Neurotrophin receptor activation rescues cognitive and synaptic abnormalities caused by mutation of the psychiatric risk gene Cacna1c

Genetic variation in CACNA1C, which encodes the alpha-1 subunit of CaV1.2 L-type voltage-gated calcium channels, is strongly linked to risk for psychiatric disorders including schizophrenia and bipolar disorder. To translate genetics to neurobiological mechanisms and rational therapeutic targets, we investigated the impact of altered Cacna1c dosage on rat cognitive, synaptic and circuit phenotypes implicated by patient studies. We show that rats hemizygous for Cacna1c harbor marked impairments in learning to disregard non-salient stimuli, a behavioral change previously associated with psychosis. This behavioral deficit is accompanied by dys-coordinated network oscillations during learning, pathway-selective disruption of hippocampal synaptic plasticity, attenuated Ca2+ signaling in dendritic spines and decreased signaling through the Extracellular-signal Regulated Kinase (ERK) pathway. Activation of the ERK pathway by a small molecule agonist of TrkB/TrkC neurotrophin receptors rescued both behavioral and synaptic plasticity deficits in Cacna1c+/- rats. These results map a route through which genetic variation in CACNA1C can disrupt experience-dependent synaptic signaling and circuit activity, culminating in cognitive alterations associated with psychiatric disorders. Our findings highlight targeted activation of neurotrophin signaling pathways with BDNF mimetic drugs as a novel, genetically informed therapeutic approach for rescuing behavioral abnormalities in psychiatric disorder. One Sentence Summary Neurotrophin receptor activation reveals that BDNF mimetic drugs have therapeutic potential to ameliorate genetic risk for psychiatric disorders.


Introduction
The major psychiatric disorders such as schizophrenia and bipolar disorder place an enormous burden on society. These conditions have not however seen the advances in mechanistic understanding and therapy realized in other areas of medicine. There is now hope that recent advances in the understanding of the genomic basis of these conditions may pave the way to the development of new therapeutics (1). Particularly promising in this respect is the demonstration of a strong association between genetic variation in voltage gated calcium channels with schizophrenia and bipolar disorder (2). Genome-wide association studies (GWAS) have consistently identified single nucleotide polymorphisms (SNPs) in CACNA1C, which encodes the pore-forming α1C subunit of CaV1.2 L-type VGCCs (L-VGCCs), as having significant association with both conditions (3,4). While schizophrenia and bipolar disorder can present differently in the clinic, both are associated with psychosis, and genomic studies have indicated a significant shared genetic architecture between the two disorders (5).
The exact molecular effects of the CACNA1C SNPs implicated by GWAS are not yet fully understood. However, risk-associated common variants in CACNA1C are intronic and are likely to act by altering gene expression. Previous studies have indicated that these risk variants may act to decrease expression of CACNA1C (6,7), including in the human hippocampus (8). The association of rare deleterious mutations in genes encoding VGCC subunits, including CACNA1C, in people with schizophrenia and other neurodevelopmental disorders, further supports the view that decreased expression of CACNA1C can contribute to disease risk (9,10).
Therefore, understanding the effects of reduced CACNA1C dosage is important in discerning how genetic variation in L-VGCCs might contribute to risk for neuropsychiatric illness. 4 CaV1.2 L-VGCCs are highly expressed the mammalian brain, including in the hippocampus (11).
Associative learning in the hippocampus is orchestrated by rhythmic neural activity in the entorhinal-hippocampal network and is underpinned by associative synaptic plasticity at hippocampal glutamatergic synapses (22)(23)(24). Induction of plasticity at these synapses relies on the coordinated activation of both postsynaptic NMDA receptors (NMDAR) and voltage-gated calcium channels including L-VGCCs (25)(26)(27).
Neuronal L-VGCCs play a key role in linking membrane depolarization to the activation of signaling pathways in particular the Extracellular-signal Regulated Kinase (ERK) signaling pathway. The ERK pathway critically regulates both early (28,29) and late stages of synaptic plasticity (30,31). L-type VGCC-dependent ERK signaling activates transcription factors including CREB that control the expression of genes required for long-lasting plasticity, including the neurotrophin BDNF (Brain Derived Neurotropic Factor) (30,31). Thus, calcium entry through L-VGCCs plays a critical role in regulating the changes in synaptic efficacy and gene expression underlying associative learning.
In the present study, we used a translationally relevant rat model, Cacna1c +/to examine the impact of reduced dosage of the psychiatric risk gene Cacna1c on hippocampus-dependent associative learning and went on to reveal the impact of reduced Cacna1c dosage on underlying synaptic plasticity, network synchronization and ERK pathway signaling in the hippocampus.
Having mapped the links between genetic variation in CACNA1C and disruptions to experience-5 dependent synaptic signaling, circuit activity and behavior we then showed that activation of the ERK signaling pathway through a small molecule agonist of the TrkB/TrkC neurotrophin receptors could rescue impairments in both behavior and synaptic plasticity. Our findings suggest that targeted activation of neurotrophin signaling pathways with BDNF mimetic drugs may be of use in treating cognitive and brain abnormalities in psychiatric disorder.

Behavioral effects of Cacna1c hemizygosity on latent inhibition
We investigated the effects of reduced CaV1.2 dosage on associative learning using rats hemizygous for a truncating mutation in exon 6 of Cacna1c gene encoding the pore-forming α1C subunit of CaV1.2 L-VGCCs (32) (Methods). Cacna1c +/rats have an approximately 50% decrease in both Cacna1c mRNA and protein in the hippocampus, mimicking the anticipated impact of deleterious mutations in CACNA1C in humans (33). We assessed the behavior of Cacna1c +/rats and wild type littermates using tests of contextual fear conditioning (CFC), as the formation of contextual fear associations is known to depend on the hippocampus (34) and to require L-VGCCs (35). To determine the ability of Cacna1c +/rats to establish and retrieve contextual fear associations we used a paradigm where animals received a single mild footshock (US, 2s, 0.5 mA) 2 min after being placed in a novel context (Fig. 1A and Supplementary Methods). Memory for the context-fear association was assessed by measuring the freezing response upon return to the conditioned context 3h (short-term memory, STM), 24h (long-term memory, LTM1), and 8 days (LTM2) after CFC training. Cacna1c +/+ and Cacna1c +/animals had equivalent levels of fear response at all recall sessions, as shown in Fig. 1A and replicated in a separate cohort (Fig. S1), indicating that reduced dosage of Cacna1c has no impact on contextual fear learning per se. 6 We next assessed the performance of Cacna1c +/rats and wild-type littermates in a paradigm of latent inhibition (LI) of contextual fear conditioning (schematics in Fig. 1B and Fig. S1). LI is the reduced ability to form conditioned associations to a stimulus (in this case, the conditioning context) as a result of pre-exposure to the stimulus alone and reflects the normal process of learning to ignore irrelevant stimuli. Notably latent inhibition has previously been found to be impaired in individuals experiencing psychosis (15,16). For LI training, animals were preexposed to the to-be-conditioned context (PE) for 4h, then subjected to CFC training 24h or 48h later (Figs. 1B, S1). In Cacna1c +/+ animals, the preexposure resulted in a robust LI of CFC manifested as reduced freezing response at long-term memory sessions (Figs. 1B, S1). In contrast, Cacna1c +/animals had a marked deficit in LI of CFC, an effect seen in two separate cohorts (Fig. 1B, S1). These results show that Cacna1c hemizygosity selectively impairs the LI of contextual fear associations. To determine whether this effect was mediated through an impact on L-VGCCs specifically in the hippocampus we infused the L-VGCC antagonist diltiazem (DTZ) into the dorsal hippocampus during context pre-exposure and observed a similar impairment in the establishment of latent inhibition (Fig. S2). Taken together, these results indicate a specific role of hippocampal CaV1.2 channels during the preexposure stage in the LI of CFC paradigm.
LI of CFC requires the ability of dorsal hippocampus to form and store context-specific memories (36). We hypothesized that the deficit in LI of CFC observed in Cacna1c +/animals reflects a disruption of dorsal hippocampal processes that form representations of a novel environment during preexposure. Therefore we investigated the impact of Cacna1c hemizygosity on two fundamental neural mechanisms proposed to support memory encoding in the 7 hippocampus: associative plasticity at CA1 pyramidal cell synapses (24) and phase-amplitude coupling between the theta and gamma oscillations of the local field potential in CA1 (22).

Cacna1c +/rats show altered plasticity in the dorsal hippocampal CA1
Formation of stable context-specific representations in the hippocampus requires strengthening of cortical excitatory inputs to CA1 pyramidal neurons. These inputs arrive indirectly through the CA3 subfield via the Schaffer collateral (SC-CA1) pathway, and directly through the temporoammonic (TA-CA1) pathway (37-40). We examined the induction of LTP at SC-CA1 and TA-CA1 synapses onto CA1 pyramidal neurons in the dorsal hippocampus using ex vivo slices from Cacna1c +/and Cacna1c +/+ rats. To induce LTP we used a theta-burst pairing protocol consisting of synaptic stimulation coincident with post-synaptic action potentials (TBP, Fig. 2A). TBP mimics neuronal activity patterns observed during learning in vivo (41) and relies on the coordinated activation of postsynaptic NMDAR and VGCCs (27,42). TBP failed to induce LTP at SC-CA1 synapses in slices from Cacna1c +/animals (Fig. 2B,D) whereas it produced robust SC-CA1 LTP in Cacna1c +/+ slices (Fig. 2C,D). In Cacna1c +/+ slices the TBP-induced LTP at SC-CA1 synapses was also blocked by L-VGCC antagonists isradipine, DTZ, and the NMDAR glycine site antagonist L-689560 H), confirming that TBP-induced LTP requires the activation of both NMDARs and L-VGCCs.
The deficit of TBP-induced LTP at SC-CA1 synapses in Cacna1c +/animals could be selective or could reflect a broad impairment of plasticity mechanisms. To disambiguate these possibilities, we tested the induction of SC-CA1 LTP by an alternative protocol of low frequency pre-synaptic stimulation paired with tonic post-synaptic depolarization (LFS-pairing, Fig. 2E). LFS-pairing induces NMDAR-dependent synaptic potentiation with no contribution from post-synaptic action 8 potentials or L-VGCCs (43). We observed a robust SC-CA1 LTP induced with LFS-pairing in both genotypes . In Cacna1c +/+ slices LFS-pairing-induced LTP was insensitive to either isradipine or DTZ but was blocked by L-689560 H) confirming that NMDAR activation but not L-VGCC activation is required for this form of LTP. In contrast to SC-CA1 synapses, TBP induced a robust LTP at TA-CA1 synapses in both genotypes  which was sensitive to L-VGCC block in wild-type slices (Fig. S3G,H).
Overall, these results show that Cacna1c hemizygosity impairs forms of LTP that depend on the activation of post-synaptic L-VGCCs during somatic action potential bursts without affecting NMDAR-dependent mechanisms of synaptic plasticity, at the SC-CA1 pathway.
CA1 neurons in Cacna1c +/rats have impaired spine Ca 2+ signaling associated with postsynaptic spike bursts LTP requires a neuronal associative signal which, in hippocampal neurons, can be provided by postsynaptic action potentials (AP) backpropagated at SC-CA1 synapses and by local dendritic spikes at the more distal TA-CA1 synapses (26,44). Therefore, we investigated the excitability of CA1 pyramidal neurons in hippocampal slices from Cacna1c +/and Cacna1c +/+ animals. The passive membrane properties, AP threshold and rheobase current were comparable in CA1 pyramidal neurons from both genotypes (Supplementary Table S1, Fig. S4A-E). However, Cacna1c +/neurons showed a significant reduction in the broadening of somatic AP waveforms during high frequency (>40Hz) AP firing when compared to wild-type cells . In Cacna1c +/+ neurons the L-VGCC blocker isradipine inhibited somatic AP broadening to levels comparable to those in Cacna1c +/neurons in the absence of the drug (Fig. S4J,K). Somatic AP broadening occurs normally during high-frequency AP bursts and reflects the slowing of AP 9 repolarization mediated by voltage-and Ca 2+ -sensitive K + channels (45,46). The impaired broadening of somatic APs in Cacan1c +/pyramidal neurons indicates that low CaV1.2 dosage alters the Ca 2+ -sensitive spike repolarization during burst firing.
Somatic AP broadening in CA1 pyramidal neurons has been proposed to facilitate the dendritic backpropagation of somatic spikes (47) necessary for the induction of associative synaptic plasticity such as 42,44) and to increase the gain of intracellular Ca 2+ signals associated with AP bursts (45) such as those occurring during the exploration of a novel environment (40,48). The reduced AP broadening during burst firing in Cacna1c +/neurons may impact on postsynaptic Ca 2+ signals triggered by APs backpropagated in dendritic spines during plasticity induction. We tested this hypothesis by comparing Ca 2+ transients elicited with highfrequency somatic AP bursts in spines located on radial oblique dendrites, in Cacna1c +/vs Cacna1c +/+ CA1 pyramidal neurons (Fig. 3A,B, S5, Supplementary Tables S2, S3).
Backpropagated APs activate spine VGCCs including the L-type (27,49). Spine Ca 2+ transients elicited by back-propagated APs (APCaTs) attenuated with increasing distance from the soma S6). However, in Cacna1c +/neurons the APCaTs in spines at 150-250 µm from the soma were consistently smaller compared to those in wild-type cells (Fig. 3D,E) and summated poorly with the number of APs per burst (Fig. 3F,G,S7).
These results show that the low CaV1.2 dosage in Cacna1c +/-CA1 pyramidal neurons alters the burst-associated spine Ca 2+ signaling at SC-CA1 synapses, which may underlie the selective deficit in synaptic plasticity induction during burst firing and impaired TBP-LTP at SC-CA1 synapses. 10 Cacna1c +/rats have reduced phase-amplitude coupling between theta and gamma oscillations of dorsal CA1 local field potential.
During the encoding of novel environments, the induction of synaptic plasticity leading to the establishment of neuronal ensembles depends on the critical timing of excitatory inputs, coordinated by the synchronization of the rhythmic activity in the CA1 subfield (22,50,51). To determine the impact of Cacna1c heterozygosity on the synchronization of neural activity in the CA1 network, we monitored the modulation of the local field potential (LFP) gamma oscillations by the phase of LFP theta oscillations (theta-gamma Phase-Amplitude Coupling (PAC)) in the dorsal CA1. In the hippocampus, theta-gamma PAC has been proposed as a mechanism for the storage and recall of object and event representations (23,52,53). In the dorsal CA1, the slow (~25-40 Hz) and fast (~ 65-140 Hz) gamma sub-bands are thought to reflect the synchronization of CA1 pyramidal neurons with the activity of CA3 neurons via SC-CA1 pathway, and that of the medial entorhinal cortex (MEC) neurons, via TA-CA1 pathway, respectively (54).
To determine the changes in theta-gamma PAC associated with the exploration of a novel environment, we recorded LFP oscillations in dorsal CA1 in rats running along a track in a familiar, then a novel environment (Figs. 4A and S8A-C). In the familiar environment, animals from both genotypes had similar phase-amplitude coupling across the gamma frequency spectrum (Fig. 4B,C, "Familiar" sub-panels) and power spectra of LFP oscillations in the 1-40 Hz range (Fig. S8D,E,. Upon switching to the novel environment, the Cacna1c +/+ rats had theta-gamma PAC levels similar to the initial response to the subsequently familiar environment (Fig. 4B,C,. However, the Cacna1c +/animals showed a significantly impaired theta-gamma PAC compared to wild type animals (Fig. 4B,C). 11 Our results indicate that Cacna1c hemizygosity results in altered network properties in the hippocampus including a deficit of coupling between CA3 and CA1 subfields. This effect may disrupt hippocampal network physiology during the encoding of contextual information and contribute to the observed deficit in LI seen in Cacna1c +/animals.
Hippocampal levels of phosphorylated ERK and phosphorylated CREB are reduced in the Cacna1c +/rats.
We next investigated molecular changes associated with altered synaptic plasticity and network activity in the hippocampus of Cacna1c +/animals. Our observation of impaired spine Ca 2+ signaling at SC-CA1 synapses in Cacna1c +/rats suggests that low dosage of Cacna1c may impact calcium-dependent signaling pathways downstream of CaV1.2 VGCCs. In particular, the ERK signaling pathway is known to couple L-VGCC activation to changes in CREB activation and gene expression required to support long-lasting plasticity (30, 31). We therefore examined (basal) levels of ERK and CREB and the levels of phosphorylated ERK (pERK) and phosphorylated CREB (pCREB) in the dorsal hippocampus using immunohistochemistry. We found a significant reduction in pERK (Figs. 5A, S9A) in all hippocampal subfields and reduced pCREB in CA1 and CA3 (Figs. 5B, S9C) in the Cacna1c +/compared to Cacna1c +/+ animals with no differences in overall ERK and CREB protein levels between the two genotypes (Fig. S9B,D). This marked decrease of hippocampal pERK and pCREB levels in Cacna1c +/rats indicates ERK pathway activation and subsequent nuclear signaling are highly sensitive to CaV1.2 levels.

Rescue of ERK signaling and associative plasticity in Cacna1c +/animals with TrkB/TrkC
agonist  We hypothesized that pharmacological activation of the ERK pathway could reverse the molecular and physiological deficits observed in Cacna1c +/animals. We targeted the ERK pathway by using a recently discovered small molecule BDNF mimetic TrkB/TrkC neurotrophin receptor co-activator (LM22B-10) previously shown to activate the ERK pathway signaling in neurons (55). Systemic administration of LM22B-10 (25 mg/kg i.p.) restored the baseline levels of pERK and pCREB activation in the dorsal hippocampus in Cacna1c +/rats to the levels seen in Cacna1c +/+ animals 60 min after injection (Fig. 5). LM22B-10 did not however alter the number of pERK-and pCREB-positive cells in wild-type hippocampus.

TrkB/TrkC agonist LM22B-10 rescues the behavioral deficits in Cacna1c +/rats
Given that LM22B-10 reverses the molecular and plasticity deficits in Cacna1c +/rats, we investigated whether TrkB/TrkC activation in the hippocampus could rescue the behavioral deficit in LI of contextual fear memory observed in these animals. To explore this possibility, we 13 initially directly infused LM22B-10 into the dorsal hippocampus during the pre-exposure stage of LI training in Cacna1c +/rats. A single infusion of LM22B-10 into the dorsal hippocampus 60 min prior to the pre-exposure rescued the LI effect when animals were tested 24 h, 7 days and 21 days after CFC training (Fig. 6E). To assess whether LM22B-10 infusion had non-specific confounding effects on hippocampal function, all rats received a second CFC in a separate novel context (context B) in the absence of any infusion 22 or 23 days later. All rats expressed equivalent levels of freezing responses to context B and normal hippocampal-dependent CFC and fear memory in the novel context (Fig. 6E,F). In addition, when LM22B-10 infused rats returned to the original conditioned context at LTM4 they showed intact and context specific LI of CFC. Thus, intrahippocampal infusions of LM22B-10 rescued LI without producing any deleterious effects on hippocampal function.
We further investigated whether intra-peritoneal administration of LM22B-10 could rescue the behavioral deficits in Cacna1c +/rats. Our results show that systemic injections of LM22B-10 through the LI procedure can also rescue the LI effect in Cacna1c +/rats (Fig. S10). Together, these findings show that activation of the ERK signaling pathway in the dorsal hippocampus during the pre-exposure phase of LI is sufficient to rescue the behavioral deficits seen in Cacna1c +/rats. Furthermore, this effect can be achieved with both intrahippocampal administration of LM22B-10 and with systemic dosing.

Discussion
The past decade has seen major advances in psychiatric genomics. These provides grounds for optimism that greater mechanistic insight can be gained into disorders such as schizophrenia and bipolar disorder facilitating the development of new therapies. The consistent association of 14 schizophrenia, bipolar disorder and related neurodevelopmental disorders with genetic variation in VGCCs, and in particular CACNA1C, makes understanding the impact of genetic variation in the associated loci of high importance in terms of advancing mechanistic understanding of these conditions. Here we show that reduced dosage of Cacna1c, which is associated with neurodevelopmental and psychiatric disorders in humans, produces a distinctive behavioral, physiological, and molecular phenotype in rodents. We also show that the impact of low dosage of Cacna1c on behavior, plasticity and signaling pathways can be reversed by the intracerebral or peripheral administration of a small molecule drug targeting the Trk family of neurotrophin receptors, highlighting a novel target for therapy in these conditions. Cognitive abnormalities have been consistently associated with a range of psychiatric presentations. Psychotic symptoms, which are pathognomonic of schizophrenia and are frequently seen in bipolar disorder, have been associated with altered learning about the associations between environmental stimuli (16,17,19,21). Such altered associative learning can result in aberrant attribution of importance (or salience) to otherwise irrelevant events (56).
One measure of such altered learning, which has been shown to be affected in individuals with psychosis, is the latent inhibition procedure (15). Here we found that low dosage of Cacna1c was sufficient to produce a marked deficit in contextual latent inhibition consistent with the impairments seen in psychotic patients. This deficit was seen despite Cacna1c hemizygous animals showing normal contextual fear conditioning. These results are consistent with previous studies of mice with homozygous genetic deletions of Cacna1c (which encodes the pore-forming subunit of CaV1.2 L-VGCCs) or Cacna1d (which encodes the pore-forming subunit of CaV1.3 L-VGCCs) which have shown that CaV1.3, but not CaV1.2, is essential for the acquisition of fear associations (35,57). In contrast there is evidence from brain-specific knock-out studies that 15 CaV1.2, but not CaV1.3, is required for context discrimination (25,58). This suggests that the encoding of the contextual information required to mediate latent inhibition may be particularly susceptible to alterations in the dosage of Cacna1c.
The generation of novel context representations is dependent on the hippocampus and involves the recruitment of CA1 pyramidal neuronal ensembles, or engrams, driven by the conjunctive excitatory afferents from the entorhinal cortex and CA3 hippocampal area (50, 52), and supported by synaptic plasticity induced during theta burst firing events (59, 60). We hypothesized that in Cacna1c +/rats, the low CaV1.2 dosage in CA1 pyramidal neurons in the dorsal hippocampus would impact on both synaptic plasticity at CA1 synapses and on the rhythmic coordination of synaptic activity in the CA1 area.
Associative synaptic potentiation can be induced by temporally coordinated pre-and postsynaptic neuronal activity and requires the activation of postsynaptic NMDARs and VGCCs (26,27,42). We found that induction of associative long-term potentiation (LTP) by coincident pre-and postsynaptic spikes (TBP-LTP) is impaired at SC-CA1 synapses in Cacna1c +/animals.
However, the SC-CA1 synapses in Cacna1c +/rats appear mechanistically capable of plasticity and can be strengthened under conditions not requiring the activation of L-VGCCs as shown by using an LFS-pairing protocol known to induce NMDAR-dependent but L-VGCC-insensitive LTP (43). In contrast to the loss of TBP-LTP at SC-CA1 synapses, induction of associative plasticity was notably preserved at TA-CA1 synapses, which represent the other major source of cortical excitatory inputs onto CA1 pyramidal neurons in the hippocampus.
In order to further investigate the causes of this selective loss of TBP-LTP at SC-CA1 synapses in the hippocampus we investigated the excitability of, and calcium signaling in, CA1 pyramidal neurons in Cacna1c +/rats. Cacna1c +/neurons showed a reduction in the frequency-dependent 16 broadening of somatic action potentials (APs) during repetitive discharge. This effect was significant at firing frequencies that occur during LTP induction with TBP and was replicated in wild-type cells by application of isradipine, confirming a role for L-VGCCs. Narrower somatic APs elicit weaker dendritic Ca 2+ signals in CA1 pyramidal neurons, when compared to broader APs (47). Our Ca 2+ imaging experiments reveal an inefficient summation of spine Ca 2+ signals during bursts of back-propagated APs at SC-CA1 synapses in Cacna1c +/neurons. The reduced somatic AP broadening and the diminished spine Ca 2+ signals during AP bursts indicate that the low Cacna1c dosage impacts on the association between postsynaptic spiking and synaptic input necessary for associative plasticity to occur at SC-CA1 synapses (27,44). The deficit in action potential-triggered Ca 2+ signaling was significant in dendritic spines in the stratum radiatum, where the majority of Schaffer collateral (SC) synapses are made (61). Our findings that TA-LTP is intact in Cacna1c +/neurons suggest that the reduced Cacna1c dosage does not affect L-VGCC-dependent mechanisms in the apical tuft, where local dendritic spikes appear more important for LTP induction (26). We propose that the selective impairment of forms of LTP that require post-synaptic spiking in addition to NMDAR activation at SC-CA1 synapses underpins the selective behavioral phenotype seen in Cacna1c +/animals.
Network synchronization, and in particular theta-gamma PAC, in area CA1 are thought to contribute to the establishment of hippocampal memory traces by providing adequate temporal coordination between pre-and postsynaptic spiking (22). We found a deficit in the phaseamplitude coupling between theta and slow-gamma oscillations in the CA1 area of Cacna1c +/rats, manifested specifically during the exploration of a novel environment, indicating an impaired coupling between CA1 neurons and excitatory input from area CA3 (52). Alterations in theta-gamma PAC across brain regions has been recently reported in animal models of psychosis 17 (23,62) and in schizophrenic patients (63,64), and therefore may reflect a broader spectrum of network deficits in psychoses. Taken together these results show that low dosage of Cacna1c impairs both plasticity and network activation in the hippocampus in a manner likely to contribute to the observed deficits in contextual learning and latent inhibition seen in Cacna1c +/rats.
At a molecular level L-VGCC-mediated Ca 2+ signaling is essential for excitation-translation coupling. Ca 2+ influx via L-VGCC imparts location and temporal specificity for activation of the ERK signaling pathway (30), necessary for mechanisms of early LTP including AMPA receptor trafficking (28, 29) and for protein synthesis-dependent forms of long-term potentiation and memory (65). Here we show that Cacna1c +/rats have markedly decreased levels of ERK and CREB activation as assessed by their phosphorylated levels in the hippocampus. To test the hypothesis that impaired ERK signaling downstream CaV1.2 contributes to the behavioral and hippocampal functional deficits in Cacna1c +/animals, we sought to activate the ERK pathway independently of L-VGCCs. Recent evidence supports the targeting of neurotrophin receptors using small molecule BDNF mimetics as potential therapeutic strategy in a range of neuropsychiatric and neurodegenerative disorders (66, 67). With a view to rescuing synaptic dysfunction and spine deficits in the Cacan1c+/-animals, we used a small molecule TrkB/TrkC co-activator LM22B-10, previously shown to activate the ERK pathway in mice in vivo (55).

Direct application of LM22B-10 on slices rescued TBP-induced LTP at SC-CA1 synapses in
Cacna1c +/rats, and direct hippocampal infusion of LM22B-10 during context preexposure was sufficient to rescue the observed behavioral deficit in LI. In addition, we were able to show that peripheral (systemic) treatment with LM22B-10 was sufficient to rescue the observed molecular and plasticity changes in the hippocampus and to restore normal LI behavior in Cacna1c +/rats. 18 These results suggest that the targeting of neurotrophin receptor pathways may represent a route to ameliorating the deficits produced by altered dosage of CACNA1C in neuropsychiatric disorders.
In conclusion by adopting a genomically informed approach to investigate pathological processes associated with genetic risk for neuropsychiatric disorders we highlight the potential of drugs targeting neurotrophin receptor signaling as novel therapeutics in major psychiatric disorders including schizophrenia and bipolar disorder. 19

Study design
We used a Cacna1c +/rat model in a multi-disciplinary, controlled laboratory experiment design with the objective to characterize the impact of reduced dosage of the psychiatric risk gene CACNA1C on behavior, synaptic and circuit function and to explore ways to rescue the observed deficits. LI of CFC was chosen as the central behavioral paradigm of the study because it explores hippocampal-dependent associative learning implicated in cognitive functions translationally relevant to psychoses (15-17, 21, 68). The in vivo and ex vivo electrophysiology and two-photon imaging experiments in the dorsal hippocampal CA1 subfield were informed by the observed context-sensitive behavioral deficits in Cacna1c +/rats. These experiments were designed to investigate the impact of Cacna1c heterozygosity on hippocampal neural circuit synchronization, synaptic plasticity, excitability, and spine Ca 2+ signaling as potential mechanisms underlying the behavioral deficit. The immunohistochemistry experiments were conducted to determine the impact of reduced Cacna1c dosage on the ERK/CREB signaling pathways downstream L-type VGCCs which has been implicated in synaptic plasticity, associative learning, and context-dependent learning (28,30,65). For the rescue experiments we administered a small molecule TrkB/TrkC receptor agonist previously characterized for ERK pathway activation activity and neurotrophic effects after systemic administration (55) either intrahippocampally or systemically (intra-peritoneal injections). When given peripherally, the agonist (or vehicle) was administered prior to each behavioral manipulation to account for nonspecific state dependent effects (69). 20 Animals.
Cacna1c hemizygous rats with Sprague-Dawley background (33) were generated from cryopreserved embryos (strain SD-Cacna1ctm1Sage-generated using zinc-finger nuclease technology, Sage Research Labs, Pennsylvania, USA) for a truncating mutation in exon 6 of

Statistical analysis.
Sample sizes were determined by power analysis based on effect sizes routinely observed in the laboratory. Animals were assigned pseudo-randomly to each experimental group. Experimenters were blind to the genotype of animals when collecting the data. Behavioral and immunohistochemistry data was analyzed with repeated measures analysis of variance (ANOVA). Sphericity was tested using Mauchly's Test and the Greenhouse-Geisser correction was applied if necessary and reported. Sources of significance were determined post-hoc using 21 Tukey's test. When drugs were used, only post treatment data was included in the ANOVA.
Pairwise comparisons with Bonferroni correction were performed when ANOVA showed significant interaction between factors. In fig. S9 two statistical outliers (> ± 2 SD) were removed from the analysis of immunohistochemistry data for basal phosphorylated CREB and phosphorylated ERK (one Cacna1c +/rat each). For synaptic plasticity experiments, comparisons were made between normalized mean EPSC amplitudes at 30-35 min after conditioning, with Test/Control pathway (within subjects factor) and genotype or drug treatment (between subjects factor). The effect of genotype or drugs on LTP induction were tested for significance using a two-way ordinal regression (cumulative link model) followed by analysis of deviance (ANODE).
Passive membrane properties were compared using two-sample Mann-Whitney U test between genotypes against the null hypothesis of no difference between genotype-specific means. AP parameters for the first five or 10 APs during spike train discharges were compared using two-     Summary of changes in normalized EPSC amplitude at 30-35 min shown in F and G: no effect of genotype (LR (1)     Behavior was video recorded with cameras (JSP Electronics Ltd, China) positioned centrally above the chambers, digitized and analyzed offline using Numeroscope software (Viewpoint, France). For each training or retrieval session, animals were individually transferred between from home cages and the testing room in the same large transport box.

In vivo electrophysiology.
Surgical Procedure. All surgery was carried out aseptically to minimize the risk of infection.
Animals were anaesthetized with isoflurane in oxygen (maintained at 1.5-2%) and head-fixed in a stereotaxic frame (Kopf model 1900). After exposing the skull surface, a 2.5mm craniotomy was made using a tungsten carbide burr (3.6mm posterior to bregma and 2.5mm lateral to the midline) and a 20-tetrode (16 recording electrodes and 4 reference electrodes) microdrive was lowered into the craniotomy (Fig. S8A). The microdrive was fixed in place using dental cement (dePuY). Animals were given buprenorphine (0.025mg/kg) following surgery, and weight and water consumption were monitored for at least one week afterwards. Over a duration of 2-4 52 weeks tetrodes were gradually lowered (~20-40µM a day) to the pyramidal cell layer of CA1 (verified by the presence of sharp-wave ripples and bursting spike activity).
Data Acquisition. Recordings were made using a Digital Lynx SX system (Neuralynx). Local field potentials were sampled at 1kHz and filtered between 0.1 and 475Hz). Animal position was monitored with an overhead camera and an LED attached to the headstage.
Training. Animals were trained to run back and forth on a 175cm linear track over a period of 1-4 weeks. During this period animals were food restricted to 85% of their initial weight.
Behavioral experiments were carried out once tetrodes were in place in the pyramidal cell layer of dorsal CA1 and animals had reached a minimum criterion of 10 complete circuits on the linear track.
Familiar Track Behavioral Protocol (Fig. S8B). Animals were placed in a sleep-box (a soundattenuating chamber) and left to rest for approximately 30 minutes prior to track running. After the pre-sleep period animals were allowed to run freely for sucrose rewards. Following the run, animals were placed back in the sleep-box for a further 30 minutes. Animal position and electrophysiological signals were continuously recorded throughout the behavioral session.
Novel Track Behavioral Protocol (Fig. S8C). Following at least one recording on the familiar track and at least a day after the familiar track session, animals were subject to the novel position behavioral protocol. The pre and post sleep protocol here did not differ to that of the familiar track behavioral sessions. For the track sessions animals were placed on the familiar track for approximately 10 minutes to take baseline recordings. The animals were then taken off the track, the track was rotated 45° and animals were immediately placed back on and recorded while they now freely ran through a different portion of the recording room. 53 Data Analysis. All data analysis of in vivo electrophysiology was carried out in MatLab (Mathworks). Local field potential signals were taken from periods of track running over 35cm/s and all analysis was performed on these segments of local field potential. To control for individual animal differences in the number of circuits on the familiar track compared to the novel track, a matched number of runs were pseudo-randomly selected from the session containing more runs.
Power Spectral Analysis. All power spectral estimates were calculated using multi-taper spectral analysis (Chronux toolbox) with a window length of 2.5s, a time-bandwidth product of 3 and 5 tapers.
Phase-Amplitude Coupling. Phase-amplitude coupling was detected using a PAC toolbox applying the modulation index measure. (Onslow et al., 2011). In brief this method combines the amplitude envelope of the local field potential filtered at a high frequency band with the phase values of the signal filtered at a lower band to form a composite signal. The mean of this signal provides a measure of the strength of coupling between the amplitude of the higher frequency band with the phase of the lower frequency band.

Ex vivo electrophysiology.
Slice preparation and recordings. Acute transverse hippocampal slices were obtained from 3 -6 months old male and female SD-Cacna1c rats after being given a lethal dose of isoflurane Passive and active electrophysiological membrane properties. Membrane properties were determined from membrane potential (Vm) recordings during 500 ms square-wave current injections in whole-cell current clamp, using the amplifier bridge circuit (71). The recordings were filtered at 6 kHz and digitized at 50 kHz. After determining the resting membrane potential (RMP), the membrane potential was set to -70 mV by a small inward current injection. The membrane time constant (τm), the hyperpolarization-activated "sag" potential (Vsag) and the input resistance (Rin) were determined current injections of -0.1 nA (schematic in Fig S4A). Vsag was 56 measured as the difference between the Vm minimum and the steady-state hyperpolarization elicited by the injected current, and expressed as a percentage of the steady-state membrane voltage (Vss); τm was determined from an exponential fit of the Vm curve between 10% -95% from baseline to the Vsag minimum; Rin was calculated according to the Ohm's law (V=I×R) using the difference between the steady-state hyperpolarization (Vss) and the resting membrane potential (RMP). Active membrane properties were determined from Vm recordings during a series of depolarizing current injections steps from 0 to +0.7 nA in 50 or 100 pA increments every 10 s, to elicit action potential (AP) firing (Fig. S4B). The rheobase current (Irh) was determined using exponential fits through strength-latency curves representing the injected current intensity as a function of spike latency (72) EGTA was omitted to avoid introducing additional Ca 2+ buffering capacity in the cells.
Fluorescence was excited with a Chameleon Ultra II Ti:Sapphire laser (Coherent) tuned at 810 nm. Ca 2+ transients elicited by back-propagated action potentials (APCaTs) were imaged in dendritic spines in the stratum radiatum using line scanning mode (74)  channel at the end of the experiment. The spines located on the same segment were attributed the same distance from soma as the parent dendritic segment. The data were classified according to 59 the somatic distance of their parent dendritic segment and grouped in three distance zones (in µm): 0 -150, 150 -250, and 250 -400. The distribution of spine locations is shown in Fig. S5.

Immunohistochemistry.
Rats were given a lethal dose of Euthatal, i.p. and transcardially perfused with iced-cold saline and 4% PFA in phosphate buffer (PB Systemic administration of drugs. LM22B-10 (Tocris Bioscience, UK) for peripheral injections was prepared as described (55).
The animals were weighted and the amount of drug necessary to deliver 25 mg/kg in 1 ml injection volume was dissolved in HCl then mixed with 5% Cremophor (Sigma-Aldrich) and 2X sterile PBS. The solution was adjusted for pH 5 then diluted with double deionized water to 1 X PBS and 2.5% Cremophor. The drug solution was injected via intra-peritoneally 60 mins prior to each training stage of LI or before immunohistochemical and ex vivo slices preparations.  Cacna1c +/rats pre-exposed 48h before CFC was similar to that seen previously with a 24 h delay between PE and CFC ( Figure 1). Data shown as means ±SEM. *p <0.05 determined by two-way repeated measures ANOVA followed by Tukey adjustment of p-values for contrasts.   Fig. S1. Animals received an i.p. bolus of either LM22B-10 (25 mg/kg; pre-exposed: LM22B-10-PE, n=6; non preexposed: LM22B-10-NPE, n=6) or vehicle (pre-exposed: Veh-PE, n=5) 60 min before each