Either elevated or reduced empathy can contribute to difficulties in social interactions and mental well-being. Disturbance of empathy is a salient feature of many neuropsychiatric conditions, particularly autism spectrum disorders (ASDs) and psychopathy (Baron-Cohen and Wheelwright, 2004Bora et al., 2008). Although there is a considerable genetic contribution to individual variability in empathy (Ebstein et al., 2010Knafo et al., 2009Rodrigues et al., 2009Warrier et al., 2017), identification of specific genes that determine individual variability in empathy has been largely limited, primarily because it is difficult to control the social context in humans.

Recent evidence shows that empathy is evolutionarily conserved from rodents to humans (de Waal and Preston, 2017Decety, 2011). Rodents possess a remarkable affective sensitivity to the emotional state of others and show empathy-related behaviors such as observational fear, emotional contagion of pain, social buffering, and prosocial helping behaviors (Ben-Ami Bartal et al., 2011Burkett et al., 2016Church, 1959Jeon et al., 2010Langford et al., 2006). In particular, observational fear has been recognized as a useful behavioral model for assessing empathic fear capacity (Debiec and Olsson, 2017Keum and Shin, 2016Meyza et al., 2017Panksepp and Lahvis, 2011Sivaselvachandran et al., 2016). In observational fear, a mouse is vicariously conditioned for fear by observing a conspecific receiving aversive foot shocks. This phenomenon, referred to as emotional state-matching or affect sharing, was measured as socially transmitted fear (Chen et al., 2009Jeon et al., 2010). Human performance in a similar observational fear learning process was correlated with trait measures of empathy (Haaker et al., 2017Kleberg et al., 2015Olsson et al., 2007), suggesting that social transfer of fear is a fundamental feature of empathy that is conserved across species (Olsson and Phelps, 2007Panksepp and Panksepp, 2013). Brain imaging studies have contributed to the understanding of the neural circuitries involved in empathy. Specifically, the anterior cingulate cortex (ACC) is involved in empathic responses of pain or fear (Olsson et al., 2007Singer et al., 2004). Likewise, the activity of the ACC is augmented in mice engaged in observational fear, and its role in the acquisition of vicarious freezing has been demonstrated using neuroanatomical lesions (Jeon et al., 2010Kim et al., 2012). However, despite the accumulating information about executive neural circuitry controlling observational fear, specific genes that determine different presentations of empathy-related behaviors are poorly understood.

To address these issues, we recently surveyed multiple inbred mouse strains and found that the vicarious freezing response was highly variable among different strains, suggesting that the innate observational fear response is under genetic control (Keum et al., 2016). By comparing a panel of genetically nearly identical 129 Steel-lineage (129S) substrains, we identified that a missense variant in the Nrxn3 gene present only in the 129S1/SvImJ (129S1) strain enhanced empathy fear. Using a combination of approaches, including cell type-specific ablation, ex vivoslice electrophysiology, and optogenetic manipulation, we demonstrate that Nrxn3-dependent somatostatin-expressing (SST+) interneurons in the ACC control the degree of social transfer of fear in mice.


The R498W Variant in Nrxn3 Causes Elevation of Observation Fear

In the observational fear task, without receiving direct aversive stimuli, a mouse (observer) is vicariously conditioned for context-dependent fear by observing another mouse (demonstrator) receiving repetitive foot shocks (Figure 1A). Mice of the 129S1 inbred strain exhibited a marked increase in observational fear (Video S1) compared with other closely related 129S substrains, including 129S2/SvPas (129S2), 129S4/SvJaeJ (129S4), and 129S8/SvEvNimrJ (129S8) (Figures 1B and 1C; Video S2). Surprisingly, we found that the level of observational fear response in 129S1 mice was significantly higher than that in any of the 17 other common inbred strains examined (Figures 1D and 1E). This elevated vicarious fear response in 129S1 mice did not significantly correlate with variation in locomotion, anxiety, or Pavlovian fear conditioning (Keum et al., 2016), leading us to hypothesize that this extreme phenotype in 129S1 mice was caused by observational fear-specific genetic variations unique to this strain.

To identify a causative variant, we performed whole-genome sequencing (WGS) of 129S1 and 129S4 strains (Table S1) and identified 32 non-synonymous coding SNPs in 23 genes that differed between 129S1 and 129S4 mice (Tables S2A and S2B). To further validate and identify 129S1-unique SNPs, we compared those 32 coding variants with published genome sequences of 17 inbred mouse strains (Keane et al., 2011). Only eight were 129S1-unique SNPs that distinguished the 129S1 strain from the other 17 strains (Table S3). We prioritized them based on the predicted consequences of the coding changes on the protein function and the mRNA abundance of the genes in the brain (STAR Methods). This analysis identified a homozygous non-synonymous SNP (rs241832271) that changes C to T at the 89,254,694 base pair (bp) on chromosome 12 as a top candidate to account for the altered behavioral phenotype of the 129S1 mice (Figure 2A). This SNP occurs in exon 6 of the neurexin 3 (Nrxn3) gene, encoding an evolutionarily conserved synaptic cell adhesion molecule that is essential for normal synapse assembly and synaptic transmission (Reissner et al., 2013Südhof, 2008). The C-to-T change produces an arginine-to-tryptophan protein change at position 498 (R498W) in the third extracellular LNS (laminin-neurexin-sex hormone binding globulin) domain of NRXN3, which is highly conserved among vertebrates (Figures 2B and 2C). Notably, this change is predicted to be deleterious (Figure S1). The R498W variant was present only in 129S1 mice; none of the other inbred strains, including wild-derived mice, shared the T allele (Figure 2D; Table S2B), suggesting that this variant is not ancestrally inherited (Yang et al., 2011). To further pursue the historical origin of this variation, we performed DNA sequencing for the variant in an additional 129 substrains—129S2, 129S8, and 129T2—representing different genetic lineages (Figure 2E). The times of separation from the founder colony of these strains have been well documented (Festing et al., 1999Simpson et al., 1997), allowing construction of a phylogenetic timeline of this Nrxn3 variant (Figure 2F). Intriguingly, the R498W variant at the Nrxn3 locus was fixed only in the 129S1 colony during selective breeding (Simpson et al., 1997) and was not present in most commercially available 129 strains. No 129S1-unique insertions or deletions (indels) or structural variants that caused protein coding changes were found.

To confirm that the R498W variant in Nrxn3 caused elevated observational fear, we introduced the C > T non-synonymous change into B6J mice using CRISPR/Cas9 genome editing (Figure 3A). As expected, knockin (KI) mice harboring a homozygous Trp498 allele (KI-Nrxn3WW) exhibited a significantly higher level of vicarious fear response than littermate wild-type (WT) mice (KI-Nrxn3RRFigure 3B). No difference was found in 24-hr memory between the two groups (Figure 3C). To further explore the effect of this coding variant on conditioned fear, we examined the KI-Nrxn3WW mice on a classical fear conditioning task but found no significant change in fear conditioning or 24-hr contextual fear memory (Figures 3D and 3E). Thus, the R498W variant in the Nrxn3 gene specifically increased the degree of behavioral response of observer mice to the distress of demonstrator mice, indicating that KI-Nrxn3WW mice phenocopied the behavior of 129S1 mice.

SST+ Neuron-Specific Deletion of Nrxn3 in the ACC Increases Observational Fear

The ACC was shown to be crucial for the acquisition of observational fear (Jeon et al., 2010Kim et al., 2012). In the cortex, information processing depends on highly interconnected microcircuits composed of excitatory glutamatergic pyramidal and γ-aminobutyric acid-releasing (GABAergic) inhibitory neurons (Figure 4A) (Tremblay et al., 2016). Although the lack of high-affinity antibodies has hindered the assessment of Nrxn3 protein expression (Reissner et al., 2013Südhof, 2008), in situ hybridization and mRNA transcriptome studies have demonstrated that Nrxn3 is highly expressed in the cortex (Aoto et al., 2013Chen et al., 2017Schreiner et al., 2014Treutlein et al., 2014). In addition, Nrxn3 shows distinct synaptic functions in different brain regions (Aoto et al., 2015). Thus, to elucidate the role of Nrxn3 in observational fear, we used a cell type-specific targeting approach to dissect its specific involvement in distinct neuronal populations. First, to selectively delete Nrxn3 in excitatory glutamatergic neurons of the forebrain, we generated conditional knockout (KO) mice by breeding mice that harbored a Nrxn3 conditional allele (Nrxn3f/f) with Emx1-Cre mice (Gorski et al., 2002) (Emx1cre/+Nrxn3f/f, designated Emx1-Nrxn3 KO). The Emx1-Nrxn3 KO mice showed levels of observational fear similar to those of their WT littermates (Figure 4B). There was also no difference in 24-hr memory between the genotypes (Figure 4C). To confirm this, we deleted Nrxn3 in putative excitatory cortical neurons by focally injecting an adeno-associated virus (AAV) expressing Cre recombinase under the control of the calcium/calmodulin-dependent protein kinase IIα promotor (AAV-Camk2α-Cre) into the ACC of Nrxn3f/f mice (Figure S2A). Similar to Emx1-Nrxn3 KO mice, observer mice with a localized Nrxn3 deletion in excitatory neurons in the ACC exhibited no difference in either acquisition or 24-hr memory of observational fear compared with control mice (Figures S2B and S3C), indicating that Nrxn3 in cortical pyramidal neurons is not critically involved in the regulation of observational fear.

Next, to examine the role of Nrxn3 in GABAergic inhibitory neuron populations, we first crossed conditional Nrxn3f/f mice with pan-GABAergic Vgat-Cre mice in which Cre recombinase is expressed under the control of the GABA vesicular transporter (Slc32a1) (Vong et al., 2011). However, we found that, when crossed to homozygosity, Vgat-Nrxn3 KO mice were not viable (live births: 20 WT, 47 heterozygous, and 0 homozygous KO), consistent with a previous report that germline Nrxn3-KO mice die at birth (Aoto et al., 2015). These results underscored the possibility that Nrxn3 plays a critical role in GABAergic synapse development. This finding prompted us to further explore the consequence of Nrxn3 ablation on the synaptic function of specific GABAergic inhibitory neuronal populations. To this end, we generated three lines of conditional KO mice lacking Nrxn3 in parvalbumin-expressing (PV+) neurons (PV-Nrxn3 KO), SST+ neurons (SST-Nrxn3 KO), or vasoactive intestinal peptide-expressing (VIP+) neurons (VIP-Nrxn3 KO) by crossing Nrxn3f/f mice with PV-Cre, SST-Cre, or VIP-Cre mice, respectively, representing the majority (>80%) of the GABAergic neuronal population in the cortex (Hippenmeyer et al., 2005Taniguchi et al., 2011Tremblay et al., 2016). Strikingly, we found that SST-Nrxn3 KO observer mice showed greatly increased vicarious freezing compared with their WT littermates (Figure 4D; Video S3). 24-hr memory was also significantly higher in SST-Nrxn3 KO mice compared with WT mice (Figure 4E). By contrast, both PV-Nrxn3 KO and VIP-Nrxn3 KO mice exhibited no difference in observational fear compared with their WT controls (Figures 4F–4I). We tested SST-Nrxn3 KO mice on a classical fear conditioning task but found no difference between the KO and WT mice (Figures S3A and S3B), highlighting the specific role of Nrxn3in SST+ neurons in the modulation of observational fear. Next, we examined whether the elevated observational fear in SST-Nrxn3 KO mice was due to the loss of Nrxn3 in SST+ neurons in the ACC, which is integral to the acquisition of vicarious fear (Jeon et al., 2010Kim et al., 2012). To this end, we bred SST-Flp;Nrxn3f/f mice, in which flippase (Flp) recombinase was selectively expressed in SST+ neurons in a Nrxn3f/f genetic background, and injected the ACC of these mice with an AAV expressing Cre recombinase in a Flp-dependent manner with a double-floxed inverted open reading frame (AAV-fDIO-Cre). This approach allowed SST+ neuron-specific deletion of Nrxn3 restricted to the ACC area (Figure 4J). Indeed, we found that this resulted in elevated vicarious fear responses (Figures 4K and 4L), a phenotype that resembled that of the SST-Nrxn3 KO mice. Thus, the lack of Nrxn3 in SST+ neurons in the ACC caused elevation of observational fear.

Reduced GABAergic Transmission from SST+ Neurons Lacking Nrxn3

SST+ interneurons primarily target distal dendrites of pyramidal cells and have a prominent role in regulating distal dendritic excitability (Chiu et al., 2013Dumitriu et al., 2007Gentet et al., 2012Urban-Ciecko et al., 2015). Thus, we measured synaptic functions in SST-Nrxn3 KO mice by performing whole-cell patch-clamp recordings in layer 2/3 (L2/3) of the right ACC in acute brain slices. Given that interneuron activity-dependent inhibition tightly modulates the output of excitatory neurons, we first measured the intrinsic excitability of putative pyramidal neurons and SST+ interneurons in SST-Nrxn3 KO mice. To identify SST+ interneurons, we crossed SST-Nrxn3 KO mice with a Cre-dependent Rosa26LSL-tdTomato (Ai14) reporter line (SSTcre/+;Nrxn3f/f;Ai14f/+) to label SST+ neurons with a red fluorescent protein (Madisen et al., 2010). All recorded putative pyramidal cells exhibited a regular adaptation firing discharge pattern, and their average frequency of action potential discharge at incremental step current injections was similar between WT and SST-Nrxn3 KO mice (Figures S4A and S4B). SST+ interneurons lacking Nrxn3also showed no difference in intrinsic excitability (Figures S4C and S4D).

See full article

Leave a Reply

Your email address will not be published. Required fields are marked *