Genetic recovery of ErbB4 in adulthood partially restores brain functions in null mice

Significance

NRG1–ErbB4 signaling is implicated in GABAergic circuit assembly during development and GABAergic transmission at adulthood. However, it is unclear whether phenotypes in the adult stage in ErbB4 mutant mice result from abnormal neural development. By using two strains of mice with temporal control of ErbB4 deletion and expression, we demonstrate that ErbB4 deletion in adult mice impaired behavior and GABA release, whereas deficits due to ErbB4 null mutation during development were alleviated by restoring ErbB4 expression at the adult stage. Together, our results indicate that NRG1–ErbB4 signaling at adulthood is critical to GABAergic transmission and behavior and suggest that restoring NRG1–ErbB4 signaling at the postdevelopmental stage might benefit relevant brain disorders.

Abstract

Neurotrophic factor NRG1 and its receptor ErbB4 play a role in GABAergic circuit assembly during development. ErbB4 null mice possess fewer interneurons, have decreased GABA release, and show impaired behavior in various paradigms. In addition, NRG1 and ErbB4 have also been implicated in regulating GABAergic transmission and plasticity in matured brains. However, current ErbB4 mutant strains are unable to determine whether phenotypes in adult mutant mice result from abnormal neural development. This important question, a glaring gap in understanding NRG1–ErbB4 function, was addressed by using two strains of mice with temporal control of ErbB4 deletion and expression, respectively. We found that ErbB4 deletion in adult mice impaired behavior and GABA release but had no effect on neuron numbers and morphology. On the other hand, some deficits due to the ErbB4 null mutation during development were alleviated by restoring ErbB4 expression at the adult stage. Together, our results indicate a critical role of NRG1–ErbB4 signaling in GABAergic transmission and behavior in adulthood and suggest that restoring NRG1–ErbB4 signaling at the postdevelopmental stage might benefit relevant brain disorders.

Schizophrenia (SZ) is a disabling mental disorder that affects ∼1% of the population worldwide (1). It alters basic brain processes of perception, emotion, and judgment to cause hallucinations, delusions, thought disorder, anhedonia, and cognitive deficits. Despite extensive efforts to study its pathophysiological mechanisms, SZ remains one of the least understood brain disorders. Recent identification of SZ susceptibility genes and studies of their functions have begun to shed light on its pathophysiology.

Neuregulin 1 (NRG1) is an EGF-domain–containing trophic factor that acts by activating ErbB tyrosine kinases including ErbB4. NRG1–ErbB4 signaling is critical for assembling GABAergic circuits. Mutant mice lacking NRG1 or ErbB4 display deficits in interneuron migration, axon and dendrite development of interneurons, and synaptogenesis onto and by interneurons (2–10). Both NRG1 and ErbB4 are expressed in the adult brain. NRG1 is produced mainly in neurons in an activity-dependent manner (11, 12). On the other hand, 99% of ErbB4-positive neurons in the adult cortex and hippocampus are GABAergic (4, 13–17), the majority of which express parvalbumin (PV) (4, 13, 15, 16). In adult animals, NRG1–ErbB4 signaling could promote GABAergic transmission and thus control pyramidal neuron activity (17–24). It was proposed that NRG1–ErbB4 signaling serves as a homeostatic mechanism to control the excitation–inhibition (E-I) balance (25, 26).

Interestingly, NRG1 and ErbB4 are SZ risk genes in diverse populations based on family trio studies, case-controlled association studies, and meta-analysis studies [see review by Mei and Nave (25)]. However, their polymorphism did not reach genome-wide significance in a recent genome-wide association study of SZ (27), perhaps as a result of allelic heterogeneity at NRG1 and ErbB4 loci, the existence of haplotypes (28–32), and/or population stratification (33–36). The following evidence supports the notion that NRG1 and ErbB4 alteration contributes to pathophysiological mechanisms of SZ. First, NRG1 and ErbB4 variants have been shown to be associated with reduced volume and decreased activation of brain regions and cognitive phenotypes (37–39) and with responses to antipsychotics treatment (40–42). Second, altered NRG1 or ErbB4 levels were detected in postmortem brain samples and peripheral blood of SZ patients (43–45). Third, mutating NRG1 and ErbB4 or altering their levels in mice could recapitulate SZ-related endophenotypes (3, 18, 24, 46–48). Finally, recent meta-analyses (49–51), including one from 2017 on >16,000 schizophrenic patients and >20,000 controls (49), identify NRG1 and ErbB4 as risk genes for SZ. Interestingly, SNP rs7598440 of ErbB4 could predict cortical or cerebrospinal fluid GABA concentration in healthy human subjects (52, 53), in agreement with critical roles of ErbB4 in the development and function of the GABA circuitry from mouse studies.

A critical question in understanding how abnormal NRG1–ErbB4 signaling alters brain functions is whether it also involves the E-I imbalance in adulthood. A related question is whether SZ-associated endophenotypes caused by ErbB4 deficiency in early development could be diminished by restoring ErbB4 expression in adult animals. These questions require temporal control of ErbB4 mutation and expression. In this paper, we developed genetic approaches; in inducible knockout (iKO) mice, ErbB4 expression is normal until tamoxifen (Tam) treatment, whereas in recovery knockout (rKO) mice, ErbB4 is absent during development but can be restored in adulthood upon Tam treatment. We studied behaviors and characterized GABAergic transmission of Tam-treated iKO and rKO mice. ErbB4 deletion in adult mice was sufficient to cause behavioral and synaptic deficits; on the other hand, behavioral and synaptic deficits observed in ErbB4 null mice could be diminished by restoring ErbB4 expression at the adult stage. These results provide compelling evidence that ErbB4 is critical for synaptic transmission and plasticity after development and suggest that restoring ErbB4 signaling could be beneficial to relevant SZ.

Results

Behavioral Deficits in Mice Lacking ErbB4 in Adulthood.

To delete ErbB4 at the adult stage, we generated iKO mice by crossing ErbB4^f/f mice with CAG::Cre-ER mice (Fig. 1A). CAG::Cre-ER mice express a fusion protein consisting of Cre recombinase and a modified ligand-binding domain of the estrogen receptor (ER) under the control of the ubiquitous promoter CAG (CMV enhancer, chicken β-actin promoter, rabbit β-globin polyA). In ErbB4^f/f mice, exon 2 is floxed, and exon 2 deletion generates a frame shift to disrupt ErbB4 expression. In the resulting CAG::Cre-ER;ErbB4^f/f (iKO) mice, expression of ErbB4 continues to be controlled by the promoter of the endogenous ErbB4 gene. iKO mice were injected with Tam, whose metabolite binds to ER and activates the Cre. Therefore, ErbB4 is expressed in iKO mice until Tam injection. We injected iKO mice with Tam or vehicle (referred to as iKO+Tam and iKO+Veh mice) at 8 wk of age and analyzed ErbB4 expression 6 wk after (Fig. 1B). As shown in Fig. 1 C and D, ErbB4 expression in the cortex and hippocampus was abolished in iKO+Tam mice compared with samples from iKO+Veh mice that were similar to wild-type mice. Thus, ErbB4 was expressed at a normal level in iKO mice during development but was ablated upon Tam induction. Next, we determined whether adult ErbB4 deletion alters mouse behavior and focused on the paradigms where ErbB4 null mice were found to be deficient (Fig. 1 E–M). As shown in Fig. 1 E–M, iKO+Veh mice behaved similarly to WT mice in all paradigms, suggesting no apparent effect of the Cre transgene. In contrast, compared with iKO+Veh mice, iKO+Tam mice were hyperactive in an open-field test as travel distance was increased (Fig. 1 E–G). They were also impaired in prepulse inhibition (PPI) (Fig. 1 H and I) and social interaction (Fig. 1 J and K) as PPI ratio was lower while time spent with stranger mice was reduced. In addition, freezing time was reduced in iKO+Tam mice in contextual fear conditioning compared with iKO+Veh mice (Fig. 1 L and M). Notice that the freezing times among groups during training were similar (Fig. 1M), suggesting a normal ability to sense or escape from foot shock. These results indicate that ErbB4 in adult mice is necessary for proper behavior. Interestingly, impairment in iKO+Tam mice in open-field (Fig. 1 E–G) and PPI tests (Fig. 1 H and I) was less severe than in ErbB4 null mice, but that in social interaction (Fig. 1 J and K) and contextual memory (Fig. 1 L and M) was similar between the two genotypes. This suggests that the ErbB4 null mutation was more damaging than adult deletion and ErbB4 in adults is more critical for selective behavioral paradigms.

Fig. 1.

Behavioral deficits in Tam-treated iKO mice. (A) Breeding diagram of iKO mice. (B) Times of Tam injection and analysis. Tam was injected once every other day for 20 d. (C and D) Diminished ErbB4 expression in iKO+Tam mice and ErbB4 null mice. From left to right, WT, ErbB4 null, ErbB4^f/f, iKO+Veh, and iKO+Tam; n = 3 mice in each group. CT, cortex; HC, hippocampus. (E) Representative travel traces of mice in an open-field test. (F) Increased travel distance (per 5 min) by iKO+Tam mice. (G) Increased total travel distance (within 30 min) by iKO+Tam mice. (H) Diagram of the PPI test. PPI (%) = 100 × (a − b)/a. (I) Impaired PPI of iKO+Tam mice. (J) Decreased preference for the social chamber by iKO+Tam mice. (K) Decreased preference for social novelty by iKO+Tam mice. (L) Diagram of contextual fear memory test. (M) Unaltered freezing response to foot shock in training session and reduced freezing time in test session by iKO+Tam mice. n = 12 per group for behavior tests; *P < 0.05, **P < 0.01, compared with ErbB4^f/f mice; ^#P < 0.05, ^##P < 0.01, compared with ErbB4 null mice; ^$P < 0.05, ^$$P < 0.01, compared with iKO+Veh mice.

Decreased Inhibitory Transmission in iKO+Tam Mice.

Morphological studies of iKO+Tam mice indicate that adult ErbB4 deletion had little effect on global structures and numbers of interneurons pyramidal neurons of the cortex and hippocampus (SI Appendix, Fig. S1). ErbB4 mutation in excitatory neurons has no effect on dendrites of pyramidal neurons (47). In agreement, iKO+Tam mice displayed a similar dendrite length and complexity of CA1 pyramidal neurons of the hippocampus (SI Appendix, Fig. S2). ErbB4 null and interneuron-specific mutation caused deficits in spines (3, 47, 55, 56) and excitatory synapses onto interneurons (3, 4, 8). To determine whether similar changes may result from postdevelopmental ErbB4 deletion, we first crossed iKO mice with Thy1-GFP mice to label spines. Quantification of GFP-labeled spines in the hippocampus CA1 region showed spine density and morphology in iKO+Tam mice similar to those in iKO+Veh mice and ErbB4^f/f mice (SI Appendix, Fig. S3 A–E). In agreement, there was no change in miniature excitatory postsynaptic current (mEPSC) frequency or amplitude of CA1 pyramidal neurons (SI Appendix, Fig. S3 F–I). These results demonstrate that adult ErbB4 has little effect on excitatory synapses between pyramidal neurons or their function. ErbB4 is present at postsynaptic sites of excitatory synapses onto inhibitory neurons and is implicated in their formation (3, 4, 8). We investigated whether these synapses are altered by adult ErbB4 deletion by using Vglut1 and PV antibodies to label, respectively, presynaptic excitatory terminals and postsynaptic dendrites of PV+ interneurons, many of which are ErbB4+ (4, 13, 15, 16). As shown in SI Appendix, Fig. S3 J and K, the numbers of Vglut1+ puncta onto PV+ dendrites were similar among different genotypes, regardless of Tam treatment. In agreement, there was no difference in mEPSC frequency or amplitude of interneurons in the stratum oriens and stratum pyramidale of the hippocampus CA1 region (SI Appendix, Fig. S3 L–O). These results indicate no detectable effect of adult ErbB4 deletion on excitatory synapses onto interneurons.

Studies of ErbB4 null or PV-specific mutation demonstrated that ErbB4 is necessary for GABAergic transmission in the hippocampus and cortex (3, 4, 17–24, 55). We next determined whether this function requires ErbB4 in adulthood by recording IPSCs (inhibitory postsynaptic currents) in CA1 pyramidal neurons (Fig. 2A). As shown in Fig. 2B, evoked IPSC (eIPSC) amplitude was reduced in iKO+Tam slices, indicating a necessary role of adult ErbB4 in maintaining GABA transmission in the hippocampus. This reduction may be caused by a reduction in GABA release and/or GABA receptor density on postsynaptic membrane. To address this question, we compared miniature IPSC (mIPSC) frequency and amplitude between iKO+Tam and iKO+Veh mice. There was no change in mIPSC amplitude, suggesting GABA receptor density was not compromised (Fig. 2D). However, mIPSC frequency was reduced (Fig. 2E), which may suggest fewer numbers of inhibitory synapses or diminished release probability. Next, we quantified the numbers of PV-labeled perisomatic inhibitory synapses onto CA1 pyramidal neurons, which were similar among different genotypes, regardless of Tam treatment (Fig. 2 F and G), suggesting that reduced mIPSC frequency may be due to a problem with release probability. To test this hypothesis, we characterized the paired-pulse ratios (PPRs) by recording eIPSCs in response to tandem stimuli with different intervals. The PPRs were increased in iKO+Tam mice compared with ErbB4^f/f and iKO+Veh mice (Fig. 2 H and I), indicating compromised GABA release probability. Besides perisomatic synapses, ErbB4+ neurons also form inhibitory synapses onto axon initial segments (AISs) of pyramidal neurons (3, 4, 16). Adult deletion seemed to have no effect on the number of these synapses (Fig. 2 J–L). Taken together, these observations suggest that ErbB4 in adult animals is necessary for GABA activity by maintaining release probability.

Fig. 2.

Decreased inhibitory transmission in Tam-treated iKO mice. (A) Recording diagram. (B) Decreased eIPSC amplitude in iKO+Tam mice. n = 7/8 neurons from three mice. (C) Representative traces of mIPSCs. (Scale bar: 20 pA/s.) (D) No change in mIPSC amplitude. (E) Decreased mIPSC frequency. n = 8/9 neurons from three mice. (F) Representative images of PV+ inhibitory synapses onto NeuN+ postsynaptic somata. (Scale bar: Upper, 5 µm; Lower, 2 µm.) (G) Quantification data of F. n = 40/42/42 somata of six mice. (H) Representative IPSC traces induced by paired-pulse stimulation with a 100-ms interval. (Scale bar: 250 pA/50 ms.) (I) Increased paired-pulse ratios in iKO+Tam mice. n = 10 neurons from four mice. (J) Diagram of inhibitory synapses onto axon initial segments. (K) Representative images of GAD67+ inhibitory synapses onto AnkG+ axon initial segments. AnkG, Ankyrin G. (Scale bar: 5 µm.) (L) Quantification data of K. n = 42/40/43 neurons of six mice. *P < 0.05, **P < 0.01, compared with ErbB4^f/f mice; ^$P < 0.05, ^$$P < 0.01, compared with iKO+Veh mice.

Diminished GABA Transmission Deficits by Restoring ErbB4 Expression in Adult Animals.

Next, we determined whether ErbB4 in adulthood is sufficient for GABA transmission by restoring ErbB4 expression in ErbB4 null mice. We generated rKO mice by inserting a loxP-NeotpA-loxP cassette into the first intron (between exons 1 and 2) of the ErbB4 gene to first produce Stop-ErbB4 mice (Fig. 3A). The cassette contained the anti-neomycin gene (for embryonic stem cell selection) and a poly-A transcription stop signal. Homozygous Stop-ErbB4 mice did not express ErbB4 in any tissues or cells and died prematurely due to cardiac deficits, like ErbB4 null mice. To prevent embryonic lethality, hemizygous Stop-ErbB4 mice were crossed with α-MHC::ErbB4 mice, which express ErbB4 specifically in developing heart cells. Finally, Stop-ErbB4;α-MHC::ErbB4 mice were crossed with the aforementioned CAG::Cre-ER mice to generate Stop-ErbB4;α-MHC::ErbB4;CAG::Cre-ER (rKO) mice (Fig. 3B). Therefore, rKO mice do not express ErbB4 in the brain or any other tissue except the heart (Fig. 3C). After Tam treatment (Fig. 3 A and B) to release the loxP-NeotpA-loxP cassette, rKO mice expressed ErbB4 in the brain (Fig. 3 D and E). Notice that although ErbB4 recovery is mediated by CAG-CreER, the promoter activity of the endogenous ErbB4 gene controls the cells in which ErbB4 is expressed. As shown in Fig. 3E, the ErbB4 level in the brain of rKO mice was restored by Tam treatment to a level that was comparable to that of control mice (CAG::Cre-ER mice).

Fig. 3.

Generation and characterization of rKO mice. (A) ErbB4 alleles of WT, Stop-ErbB4, and rKO+Tam mice. (B) Breeding diagram of rKO mice. (C) Expression of ErbB4 in Stop-ErbB4;α-MHC::ErbB4 mice. (D and E) Restored ErbB4 expression in the cortex and hippocampus of rKO+Tam mice. CT, cortex; HC, hippocampus; from left to right: control, ErbB4 null, iKO+Veh double, iKO+Tam. n = 4 mice; **P < 0.01, compared with control, ^$$P < 0.01, compared with rKO+Veh mice.

ErbB4 null mutation reduces the number of PV+ interneurons in the cortex (5, 7) and hippocampus (54). In agreement, vehicle-treated rKO mice (that did not express ErbB4, Fig. 3E) displayed fewer PV+ neurons in the cortex (SI Appendix, Fig. S4 A–C) and hippocampus (SI Appendix, Fig. S4 D and E). The reduced number of PV+ interneurons remained unchanged after Tam treatment (SI Appendix, Fig. S4 B–E), indicating that interneuron migration deficit was not rescued by restoring ErbB4 expression in adult animals. The result is not unexpected because interneuron migration was completed before Tam treatment.

GABA transmission in the cortex and hippocampus is reduced in ErbB4 mutant mice (3, 4, 17–24, 55). We next determined whether GABA release deficits due to ErbB4 mutation could be rescued by restoring ErbB4 expression in adult animals. As shown in Fig. 4B, eIPSC amplitude was reduced in hippocampal slices of rKO+Veh mice, in agreement with previous reports (18, 19, 21, 25). Remarkably, the reduction was mitigated in rKO+Tam slices (Fig. 4B), indicating that restoring ErbB4 is able to enhance GABAergic transmission. mIPSC frequency was also reduced in rKO+Veh slices (Fig. 4 C–E), in agreement with previous studies (3, 4, 21, 23). Remarkably, after Tam treatment, the reduction was diminished (i.e., in rKO+Tam slices) (Fig. 4 C–E), indicating a rescue effect by adult ErbB4 expression. To determine whether this effect resulted from increased numbers of inhibitory synapses, we quantified the number of perisomatic inhibitory synapses onto CA1 pyramidal neurons. As shown in Fig. 4 F and G, there was no difference between Tam- and Veh-treated rKO mice, suggesting that the rescue effect on mIPSC frequency may be mediated by improved release probability. This notion was supported by recovered PPRs (Fig. 4 H and I). As with perisomatic synapses, adult recovery of ErbB4 had little effect on the number of inhibitory synapses onto AIS of pyramidal neurons, which remained low in Tam-treated rKO mice (Fig. 4 J–L). mIPSC amplitude was not changed regardless of genotypes and treatment (Fig. 4E). Taken together, these observations suggest that restoring ErbB4 in adult animals is able to rescue functional deficits of GABA transmission without increasing the number of interneurons or inhibitory synapses.

Fig. 4.

Diminished GABA transmission deficits in Tam-treated rKO mice. (A) Recording diagram. (B) Diminished reduction of eIPSC amplitude in rKO+Tam mice. n = 7/8 neurons from three mice. (C) Representative traces of mIPSCs. (Scale bar: 20 pA/s.) (D) Diminished reduction of mIPSC frequency in rKO+Tam mice. (E) No difference in mIPSC amplitude among groups. n = 8/9 neurons from three mice. (F) Representative images of PV+ inhibitory synapses onto NeuN+ postsynaptic somata. (Scale bar: Upper, 5 µm; Lower, 2 µm.) (G) No effect of Tam treatment on decreased number of inhibitory synapses. n = 45/45/47 somata of seven mice. (H) Representative IPSC traces induced by paired-pulse stimulation with a 100-ms interval. (Scale bar: 250 pA/50 ms.) (I) Recovered paired-pulse ratios in Tam-treated rKO mice. n = 9/10 neurons of four mice. (J) Diagram of inhibitory synapses onto axon initial segments. (K) Representative images of GAD67+ inhibitory synapses onto AnkG+ postsynaptic axon initial segments. AnkG, ankyrin G. (Scale bar: 5 µm.) (L) No effect of Tam treatment on decreased number of inhibitory synapses. n = 45/44/47 neurons of seven mice. *P < 0.05, **P < 0.01, compared with control, ^$P < 0.05, ^$$P < 0.01, compared with rKO+Veh mice.

No Effect on Excitatory Synaptic Deficits by Adulthood ErbB4 Expression.

As stated above, ErbB4 mutation has no effect on dendrites of pyramidal neurons in vivo (47); in agreement, rKO+Veh mice showed a similar dendrite number and complexity (SI Appendix, Fig. S5). To determine whether ErbB4 expression may rescue deficits of excitatory synapses, we examined their morphological and functional properties in Veh- and Tam-treated rKO mice. As shown in SI Appendix, Fig. S6 A–E, spines of CA1 pyramidal neurons, in particular mushroom-shaped spines, were reduced in Veh-treated rKO mice (that did not express ErbB4), in agreement with previous reports (3, 47, 55, 56), suggesting that ErbB4 is required for spine development. However, the spine deficits were not diminished by ErbB4 recovery (i.e., in rKO+Tam mice). mEPSC frequency of CA1 pyramidal neurons was reduced in rKO+Veh mice compared with control (SI Appendix, Fig. S6 F–I), in agreement with reduced spine number. There was no difference between mEPSC frequencies of Veh- and Tam-treated rKO mice (SI Appendix, Fig. S6 F–I), indicating that adult ErbB4 expression was unable to rescue deficits of excitatory synapses onto pyramidal neurons. The number of excitatory synapses onto interneurons is reduced after early ErbB4 mutation (3, 4, 8), as observed in rKO+Veh mice (SI Appendix, Fig. S6 J and K). This was associated with a reduction in mEPSC frequency of interneurons (SI Appendix, Fig. S6 L–O). Both of these deficits (reduced number of excitatory synapses and mEPSC frequency) remained in rKO+Tam mice (SI Appendix, Fig. S6 J–O), indicating that they could not be rescued by adult ErbB4 expression.

Mitigated Behavioral Deficits by Restoring ErbB4 Expression in Adult Mice.

To determine whether adult ErbB4 expression mitigates behavioral deficits caused by ErbB4 null mutation during development, we characterized behaviors of Veh- and Tam-treated rKO mice. Compared with control mice, rKO+Veh mice were hyperactive in an open field as travel distance was increased (Fig. 5 A–C). They were also impaired in PPI (Fig. 5 D and E) and contextual fear conditioning (Fig. 5 F and G). Remarkably, these behavioral deficits were ameliorated by Tam treatment in rKO mice. Compared with Veh-treated rKO mice, Tam-treated rKO mice were less hyperactive in open field (Fig. 5 A–C), displayed better PPI (Fig. 5 D and E), and showed increased freezing time in contextual fear conditioning (Fig. 5 F and G), indicating improved behavioral scores by restoring ErbB4 in adult mice. Notice that the recovery from behavioral deficits was incomplete; that is, Tam-treated rKO mice remained deficient in these paradigms compared with control mice. These results suggest that restoring ErbB4 expression in adulthood mitigates but does not eliminate behavioral deficits caused by ErbB4 loss of function during development.

Fig. 5.

Mitigated behavioral deficits in Tam-treated rKO mice. (A) Representative travel traces of mice in an open-field test. (B) Mitigation of increased travel distance (per 5 min) in rKO+Tam mice. (C) Mitigation of increased total travel distance (within 30 min) in rKO+Tam mice. (D) Diagram of the PPI test. PPI (%) = 100 × (a − b)/a. (E) Mitigation of impaired PPI in rKO+Tam mice. (F) Diagram of contextual fear memory test. (G) Mitigation of reduced freezing time in rKO+Tam mice. n = 12 mice; *P < 0.05, **P < 0.01, compared with control; ^$P < 0.05, ^$$P < 0.01, compared with rKO+Veh mice.

Discussion

This study determined the role of ErbB4 in the adult brain by utilizing mouse lines that enable temporal control of ErbB4 deletion or expression. We showed that adult deletion of ErbB4 caused behavioral deficits and reduced IPSCs in hippocampal slices, indicating compromised GABAergic transmission. However, it had no effect on numbers of interneurons, inhibitory synapses onto excitatory neurons, and excitatory synapses in the hippocampus. These results suggest that ErbB4 is critical to GABAergic transmission in adult mice. On the other hand, rKO mice did not express ErbB4 during development and displayed behavioral deficits similar to those of ErbB4 null mutant mice, including hyperactivity in an open field and impaired PPI and contextual fear memory. Remarkably, these deficits were diminished after Tam treatment, indicating that deficits caused by ErbB4 deletion during development could be mitigated by restoring ErbB4 expression at the adult stage. Notice that numbers of PV+ interneurons and inhibitory and excitatory synapses remained depressed after Tam treatment, suggesting that restoring ErbB4 in adult animals was unable to diminish structural deficits caused by ErbB4 loss of function during development. Together, these observations support a working hypothesis that restoring ErbB4 at adulthood could improve GABAergic transmission even on compromised circuits and thus diminish behavioral deficits that are caused by developmental ErbB4 mutation.

ErbB4 is expressed in developing and mature interneurons (4, 13–17) and in mature interneurons, is present in somata, dendrites, and, arguably, axons. ErbB4 is necessary for interneuron migration from ganglion eminences to the cerebral cortex and forming synapses onto and from pyramidal neurons (25, 26, 57). In the adult cortex and hippocampus, ErbB4 is almost exclusively expressed in GAD67+ interneurons, not in pyramidal neurons (4, 13–16), and has been implicated in maintaining GABAergic transmission and synaptic plasticity (3, 4, 17–24, 55). ErbB4 null mutant mice displayed fewer numbers of PV+ interneurons (5, 7, 54) and compromised GABA release (17, 19, 20, 23). Adult ablation of ErbB4 impaired GABAergic transmission, in further support of the hypothesis that GABAergic transmission requires ErbB4. However, the number of PV+ neurons was similar in iKO and control mice, suggesting that adult ErbB4 is not required for the migration and survival of PV+ interneurons within the experimental period. Similarly, restoring ErbB4 expression in adults improved GABAergic transmission compared with that in ErbB4 null mice but was unable to increase the number of PV+ interneurons (because interneuron migration had already finished). Molecular mechanisms by which ErbB4 regulates interneuron migration, synapse formation onto pyramidal neurons and interneurons, and GABA releases are complex. For example, maintaining GABA release requires kinase activity, whereas interneuron migration and synapse formation could also be mediated by kinase-independent, cell-adhesion-dependent mechanisms (6, 58). Future studies are warranted to determine which mechanism is involved in mitigating deficits caused by ErbB4 null mutation. Evidently, ErbB4 level or promoter activity is dominantly expressed in interneurons (4, 13–17). However, ErbB4 knockdown was shown to reduce spines in vitro (56), and pyramidal neuron-specific deletion of ErbB4 could reduce the number of spines of pyramidal neurons in vivo (47, 59). These results suggest a role of ErbB4 in pyramidal neurons. Our data are unable to determine whether phenotypes were due to loss of ErbB4 in interneurons and/or pyramidal neurons. For example, loss of spines could be due to the loss of ErbB4 in pyramidal neurons or a compensatory mechanism due to ErbB4 loss in interneurons (3, 55, 60).

Earlier studies suggested NRG1–ErbB4 signaling is involved in controlling E-I balance by maintaining GABAergic transmission. First, NRG1 increases mIPSCs and eIPSCs in cortical and hippocampal slices (17, 18, 20, 24). This is likely due to increased GABA release probability (17, 23, 24). Second, this effect requires ErbB4 because it is diminished by pharmacological inhibition of the kinase activity of ErbB4 or genetic deletion of the ErbB4 gene in null or interneuron-specific mutant mice (17, 18, 21, 24). Third, treating brain slices with an NRG1-neutralizing peptide decreases mIPSCs and eIPSCs (17, 18, 20, 21, 23, 24, 61). Fourth, NRG1 neutralization and/or ErbB4 mutation impair behaviors in various paradigms involving prefrontal cortex, the hippocampus, and the amygdala (3, 15, 18–24), where ErbB4 is enriched. Results from iKO and rKO mice add to the notion that ErbB4 is critical to E-I balance in the adult brain.

In summary, our study demonstrates a role of ErbB4 signaling at the adult stage in maintaining normal GABAergic transmission, E-I balance, and behaviors. The results suggest that abnormal ErbB4 function in adult animals could contribute to the pathophysiological mechanism of relevant disorders. A mouse model mimicking high NRG1 levels in patients (32, 62) displayed SZ-related deficits (46). Such deficits could be rescued by reducing the NRG1 level at the adult stage (46). Together, these observations suggest that relevant patients could benefit from therapeutic approaches of optimizing NRG1–ErbB4 signaling.

Materials and Methods

Animals, electrophysiology, immunofluorescence, Western blot, Golgi staining, analysis of spine morphology, behavioral tests, and statistical analyses are described in SI Appendix. Experimental procedures were approved by the Institutional Animal Care and Use Committees of Augusta University and Case Western Reserve University.