ErbB4 activity in hippocampal-prefrontal synchrony and top-down attention in rodents

Summary

Top-down attention is crucial for meaningful behaviors and impaired in various mental disorders. However, its underpinning regulatory mechanisms are poorly understood. We demonstrate that the hippocampal-prefrontal synchrony associates with levels of top-down attention. Both the attention and the synchrony are reduced in mutant mice of ErbB4, a receptor of neuregulin-1. We used chemical genetic and optogenetic approaches to inactivate ErbB4 kinase and ErbB4+ interneurons, respectively, both of which reduce GABA activity. Such inhibitions in hippocampus impair both the hippocampal-prefrontal synchrony and top-down attention whereas those in prefrontal cortex alter only the attention, but not the synchrony. These observations identify a role of ErbB4-dependent GABA activity in hippocampus in synchronizing the hippocampal-prefrontal pathway and demonstrate that acute, dynamic ErbB4 signaling is required to command top-down attention. Because both neuregulin-1 and ErbB4 are susceptibility genes of schizophrenia and major depression, our study contributes to a better understanding of these disorders.

In Brief

Top-down attention is important for cognitive function, but its underpinning mechanisms are not well understood. Tan et al. demonstrate that the attention correlates with the hippocampal-prefrontal synchrony, both of which require acute, dynamic activity of ErbB4.

Introduction

Attention is a process that selectively concentrates on one among many stimuli or trains of thought and ignores other perceivable information. It is crucial for thought organization and meaningful behaviors (Baluch and Itti, 2011; Corbetta and Shulman, 2002; Moore and Zirnsak, 2017). People with attention deficits may have difficulty to focus or display compulsive behaviors (Mueller et al., 2017). Attention deficits are symptoms of neuropsychiatric disorders including schizophrenia, bipolar disorder, attention-deficit hyperactivity disorders (ADHD), and Alzheimer’s disease (Chen et al., 1998; Clark et al., 2002; Perry and Hodges, 1999). Attention falls into two categories: bottom-up and top-down attention. Bottom-up attention is driven by external salient stimuli and requires perspective subcortical and cortical regions that process sensory information. Deployed fast, it is important for quick reaction and danger avoidance (Baluch and Itti, 2011; Hopfinger et al., 2000). Top-down attention is volitionally initiated by inner thought and task demands (Baluch and Itti, 2011; Buschman and Miller, 2007; Engel et al., 2016; Hopfinger et al., 2000; Saalmann et al., 2007). It is controlled by the prefrontal cortex (PFC) although recent data suggest involvement in subcortical brain structures (Baluch and Itti, 2011; Goldfarb et al., 2016; Shipp, 2004). However, underlying molecular mechanisms and circuits remain unclear.

Interneurons, comprising 15-20% of total neurons in the brain, control the excitability of neurons they synapse onto (Isaacson and Scanziani, 2011). Via feedforward and feedback inhibition, interneurons increase the computational power of cortical networks and facilitate the processing and flow of information within and between brain regions (Isaacson and Scanziani, 2011; Tremblay et al., 2016). Interneurons are classified into different types based on morphology, electrophysiological characteristics and molecular features (Wamsley and Fishell, 2017). For example, parvalbumin-positive (PV+) basket and chandelier cells synapse onto somas and axonal initial segments of pyramidal neurons, respectively, to control their firing and thus the excitation/inhibition (E/I) balance (Hu et al., 2014). In mice, attention is associated with increased activity of PFC PV+ interneurons (Kim et al., 2016). Injecting bicuculline into hippocampus attenuated attention (McGarrity et al., 2016). These results raise a critical question whether GABA activity in the hippocampus controls the function of PFC and if yes, how.

Neuregulin-1 (NRG1) is a trophic factor that acts by stimulating the ErbB4 tyrosine kinase. The NRG1/ErbB4 signaling has been implicated in the assembly of the GABAergic circuitry (Mei and Nave, 2014). NRG1 and ErbB4 are expressed in adult brains and are necessary for the E/I balance by promoting or maintaining GABA release (Chen et al., 2010; Del Pino et al., 2013; Fazzari et al., 2010; Wen et al., 2010; Woo et al., 2007). Both NRG1 and ErbB4 are susceptibility genes of schizophrenia, bipolar disorders and major depression (Mei and Nave, 2014).

We investigated the functional connectivity of the hippocampal-prefrontal pathway in free-moving mice being tested for top-down attention. We found that this pathway is better synchronized when attention level is high. Noticeably, both attention and the synchrony are reduced in mice lacking ErbB4. We explored consequences of chemical genetic inhibition of ErbB4 kinase activity and optogenetic inactivation of ErbB4+ interneurons in different brain regions of adult mice. Our results identify a role of ErbB4-dependent GABA activity in hippocampus in synchronizing the hippocampal-prefrontal pathway and demonstrate that acute, dynamic ErbB4 signaling is required to command top-down attention. Because both neuregulin-1 and ErbB4 are susceptibility genes of schizophrenia and major depression, our study contributes to a better understanding of these disorders.

RESULTS

Hippocampal-prefrontal synchrony positively correlates with attention

To investigate mechanisms underlying attention, mice were subjected to 5-choice serial reaction time task (5-CSRTT) (Figure 1A), a well-established test of sustained attention with high construct validity (Bari et al., 2008; Kim et al., 2016; Robbins, 2002). Food-deprived mice were presented with a brief visual stimulus (cue) in one of five apertures and required to identify the lit aperture. If it is correctly identified, the mouse receives a food reward pellet (hereafter as “correct” trial). “Error” responses include incorrect, omission, and premature (Figure S1A–C, videos S1–4) (see Methods), which are considered to reflect disturbances in attentional processing and executive functioning (Bari et al., 2008; Robbins, 2002). Mice were trained by a 7-stage protocol to gradually reduce cue time to 0.8 sec and limited hold time to 5 sec and increase the inter-trial intervals (ITI) to 3-5 sec, until mice accomplished >50 correct trials, >80% accuracy and <40% omission (Figure S1A). The average training time for wild-type mice was 21 ± 5 days.

Figure 1. Positive correlation between correct trials and the hippocampal-prefrontal synchrony.

(A) Schematic diagram of 5-CSRTT and in vivo recording.

(B) Diagram of a mouse brain with electrode positions.

(C) Coherence between PFC and hippocampal LFPs in correct (blue) and error trials (orange). Area in rectangle is enlarged as C’.

(D) Statistical analysis of data in C (n=14 tests, 7 mice; paired t-test, delta: ***P<0.001; theta: ***P<0.001; beta: p=0.3921; gamma: p=0.4842; ns: not significant).

(E) Examples of in-vivo recording traces showing spikes recorded from a prefrontal neuron (vertical lines) in relation to raw hippocampal LFPs (red) and filtered theta oscillation (black).

(F, H) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta (F) or theta (H) oscillations in correct (top) and error trials (bottom).

(G, I) Decreased PPC values at delta (G) or theta (I) bands in error trials (n=93 units, 7 mice; paired t-test, **P=0.0014 (G); ***P<0.001 (I)).

Pooled data are shown as mean ± SEM. See also Figure S1.

Neuronal activities (local field potentials or LFPs and spikes) in PFC and ventral hippocampus (vHPC) were recorded by implanted tetrodes in free-moving mice during 5-CSRTT (Figure 1B, Figure S1D, E). During the recording, ITIs varied pseudorandomly among 3, 4, or 5 sec to enhance the attentional load and to eliminate prediction of stimulus onset by self-pacing strategy (Bari et al., 2008; Kim et al., 2016). Neuronal activity during the 3-sec period before poking the cue aperture were analyzed. Because there was no cue to be associated with a premature response, our analysis did not include recordings from premature trials. Two independent measurements were used to determine the hippocampal-prefrontal synchrony: the coherence of prefrontal and hippocampal LFPs and the phase-locking of prefrontal spikes to hippocampal LFPs. We found that correct trials associated with an increase in the coherence of PFC and vHPC LFPs in delta (0.5-4 Hz) and theta (4-12 Hz) frequency ranges (Figure 1C, D), suggesting that the two brain regions are better synchronized when the attention level is high. The coherence at beta and gamma frequencies was not increased in correct trials (Figure 1C, D), in agreement with earlier reports that low frequency synchrony marks the functional connectivity between the hippocampus and the prefrontal cortex (Fries, 2015; Jones and Wilson, 2005; Sigurdsson et al., 2010; Spellman et al., 2015). Because vHPC neurons project to the PFC (Jay et al., 1989; Swanson, 1981) we next determined whether the firing of prefrontal neurons are modulated by vHPC neurons by determine whether the spikes of PFC neurons were phase-locked with vHPC LFPs at different frequencies (Figure 1E). As shown in Figure 1F–I, phase-locking was stronger in correct trials, when attention level were high, than in error trials, when attention level were low. The phase-locking intensity was measured by pairwise phase consistency (PPC) which is based on the average pairwise circular distance (Vinck et al., 2010). This phase-locking effect was observed at delta and theta frequencies where the two brain regions demonstrate increased coherence (Figure 1C, D). Consistently, phase-locking was not observed at gamma range where correct trials were not associated with increased coherence (Figure S1F, G). Together, these results indicate that during attention, neural activities in vHPC and PFC are synchronized and suggest that vHPC may be important in regulating attention.

ErbB4 is required for attention and associated with hippocampal-prefrontal synchrony

Local interneuron circuits control the neural activity of individual brain regions (Hu et al., 2014; Tremblay et al., 2016). The NRG1-ErbB4 signaling is known to regulate the E/I balance in the vHPC and PFC by promoting GABA release from a subset of PV+ interneurons (Chen et al., 2010; Del Pino et al., 2013; Mei and Nave, 2014; Wen et al., 2010; Woo et al., 2007). To determine whether NRG1-ErbB4 signaling is involved in attention and in regulation of the hippocampal-prefrontal synchrony, ErbB4 null mice were subjected to 5-CSRTT and in vivo recording. To prevent embryonic lethality, ErbB4 null mice were crossed with αMHC-ErbB4 transgenic mice and resulting ErbB4 null;αMHC-ErbB4 (hereafter referred as ErbB4−/−) were characterized (Tidcombe et al., 2003). ErbB4−/− mice were slower, compared with control littermates, in learning 5-CSRTT task, as it took longer for them to reach the progression criteria of each training stage (Figure 2A, Figure S2A). Within ~25 days of training, ErbB4−/− mice were able to reach an accuracy rate [correct trials/(correct trials + incorrect trials)] similar to control mice (Figure 2A). This suggests that ErbB4−/− mice were able to learn and practice 5-CSRTT rules. However, the correct ratio [correct trials/(correct trials + incorrect trials + omission trials)] was lower in ErbB4−/− mice than control mice during training (Figure 2B); and at testing stage, ErbB4−/− mice had higher omission error ratio [omission trials/(correct trials + incorrect trials + omission trials + premature trials)], but not incorrect or premature error ratio (Figure 2C). Moreover, the latency to correct responses was increased in ErbB4−/− mice (Figure S2B). No difference was detected in the latency to reward or latency to incorrect between ErbB4−/− and control mice (Figure S2B), suggesting that ErbB4 mutation had little effect on the motivation for food or mobility in 5-CSRTT (Amitai and Markou, 2010; Kim et al., 2016; Robbins, 2002). Taking together, these results suggest that attention requires ErbB4 signaling.

Figure 2. vHPC-PFC synchrony reduction and attention deficits in ErbB4−/− mice.

(A) Accuracy rates of indicated mice in 5-CSRTT training (Control (Ctrl), n = 13 mice; ErbB4−/−, n = 13 mice; Two way ANOVA, p<0.001).

(B) Lower correct ratios of ErbB4−/− mice, compared with control during 5-CSRTT training (Control, n = 13 mice; ErbB4−/−, n = 13 mice; Two way ANOVA, p=0.0149).

(C) Increased omission ratios but not incorrect or premature ratios of ErbB4−/− mice (Control, n = 13 mice; ErbB4−/−, n = 13 mice; t-test, incorrect: p=0.1856; omission: ***p<0.001; premature: p=0.964; ns: not significant).

(D) Coherence between PFC and vHPC LFPs in control (purple) and ErbB4−/− (green) mice (n=11 tests (7 mice); two way ANOVA, p<0.001; post hoc, Bonferroni, delta: ***P<0.001; theta: ***P<0.001; beta: ns, p=0.992; gamma: ns, p=0.5704; ns: not significant).

(E, G) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta (E) or theta (G) oscillations in control (top) or ErbB4−/− (bottom) mice.

(F, H) Reduced delta (F) or theta (H) PPC in ErbB4−/− mice (control, n=96 units, 7 mice; ErbB4−/−, n=90 units, 7 mice; t-test, *p=0.0223 for F and *p=0.0238 for H).

Pooled data are shown as mean ± SEM. See also Figure S2.

To determine whether ErbB4 mutation alters the synchrony between vHPC and PFC, we compared LFP coherence and spike firing phase-locking between ErbB4−/− and control mice. ErbB4−/− mice displayed an increased LFP power in gamma band (Figure S2C–H), consistent with a previous report of deleting ErbB4 from PV+ neurons (Del Pino et al., 2013). In addition, ErbB4−/− mice showed a reduction in the synchrony between vHPC and PFC LFPs in delta and theta (not gamma) frequencies, compared with control littermates (Figure 2D). The PFC spike phase-locking to hippocampal LFPs was reduced in ErbB4−/− mice, compared with controls (Figure 2E–H) although PFC neurons of the two genotypes had similar basic electrophysiological properties (Figure S2I). These results suggest that the hippocampal-prefrontal synchrony is compromised by ErbB4 mutation. Next, we determined whether ErbB4 mutation altered attention-associated synchrony by subjecting mice to 5-CSRTT. In control mice, the coherence of vHPC-PFC LFPs was higher with correct trials, compared with error trials at delta and theta bands (Figure 3A–C). In ErbB4−/− mice, correct trials were also associated with higher coherence than error trials (Figure 3A–C); however, the coherence reduction in error trials at delta and theta bands was significantly diminished in ErbB4−/− mice (Figure 3D, E). In accord, correct trials were also associated with higher PPC values than error trials in ErbB4−/− mice in delta and theta bands; however, the difference of PPC values between correct and error trials was reduced in ErbB4−/− mice (Figure 3F–K). These results indicate that ErbB4 is required for attention-associated synchrony between vHPC and PFC and suggest that the hippocampal information flow to the PFC is important for attention.

Figure 3. Diminished attention-associated vHPC-PFC synchrony in ErbB4−/− mice.

(A) Coherence between PFC and vHPC LFPs of correct and error trials of the two genotypes.

(B, C) Reduced delta (B) or theta (C) coherence in error trials in both genotypes (n = 15 tests, 7 mice; paired t-test; for B, control: ***P<0.001; ErbB4−/−: **p=0.0089; for C, ***P<0.001; ErbB4−/−: **p=0.0041).

(D, E) Diminished difference (diff.) in delta (D) or theta (E) coherence between correct and error trials in ErbB4−/− mice, compared with control mice (t-test, ***p<0.001).

(F, I) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta (F) or theta (I) oscillations in correct (top) and error (bottom) trials in control mice (left panels) or ErbB4−/− mice (right panels).

(G, J) Reduced delta (G) or theta (J) PPC values in error trials in both genotypes (control: n = 96 units, 7 mice; ErbB4−/−: n = 90 units, 7 mice; for G: paired t-test, control: **P=0.0015, ErbB4−/−: *P=0.0313; for J: control: ***P<0.001; ErbB4: ns: not significant, P = 0.5909).

(H, K) Diminished difference (diff.) in delta (H) or theta (K) PPC values between correct and error trials in ErbB4−/− mice (t-test, for H: *p=0.0398; for K: *p = 0.0469).

(L, M) (top) Heatmaps showing normalized phase-locking values (PPC) for each PFC neuron in control mice (L) or ErbB4−/− mice (M) during correct trials (left) or error trials (right). (bottom) Histograms showing the distribution of the maximal PPC value for each neuron in control mice during correct trials (left) or error trials (right). Triangles indicate mean lag value across the population. In L, for correct trials: mean = −15.09, two-tailed signed-rank test, *p = 0.0204; for error trials: mean = −4.18, two-tailed signed-rank test, ns, p = 0.7769; in M, for correct trials: mean = −8.32, two-tailed signed-rank test, ns, p = 0.225; for error trials: mean = −6.489, two-tailed signed-rank test, ns, p = 0.4513).

Pooled data are shown as mean ± SEM.

To further test this hypothesis, phase-locking data were subjected to lag analysis (Bolkan et al., 2017; Likhtik et al., 2014; Spellman et al., 2015). We focused theta bands which showed largest difference between correct and error trials. The mean maximum PPC of spike times during correct trials occurred significantly before PFC neurons fired in control mice (Figure 3L), suggesting that the maximum phase locking occurs prior to PFC neuron firing when the attention level was high. However, the mean maximal phase locking to the vHPC happened around the PFC spike time in error trials of control mice and in both correct and error trials of ErbB4−/− mice, alluding to deficits in hippocampal signaling to the PFC (Figure 3L, M). Taking together, these data support the hypothesis that hippocampal information flows to the PFC for attention and ErbB4 activity plays an important role in regulating attention-associated vHPC-PFC synchrony.

Acute, chemical-genetic inhibition ErbB4 impairs attention and the hippocampal-prefrontal synchrony

The impaired attention and reduced hippocampal-prefrontal synchrony in ErbB4−/− mice may result from improper neural development and/or E/I imbalance in adult animals. To address this question and to determine which one is the cause, we aimed to inhibit ErbB4 pharmacologically. Unfortunately, this approach requires a specific inhibitor of ErbB4, which is not available. To this end, we took a chemical genetic strategy (Bishop et al., 2000; Cao et al., 2008; Chen et al., 2005) to enlarge the ATP binding pocket of ErbB4 by mutating the conserved threonine 796 to glycine (Figure S3A, B). This T796G mutation enabled specific inhibition by the bulky inhibitor 1NM-PP1 in HEK293 cells (Figure S3C–F). This inhibitory effect is specific because 1NM-PP1 had no effect on NRG1 activation of WT ErbB4 (Figure S3E, F). We then generated T796G knock-in mutant mice. Unlike ErbB4 null mice that die in embryos, T796G mice are viable and fertile, with comparable body weight and brain size with control mice (Figure S4A, B). Also unlike ErbB4−/− mice that display a reduction in PV+ interneurons in the hippocampus and PFC (Flames et al., 2004), T796G mice exhibit similar numbers of neurons and PV+ interneurons in these brain regions (Figure S4C–F). These data suggest that the T796G point mutation had little effect on neural development.

To investigate whether ErbB4 kinase activity is necessary for attention, T796G mice were trained for 5-CSRTT. As shown in Figure 4A, T796G mice were able to reach progression criteria during training and achieve same accuracy, compared to control mice, indicating that T796G mice had no problem in acquiring 5-CSRTT task (Figure 4A, Figure S5A). When 1NM-PP1 was administered systemically (i.p.) at 0.1 μg/g bodyweight, a dose that inhibits ATP-binding pocket-enlarged TrkB and CaMKII (Cao et al., 2008; Chen et al., 2005), tyrosine-phosphorylated ErbB4 in the brain was reduced within 15 min, indicating in-vivo inhibition of ErbB4. The inhibition peaked around 30 min after injection (Figure 4B), in a time course similar to CaMKII chemical genetic mutant (Cao et al., 2008). Fully trained T796G mice were injected with 1NM-PP1 and 15 min later, subjected to 5-CSRTT. Remarkably, 1NM-PP1 increased the omission error ratio (but not incorrect or premature ratio) (Figure 4C) and the latency to correct, rather than latency to incorrect and latency to reward (Figure S5B). Notice that 1NM-PP1 had no effect on 5-CSRTT performance of control mice (Figure 4C and Figure S5B), indicating the specificity of the effect. These data demonstrate that acute inhibition of NRG1/ErbB4 signaling impairs attention.

Figure 4. Acute inhibition of ErbB4 impairs attention and the vHPC-PFC synchrony.

(A) Similar accuracy rates between T796G and control mice during 5-CSRTT training (control, n = 11 mice; T796G, n = 11 mice; Two way ANOVA, p=0.9062).

(B) Time course of 1NM-PP1 inhibition of ErbB4 phosphorylation after i.p. injection (repeated 3 times, t-test, 5 min: *p=0.0392; 15 min: **p=0.0042; 30 min: ***p<0.001; 45 min: **p=0.0022; 60 min: *p=0.0179; 75 min: ns, not significant, p=0.082).

(C) Increased omission ratios but not incorrect or premature ratios of T796G mice, but not control mice, after 1NM-PP1 treatment (i.p.; control: n=11 mice, two-way ANOVA, incorrect: ns, p=0.9171; omission: ns, p= 0.7435; premature: ns, p=0.9563; T796G: n=11 mice, two-way ANOVA, incorrect: ns, p=0.9095; omission: ***p<0.001; premature: ns, p=0.9301; ns: not significant).

(D, I) Coherence between prefrontal and hippocampal LFPs after i.p. injection of vehicle (grey) and 1NM-PP1 (red) in T796G mice (D) or control mice (I). For D, n=11 tests, 6 mice; two-way ANOVA, ***p<0.001; post hoc, Bonferroni, delta: ***P<0.001; theta: ***P<0.001; beta: ns, p=0.9936; gamma: ns, p=0.1854. For I, n=11 tests; 6 mice; two way ANOVA, ns, p=0.9991; ns: not significant).

(E, J) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta oscillation in vehicle (top) or 1NM-PP1 (bottom) injection in T796G mice (E) or control mice (J).

(F, K) Reduced delta PPC in 1NM-PP1 injection in T796G mice (F), but not control mice (K). For F, vehicle, n=94 units, 6 mice; 1NM-PP1, n=99 units, 6 mice; t-test, ***p<0.001. For K, vehicle, n=89 units, 6 mice; 1NM-PP1, n=95 units, 6 mice; t-test, ns: not significant, p=0.89.

(G, L) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal theta oscillation in vehicle (top) or 1NM-PP1 (bottom) injection in T796G mice (G) and control mice (L).

(H, M) Reduced theta PPC in 1NM-PP1 injection in T796G mice (H), but not control mice (M). For H, vehicle, n=94 units, 6 mice; 1NM-PP1, n=99 units, 6 mice; t-test, **p=0.007581. For M, vehicle, n=89 units, 6 mice; 1NM-PP1, n=95 units, 6 mice; t-test, ns: not significant, p=0.8241.

Pooled data are shown as mean ± SEM. See also Figure S3-5.

To determine whether the hippocampal-prefrontal synchrony requires ErbB4 activity, we looked at LFP coherence and spike firing phase-locking 15 min after 1NM-PP1 administration. Compared with vehicle injection, 1NM-PP1 reduced the synchrony in delta and theta (but not gamma) frequencies (Figure 4D) and decreased PFC spike phase-locking to hippocampal LFPs (Figure 4E–H). Notice that 1NM-PP1 had no effect on basic electrophysiological properties of T796G PFC neurons (Figure S5C). Moreover, 1NM-PP1 had no effects on both LFP coherence and spike phase-locking of control mice (Figure 4I–M, Figure S5D). These results suggest that acute, dynamic ErbB4 kinase activity is required for the hippocampal-prefrontal synchrony. Next, we determined whether ErbB4 kinase is required for attention-associated synchrony. In vehicle-treated mice, as observed earlier, the coherence of vHPC-PFC LFPs was higher with correct trials at delta and theta bands, compared with error trials (Figure 5A–E). In 1NM-PP1-injected mice, correct trials were associated with higher coherence than error trials; however, the coherence reduction in error trials at delta and theta bands was significantly diminished in 1NM-PP1-injected mice, compared with mice injected with vehicle (Figure 5D, E). Although correct trials were associated with higher PPC values in 1NM-PP1-treated mice, however, the difference of PPC values between correct and error trials was reduced, compared with vehicle-injected mice (Figure 5F–K). These results suggest that acute ErbB4 activity is required for proper attention behavior and associated vHPC-PFC synchrony.

Figure 5. Diminished attention-associated hippocampal-prefrontal synchrony by 1NM-PP1.

(A) Coherence between prefrontal and hippocampal LFPs of correct and error trials in vehicle- and 1NM-PP1-treated T796G mice.

(B, C) Reduced delta (B) and theta (C) coherences in error trials in vehicle- and 1NM-PP1-treated T796G mice. Vehicle: n=13 tests (7 mice), paired t-test, delta: ***p<0.001; theta: ***p<0.001. 1NM-PP1: n=12 tests (7 mice), paired t-test, delta: *p=0.0151; theta: ***p<0.001.

(D, E) Diminished difference (diff.) between correct and error trials in delta (D) or theta (E) coherence in 1NM-PP1-treated T796G mice, compared with vehicle-treated T796G mice (t-test, for D, *p=0.0254; for E, *p=0.0236).

(F, I) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta (F) or theta (I) oscillations in correct (top) and error (bottom) trials in T796G mice injected with vehicle (left panels) or 1NM-PP1 (right panels).

(G, J) Reduced delta (G) or theta (J) PPC values in error trials, compared with correct trials, in vehicle-and 1NM-PP1-injected T796G mice (vehicle: n = 94 units, 7 mice; paired t-test, delta: ***p<0.001; theta: ***p<0.001. 1NM-PP1: n = 99 units, 7 mice, paired t-test, delta: ***p<0.001; theta: *p=0.0101).

(H, K) Diminished difference (diff.) in delta (H) or theta (K) PPC values between correct and error trials in 1NM-PP1-treated T796G mice, compared with vehicle-injected T796G mice (t-test, delta: **p=0.0029, theta: *p=0.0158).

Pooled data are shown as mean ± SEM.

Hippocampal, but not prefrontal ErbB4 activity for the hippocampal-prefrontal synchrony in attention

To determine which brain region is critical for the synchrony during 5-CSRTT, 1NM-PP1 was injected to PFC or vHPC; sites of injection were confirmed by postmortem morphology study (Figure S6A, B). 1NM-PP1 injection (50 μM, 0.2 μl) into the vHPC suppressed the power of vHPC LFPs at gamma bands (Figure S6C, D). It also reduced the hippocampal-prefrontal coherence at delta and theta frequencies (Figure 6A), PFC spikes that were phase-locked with vHPC LFPs, and PPC values (Figure 6B–E, Figure S6E). On the other hand, 1NM-PP1 injection into PFC reduced the power of PFC LFPs at gamma bands (Figure S6F, G), but had little effect on the hippocampal-prefrontal synchrony (Figure 6F), PPC values, and PFC spikes that were phase-locked with vHPC LFPs (Figure 6G–J, Figure S6H). These results demonstrate that acute ErbB4 inhibition in hippocampus or PFC impaired local LFP power. However, only ErbB4 activity in the hippocampus is necessary for attention-associated hippocampal-prefrontal synchrony. At behavioral level, 1NM-PP1 injection into either vHPC or PFC increased the omission error ratio (Figure 6K, L), suggest that ErbB4 in these two regions is necessary for attention. Interestingly, 1NM-PP1 injection into the vHPC increased the latency to correct response (Figure S6I); however, this effect was not observed in T796G mice where 1NM-PP1 was injected into the PFC (Figure S6J). Latency to incorrect response had no change in both conditions (Figure S6I, J). Because 1NM-PP1 injection into the vHPC and PFC had little effect on the latency to reward (Figure S6I, J), excluding a problem with motivation or mobility. Together, these observations suggest that ErbB4 in both regions is necessary for attention.

Figure 6. Hippocampal ErbB4 activity is required for attention and vHPC-PFC synchrony.

(A) Reduced hippocampal-prefrontal synchrony in T796G mice by vHPC injection of 1NM-PP1. (n=12 tests (6 mice); two way ANOVA, ***p<0.001; post hoc, Bonferroni, delta: ***P<0.001; theta: ***P<0.001; beta: ns, p=0.941; gamma: ns, p=0.9514; ns: not significant). Inset: Schematic diagram of a mouse brain showing positions of recording electrodes and drug infusion cannula.

(B) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta oscillation in T796G mice after vHPC vehicle (top) or 1NM-PP1 (bottom) injection.

(C) Reduced delta PPC by vHPC 1NM-PP1 injection, compared with vehicle injection (vehicle: n=102 units (6 mice); 1NM-PP1: n=95 units (6 mice); t-test, ***p<0.001).

(D) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal theta oscillation in T796G mice after vHPC injection with vehicle (top) or 1NM-PP1 (bottom).

(E) Reduced theta PPC by vHPC 1NM-PP1 injection, compared with vehicle injection (vehicle: n=102 units (6 mice); 1NM-PP1: n=95 units (6 mice); **p=0.0019).

(F) Little effect of local PFC injection of 1NM-PP1 on the hippocampal-prefrontal synchrony (n=12 tests (6 mice); two way ANOVA, ns: not significant, p=0.9991). Inset: Schematic diagram of a mouse brain showing positions of recording electrodes and drug infusion cannula.

(G) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta oscillation in T796G mice after PFC injection with vehicle (top) or 1NM-PP1 (bottom).

(H) No effect on delta PPC by PFC 1NM-PP1 injection, compared with vehicle injection (vehicle: n=99 units (6 mice); 1NM-PP1: n=94 units (6 mice); t-test, ns: not significant, p=0.9408).

(I) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal theta oscillation in T796G mice after PFC injection with vehicle (top) or 1NM-PP1 (bottom).

(J) No effect on theta PPC by PFC 1NM-PP1 injection, compared with vehicle injection (vehicle: n=99 units (6 mice); 1NM-PP1: n=94 units (6 mice); t-test, ns: not significant, p=0.9795).

(K) Increased omission ratio but not incorrect or premature ratios in T796G mice by local vHPC injection of 1NM-PP1 or bicuculline (n=11 mice, two-way ANOVA, incorrect: ns, p=0.7198/0.91/0.9501, respectively; omission: ***p< 0.001; premature: ns, p=0.8317/0.7328/0.9531, respectively; ns: not significant).

(L) Increased omission ratio but not incorrect or premature ratios in T796G mice by local PFC injection of 1NM-PP1 or bicuculline (n=11 mice, two-way ANOVA, incorrect: ns, p=0.6861/0.8652/0.9196, respectively; omission: ***p< 0.001; premature: ns, p=0.865/0.987/0.9347, respectively; ns: not significant).

Pooled data are shown as mean ± SEM. See also Figure S6.

Acute inhibition of ErbB4 impairs GABA transmission

We next investigated underlying mechanisms of hippocampal ErbB4 for the hippocampal-prefrontal synchrony and attention. Because T796G mice are normal in lamination and numbers of neurons (including interneurons) and because the effect of 1NM-PP1 is acute, we examined GABA transmission in the hippocampus. Previous studies indicate that ErbB4 is required for maintaining GABA activity in the hippocampus (Chen et al., 2010). In agreement, treatment of 1NM-PP1 of cultured hippocampal neurons from T796G mice, which inhibited ErbB4 activation (Figure S7), reduced the frequency of miniature inhibitory postsynaptic currents (mIPSC) and the amplitude of evoked IPSC (eIPSC) of CA1 pyramidal neurons (Figure 7), but had no effect on miniature excitatory postsynaptic currents (mEPSCs) (Figure S8A, B). The inhibitory effect on mIPSC had the following characteristics. First, it occurred within 15 min of treatment and was reversible (Figure 7A–D). Second, it was occluded by PD168393, a pharmacological inhibitor of ErbB4 (Figure S8C, D), indicating that the effect of 1NM-PP1 is mediated by inhibiting ErbB4. Third, 1NM-PP1 had no effect on mIPSC or eIPSC in hippocampal slices of control mice (Figure 7C–F), indicating 1NM-PP1 is specific for T796G ErbB4. These results indicate that 1NM-PP1, via inhibiting ErbB4, specifically reduces GABA transmission. Mechanistically, 1NM-PP1 had no effect on mIPSC amplitude (Figure 7B, D), but increased the paired-pulse ratio (PPR) (Figure 7G, H), suggesting a decrease in presynaptic GABA release probability. As shown previously, exogenous NRG1 increases mIPSC frequency (Chen et al., 2010); this effect was blocked by 1NM-PP1 (Figure S8E, F). Similar inhibitory effect of 1NM-PP1 on GABA activity was observed in prefrontal slices of T796G mice (data not shown).

Figure 7. 1NM-PP1 inhibits hippocampal GABAergic transmission in T796G mice.

(A) Diagram of recording a pyramidal neuron (grey) that was innervated by an interneuron (yellow).

(B) Represent traces of mIPSC before, during and after 1NM-PP1 application.

(C) Time course of 1NM-PP1 inhibition of mIPSC frequency on hippocampal slices from T796G mice, but not control mice (control n=6 neurons (3 mice); T796G n=9 neurons (4 mice); one way ANOVA, ***P<0.001).

(D) No effect of 1NM-PP1 on mIPSC amplitude on hippocampal slices of either genotype (control n=6 neurons (3 mice); T796G n=9 neurons (4 mice); one way ANOVA, ns, p=0.9935).

(E) Represent traces of eIPSC before, during and after 1NM-PP1 application in control and T796G mice.

(F) 1NM-PP1 inhibition of eIPSC amplitude in hippocampal slices from T796G (n=6 neurons (3 mice), t-test, **P=0.0011; *p=0.0294) but not control mice (n=7 neurons (3 mice), t-test, ns, p=0.8534; ns, p=0.9675; ns: not significant).

(G) Represent traces of paired-pulse (interval 0.25s) stimuli.

(H) Increased eIPSC paired-pulse ratio by 1NM-PP1 (n=6 neurons (3 mice), one way ANOVA, **p=0.0013).

Pooled data are shown as mean ± SEM. See also Figure S7, S8.

As result of inhibited GABA activity, 1NM-PP1 increased the excitability of pyramidal neurons (Figure S8G, H). In accord, 1NM-PP1 facilitated long-term potentiation (LTP) induction in T796G hippocampal slices (Figure S8I), but had little effect on slices from control mice (Figure S8J). This effect was occluded by GABA_A receptor antagonist bicuculline (Figure S8K, L), suggesting it is likely to be mediated through reducing GABA transmission. Together, these results support the hypothesis that 1NM-PP1 injection in the hippocampus disrupts the hippocampal-prefrontal synchrony and attention by inhibiting local ErbB4 activity and GABA activity. On the other hand, ErbB4 activity and its regulation of GABA activity in the PFC are required for attention but not the hippocampal-prefrontal synchrony.

To further test the involvement of ErbB4-regulated GABA activity in attention, we injected bicuculline into hippocampus and PFC, respectively. As shown in Figure 6K and 6L, bicuculline injection (10 μM, 0.2 μl) into either region increased error ratio, as observed in 1NM-PP1 injected mice. However, in the presence of bicuculline, 1NM-PP1 was unable to further increase the omission error ratio (Figure 6K, L). The occlusion effect of bicuculline provides further evidence that 1NM-PP1 regulates attention by inhibiting ErbB4-regulated GABA activity. On the other hand, bicuculline injection into the vHPC, not PFC, increased the latency to correct response, and displayed similar occlusion effect on 1NM-PP1 (Figure S6I, J). Both 1NM-PP1 and biccuculline had no effect on the latency to incorrect response or latency to reward (Figure S6I, J). A parsimonious interpretation of these results is that GABA activities in hippocampus and PFC, both controlled by ErbB4, are critical for attention. Moreover, the subcortical structure hippocampus regulates attention by promoting the hippocampal-prefrontal synchrony.

Hippocampal ErbB4+ interneuron circuits for the hippocampal-prefrontal synchrony and attention

Next, we determined whether the hippocampal-prefrontal synchrony requires proper activity of ErbB4+ interneurons. Archaerhodopsin (ArchT), a light-driven proton pump which enables robust and reversible photoinhibition of neuronal action potential firing (Chow et al., 2010), was expressed in ErbB4+ interneurons by crossing ArchT^flox/flox mice with ErbB4-CreER^T2 mice that express Cre under the control of the endogenous ErbB4 promoter (Bean et al., 2014). Resulting ErbB4-CreER^T2;ArchT^flox/flox (ErbB4-ArchT) mice were characterized by slice recording. Action potentials of ErbB4+ interneurons elicited by current injection were reduced in frequency by yellow light (590 nm) stimulation (Figure S9A). This effect was not observed in ErbB4- neurons (Figure S9B); in contrast, some ErbB4- neurons may be excited by yellow light stimulation (Figure S9C). In 5CSRTT behavior test, ErbB4-ArchT mice had similar learning time course with control ArchT^flox/flox mice (Figure S9D). Fully trained ErbB4-ArchT mice were implanted with optoelectrodes into vHPC or PFC (Figure S9E, F). Activation of ArchT in vivo inhibited firing rates of action potentials of a set of neurons (presumably ErbB4+ interneurons, ~20.7%) (Figure 8A, S9G) and increased firing rates of some neurons (presumably target neurons of ErbB4+ interneurons, ~43.2%) (Figure 8B, S9G). Neurons whose firing rates were not changed may represent those that did not receive direct input from ErbB4+ interneurons (~36.1%) (Figure S9G).

Figure 8. Distinct regulations by vHPC and PFC ErbB4+ neurons on the hippocampal-prefrontal synchrony and attention.

(A, B) Representative in vivo multi-unit raster in ErbB4-CreER^T2;ArchT mice in response to yellow LED illumination.

(C) No effect on the hippocampal-prefrontal synchrony-by yellow LED illumination at the PFC (n=12 tests (6 mice); two way ANOVA, ns: not significant, p=0.9991). Inset: Schematic diagram of a mouse brain showing positions of optoelectrodes and tetrodes.

(D) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta oscillation without (top) or with (bottom) LED stimulation at the PFC.

(E) No effect on delta PPC by PFC LED illumination, compared with no LED illumination (n=83 units (6 mice), t-test, ns: not significant, p=0.8804).

(F) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal theta oscillation without (top) or with (bottom) LED stimulation at the PFC.

(G) No effect on theta PPC by PFC LED illumination, compared with no LED illumination (n=83 units (6 mice), t-test, ns: not significant, p=0.4752).

(H) Increased omission (Omis.) ratio but not incorrect (Incor.) or premature (Prem.) ratio by LED illumination in the PFC (n=12 mice, two-way ANOVA, incorrect: ns, p=0.9897; omission: ***p< 0.001; premature: ns, p=0.5986; ns: not significant).

(I) Reduced hippocampal-prefrontal synchrony by yellow LED illumination at the vHPC (n=86 units (6 mice); two way ANOVA, ***p<0.001; post hoc, Bonferroni, delta: ***P<0.001; theta: ***P<0.001; beta: ns, p=0.933; gamma: ns, p=0.9715; ns: not significant). Inset: Schematic diagram of a mouse brain showing positions of optoelectrodes and tetrodes.

(J) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal delta oscillation without (top) or with (bottom) LED stimulation at vHPC.

(K) Reduced delta PPC by vHPC LED illumination, compared with no LED illumination (n=86 units (6 mice), t-test, ***p<0.001).

(L) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal theta oscillation without (top) or with (bottom) LED stimulation at vHPC.

(M) Reduced theta PPC by vHPC LED illumination, compared with no LED illumination (n=86 units (6 mice), t-test, ***p<0.001).

(N) Increased omission (Omis.) ratio but not incorrect (Incor.) or premature (Prem.) by LED illumination in the vHPC (n=12 mice, two-way ANOVA, incorrect: ns, p=0.8092; omission: ***p< 0.001; premature: ns, p=0.8709; ns: not significant).

Pooled data are shown as mean ± SEM. See also Figure S9.

As shown in Figure S9H and S9I, the power of PFC LFPs was reduced at gamma band by yellow light stimulation of the PFC. The stimulation had little effect on the coherence between vHPC and PFC (Figure 8C), PPC values (Figure 8D–G), or PFC spikes that were phase-locked with vHPC LFPs (Figure 8D, F). These results suggest that inhibiting PFC ErbB4+ interneurons disrupts local LFP power, but did not alter the hippocampal-prefrontal synchrony. Concomitantly, the omission error ratio but not incorrect or premature error ratio was increased (Figure 8H), indicating that attention requires the activity of PFC ErbB4 interneurons. Light stimulation of PFC did not change the latencies to correct, incorrect and reward response (Figure S9J).

On the other hand, yellow light stimulation of vHPC decreased LFP power at gamma band (Figure S9K, L) as well as the vHPC-PFC synchrony as evidenced by reduced coherence at delta and theta bands (Figure 8I), decreased PPC values and fewer PFC spikes that were phase-locked with vHPC LFPs (Figure 8J–M). These observations indicate that the hippocampal-prefrontal synchrony requires the proper activity of ErbB4+ interneurons in the hippocampus. Finally, omission ratios and latency to correct response, rather than incorrect or premature ratios, were both increased in ErbB4-ArchT mice whose vHPC was stimulated (Figure 8N, S9M). Light stimulation of both PFC and vHPC had no effect on the latency to incorrect and latency to reward (Figure S9J, M), excluding a problem with motivation or mobility. As control, light stimulation had no effect on 5-CSRTT performance in ArchT^flox/flox mice (Figure S9N, O), indicating that the effect is dependent on ArchT in ErbB4+ interneurons. These results support the model that ErbB4+ interneuron circuits in the hippocampus are necessary for attention by promoting the hippocampal-prefrontal synchrony.

DISCUSSION

Studies of various species have identified DLPFC in human and monkey and PFC in rodents as a key region to control attention and to modulate neural responses in subcortical structures (Baluch and Itti, 2011; Li et al., 2010). For bottom-up attention, sensory signals are processed first in perspective centers and relayed to DLPFC or PFC to initiate attention. On the other hand, top-down attention, which is critical for goal-driven behavior, arises from the PFC in mice (Baluch and Itti, 2011; Buschman and Miller, 2007; Kim et al., 2016; Li et al., 2010). Recent data suggest that top-down attention may involve subcortical structures (Baluch and Itti, 2011; Goldfarb et al., 2016); however, a causal link of subcortical areas to the PFC has not been well established. Neuronal activities of PFC and hippocampus are synchronized (Jones and Wilson, 2005); this synchrony has been implicated in spatial working memory (Spellman et al., 2015) and is impaired in mouse models of brain disorders (Sigurdsson et al., 2010). We show here that PFC and hippocampus are better synchronized, at theta and delta bands, when mice are making a correct choice in 5-CSRTT, suggesting that the hippocampal activity may regulate the function of PFC during attention process.

Both PFC and vHPC are important in attention (Buschman and Miller, 2007; Goldfarb et al., 2016; Kim et al., 2016; McGarrity et al., 2016). Modulation of synaptic transmission from PV+ interneurons regulates top-down attention (Kim et al., 2016) whereas injecting a GABA antagonist into the hippocampus impairs attention (McGarrity et al., 2016). Here we used two approaches to specifically disrupt hippocampal GABA activity. First, taking the advantage of the chemical genetic mutant mice, we inhibited ErbB4 kinase activity specifically in the hippocampus by locally injecting 1NM-PP1 into adult T796G mice and show that 1NM-PP1 reduces the hippocampal-prefrontal synchrony and impaired attention. This effect is specific because ErbB4 is the only kinase in the entire brain or body in T796G mice that can be inhibited by 1NM-PP1. Second, in an optogenetic approach, we specifically inhibited ErbB4+ interneurons in the hippocampus by locally delivered yellow light. This also reduces the synchrony and impairs attention. In contrast, 1NM-PP1 injection into the PFC or optogenetic inhibition of ErbB4+ interneurons in the PFC impaired attention, but had no effect on the synchrony. These results provide compelling evidence that ErbB4+ interneurons and acute, dynamic ErbB4 activity in these neurons in the hippocampus play a critical role in regulating the function of PFC. In support of this model, earlier work indicates that pyramidal neurons in the hippocampus project to the PFC in mice, and this projection is not reciprocal (Jay et al., 1989; Swanson, 1981).

The correct trials that ErbB4−/− and 1NM-PP1-treated T796G mice performed were >40% at most of the time, more than ~20% which is predicted as result of “guess”. These data suggest that these mice remain able to perform the task, but with less engagement of hippocampal-PFC circuitry. Intriguingly, after correction for multiple comparisons with Bonferroni test, the two-way ANOVA analysis indicated that the vHPC-PFC synchrony at gamma bands was not altered significantly by ErbB4 mutation and 1NM-PP1 inhibition of ErbB4 (via systemic and hippocampal injection) (Figure 2D, 4D and 6A). Nevertheless, the synchrony reduction was significant at low (30-60 Hz) gamma, but not at high (60-100 Hz) gamma bands (data not shown). This reduction may contribute to attention deficits, as suggested for theta and delta bands.

Our finding of increased gamma oscillation in both hippocampus and PFC in ErbB4 null mutant mice is consistent with previous studies that specifically deletion of NMDAR NR1 subunits or ErbB4 in PV+ interneurons (Carlen et al., 2012; Del Pino et al., 2013; Korotkova et al., 2010). However, gamma oscillation was reduced by inhibiting ErbB4 by 1NM-PP1 (by either systemic or hippocampal injection) or inhibiting ErbB4+ neurons by optogenetics (Figure S6C, S6F, S9H and S9K). These effects are in general agreement with reports that acute activation of PV+ interneurons increases gamma oscillation (Sohal et al., 2009) and that NRG1 increases gamma oscillation of hippocampal slices in vitro (Fisahn et al., 2009). The different effects between ErbB4 null mutation and acute inhibition on gamma oscillation are likely to be caused by circuit deficiency (due to loss of ErbB4 during development) and associated problems with synaptic transmission in ErbB4 mutant mice. T796G mice are normal in neural development (Figure S4), and the effect of 1NM-PP1 on oscillation is likely mediated by acute ErbB4 inhibition.

Interneurons reduce neuronal excitability and provide a complex modulation system to control the output of a local circuit (Hu et al., 2014; Isaacson and Scanziani, 2011). A majority of ErbB4+ interneurons in the PFC and hippocampus are PV+ (Bean et al., 2014). PV+ interneurons have two major types: basket cells that form perisomatic synapses onto the somas of pyramidal neurons and chandelier interneurons that form synapse on the AISs of pyramidal neurons, both of which control the firing and output of pyramidal neurons (Hu et al., 2014). Pharmacological and genetic studies indicate that NRG1 promotes GABA release in an ErbB4-dependent manner and thus represses the neural activity in the hippocampus and PFC (Chen et al., 2010; Wen et al., 2010; Woo et al., 2007). Importantly, neutralizing endogenous NRG1, pharmacological inhibition of ErbB4 and genetic ablation of the ErbB4 gene, reduce GABAergic transmission (Chen et al., 2010; Pitcher et al., 2008; Wen et al., 2010; Woo et al., 2007), indicating a critical role of ErbB4 in maintaining homeostatic GABA activity. In anesthetized mice, ErbB4 knockout in interneurons regulates gamma oscillation in the dorsal hippocampus and PFC and reduced the synchrony between the two regions (Del Pino et al., 2013). In these studies, however, it was difficult to exclude off-target effects of pharmacological inhibitors and effects of impaired GABAergic circuits due to ErbB4-loss during development. Here we demonstrate that chemical-genetic inhibition of ErbB4 in adult T796G mice reduced GABA transmission (Figure 7A–D), providing compelling evidence that homeostatic GABA activity depends on acute, dynamic ErbB4 activity.

In addition to PV+ interneurons, ErbB4 is also expressed in interneurons positive for CCK or nNOS in the hippocampus and PFC (Del Pino et al., 2017; Neddens and Buonanno, 2010). Delta and Theta power was reduced when Erbb4 was mutated in CCK interneurons (Del Pino et al., 2017). ErbB4 and SOM double positive interneurons are rare in the hippocampus (Bean et al., 2014; Neddens and Buonanno, 2010), but abundant in TRN where they regulate bottom-up attention (Ahrens et al., 2015). In visual cortex, ErbB4 is mainly present in PV+ and VIP+ cells, and less abundant in SOM+ cells where it regulates formation of dominant column and temporal precision of pyramidal neuron activity and network (Batista-Brito et al., 2017; Sun et al., 2016). Whether these interneurons contribute to attention and if so, how, warrant future studies.

In sum, we provide evidence that attention requires acute, NRG1/ErbB4-regulated GABA activity in both PFC and hippocampus. It is the hippocampus where ErbB4 activity is necessary for the hippocampal-prefrontal synchrony. Therefore, hippocampal ErbB4 and ErbB4+ interneurons play an important role in regulating the activity of PFC. Attention deficit is a common symptom of various neurological and psychiatric disorders including schizophrenia, ADHD, and Alzheimer’s disease (Chen et al., 1998; Clark et al., 2002; Perry and Hodges, 1999). Given that ErbB4 and its ligand NRG1 are both susceptibility genes of schizophrenia, bipolar disorders and major depression (Del Pino et al., 2013; Mei and Nave, 2014), our findings provide insight into potential pathophysiological mechanisms of these brain disorders. T796G mice could be a useful tool to study brain functions by inhibiting ErbB4 and GABA transmission in a temporally and spatially-controlled manner.

STAR Methods

Detailed methods are provided in the online version of this paper and including the following:

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-phosphorylated-ErbB4	Cell signaling	Cat# 4757S; RRID: AB_2099987
Rabbit polyclonal anti-pan-ErbB4	Dr. Cary Lai	0618
Rabbit polyclonal anti-PV	Swant	Cat# PV 25; RRID: AB_10000344
Mouse monoclonal anti-NeuN	Neuromics	Cat# MO22122
Chemicals, Peptides, and Recombinant Proteins
1NM-PP1	EMD Milipore	Cat# 529581
PD168393	Sigma-Aldrich	Cat# PZ0285
NRG1	Holmes et al. 1992	rHRG β177-244
Experimental Models: Organisms/Strains
Mouse: T796G	This paper	N/A
Mouse: ErbB4-CreERT2	The Jackson Laboratory	Jax No.: 012360
Mouse: Ai40	The Jackson Laboratory	Jax No.: 021188
Software and Algorithms
MATLAB	MathWorks	RRID: SCR_001622
NeuroExplorer	Nex Technologies	RRID: SCR__001818
Offline Sorter	Plexon	RRID: SCR_000012
pClamp 9.2	Molecular Devices	http://mdc.custhelp.com/app/answers/detail/a_id/18826/related/1

Contact for reagent and resource sharing

Further information and requests for reagents and resources should be directed to and will be fulfilled by Lin Mei (lin.mei@case.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

Mice were maintained on a 12-h light-dark cycle with ad libitum access to food and water. Male mice, 2-5 months of age, were used in the study. ErbB4 null mutant mice were crossed with αMHC-ErbB4 transgenic mice that express ErbB4 in the heart to generate ErbB4 null; aMHC-ErbB4 (ErbB4−/−) mice that were described previously (Woo et al., 2007). Erbb4 T796G knock-in mice were generated by homologous recombination in mouse ES cells. To generate the targeting construct, the 5′ and 3′ arms were PCR-amplified from C57BL/6J BAC DNA clone and subcloned into a PKI vector containing a loxP-neo-loxP cassette for positive selection and a diptheria toxin A negative-selection (DTA) cassette for negative selection. Threonine (T) 796 was encoded by the nucleotides in the exon 20 of the mouse Erbb4 gene and was mutated to glycine (G) by site-directed mutagenesis. The targeting constructs were confirmed by sequencing. Linearized targeting construct was electroporated into 129S6-C57BL/6J hybrid ES cells, which were subsequently selected with G418. Correct targeting of ES cells was initially screened by Southern analysis and then confirmed by direct sequencing of PCR products amplified from the mutated alleles. The positive ES cell clones were microinjected into C57BL/6J blastocysts to generate chimeric mice. Male chimeric animals were mated to C57BL/6J wild-type females for germline transmission of the targeted allele. The F1 offspring were crossed with CMV-Cre mice to remove the neomycin resistance cassette. The primers for genotyping wild type Erbb4 allele were as follows: forward, 5′-GATCT GCAGA TCAAT TCAC-3′, reverse, 5′-GCCAA CCAAC TGGAT AGTG-3′. The primers for genotyping the T796G Erbb4 allele were as follows: forward, 5′ –AACTG AATTC ACTTT GTGG-3′, reverse, 5′-CTGTA GCAGC AACAA TAGC-3′. Mice were maintained as homozygotes in a mixed 129S6 and C57BL/6J background.

ErbB4-ArchT mice were generated by crossing ErbB4-CreER^T2 mice (Jackson Lab, stock number 012360) with Ai40 mice (Jackson Lab, stock number 021188). In ErbB4-CreER^T2 mice, the endogenous ErbB4 gene is intact and CreER^T2 fusion protein is expressed under the control of the endogenous promoter/enhancer regions of the ErbB4 gene. Ai40 mice express an Archaerhodopsin-3/EGFP fusion protein (ArchT-EGFP) that is restricted by the floxed-STOP cassette. When ErbB4-ArchT mice were injected with tamoxifen to induce Cre-mediated excision of the STOP signal in the floxed-STOP cassette, ArchT-EGFP is expressed in ErbB4+ cells. Animal protocols have been approved by the Institutional Animal Care and Use Committee (IACUC) of Augusta University and Case Western Reserve University, all experiments conformed to the guidelines of the National Institutes of Health (NIH).

METHOD DETAILS

Behavioral analysis

For 5-CSRTT, mice were food-deprived to maintain 85% (22.6 ± 1.4 g) of the body weight of mice with ad libitum access to food (27.3 ± 1.6 g). In this task, mice were trained to report the location of a visual stimulus in order to receive a food reward. The 5-CSRTT operant chambers (Med-Associates, USA) were equipped with 1 house light and 5 stimulus (cue) presentation holes with internal light-emitting diodes (LED). The holes are equipped with an infrared sensor to detect nose pokes by mice. The reward port, situated on the rear wall, has a reward magazine to dispense food pellets (20 mg/pellet) and an infrared sensor to detect the insertion of mouse nose. One food pellet was dispensed when a correct choice was made. During 4-day habituation, mice were first handled by an investigator for 5 min and placed in an operant chamber. Two food pellets were placed in each stimulus hole and 10 pellets were placed in the reward magazine, to guide mice to explore operant chambers for 20 min. During training, the LED of one of the 5 cue holes will be on and mice were trained to identify the lit hole by nose poke to be rewarded with a food pellet. Training was divided into 7-successive stages each with specific criteria, which had to be met for two consecutive days before progression into the next stage (Figure S1A). During the training, LED-on duration was reduced from 30 to 0.8 sec and the limited hold was reduced from 30 to 5 sec. When the response was correct, mice were allowed 8 sec for food pellet consumption. The time from cue onset to first registered nose poke of correct response was defined as latency to correct. The time between nose poke into the cue hole and reward port trigger in a correct trial was defined as the latency to reward. Nose pokes into one of the other four holes that were not lit were scored as incorrect response. Failure to report the cue location within the limited hold (5 sec for fully trained mice) was scored as an omitted trial (omission), and a nose poke before LED/cue presentation was scored as a premature response. Incorrect, premature and omitted responses resulted in a 5-sec darkness timeout, during which a new trial could not be initiated. A typical training session lasted 30 min or had 100 trials, whichever came first. The inter trial intervals (ITI, time from trial start to cue onset) was 2 sec for stages 1 and 2 and 5 sec for stages 3 and 4. At stages 5-7, the ITIs were 3, 4, or 5 sec that were presented pseudorandomly to increase attentional load. All sessions were videotaped.

Electrode, fiber, and cannula implantation

Mice were anaesthetized with ketamine/xylazine (Sigma, 0.1 ml/10 g bodyweight, i.p.), shaved on the skull and positioned on a stereotaxic apparatus. After antiseptic treatment, the scalp was removed, and the exposed skull area was cleared using 1% H₂O₂. For in-vivo recording, craniotomy was performed unilaterally above the prelimbic PFC (1.8 mm anterior, 0.4 mm lateral, 1.4 mm ventral), and vHPC (3.1 mm posterior, 3.0 mm lateral, 3.9 mm ventral). Two sets of tetrodes were implanted in the PFC and vHPC, respectively. Tetrodes were made from 13-μm-diameter platinum (with 10% iridium) fine wire (California Fine Wire Company, USA) and were attached to a movable screw microdrives on a custom-made frame (Lin et al., 2006). Skull screws were placed over the cerebellum and olfactory bulb as ground and reference, respectively. For optogenetics in combination of in-vivo recording, two tetrodes were glued to a ferrule-bound optical fiber (Thorlabs, 200-μM-diameter core, 0.39 NA) where tetrodes were arrayed semi-circularly around the lateral edge of the fiber and the fiber tip was positioned 300–500 μm dorsal to those of tetrodes. For 1NM-PP1 local injection, a cannula (Plastic One, 20 gauge, USA) was glued to tetrodes, with its tip positioned 300-500 dorsal to tetrode tips. Tetrodes with optic fiber and cannula were implanted as described above. To verify the locations of tetrodes, electro thermolytic lesions were induced by electric current (50 μA, 20 s) at the end of recording and brain samples were examined histologically for lesion locations. To verify the locations of cannula and 1NM-PP1 diffusion, Nile red was injected at the end of recording and brain samples were examined histologically. Data were discarded when tetrodes were not in correct locations.

In vivo recording and optogenetics

After surgery, mice were allowed for recovery at least 1 week before recording. Spikes and LFP signals were recorded simultaneously from the vHPC and PFC. Movable microdrives were advanced at a rate of 30 μm per day to record a spectrum of neurons. Signals were amplified; band-pass filtered (0.5-1,000 Hz for LFPs and 0.6-6 kHz for spikes) and digitized using the OmniPlex Neural Data Acquisition System (Plexon, USA). LFPs were collected at a rate of 1 kHz, while spikes, detected at adjustable online thresholds, were collected at 40 kHz. Units were clustered using the offline sorter MClust (MClust-4.4, A. D. Redish et al.) and specifically analyzed with a semi-automated super paramagnetic clustering (SPC) algorithm (“SPC”, B. Hasz). SPC is a non-parametric clustering method that is more appropriate for spikes sorting when the numbers and structures of clusters of neurons are unknown (Blatt et al., 1996). Two principal components for sorting were voltage peak and energy from each channel (Spellman et al., 2015). Clusters were accepted, merged or eliminated based on visual inspection of feature segregation, waveform distinctiveness and uniformity, stability across recording session, and inter-spike interval distribution (Spellman et al., 2015).

For optogenetics, ArchT expression was induced after surgery by tamoxifen (20 mg/kg bodyweight, i.p., 3 times every other day). Two weeks after the first injection of tamoxifen, mice were subjected to in-vivo recording or 5-CSRTT. Yellow light (590 nm) was delivered at 0.04 Hz (10 sec on, 15 sec off) from a LED light source (Plexon, USA). The frequency of the light was controlled by a stimulator (Master 8, Israel). Light intensity was measured with a power meter (Thorlab, USA) and maintained at 4-6 mW/mm² for recording. For 5-CSRTT, the light was delivered at the start of trial and terminated before cue emerges, during which mice were believed to have the highest attention level (Amitai and Markou, 2010; Kim et al., 2016; Robbins, 2002). Data were compared between light on and light off within the same animal.

Power and coherence analysis

Data were imported into MATLAB (MathWorks) for analysis using custom-written scripts. LFP recordings were processed in Neuroexplorer (Nex Technologies, USA). Fourier Transform of LFPs was performed and data was then binned with bin size of

$$\mathit{\text{Bin Size}}=\frac{1}{\left(2\times \mathit{\text{Maximum Frequency}}\right)}$$

and the number of bins equal to two times the number of frequency values, which was set to 1024 (Stujenske et al., 2014). The coherence for frequencies, f was calculated for PFC and vHPC, x(t) and y(t), as

$${\mathit{\text{Coherence}}}_{xy}\left(f\right)=\frac{{\left|S_{xy}\left(f\right)\right|}^2}{S_{xx}\left(f\right)S_{yy}\left(f\right)}$$

where S_xy(f) is the cross spectrum product of PFC and vHPC and S_xx(f) and S_yy(f) the auto-spectrum products from the two brain regions, respectively. The coherence was processed by Gaussian filter with a bin size of 3. Power spectrum data of each brain regions was calculated by normalizing auto-spectrum products so that the values are equal to the mean squared values from the rate histogram that were calculated from the overall data. Normalized data was then converted to logarithmic scale and filtered by Gaussian filter with bin size of 3.

Phase locking analysis

Hippocampal LFPs were digitally filtered using a band-pass filter, with filter order equivalent to the sampling frequency, with a zero-phase delay (filtfilt in MATLAB) to isolate frequencies in intended regions (0.5 Hz to 4 Hz for delta range, 4 Hz to 12 Hz for theta range and 30 Hz to 100 Hz range for gamma) (Spellman et al., 2015). Zero-phase delay filter filters both in the forward and reverse direction to avoid phase distortion and preserve temporal aspects of the data (Oppenheim et al., 1999). Filtered data was converted to corresponding phases by the Hilbert transform (hilbert function in MATLAB), which calculates real and imaginary parts of the data to obtain instantaneous amplitudes and phases at any given time point (Oppenheim et al., 1999). PFC spikes at individual time points were assigned to hippocampal LFP phases from the Hilbert transform. Only neurons with over 100 spikes were used in calculation to limit sample-size bias (Mukai et al., 2015; Tamura et al., 2016). To assess phase-locking, Pairwise Phase Consistency (PPC) values were calculated from the following two equations (Vinck et al., 2010).-First, the average pairwise circular distance (D) of a unit’s phase is calculated as:

$$D=\frac{2}{N\left(N-1\right)}{\displaystyle {\sum }_{j=1}^{N-1}{\displaystyle {\sum }_{k=\left(j+1\right)}^{N}d}}\left(\theta j,\theta k\right),$$

where d (θj, θk) was the absolute angular distance between two spike’s phases and N was the overall spike number. The circular distance function was used to find the angular distance (Anastassiou et al., 2011; Berens, 2009). PPC was calculated by normalizing D as follows:

$$\mathit{PPC}=\frac{\pi -2D}{\pi }.$$

Directionality analyses

Directionality between the PFC and vHPC was revealed by lag analysis that based on PPC of PFC spikes to vHPC LFPs (Bolkan et al., 2017; Likhtik et al., 2014; Spellman et al., 2015). PFC spike times from the three second behavioral time periods were obtained and the theta PPC value were successively calculated when shifting PFC spikes in 10-ms steps within ± 100-ms (Bolkan et al., 2017). The time of maximal PPC was found for each neuron within the 200-ms time period. The Wilcoxon signed-Rank test was used to determine which region was leading during correct and error trials (Bolkan et al., 2017).

Slice preparation

Mice (4-5 weeks old) were anesthetized with isoflurane and were subjected to cardiac perfusion with ice- cold oxygenated (95% O₂ and 5% CO₂) artificial cerebrospinal fluid (ACSF) before decapitation. Brains were removed rapidly and placed into ice-cold oxygenated ACSF containing (in mM): 110 choline Cl, 3.5 KCl, 0.5 CaCl₂, 7 MgCl₂, 1.3 NaH₂PO₄, 25 NaHCO₃, and 20 glucose. The hippocampal formation was dissected, transverse slices (250-300 μM) were cut on a vibratome (VT1200, Leica) and allowed to recover at 32°C for 1 h before recording. The ACSF used for recovery and recording contained (in mM): 125 NaCl, 3.5 KCl, 2 CaCl₃, 1.3 MgCl₂, 1.3 NaH₃PO₄ 25 NaHCO₃, and 10 glucose. Individual slices were transferred to a submerged recording chamber and continuously perfused with the ACSF (3.0 mL/min) at 32°C. Slices were visualized under a microscope (IX51WI, Olympus) using infrared video microscopy and differential interference contrast optics.

Electrophysiology recording

Intemeurons and pyramidal neurons in the hippocampus were identified based on location, shape, and firing properties. Patch electrodes were made from borosilicate glass capillaries (B-120-69-15, Sutter Instruments) with a resistance in the range of 2.5-4 MΩ. The internal solution for IPSC recording contained (in mM): 100 CsCH3SO3, 40 CsCl, 10 HEPES, 4 MgCl2, 4 Na2ATP, 0.4 Na3GTP, 10 Tris-phosphocreatine and 0.2 EGTA. For mIPSC recording, TTX (1 μM), CNQX (20 μM) and DL-AP5 (50 μM) were added to the recording ACSF to block action potential and glutamatergic transmission, respectively. For eIPSC recording, TTX was excluded. The internal solution for mEPSC recording contained (in mM): 125 K-gluconate, 15 KCl, 10 HEPES, 4 MgCl2, 4 Na2ATP, 0.4 Na3GTP, 10 Tris-phosphocreatine and 0.2 EGTA. TTX (1 μM) and bicuculline (10 μM) were added to the recording ACSF to block action potential and GABAergic transmission, respectively. Recordings were made with an Axon 700A patch-clamp amplifier and 1320A interface (Axon Instruments). The signals were filtered at 2 kHz using amplifier circuitry, sampled at 10 kHz and analyzed using Clampex 9.0 (Axon Instruments).

Neuronal cell culture

Hippocampal neurons were prepared from E16-17 mouse embryos as previously described (Woo et al., 2007). Neurons were cultured in Neurobasal medium (Invitrogen) supplemented with 1 × B27 (Invitrogen), penicillin/streptomycin (100 U per ml and 100 μg per ml, respectively) and 2 mM glutamine. For western blot, neurons were seeded at a density of 2 × 10⁶ neurons per well of a six-well plate coated with poly-L-lysine (Sigma).

Immunohistochemistry and confocal imaging

Adult (2 month) T796G mice were anesthetized with isoflurane and perfused transcardially with 0.9% NaCl followed by 4% PFA dissolved in 0.1M phosphate-buffered saline (PB, pH 7.4). Brains were removed and post fixed in 4% PFA for 4-6 h and immersed in 30% sucrose containing 0.1 M PB overnight at 4°C. Coronal sections (30 μm) were cut on a cryostat microtome (CM1900, Leica). Sections were treated with 0.2% Triton X-100 for 30 min and blocked in 10% donkey serum for 1 h. Slices were stained with anti-NeuN antibody (1:1000, monoclonal, Neuromics) and anti-PV antibody (1:1000, polyclonal, Swant) for 24 h at 4°C. Sections were imaged under a confocal microscope (Nikon) using a 20×/0.7 objective lens (Nikon). Images from each channel were collected in multitrack modes to avoid crosstalk between channels.

LTP induction and recording

For LTP induction, field EPSPs (fEPSPs) were recorded from the dendrites of CA1 pyramidal cells using a glass electrode (3–5 MΩ) filled with extracellular solution. Constant current pulses (20–100 μA, 100 μs, at 0.05 Hz) were generated by Master-8 (AMPI) and applied through concentric electrodes (World Precision Instruments) placed at the stratum radiatum to induce fEPSPs. The stimulus intensity was set to generate fEPSP with an amplitude that was 30–40% of the maximum response. After 10 min steady baseline, LTP was induced by tetanic stimulation which consisted 100 pulses at 100 Hz and repeated twice at 20-s intervals.

Chemicals and antibodies

The recombinant NRG1 was prepared as described previously (Huang et al., 2000; Woo et al., 2007). 1NM-PP1 was purchased from EMD Milipore (product number 529581). PD168393 was purchased from Sigma-Aldrich (produce number PZ0285). The rabbit anti-phosphorylated-ErbB4 (Tyrosine 1248) antibody was from Cell Signaling (product number 4757S). Dr. Cary Lai kindly supplied the rabbit anti- pan-ErbB4 polyclonal antibody (0618). The rabbit anti-PV antibody was from Swant (product number PV25). The mouse anti-NeuN antibody was from Neuromics (product number MO22122).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical Analysis

Two-tailed paired or un-paired Student’s t-test was used for analyzing data sets with two groups as indicated in figure legends. One-way or two way ANOVA was used for data sets with repeated measures. Bonferroni correction was used for multiple comparisons. Wilcoxon signed-Rank test was used for directionality analysis. P<0.05 is considered significant. Data were represented with means ± SEM.

Supplementary Material

1

Figure S1. 5-CSRTT introduction and electrode-position verification. Related to Figure 1. (A) Tables show the division of training stage and the parameters, progression criteria of each stage. (B) Schematic diagram of the possible trial sequence of 5-CSRTT. (C) Schematic diagram of the time course of each possible trials. (D) Left: Electrode placements in control mice in the prefrontal cortex. Right: Location of electrode tips in the PFC. DIC image shows lesions made by electronic heat (arrow). Scale bar: 1 mm. (E) Left: Electrode placements in control mice in the hippocampus. Right: Location of electrode tips in the vHPC. DIC image shows lesions made by electronic heat (arrow). Scale bar: 1 mm. (F) Representative distributions of preferred phases of a prefrontal unit in relation to hippocampal gamma oscillations in correct (top) and error trials (bottom). (G) No change in PPC values at gamma band in error trials (n = 93 units (7 mice), paired t-test, ns, not significant (p=0.8298)). Pooled data are shown as mean ± SEM.

Figure S2. Reduced hippocampal-prefrontal synchrony and attention deficit in ErbB4−/− mice. Related to Figure 2. (A) Days taken for ErbB4−/− and control mice to reach criterion in each training stage of 5-CSRTT (control, n=13 mice; ErbB4−/−, n=13 mice; two way ANOVA, ***p<0.001; stage 1: ***p<0.001; stage 2: ns, not significant, p=0.0666; stage 3: **p=0.0079; stage 4: **p=0.0081; stage 5: ***p<0.001; stage 6: ***p<0.001). (B) Latencies to correct, incorrect and reward of 5-CSRTT in control (purple) and ErbB4−/− (green) mice (control, n = 13 mice; ErbB4−/−, n = 13 mice; t-test, latency to correct: ***p<0.001; latency to incorrect: ns, p=0.7889; latency to reward: p=0.4816). (C) Left: Electrode placements in control mice (purple circle) and ErbB4−/− mice (green circle) in the hippocampus. Right: Location of electrode tips in the vHPC. DIC image shows lesions made by electronic heat (arrow). Scale bar: 1 mm. (D) Left: Electrode placements in control mice (purple circle) and ErbB4−/− mice (green circle) in the prefrontal cortex. Right: Location of electrode tips in the PFC. DIC image shows lesions made by electronic heat (arrow). Scale bar: 1 mm. (E, F) Comparison of local field potential power between control and ErbB4−/− mice in vHPC (E, for both control and ErbB4−/− n=12 tests (6 mice); two way ANOVA, *p=0.0352; post hoc, Bonferroni, delta: ns, p=0.9996; theta: ns, p=0.9516; beta: ns, p=0.9541; gamma: *p=0.0163. F, t-test, delta: ns, p=0.6361; theta: ns, p=0.7804; beta: ns, p=0.6329; gamma: *p=0.0107). (G, H) Comparison of local field potential power between control and ErbB4−/− mice in PFC (G, for both control and ErbB4−/− n=12 tests (6 mice); two way ANOVA, *p=0.0157; post hoc, Bonferroni, delta: ns, p=0.9991; theta: ns, p=0.9971; beta: ns, p=0.9461; gamma: **p=0.0075. H, t-test, delta: ns, p=0.9682; theta: ns, p=0.833; beta: ns, p=0.4509; gamma: **p=0.0026). (I) Distribution of spike waveform properties (half width of valley and trough-to-peak duration) in comparison of firing rate. Note large overlap between units in control and ErbB4−/− mice. Inset: Represent trace of a spike showing how half width and trough-to-peak duration were calculated. Pooled data are shown as mean ± SEM.

Figure S3. 1NM-PP1 block phosphorylation of ErbB4 in vitro. Related to Figure 4. (A) Diagram of the chemical-sensitive mutant protein kinases and structures of PP1 and its analog 1NM-PP1. (B) Alignment of the amino acid sequence of various tyrosine kinase. Note that threonine (T) is a conserved residue. (C, D) Dose dependent inhibition of 1NM-PP1 on NRG1 induced T796G ErbB4 phosphorylation (repeated 3 times, 0 μM: *p=0.0261; 1 μM: ns, not significant (p=0.8933); 3 μM: *p=0.0184; 10 μM: ***p<0.001). T796G-ErbB4-GFP was transfected into HEK 293 cells then NRG1 induced phosphorylation was examined with Western blot. (E, F) No effect of 1NM-PP1 on WT-ErbB4 (repeated 3 times, 0 μM: **p=0.0045; 1 μM: **p=0.00144; 3 μM: **p=0.0019; 10 μM: **p=0.0016). WT-ErbB4- GFP was transfected into HEK 293 cells then NRG1 induced phosphorylation was examined with Western blot. Pooled data are shown as mean ± SEM.

Figure S4. T796G mice developed normally. Related to Figure 4. (A) Adult T796G and control littermate have the same body size. Scale bar: 1 cm. (B) Adult T796G and control littermate have the same brain size. Scale bar: 0.5 cm. (C) Representative prefrontal cortex confocal images of immunostaining for NeuN and PV in control and T796G mice. Scale bar: 200 μm. (D) Representative hippocampal confocal images of immunostaining for NeuN and PV in control and T796G mice. Scale bar: 500 μm. (E, F) statistical analysis of prefrontal and hippocampal NeuN+ (E, 6 images (2 mice), t-test, ns, not significant (p=0.6699)) and PV+ (F, 6 images (2 mice), t-test, ns, not significant (p=0.6276)) cell numbers in control and T796G mice. Pooled data are shown as mean ± SEM.

Figure S5. Acute inhibition of ErbB4 impairs attention and the hippocampal-prefrontal synchrony in T796G but not control mice. Related to Figure 4. (A) Days taken for control and T796G mice to reach criterion in each training stage of 5-CSRTT (control, n=11 mice; T796G, n=11 mice; two way ANOVA, ns, not significant (P=0.6143); stage 1: ns, not significant (P=0.4564); stage 2: ns, not significant (P=0.1395); stage 3: ns, not significant (P=0.2978); stage 4: ns, not significant (P=0.6309); stage 5: ns, not significant (P=0.6657); stage 6: ns, not significant (P=0.7793)). (B) Latency to correct (Cor.), incorrect (Incor.) and reward (Rew.) of 5-CSRTT in control and T796G mice (i.p.; control: n=11 mice, two-way ANOVA, latency to correct: ns, p=0.8404; latency to incorrect: ns, p= 0.96; latency to reward: ns, p=0.6913; T796G: n=11 mice, two-way ANOVA, latency to correct: ***p<0.001; latency to incorrect: ns, p= 0.8704; latency to reward: ns, p=0.8014). (C) Distribution of PFC spike waveform properties (half width and trough-to-peak duration) in comparison of firing rate. Note the large overlap between vehicle and 1NM-PP1 injection in T796G mice. (D) Distribution of PFC spike waveform properties (half width and trough-to-peak duration) in comparison of firing rate. Note the large overlap between vehicle and 1NM-PP1 injection in control mice. Pooled data are shown as mean ± SEM.

Figure S6. Local inhibition of ErbB4 signaling causes attention deficit in T796G mice. Related to Figure 6. (A) Left: Cannula placements in the hippocampus. Right: DIC plus fluorescence images show the diffusion of Nile red within experiment time scale in the hippocampus. Scale bar: 1 mm. (B) Left: Cannula placements in the PFC. Right: DIC plus fluorescence images show the diffusion of Nile red within experiment time scale in the PFC. Scale bar: 1 mm. (C, D) Effects of 1NM-PP1 local injection on power of local field potential in vHPC (C, for both vehicle and 1NM-PP1 n=12 tests (6 mice); two way ANOVA, *p=0.0226; post hoc, Bonferroni, delta: ns, not significant (P=0.4535); theta: ns, not significant (P=0.3254); beta: ns, not significant (p=0.999); gamma: **p=0.0062. D, t-test, delta: ns, not significant (P=0.8846); theta: ns, not significant (P=0.96); beta: ns, not significant (P=0.9839); gamma: *p=0.0339). (E) Distribution of PFC spike waveform properties (half width and trough-to-peak duration) in comparison of firing rate. Note the large overlap between vehicle and 1NM-PP1 in local vHPC injection. (F, G) Effects of 1NM-PP1 local injection on power of local field potential in PFC (F, for both vehicle and 1NM-PP1 n=12 tests (6 mice); two way ANOVA, ***p<0.001; post hoc, Bonferroni, delta: ns, not significant (P=0.5231); theta: ns, not significant (P=0.2743); beta: ns, not significant (p=0.999); gamma: ***p<0.001. G, t-test, delta: ns, not significant (P=0.8474); theta: ns, not significant (P=0.9462); beta: ns, not significant (P=0.9415); gamma: *p=0.0355). (H) Distribution of PFC spike waveform properties (half width and trough-to-peak duration) in comparison of firing rate. Note the large overlap between vehicle and 1NM-PP1 in local PFC injection. (I, J) Latencies to correct, incorrect and reward after bilaterally local injection of 1NM-PP1 in vHPC (I, n=11 mice, two-way ANOVA, latency to correct: ***p<0.001; latency to incorrect: ns, p= 0.8141/0.3859/0.8444, respectively; latency to reward: ns, p=0.9332/0.9453/0.9041, respectively) PFC (J, n=11 mice, two-way ANOVA, latency to correct: ns, p=0.3051/0.326/0.389, respectively; latency to incorrect: ns, p= 0.9491/0.9381/0.9453; latency to reward: ns, p=0.7273/0.6509/0.9478, respectively). Pooled data are shown as mean ± SEM.

Figure S7. Inhibition of 1NM-PP1 on NRG1 induced phosphorylation of ErbB4 in neurons cultured from T796G mice. Related to Figure 7. (A, B) Dose dependent inhibition of 1NM-PP1 on NRG1 induced T796G-ErbB4 phosphorylation (repeated 3 times, t-test, 0 μM: **p=0.0015; 0.1 μM: **p=0.0038; 1 μM: *p=0.0324). Cortical neurons from T796G mice were cultured and treated with NRG1 then phosphorylation was examined with Western blot. (C, D) No effect of 1NM-PP1 on NRG1 induced phosphorylation of wild type ErbB4 (repeated 3 times, t-test, 0 μM: **p=0.0059; 0.1 μM: ***p<0.001; 1 μM: ***p<0.001). Cortical neurons from wild type control littermate were cultured and treated with NRG1 then phosphorylation was examined with Western blot. Pooled data are shown as mean ± SEM.

Figure S8 1NM-PP1 specifically target to ErbB4 in T796G mice. Related to Figure 7. (A, B) 1NM-PP1 has no effect on mEPSC frequency (A, for control and 1NM-PP1 each n=6 neurons (3 mice), t-test, ns, not significant (p=0.6897)) and amplitude (B, for control and 1NM-PP1 each n=6 neurons (3 mice), t- test, ns, not significant (p=0.1909)). (C, D) 1NM-PP1 has no further inhibition effects on mIPSC after pre-treatment with non-specific ErbB4 kinase inhibitor PD168393 (C, frequency, n=5 neurons (3 mice), t- test, 10 min: p=0.6727; 15 min: p=0.1611; 20 min: p=0.2596; 25 min: *p=0.0326; 30 min: **p=0.0058; 35 min: *p=0.0158; 40 min: **p=0.0014; 45 min: ***p<0.001; 50 min: **p=0.0075; 55 min: **p=0.0048; 60 min: ***p<0.001; 65 min: ***p<0.001; 70 min: ***p<0.001. D, amplitude, n=5 neurons (3 mice), t- test, ns, not significant, 10 min: p=0.7144; 15 min: p=0.3068; 20 min: p=0.3949; 25 min: p=0.2449; 30 min: p=0.3734; 35 min: p=0.2694; 40 min: p=0.1691; 45 min: p=0.3846; 50 min: p=0.4813; 55 min: p=0.5414; 60 min: p=0.5017; 65 min: p=0.6068; 70 min: p=0.5078). (E, F) 1NM-PP1 blocks NRG1 induced increase of mIPSC frequency (E, each conditions n=5 neurons (3 mice), t-test, NRG1: **p=0.0033; NRG1 plus 1NM-PP1: ns, not significant (p=0.6143)) but not amplitude (F, each conditions n=5 neurons (3 mice), t-test, NRG1: ns, not significant (p=0.8887); NRG1 plus 1NM-PP1: ns, not significant (p=0.1703)) in T796G mice. (G, H) 1NM-PP1 increases pyramidal neuron excitability in T796G mice (n=6 neurons, t-test, *p=0.04336/0.0507/0.0465/0.0477, respectively). (I, J) 1NM-PP1 promote LTP induction in T796G mice (I, n=6 slices (3 mice)) but not control littermate (J, n=7 slices in the presence of 1NM-PP1; n=8 slices in control conditions (3 mice)). (K) 1NM-PP1 has no further effect on LTP induction in the presence of GABA_A receptor antagonist bicuculline (n=8 slices (4 mice)). (L) Summary of averaged LTP amplitude measured at last 5 min (t-test, T796G: *p=0.0203; control: ns, p=0.7352; T796G + bicuculline: ns, p=0.8218). Pooled data are shown as mean ± SEM.

Figure S9. Distinct regulations by vHPC and PFC ErbB4+ neurons on the hippocampal-prefrontal synchrony and attention. Related to Figure 8. (A-C) Representative traces showing responses of ErbB4+ neurons (A) and ErbB4− neurons (B, C) to yellow light stimulation in slice. (D) Similar accuracy rates between ErbB4-ArchT and ArchT^flox/flox mice during 5-CSRTT training (ArchT^flox/flox, n=12 mice; ErbB4-ArchT, n=12 mice; two way ANOVA, ns, not significant (P=0.9779). (E) Left: Optoelectrode placements in the prefrontal cortex. Right: Location of optoelectrode tips in the prefrontal cortex. DIC image showed lesions made by optical fibers (arrow). Scale bar: 1 mm. (F) Left: Optoelectrode placements in the hippocampus. Right: Location of optoelectrode tips in the hippocampus. DIC image showed lesions made by optical fibers (arrow). Scale bar: 1 mm. (G) Summary of in-vivo optogenetic data showing the proportion of neurons that were inhibited, excited, or having no change in spike firing during yellow light stimulation. (H, I) Effects of LED light illumination on power of local field potential in PFC (H, for both control and light illumination n=12 tests (6 mice); two way ANOVA, **p=0.0079; post hoc, Bonferroni, delta: ns, not significant (P=0.999); theta: ns, not significant (P=0.999); beta: ns, not significant (p=0.999); gamma: **p=0.0057. I, t-test, delta: ns, not significant (P=0.5146); theta: ns, not significant (P=0.8371); beta: ns, not significant (P=0.5283); gamma: ***p<0.001). (J) Bilaterally yellow light illumination in PFC of ErbB4-ArchT mice have no effect on latencies to correct, incorrect responses or reward (n=12 mice, two-way ANOVA, latency to incorrect: ns, p=0.9604; latency to reward: ns, p=0.421). (K, L) Effects of LED light illumination on power of local field potential in vHPC (K, for both control and light illumination n=12 tests (6 mice); two way ANOVA, **p=0.0032; post hoc, Bonferroni, delta: ns, not significant (P=0.999); theta: ns, not significant (P=0.999); beta: ns, not significant (p=0.999); gamma: **p=0.0015. L, t-test, delta: ns, not significant (P=0.7206); theta: ns, not significant (P=0.8962); beta: ns, not significant (P=0.8348); gamma: *p=0.0275). (M) Bilaterally yellow light illumination in vHPC in ErbB4-ArchT mice have no effect on latencies to incorrect or reward (n=12 mice, two-way ANOVA, latency to incorrect: ns, p=0.2936; latency to reward: ns, p=0.3176), but increases latency to correct (two-way ANOVA, ***p<0.001). (N, O) Bilaterally yellow light stimulation in both PFC and vHPC in ArchT^flox/ilox mice has no effect on error ratios (N, n=12 mice, two-way ANOVA, incorrect (Incor.): ns, p=0.7893; omission (Omis.): ns, p=0.9743; premature (Prem.): ns, p=0.9434), and latencies to correct, incorrect and reward (O, n=12 mice, two-way ANOVA, latency to correct: ns, p=0.9439; latency to incorrect: ns, p=0.7291; latency to reward: ns, p=0.1337). Pooled data are shown as mean ± SEM.

Video S1. Correct trials of 5-CSRTT. Related to Figure 1. Representative video showing correct trials performed by a control mouse with headstage mounted.

Video S2. Incorrect trial of 5-CSRTT. Related to Figure 1. Representative video showing incorrect trials performed by a control mouse with headstage mounted.

Video S3. Omission trial of 5-CSRTT. Related to Figure 1. Representative video showing omission trials performed by a control mouse with headstage mounted.

Video S4. Premature trial of 5-CSRTT. Related to Figure 1. Representative video showing premature trials performed by a control mouse with headstage mounted.

Highlights.

• Hippocampal-prefrontal synchrony positively correlates with top-down attention level

• Acute ErbB4 signaling regulates top-down attention

• ErbB4-dependent GABA activity in hippocampus synchronizes PFC neurons in attention