Map2k7 Haploinsufficiency Induces Brain Imaging Endophenotypes and Behavioral Phenotypes Relevant to Schizophrenia

Abstract c-Jun N-terminal kinase (JNK) signaling contributes to functional plasticity in the brain and cognition. Accumulating evidence implicates a role for MAP kinase kinase 7 (MAP2K7), a JNK activator encoded by the Map2k7 gene, and other JNK pathway components in schizophrenia (ScZ). Mice haploinsufficient for Map2k7 (Map2k7+/− mice) display ScZ-relevant cognitive deficits, although the mechanisms are unclear. Here we show that Map2k7+/− mice display translationally relevant alterations in brain function, including hippocampal and mesolimbic system hypermetabolism with a contrasting prefrontal cortex (PFC) hypometabolism, reminiscent of patients with ScZ. In addition Map2k7+/− mice show alterations in functional brain network connectivity paralleling those reported in early ScZ, including PFC and hippocampal hyperconnectivity and compromised mesolimbic system functional connectivity. We also show that although the cerebral metabolic response to ketamine is preserved, the response to dextroamphetamine (d-amphetamine) is significantly attenuated in Map2k7+/− mice, supporting monoamine neurotransmitter system dysfunction but not glutamate/NMDA receptor (NMDA-R) dysfunction as a consequence of Map2k7 haploinsufficiency. These effects are mirrored behaviorally with an attenuated impact of d-amphetamine on sensorimotor gating and locomotion, whereas similar deficits produced by ketamine are preserved, in Map2k7+/− mice. In addition, Map2k7+/− mice show a basal hyperactivity and sensorimotor gating deficit. Overall, these data suggest that Map2k7 modifies brain and monoamine neurotransmitter system function in a manner relevant to the positive and cognitive symptoms of ScZ.


Introduction
c-Jun N-terminal kinase (JNK) signaling plays a key role in synaptic plasticity and neuronal communication. 1 In humans, mutations in JNK genes lead to intellectual disability. 2 We have identified a common variant in the MAP2K7 gene (encoding MAP kinase kinase 7 (MAP2K7), which activates JNKs) that shows a strong association with schizophrenia (ScZ). 3 Genes in this signaling pathway are independently associated with ScZ (supplementary figure S1). 4,5 Hence, strong evidence implicates the pathways modulated by MAP2K7 in ScZ. Furthermore, MAP2K7 transcript levels are reduced in the prefrontal cortex (PFC) of ScZ patients 3 and Map2k7 haploinsufficient mice (Map2k7 +/− mice) show deficits in PFC-dependent tasks. 3,6 Despite these observations we have a poor understanding of how MAP2K7 mutations increase the risk of developing ScZ.
Identifying neurobiological biomarkers with translational parallels in patients and rodent models is a major objective in psychiatric drug discovery. For ScZ the greatest progress has arguably been made for cognitive symptoms. [7][8][9] Developing translatable tests for positive and negative symptoms is more challenging. Prepulse inhibition (PPI) is widely promoted as a crossspecies measure recruiting similar neural circuitry and reflecting neural domains involved in positive and cognitive symptoms. 10,11 Impaired PPI is robustly observed in ScZ and correlates with positive symptomatology. [12][13][14][15] Other rodent behaviors deemed relevant to positive symptoms include hyperlocomotion, potentially indicative of dopaminergic hyperfunction. 11 Although the translational relevance of hyperlocomotion is debated, it is influenced by ScZ risk genes 16,17 and is present in patients. 18 ScZ patients exhibit a characteristic pattern of brain imaging (functional and structural) and neurochemical abnormalities. In chronic ScZ functional abnormalities include reduced PFC activity (hypofrontality). 19,20 However, during acute psychosis patients show a different profile with elevated ventral striatum (nucleus accumbens) and hippocampus/temporal cortex activity. [21][22][23] Hippocampal/temporal cortex hyperactivity is also seen in chronic ScZ and in those at high risk of psychosis. 24 Many studies have characterized alterations in functional brain network and regional connectivity in ScZ, with complex and contradictory findings. In general these studies support enhanced functional connectivity, potentially limited to specific neural systems, during early stages of the disorder 25-28 with a contrasting reduction in connectivity in chronic ScZ. [29][30][31] Dysconnectivity in ScZ may be both disease time-course and neural system dependent. For example, with regard to neural systems, increased accumbens-PFC, 32,33 reduced ventral tegmental area (VTA)-thalamus, 34 reduced VTA-insular cortex, 35 and increased hippocampal-PFC connectivity 36 have been reported. Equivalent phenotypes are rarely studied in rodent models. However, we have shown hypofrontality in rats following subchronic NMDA-R antagonist administration 37,38 and in mice with truncated Disc1. 39 These models also show brain network connectivity alterations relevant to ScZ. [38][39][40] These phenotypes provide mechanistic insight and offer translational biomarkers against which the efficacy of novel therapeutics can be tested. 41 Evidence supports both glutamatergic and monoamine (dopamine, serotonin, and noradrenaline) system dysfunction in ScZ, [42][43][44][45] with glutamatergic/NMDA-R hypofunction and monoamine hyperfunction, particularly for dopamine, supported. These disturbances have also been linked to the dysfunction of ScZ risk genes, including Disc1. 39,46 Characterizing cerebral metabolic responses to drugs that challenge these systems provides insight into their in vivo function and the mechanisms by which ScZ risk genes mediate their effect.
Here we test the hypothesis that Map2k7 +/− mice exhibit alterations in cerebral metabolism and functional brain network connectivity relevant to ScZ. To elucidate the impact of Map2k7 haploinsufficiency on in vivo NMDA-R/glutamate and monoamine system function we characterize the cerebral metabolic response to ketamine and dextroamphetamine (d-amphetamine), drugs that induce ScZ-like symptoms in humans. In light of the results obtained, we also characterize the impact of d-amphetamine and ketamine on PPI and locomotor activity (LMA) in Map2k7 +/− mice.

Animals
Mice heterozygous for Map2k7 (Map2k7 +/− ) were produced as previously described 47 and backcrossed onto the C57BL/6 background. 6 Animals were house under standard conditions (21 o C, 45%-65% humidity) with a 12-hour light/dark cycle (lights on 08:00). Experiments were conducted in accordance with the UK Animals (Scientific Procedures) Act 1986. 14 C-2-Deoxyglucose Functional Brain Imaging 14 C-2-Deoxyglucose ( 14 C-2-DG) brain imaging was completed in Map2k7 +/− and wild-type (WT) littermate mice as previously described. 39,48 Alterations in local cerebral glucose utilization (LCGU) in brain regions of interest (RoI) were analyzed using ANOVA with sex and genotype as independent variables. Significance was set at P < .05.

Functional Brain Network Analysis
Brain network properties were determined through analysis of the data from saline-treated animals. Global brain network properties, including mean degree (<k>), the clustering coefficient (C p ) and average path length (L p ) were determined as previously described. 39,40,48 Regional importance was defined through centrality analysis (degree [K i ], betweenness [B i ], closeness [C i ], and eigenvector centrality [E i ]) as previously described. 39,40,48 The standardized z-score centrality measure for each RoI was calculated by comparison to 11 000 calibrated Erdös−Rényi networks. The significance of the z-score difference between experimental groups was determined by comparison to 55 000 random permutations of the data. Significance was set a P < .05. A composite z-score, across all centrality measures, was also determined with a difference >1.96 or < −1.96 considered significant. A detailed overview of the network analysis approaches used is included in the supplementary material.

Analysis of Regional Functional Connectivity
Regional functional connectivity was analyzed using partial least squares regression (PLSR) as previously described 38,39 using the PLS package 51 in R (R Core Team, 2018). "Seed" region connectivity to all other RoI (57) was defined by the variable importance to the projection (VIP) statistic. The VIP, SD, and confidence interval (CI) were estimated by jack-knifing. A significant connection between regions exist if the 95% CI of the VIP statistic >0.8. 52 Genotype-induced alterations in the VIP statistic were statistically determined by the standardized z-score (z-score > 1.96 or < −1.96 considered significant).

Prepulse Inhibition
PPI was measured in SR-LAB chambers (San Diego Instruments) as previously described. 53 The impact of 5 mg/kg d-amphetamine i.p. (WT: male n = 11, female n = 6; Map2k7 +/− : male n = 11, female n = 12) and 25 mg/kg ketamine i.p. (WT: male n = 5, female n = 4; Map2k7 +/− : male n = 3, female n = 4) on % PPI was determined, as these doses disrupt PPI in WT mice. 54,55 Drug and vehicle treatment were counterbalanced across days in each experiment. Thereafter % PPI was calculated [(startle reactivity at 120 dB -startle reactivity with prepulse) / startle reactivity at 120 dB × 100] and analyzed by ANOVA with genotype and sex as between subjects factors, prepulse intensity and treatment as within subjects factors and each mouse nested within genotype and sex. Pairwise comparisons were made using Fisher's post hoc test.

Locomotor Activity
LMA was monitored in black opaque Perspex XT arenas (40 × 40 × 40 cm) lit from below by infrared LEDs. The impact of d-amphetamine and ketamine on LMA was determined, with drug and vehicle injection counterbalanced across days in each experiment. Each mouse was placed in the center of the arena and allowed to explore for 15 minutes (habituation). Mice were then treated with 3 mg/ kg d-amphetamine i.p. (WT: male n = 5, female n = 4; Map2k7 +/− : male n = 3, female n = 4), 20 mg/kg ketamine i.p. (WT: male n = 5, female n = 5; Map2k7 +/− : male n = 5, female n = 5) or saline and placed in the arena for the test period (ketamine, 30 min; d-amphetamine, 60 min). Total distance moved (cm) was measured using EthoVision XT (Noldus Information Technology) and analyzed by ANOVA with genotype, sex and treatment as between subjects factors and each mouse nested within genotype and sex. Pairwise comparisons were made using Fisher's post hoc test.

Map2k7 Haploinsufficient Mice Show Abnormal Functional Brain Network Connectivity Supporting Hippocampal and Septum/DB Hyperconnectivity and Mesolimbic Hypoconnectivity
Globally, functional brain networks in Map2k7 +/− mice had a similar number of connections (<k>) and showed a similar level of clustering (C p ) to those in WT animals. However, average path length (L p ) is significantly reduced (P = .0123) in Map2k7 +/− mice (figure 2), supporting an altered network structure that promotes more efficient information transfer across brain networks in these animals.
Multiple hippocampal subfields were identified as having significantly greater centrality in Map2k7 +/− mice (table 1). Multiple regions of the septum/diagonal band of Broca (septum/DB) and amygdala, along with the dorsal reticular thalamus (dRT) and multiple PFC subfields, were also identified as having increased centrality in Map2k7 +/− mice. These findings support the hyperconnectivity of these neural systems in Map2k7 +/− mice, that contributes to the global hyperconnectivity (decreased L p, figure 2) seen in these animals. Significant decreases in centrality in Map2k7 +/− mice were also identified, limited to the mesolimbic system and retrosplenial cortex (RSC, table 1). Full centrality data shown in supplementary table S2.
When the ventral hippocampus DS and ML were considered as seed regions there was evidence that the connectivity of the septum/DB, which provides direct innervation to the hippocampus, 56 was abnormally increased in Map2k7 +/− mice (figure 2), consistent with the increased centrality of these regions in these animals. The connectivity of DS and ML to the medial PFC (infralimbic cortex, IL) was also significantly increased in Map2k7 +/− mice. There was also evidence for enhanced connectivity between hippocampal subfields in Map2k7 +/− mice, specifically for the DS, a primary output of the hippocampus. 57  ) and the serotonergic dorsal raphé (DR). In contrast to these broad increases in connectivity, selective losses in DS and ML connectivity were also found, including lost connectivity to the mesolimbic system (nucleus accumbens core [NaC], nucleus accumbens shell [NaS]) and RSC, mirroring the decreased centrality of these regions (table 1).
When the DLO was considered as the seed region abnormal functional connectivity between this PFC subfield and the hippocampus (DH-CA1, DH-CA2, VH-CA2) was found in Map2k7 +/− mice. In addition, the DLO showed abnormal functional connectivity to the mesolimbic system (VTA and NaS) in Map2k7 +/− mice. By contrast, connectivity of the DLO to its projecting thalamic nuclei (AM, anteroventral nucleus [AV], mediodorsal nucleus [MD]) was lost in Map2k7 +/− mice.
When the VTA was considered as the seed, lost connectivity to other mesolimbic regions (NaS), the basal ganglia (dorsolateral striatum [DLST], globus pallidus [GP], substantia nigra pars compacta [SNC]) and thalamic nuclei (AM, AV, ventrolateral nucleus [VL], dRT) was found in Map2k7 +/− mice. Significant increases in VTA connectivity in Map2k7 +/− mice were limited. However, new abnormal connectivity between the DR and VTA was found, suggesting that serotonergic regulation of the VTA may be altered in Map2k7 +/− mice (figure 2).
MG functional connectivity was also altered in Map2k7 +/− mice, including increased connectivity to the mesolimbic system (NaC and NaS) and lost connectivity to the basal ganglia (DLST, GP, SNR) and thalamic nuclei (dRT, CM and VL).
Full PLSR data are included in supplementary tables S3-S7.

The Cerebral Metabolic Response to Ketamine Is Not Altered in Map2K7 +/− Mice
In accordance with published data subanesthetic ketamine increased LCGU in the PFC, hippocampus and nucleus accumbens and induced hypometabolism in the reticular thalamus and selected neuromodulatory nuclei. 39,49 We found no evidence that the response to ketamine was significantly altered in Map2k7 +/− mice ( figure 3). This suggests that NMDA-R/glutamate system function is not significantly disrupted in Map2k7 +/− mice. Full data are shown in supplementary table S8.

The Cerebral Metabolic Response to d-amphetamine Is Reduced in Map2K7 +/− Mice
In accordance with published data d-amphetamine increased LCGU in the thalamus, hippocampus, PFC, and basal ganglia in WT mice. 50,58 d-amphetamine also increased LCGU in neuromodulatory nuclei including the raphé (DR; median, MR) and locus coeruleus (LC), while decreasing LCGU in the amygdala.
We found widespread evidence that the LCGU response  altered inter-regional connectivity in Map2k7 +/− mice. Blue denotes lost and red denotes gained connectivity. Brain sections modified from the Allen atlas (mouse.brain-map.org/static/atlas).

Male Map2k7 +/− Mice Show a PPI Deficit and Attenuated Responses to d-amphetamine
As d-amphetamine disrupts PPI and induces hyperlocomotion 59,60 we assessed behavioral sensitivity to d-amphetamine in Map2k7 +/− mice. Sex-dependent effects were observed, whereby male Map2k7 +/− mice showed reduced baseline PPI relative to WT males, with no baseline genotype difference in females (figures 5a and 5b). d-amphetamine significantly reduced PPI in WT male but not WT female mice (figures 5a and 5b), limiting the detection of an attenuated d-amphetamine response in female Map2k7 +/− mice. Nevertheless, the impact of damphetamine on PPI was significantly attenuated in male Map2k7 +/− mice (figure 5a), and remained significant when data from both sexes was pooled (figures 5c and 5f). In parallel, d-amphetamine's stimulatory effects on LMA was significantly attenuated in male (F (1,353) = 1763, P < .001) but not female Map2k7 +/− mice (figure 5j).

Discussion
Map2k7 +/− mice show brain imaging endophenotypes and behavioral phenotypes relevant to ScZ. Behaviorally this includes a deficit in sensorimotor gating and hyperlocomotor activity, relevant to the positive symptoms of ScZ. [12][13][14][15] Previous studies have identified cognitive and reward processing deficits in Map2k7 +/− mice, also relevant to ScZ. 3,6 The brain metabolism and network connectivity deficits identified in Map2k7 +/− mice align with these behavioral deficits. Moreover, Map2k7 +/− mice show altered cerebral metabolism and behavioral responses to d-amphetamine but not ketamine, supporting monoaminergic system but not NMDA-R/glutamate system dysfunction as a consequence of Map2k7 haploinsufficiency.
The endophenotype concept continues to evolve in psychiatric genetics. [63][64][65] Key features include the following: (1) heritability and cosegregation with a disorder, or with a symptom domain that may cross diagnostic boundaries. Evidence for the association of MAP2K7 gene variants that result in decreased MAP2K7 expression with ScZ 3 supports the construct validity of Map2k7 +/− mice. (2) The endophenotype being part of the biological disease process (eg, functional or structural brain changes) and relating to the symptoms of the disease. We discuss below the strong translational alignment between the imaging endophenotypes and behavioral deficits identified in Map2k7 +/− mice and those seen in ScZ. We also discuss the alignment of the imaging endophenotypes seen in these animals with their behavioral deficits.

Map2k7 +/− Mice Show ScZ-Relevant Alterations in Cerebral Metabolism and Brain Network Connectivity
The alterations in LCGU identified in Map2k7 +/− mice parallel those observed in ScZ. This includes DLO hypometabolism and hippocampal hypermetabolism, mirroring orbitofrontal hypometabolism 19,66 and hippocampal/temporoparietal hypermetabolism in ScZ. [21][22][23][24] Intriguingly, we have previously identified DLO hypometabolism in genetic 39 and pharmacological 38 rodent models relevant to the disorder. The alterations in functional brain network connectivity present in Map2k7 +/− mice also parallel those in ScZ. Although the decreased path length (figure 2) and increased regional centrality (table 1) in Map2k7 +/− mice contrasts with that reported in chronic ScZ, [29][30][31] it does parallel the increased connectivity seen in first episode psychosis, 26 patients with prominent auditory hallucinations 21 and those with childhood onset ScZ. 28 Increased global efficiency of brain network connectivity has also been reported in chronic ScZ. 67 We also found evidence for increased PFC connectivity in Map2k7 +/− mice, paralleling that reported in early but not chronic ScZ and in those at high risk of developing the disorder. 25 Increased PFC connectivity is also found after ketamine administration, 25,48,49 a translational model relevant to ScZ, and is present in other genetic mouse models. 39 This suggests that specific neural systems, such as the PFC, may show increased connectivity at early disease stages or in those at risk of developing ScZ, and that Map2k7 +/− mice may have particular translational relevance to these populations.

Brain Imaging Endophenotypes in Map2k7 +/− Mice Align With Their Translational Behavioral Deficits
Map2k7 +/mice show a deficit in PFC-dependent working memory 3,6 and, in this study, hypofrontality (figure 1). As MAP2K7 expression levels are decreased in the PFC in ScZ 3 and Map2k7 modulates PFCdependent cognitive processes and metabolism, Map2k7 dysfunction may contribute to the hypofrontality 19 and working memory deficits 73 seen in ScZ. In addition, Map2k7 +/− mice show altered reward processing 6 that may relate to their altered mesolimbic system function (figures 1 and 2). In humans, mesolimbic system connectivity is linked to trait anhedonia 74 suggesting that the mesolimbic dysconnectivity and reward processing deficits seen in Map2k7 +/− mice may have relevance to the anhedonia and altered reward processing present in patients. 75,76 We detected reduced PPI in male, but not female, Map2k7 +/− mice. Impaired PPI is reliably observed in ScZ, may be more pronounced in male patients, 77 and correlates with positive symptoms. [12][13][14][15] PPI is a multisystem phenomenon, involving many brain regions including the RT, MD, PFC, and hippocampus, 78 figure 2). The septum/ DB regulates hippocampal activity, including the regulation of hippocampal theta waves. 79,80 These waves are dysfunctional in ScZ, modulate hippocampal-PFC connectivity 81 and are influenced by other ScZ risk genes. 70 The septum also modulates VTA dopaminergic neuron activity via the hippocampus and influences the response to amphetamine. 82 Intriguingly, we found increased hippocampal and VTA activity in Map2k +/− mice (figure 1), suggesting that increased septal-hippocampal connectivity (figure 2) may drive increased hippocampal-VTA circuit activity in these animals. Previous studies have shown that lesioning 83 or modulating activity in the septum 84 alters the response to amphetamine. Given the key role of the septo-hippocampal pathway in PPI, 85 and that the impact of d-amphetamine on hippocampal and VTA activity is attenuated in Map2k7 +/− mice (figure 4), dysfunction in this circuitry may contribute to the baseline PPI deficits and the attenuated impact of damphetamine on PPI in Map2k7 +/− mice. Intriguingly, the impact of amphetamine on cerebral metabolism and PFC dopamine release is attenuated in ScZ, mirroring the observations in Map2k7 +/− mice. 86,87 Overall, the data suggest that Map2k7 +/− mice display a range of brain imaging and behavioral deficits that have translational relevance to ScZ. The dysfunction appears to be relevant to a broad range of ScZ symptoms, against which the efficacy of novel therapeutics can be tested. Brain imaging endophenotypes in these animals with translational relevance to positive symptoms may be particularly useful because of the paucity and ambiguity of current behavioral phenotypes for these symptoms.

Supplementary Material
Supplementary data are available at Schizophrenia Bulletin online.

Funding
This work was supported by the Medical Research Council (award no. MR/K501335/1 to Dr Openshaw).