Neuropsin cleaves EphB2 in the amygdala to control anxiety.

A minority of individuals experiencing traumatic events develop anxiety disorders. The reason for the lack of correspondence between the prevalence of exposure to psychological trauma and the development of anxiety is unknown. Extracellular proteolysis contributes to fear-associated responses by facilitating neuronal plasticity at the neuron-matrix interface. Here we show in mice that the serine protease neuropsin is critical for stress-related plasticity in the amygdala by regulating the dynamics of the EphB2-NMDA-receptor interaction, the expression of Fkbp5 and anxiety-like behaviour. Stress results in neuropsin-dependent cleavage of EphB2 in the amygdala causing dissociation of EphB2 from the NR1 subunit of the NMDA receptor and promoting membrane turnover of EphB2 receptors. Dynamic EphB2-NR1 interaction enhances NMDA receptor current, induces Fkbp5 gene expression and enhances behavioural signatures of anxiety. On stress, neuropsin-deficient mice do not show EphB2 cleavage and its dissociation from NR1 resulting in a static EphB2-NR1 interaction, attenuated induction of the Fkbp5 gene and low anxiety. The behavioural response to stress can be restored by intra-amygdala injection of neuropsin into neuropsin-deficient mice and disrupted by the injection of either anti-EphB2 antibodies or silencing the Fkbp5 gene in the amygdala of wild-type mice. Our findings establish a novel neuronal pathway linking stress-induced proteolysis of EphB2 in the amygdala to anxiety.

Fear helps organisms recognize, memorize and predict danger, thereby promoting their survival. However, severe stress can trigger maladaptive forms of neuronal remodelling leading to generalization of fear and high anxiety5.

Traumatic events are memorized due to the capacity of synaptic connections and the surrounding matrix to undergo experience-dependent functional or morphological changes1, 6. Extracellular proteases are strategically poised to remodel the neuron-extracellular matrix interface and facilitate fear and anxiety2-4. Eph-receptor tyrosine kinases constitute an important group of molecules subject to modulation by extracellular proteases7. While Ephs promote neuronal plasticity8, 9 their involvement in behavioural responses to environmental stimuli is not clear.

Neuropsin is a serine protease uniquely positioned to facilitate stress-induced plasticity due to its high expression in the amygdala and hippocampus10. To investigate if neuropsin and Ephs co-localize we performed immunohistochemistry. Consistent with previous reports10, 11 we found robust expression of both neuropsin and EphB2 in the amygdala (Figure 1 and Suppl. Fig. 1) and the hippocampus (not shown). Double immunohistochemistry revealed high levels of neuropsin co-localizing with EphB2-rich clusters on amygdala neurons (Figure 1a).

Figure 1

Neuropsin and EphB2 colocalize in neurons of the basolateral complex of the amygdala

(a) Double immunohistochemistry showed neuropsin (green) and EphB2 (red) co-localize in lateral amygdala neurons (arrows show EphB2-rich clusters at neuropsin detection sites). Cells were highlighted with TOTO-3 stain. (b) Triple immunohistochemistry confirmed the presence of neuropsin/EphB2-rich clusters on the neuronal surface and low degree of co-localization with cytoplasmic Fkbp51. (c, d) Western blotting revealed amydala neuropsin upregulation following 6 hour restraint stress (F(2, 7) =8.81; p<0.05). Digits inside columns indicate n number. *p<0.05. Results are shown as mean±SEM.

To assess whether Ephs are modulated by neuropsin we treated SH-SY5Y cells with neuropsin and measured the levels of Eph receptors by Western blotting. We found that neuropsin (but not other proteases; Suppl. Fig. 2) cleaved EphB2 (decrease by 41%, p<0.001), while the levels of other Ephs or their ligand ephrinB2 remained unchanged (Figure 2a, b and Suppl. Fig. 3a). When we expressed either GFP-tagged EphB2, GFP-tagged EphA4 or unlinked GFP in SH-SY5Y cells (Suppl. Fig. 4 and 5) and treated them with neuropsin we saw a similar decrease in the EphB2-associated signal (Suppl. Fig. 5; p<0.05).

Figure 2

Neuropsin cleaves EphB2 and regulates its expression both in vitro and in the amygdala after stress

(a, b) EphB2-S band density in SH-SY5Y cells decreased upon 15 min neuropsin treatment (F(3,18)=11.24; p<0.001). Neuropsin did not cleave other molecules of the same class (b and Suppl. Fig. 3a). (c) Exposure of EphB2-GFP-transfected SH-SY5Y or HEK293 cells to neuropsin (15 or 45 min) resulted in the appearance of a ~70kDa N-terminal EphB2 fragment in the medium (Suppl. Fig. 6). (d, e) A 2-fold increase in the membrane-associated EphB2 in neuropsin−/− (F(3, 12)=6.4; p<0.05 vs non-stressed) but not wild-type mice was observed after stress (p<0.05 vs stressed neuropsin−/− mice). (f) qRT-PCR revealed a 2-fold upregulation of the EphB2 gene expression following 6 hour stress (F(3, 23)=13.48; p<0.001), not observed in neuropsin-deficient animals (p<0.001 vs. stressed wild-type mice). EphB2-S and EphB2-L describe short and long splice variants, respectively. NP – neuropsin. Digits inside columns indicate n number. *p<0.05; **p<0.01; ***p<0.001. Results shown as mean±SEM.

When we used the above protocol to examine the composition of the SH-SY5Y or HEK293 cell culture medium following the application of neuropsin, we found a new ~70kDa extracellular fragment of EphB2 released into the media (Figure 2c) whose size was consistent with neuropsin cleaving EphB2 close to the cell membrane (Suppl. Fig. 6a, b). Next, we subjected wild-type and neuropsin−/− mice to restraint stress to activate the basolateral complex of the amygdala12. Neuropsin levels increased by 50% after stress and gradually normalized during recovery in this brain region (Figure 1c, d; p<0.05). Western blotting revealed a 2-fold increase in membrane-associated amygdalar EphB2 levels after 15 minutes of restraint stress in neuropsin−/− mice (Figure 2d, e and Suppl. Fig. 7; p<0.05) indicative of new EphB2 receptors being incorporated into the membrane. This increase was not observed in wild-type animals consistent with neuropsin-mediated EphB2 cleavage during stress. Cleavage of EphB2 in the amygdala of wild-type mice was followed by a 2-fold raise in the expression of the EphB2 gene (Figure 2f; p<0.001). The cleavage was substrate-specific because stress did not alter the levels of either ephrinB2 (Suppl. Fig. 9a) or a presynaptic neuropsin substrate, NCAM-L113 (Suppl. Fig. 3b,c).

To determine the structural basis of neuropsin-specific EphB2 cleavage we analysed the fibronectin type III domain of EphB2 (Suppl. Fig. 8) looking for similarities with the previously published neuropsin cleavage sequences14. We found a critical amino acid pair Gly-Arg at position 517 of EphB2 but not EphB1, EphB6 or EphA4. Consistent with our experimental findings cleavage of EphB2 at this site would result in the release of a ~70kDa extracellular fragment.

The EphB2 receptors cluster and associate with NMDA receptors at excitatory synapses15-17. Indeed, co-immunoprecipitation revealed NR1 bound to EphB2 in the amygdala (Figure 3a). Restraint stress reduced the amount of EphB2 associated with NR1 at 15 minutes by 42% (Figure 3a, b; p<0.05) while NR1 levels were not altered (Suppl. Fig. 9b-e). Stress-induced decrease in the EphB2/NR1 association was not observed in neuropsin−/− mice but restored by intraamygdalar administration of neuropsin into these animals (Figure 3a, b), consistent with neuropsin cleaving the extracellular portion of EphB2 during stress and triggering its dissociation from NR1. These results, together with stress-induced EphB2 membrane trafficking (Figure 2d, e), suggest that neuropsin increases the dynamics of EphB2/NR1 interaction following stress.

Figure 3

Neuropsin regulates the dynamics of EphB2/NR1 interaction and controls the expression of Fkbp5

(a, b) EphB2 immunoprecipitation (before and after 15 minutes of restraint stress) from amygdalae revealed dissociation of EphB2/NR1 complexes in wild-type (F(3, 19)=4.2; p<0.05), but not neuropsin−/− mice. EphB2/NR1 dissociation was restored in neuropsin−/− mice by intraamygdalar neuropsin injections (F(3, 13)=4.7; p<0.05). Microarray analysis of wild-type and neuropsin−/− amygdalae revealed differential expression of Fkbp5 (heatmap in c, Suppl. Fig. 10). Exon-specific Fkbp5 probes showed an upregulation of the whole transcript (d) confirmed by qRT-PCR (e; F(3, 12)=72.15; p<0.001). qRT-PCR revealed attenuated stress-induced upregulation of Fkbp5 in neuropsin−/− mice (e; p<0.01 for wild-type after stress vs. neuropsin−/− after stress) rescued by intraamygdalar neuropsin injections (F(3, 14)=9.2; p<0.01). (f, g) Fkbp51 protein levels were upregulated in wild-type mice (F(3, 14)=8.95; p<0.001) but not in neuropsin−/− mice by stress. (h) Neuropsin-mediated upregulation of Fkbp5 in amygdala neuronal cultures (F(4, 29)=19.04; p<0.0001) was blocked by anti-EphB2 antibody and mimicked by stimulation of NMDA receptors (i; p<0.05). *p<0.05; **p<0.01; ***p<0.001. Results are shown as mean±SEM. NP- neuropsin. Digits inside columns indicate n number.

Regulating EphB2/NMDA-receptor interaction results in modulation of the expression of NMDA receptor-dependent genes facilitating synaptic plasticity17. To examine if neuropsin-mediated regulation of the EphB2/NR1 assembly affect gene expression in the amygdala we used microarrays in neuropsin−/− and wild-type mice. We found 19 differentially expressed transcripts with a marked upregulation of the Fkbp5 gene (Figure 3c, d; Suppl. Fig. 10, 11; p<0.0005). This gene encodes Fkbp51 protein regulating glucocorticoid receptor sensitivity. Fkbp5 has been implicated in the development of anxiety, depression and post-traumatic stress disorder (PTSD)18-20. Quantitative RT-PCR confirmed an increase in the Fkbp5 gene expression in the amygdalae of stress-naïve neuropsin−/− animals (Figure 3e; p<0.05).

The extent of upregulation of Fkbp5 mRNA shortly after trauma correlates with the development of PTSD21. If neuropsin regulates Fkbp5 gene expression then the magnitude of its stress-related regulation should be altered in neuropsin-deficient mice. When we analysed stress-induced Fkbp5 gene expression we found a 21-fold upregulation in wild-type amygdalae (Figure 3e; p<0.001) but an attenuated upregulation in neuropsin−/− mice. The increase in the Fkbp5 gene expression was accompanied by a 2-fold upregulation of Fkbp51 protein levels in wild-type mice but not neuropsin−/− animals (Figure 3f, G; p<0.05). These results indicate that neuropsin is a key regulator of the Fkbp5 gene and protein expression.

Neuropsin is an extracellular protease and thus unlikely to alter the expression of the Fkbp5 gene directly. Although the Fkbp5 gene can be regulated by glucocorticoids (Suppl. Fig. 12), the above differences in Fkbp5 expression after stress cannot be attributed to corticosterone levels (Suppl. Fig. 13). Interference with EphB2 signalling has recently been linked to the regulation of the Fkbp5 gene22. Indeed, when we mimicked stress in vitro by adding corticosterone into neuronal amygdala cultures, neuropsin-mediated upregulation of Fkbp5 was hindered by anti-EphB2 antibody (Figure 3h; p<0.001) and imitated by NMDA receptor stimulation (Figure 3i; p<0.05).

To address the effect of neuropsin on NMDA receptors directly we measured evoked NMDA/AMPA current ratio in principal neurons of the basal amygdala in wild-type and neuropsin −/− mice. We found that, unlike in the hippocampus13, the NMDA current was markedly reduced by the deletion of the neuropsin gene resulting in a ~50% drop in the NMDA/AMPA ratio (Figure 4a-c; p<0.01).

Figure 4

Neuropsin controls NMDA receptor current, E-LTP and stress-induced anxiety

(a-c) Whole-cell recordings from basal nucleus neurons of neuropsin −/− animals demonstrated lower NMDA currents compared to wild-type mice. Induction of LTP in the lateral-basal pathway (d) using a strong (e) or weak (f) protocol revealed an impairment of E-LTP in neuropsin−/− mice. The elevated plus maze test following acute or chronic restraint stress demonstrated lack of anxiety in neuropsin −/− mice as indicated by the number of entries into open arms (g). General locomotor activity was similar in both genotypes (h, i). The behavioural phenotype was reversed by bilaterally injecting neuropsin back into the amygdala of neuropsin−/− mice (j). Stress-induced anxiety in wild-type animals was disrupted by blocking EphB2 (k) or silencing the Fkbp5 gene (l) in the amygdala. NP – neuropsin; LA-lateral amygdala; BLA – basal amygdala; CA – central amygdala; MeA – medial amygdala. Digits inside columns or near symbols indicate n number. 6hS – six hour stress, 21dS – 21 days of daily restraint. *p<0.05, **p<0.01, ***p<0.001. Results shown as mean±SEM.

We next asked if the neuropsin pathway affects neuronal plasticity in the amygdala. We induced early (E-LTP) or sustained (L-LTP) long term potentiation in the amygdala lateral-basal pathway of wild-type and neuropsin−/− mice. While basal synaptic responses were not altered (Suppl. Fig. 14a), E-LTP was impaired in neuropsin−/− mice (Figure 4d-f and Suppl. Fig. 14b, c; p<0.001 vs. wild-type at 20 min post-tetanus). These changes temporally correlated with neuropsin-mediated cleavage of EphB2, its dynamic interaction with NR1 and with the involvement of NMDA receptors in E-LTP in the lateral-basal pathway (Suppl. Fig. 15).

To examine if the neuropsin pathway alters behavioural signatures of stress we subjected wild-type and neuropsin−/− mice to acute or chronic stress and measured anxiety in the elevated-plus maze (Figure 4g-i). We found that stress caused a decrease in the number of entries of wild-type mice into open arms, indicative of high anxiety levels2, 3. In contrast, after stress, neuropsin−/− mice did not develop anxiety (Figure 4g; p<0.05). Closed arm entries, the total number of entries (Figure 4h-i and Suppl. Fig. 16a, b) as well as general locomotor activity measures (Supp. Fig. 17d) were similar between the genotypes as previously reported23. Furthermore, neuropsin−/− mice demonstrated an anxiolytic phenotype in the open field test confirming a general role of neuropsin in regulating anxiety (Suppl. Fig. 17a-c). While this effect is consistent with functional deficits in the NMDA receptor function (Figure 4a-c) and E-LTP (Figure 4d-f) observed in neuropsin−/− mice it cannot be excluded that additional mechanisms, such as abnormal dendritic plasticity, may contribute to the lack of anxiety observed in neuropsin−/− mice, particularly after long-lasting stress24.

To examine whether the effect of neuropsin was acute and not associated with the lack of the protease during development we bilaterally injected neuropsin into the amygdalae of neuropsin −/− mice (Suppl. Fig. 18). The neuropsin injection restored stress-induced anxiety in these animals (Figure 4j; p<0.001). The development of anxiety was hindered by blocking EphB2 in the amygdala of wild-type mice (Figure 4k; p<0.001) consistent with neuropsin interacting with EphB2 to facilitate stress-induced behavioural changes. Similarly, stress-induced anxiety was blocked by silencing the Fkbp5 gene expression in this brain region (Figure 4l and Suppl. Fig. 19) consistent with the downstream role of Fkbp5 in the neuropsin pathway.

Our studies favour a model where, after stress, both corticosterone-induced and neuropsin-mediated components converge to modulate the Fkbp5 gene expression and trigger anxiety (Suppl. Fig. 20). Neuropsin cleaves the extracellular portion of EphB2 and facilitates the dynamic interaction of EphB2 with the NR1 subunit of the NMDA receptor. The resulting enhancement of the NMDA current causes an upregulation of Fkbp5 and promotes the development of anxiety. This novel pathway, highlighting the ability of Eph and NMDA receptors to respond to activity dependent signals from the extracellular milieu, opens new possibilities for treatment of stress-associated disorders, including various forms of anxiety disorders.

Methods Summary

Restraint stress was performed by placing the mice in wire mesh restrainers while control animals were left undisturbed. Anxiety was measured using the elevated-plus maze by counting the number of entries to closed or open arms during 5 min. Intraamygdala injections were performed through bilaterally implanted cannulae and were followed by restraint stress in plexiglass tubes. Fkbp5 gene was silenced by intraamygdala injection of lentiviral shRNA construct followed by behavioural assessment two weeks later. LTP was recorded from the lateral-basal pathway and whole-cell recordings made from basal amygdala neurons. Data were analyzed by Student t-test or ANOVA followed by Tukey’s post-test. P values of less than 0.05 were considered significant.