Prophylactic Valproic Acid Treatment Prevents Schizophrenia-Related Behaviour in Disc1-L100P Mutant Mice

Background Schizophrenia is a neurodevelopmental disorder with onset early in adulthood. Disrupted-In-Schizophrenia-1 (DISC1) is a susceptibility gene for schizophrenia and other psychiatric disorders. Disc1-L100P mutant mice show behaviors relevant to schizophrenia at 12 weeks, but not at 8 weeks of age, and may be useful for investigating the onset of schizophrenia in early adulthood. Methods We investigated whether early valproic acid treatment would prevent behavioral, cellular and gene expression abnormalities in Disc1-L100P mutants. Results Valproic acid prevented hyperactivity and deficits in prepulse inhibition and latent inhibition in Disc1-L100P mice. Genome-wide transcription profiling identified Lcn2 (lipocalin2) transcripts as being elevated by the Disc1 mutation and corrected by valproate. Disc1-L100P mice also had increased glial cell numbers in the subventricular zone, which was normalized by valproate. Genetic deletion of Lcn2 normalized glial cell numbers and behavior in Disc1-L100P mutants. Conclusions Pharmacological treatments are a feasible way of preventing abnormal behaviour in a genetic model of schizophrenia. Lcn2 is a potential novel drug target for early intervention in schizophrenia.


Introduction
Schizophrenia is a neurodevelopmental disorder in which genetic and environmental factors disturb brain development, but the symptoms typically do not emerge until early adulthood [1]. Disrupted-In-Schizophrenia-1(DISC1) is a risk gene for schizophrenia and other mental illnesses, [2] encoding a scaffold protein [3] that is critical for adult neurogenesis [4,5], neuronal migration, dendrite maturation and synaptogenesis [6,7]. Better understanding of the molecular mechanisms by which DISC1 variation alters neurodevelopment could lead to new treatment targets for neuropsychiatric illness.
There are a number of mouse models with DISC1 mutations or altered DISC1 expression [2], demonstrating that the neurobiological and behavioral effects of DISC1 perturbation are time dependent [8,9]. For example, transient expression of the DISC1 C-terminus on postnatal day 7 had effects not seen when expression was induced in adulthood [8]. The timing of expression of truncated DISC1 also has distinct behavioral effects [9].
Although behavioral impairments in various mutant DISC1 mouse models have been observed in adult mice [10], the behavior of these mice as young adults has not yet been investigated.
Because the Disc1-L100P mouse has many features consistent with schizophrenia, we sought to determine whether there was also an early adult onset of behavioral abnormalities. We found the behavior of 8 week-old mice to be normal [11,13,14], so we hypothesized that early intervention could rectify neurodevelopment and prevent abnormal behaviors emerging later in adulthood (12 weeks of age). Valproic acid (2-propylpentanoic acid) [16], given to Disc1-L100P mutant mice in early adulthood, prevented the emergence of schizophrenia-related behaviors later on. We profiled gene transcription in brain and identified higher Lcn2 (Lipocalin2) transcript levels in Disc1-L100P mice that were normalized by valproic acid. Lcn2 is co-expressed with glial fibrillary acidic protein (GFAP) in brain, and the increased number of glial cells in Disc1-L100P mutants was corrected by valproic acid. Genetic ablation of Lcn2 [17] normalized both behavior and glial cell numbers in Disc1-L100P mutants. Our work demonstrates that pharmacological intervention can prevent the onset of schizophrenia-related behaviors and suggests Lcn2 as a novel drug target for preventive treatment of schizophrenia.

Ethics Statement
All animal procedures were approved by the Toronto Centre for Phenogenomics (TCP) Animal Care Committee (AUP number 12-0025a-H) and followed the requirements of the Province of Ontario Animals for Research Act and the Canadian Council on Animal Care.
Latent Inhibition (LI). Assessed as previously described [11]. Mice were trained to drink in the experimental chamber for 5 days, 15 minutes per day. The LI procedure consisted of Preexposure, Conditioning, Lick Retraining and Test sessions. Pre-exposure: The pre-exposed (PE) mice received 40 white noise presentations (60 s inter-stimulus interval). The non-pre-exposed (NPE) mice were confined to the chamber for the same time without receiving the stimuli. Conditioning: All mice received fear conditioning to the noise stimulus. Two noise-shock pairings were used (10 s 85 dB white noise; 1 s 0.37 mA shock). Test: The noise was activated between licks 75-101. The following times were recorded: time to first lick, time to complete licks 50-75 (before noise onset: A period) and time to complete licks 76-101 (after noise onset: B period). The suppression ratio was calculated as A/(A+B).

Drug Administration Schedule
Valproic acid sodium salt (Sigma-Aldrich, Canada) in 0.9% NaCl was injected intraperitoneally (i.p) at 200 mg/kg, twice daily for 14 days [19]. Behavioral testing started 20 hours after the last injection (LI sessions were between injections), using independent cohorts for each test and for microarrays, Western blotting, and immunohistochemistry. Mice for tissue analysis were killed 20 hours after the last valproate injection, either by cervical dislocation, or anaesthetized with pentobarbital.

Preparation of Total RNA
Gross brain dissections were performed to obtain brain stem (BS), frontal cortex (FC), hippocampus (HIP), and striatum (STR) samples, from which total RNA was extracted as described previously [20]. Samples were mechanically homogenized and lysed using TrizolH Reagent (Invitrogen, Carlsbad, CA). RNA was precipitated with 70% ethanol, and purified and treated with DNAase using the Absolutely RNA MiniprepH kit (Stratagene, La Jolla, CA).

Target Preparation and Microarray Hybridization
Microarray hybridization and scanning were done at the Génome Québec Innovation Centre (Montreal, QC). RNA samples were quantified using the NanoDrop ND-1000 UV-Vis Spectrophotometer (Thermo Fisher Scientific, Waltham, MA), and evaluated using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA). Complementary RNA (cRNA) was amplified and biotinylated using the Illumina TotalPrep Amplification Kit for 3-6 replicates within each treatmentgenotype combination and hybridized to MouseRef-8 v2.0 Expression BeadChips (.25,000 probes; Illumina, San Diego, CA). BeadScan software (version 3,2006) was used to identify bead positions and to extract raw data.

Microarray Analysis
Each brain region was analyzed independently and identically. A two-factor, two-level design was used, with the factors being drug treatment and genotype. Raw array data were loaded into the beadarray package (v1.10.0) of BioConductor [21]. Following BASH analysis to remove spatial artifacts [22], data were preprocessed using Edwards background correction [23] and normalized using variance-stabilization (vsn package of BioConductor v3.8.0) [24]. A two-level, two-factor general linear model was then fit to the pre-processed expression values from each bead type using the limma package (v2. 16.3) of BioConductor, followed by an empirical Bayes moderation of the standard error and a falsediscovery rate (FDR) adjustment for multiple-testing. Genes were selected at a 10% FDR threshold in any contrast. Both normalized and pre-processed array data are available in GEO (GSE17735). Data were visualized using unsupervised machine-learning [25]. The mean normalized signal intensities for each (tissue, genotype, treatment) tuple were collated into a single matrix, with rows as genes and columns as conditions. Rows were selected based on sequential variance (0.1, 0.25, 0.5, 1.0) or signal intensity (7,8,9,10,11,12) thresholds. This matrix was mean-centered and root-mean-square-scaled, and then subjected to agglomerative hierarchical clustering using complete linkage with Pearson's correlation as a distance metric. Heatmaps were generated in the R statistical environment (v2.8.1) using the lattice (v0. [17][18][19][20] and lattice Extra (v0.5-4) packages. Functional analysis was performed using the NIAID's DAVID resource (v2008).

Western Blotting
Tissues were homogenized in RIPA buffer (Sigma-Aldrich). After centrifugation at 10,000 g at 4uC for 20 min, the supernatant  (1:2000). Immune complexes were detected using appropriate peroxidase-conjugated secondary antibodies and a chemiluminescent reagent (ThermoScientific).

Confocal Microscopy
After pentobarbital anesthesia and perfusion as above, whole brain was immersed in 4% PFA for 24 hours, and 100 mm sagittal slices were cut on a vibratome. Next, slices were immersed in 4% BSA, 0.1% Triton-X 100, PBS solution for 15 minutes. Primary antibodies (rabbit anti-LCN2 polyclonal, Santa Cruz Biotechnology, 1:100) and mouse anti-GFAP (1:250) were incubated with free-floating slices for 48 hours at 4uC. Slices were washed 3610 minutes in blocking-permeabilizing solution and incubated with secondary antibodies (rabbit LCN2-CY5, 1:200 or mouse GFAP-CY2, 1:200) for 2 hours at 20uC. Following incubation, slices were washed again (3610 minutes) and mounted on glass slides. For quantitative studies 3-5 mice per genotype and 4 slices per mouse were examined (Zeiss LSM 510 confocal microscope). All images were acquired with fixed exposure settings and analysed using Nikon EZ-C1 FreeViewer v3.9.

Statistical Analysis
The effects of the Disc1-L100P mutation, age, valproate treatment and Lcn2 deletion on mouse behaviors, cell counts, and mRNA or protein levels were evaluated by ANOVA or unpaired t-test. Because there were no significant sex effects, data for both sexes were combined. Significant main effects or interactions were followed by the Fisher's least significance difference (LSD) post-hoc test with significance set at p,0.05.

Disc1-L100P Mice Behave Abnormally at 12 but not at 8 Weeks of Age
We assessed behavior of Disc1-L100P mutants at 8 weeks of age, when PPI and LI normally develops [26,27,28], and later in adulthood (12 weeks of age). We focused on behaviors relevant to schizophrenia: hyperactivity, PPI, and LI [29]. Disc1-L100P mutants were hyperactive at 12 but not 8 weeks of age compared to WT mice (Figure 1a). A mild PPI deficit in 8 week-old Disc1-L100P mice became more pronounced at 12 weeks (Figure 1b). Disc1-L100P mice had normal Acoustic Startle Response (ASR) at 8 weeks of age, which decreased significantly by 12 weeks of age (Figure 1c). Similarly, young adult Disc1-L100P mutant mice showed LI, but had disrupted LI at 12 weeks of age, whereas WT mice showed robust LI at both ages (Figure 1d).

Valproic Acid Prevented Schizophrenia-related Behavior in Disc1-L100P Mice
Valproic acid treatment between 10 and 12 weeks of age prevented hyperactivity, PPI deficits and ameliorated ASR, but did not immediately correct LI deficits in Disc1-L100P mice (Figure 2a-c and Table S1). All behaviours in valproic acidtreated Disc1-L100P mice were comparable with WT mice, three weeks after the last dose of valproic acid (Figure 2d-f). LI actually improved in the three weeks after valproic acid treatment stopped. These data show that valproic acid treatment had lasting effects on behavior, even after the medication had been discontinued.

Transcriptomic Analysis
The Disc1-L100P mutation had the most pronounced effect on gene expression in the hippocampus (61 genes), with 13 genes showing altered expression in brain stem, but no changes in gene expression in striatum or frontal cortex (Table 1 and Figure 3a). Functional annotation of genes in the hippocampus revealed that most are involved in cell proliferation (23%) and cytoskeleton/cell shape (15%). Valproic acid significantly altered expression of 3 genes in hippocampus (Hist1h1c, Hist1h2be and Clcn2) and 2 genes in striatum (Egr2 and Fosb). Among 61 genes altered by the Disc1-L100P mutation, valproate corrected the expression of 13 genes (21.3%) in hippocampus and Lcn2 in the brain stem. Expression of 14 genes was changed only in valproate-treated Disc1-L100P mutants. The specific genes are listed in Tables S2-S6.

Valproic Acid Corrected Increased Cell Proliferation in Disc1-L100P Mutants
Given that DISC1 [4,5,6] and LCN2 [30] are both involved in cell proliferation and that cell proliferation was the largest functional category of altered transcripts (23%), we performed immunohistochemistry on the cell proliferation marker Ki67 in vehicle/valproic acid-treated Disc1-L100P and WT mice. As shown in Figure 4a-b, vehicle-treated Disc1-L100P mutants had more Ki67+ cells in olfactory bulbs (OB), rostral migratory stream (RMS), and subventricular zone (SVZ), but not sub-granular zone (SGZ) of hippocampus. These are the main areas for adult neurogenesis [31]. Valproic acid no effect on Ki67+ cell counts in WT mice.

Disc1-L100P Mutation Increased Glia but not Neurons in SVZ
To identify the proliferating cells in Disc1-L100P mutants, we performed immunohistochemistry with antibodies against neuronal nuclear protein (NeuN) and glial fibrillary acidic protein (GFAP). As shown in Figure 2c-d, there was a significant increase in GFAP-positive cells in OB, RMS and SVZ in Disc1-L100P mutants compared to WT mice but not in the hippocampus. Most post-mortem studies in schizophrenia report no or minimal gliosis [32], but some find gliosis in periventricular structures [33]. There was no difference in the number of NeuN-positive cells in all brain areas examined between the genotypes. We performed TUNEL staining to assess cell death and found no differences between Disc1-L100P and WT littermates. We also assessed cell proliferation in young animals and did not find significant differences between genotypes in the brain regions studied (Table S7).

Genetic Deletion of Lcn2 Corrected Excess GFAP + Cells and Normalized Behavior in Disc1-L100P Mutant Mice
We sought to assess the role of Lcn2 in mediating excess glial cells and schizophrenia-related behavior by crossing mice that lack both alleles of Lcn2 to either WT mice or Disc1-L100P mutants [11,17]. LCN2 is an autocrine mediator of reactive astrocytosis [34], which could be related to glial proliferation. Immunostaining with LCN2 and GFAP antibodies revealed their co-expression in OB, RMS, SVZ, cortex, hippocampus, and brain stem (Figure 5a). The expression levels of the two genes were positively correlated (r = 0.53, p,0.01) (Figure 5b). Disc1-L100P mutation increased LCN2 and GFAP proteins in the SVZ (Figure 5c-d) as well as in OB, RMS but not in the subgranular zone (SGZ). Genetic inactivation of Lcn2 reduced the excess GFAP + cells in SVZ (Figure 5c-d), OB and RMS with no effect in SGZ in Disc1-L100P mutant mice (Table S8). Finally, we assessed effect of genetic inactivation of Lcn2 on schizophrenia-related behaviors. Ablation of both copies of Lcn2 normalized locomotion, PPI, LI and ASR abnormalities of Disc1-L100P mutants to the level of WT control mice (Figure 5e-g).

Discussion
Our results show that: (1) abnormal behaviors in Disc1-L100P mutant mice are present at 12 weeks but not 8 weeks of age; (2) treatment with valproic acid prevented the emergence of abnormal behaviors in Disc1-L100P mutants; (3) 23% of transcripts altered in Disc1-L100P mice are implicated in cell proliferation, including Lcn2; (4) there were more glial cells in the SVZ, RMS and OB but not in SGZ of Disc1-L100P, which was corrected by valproic acid; (5) genetic inactivation of Lcn2 corrected excess glial cells and schizophrenia-related behavior in Disc1-L100P mutants ( Figure 6). We have identified a novel function for Lcn2 as a regulator of glial proliferation in the SVZ and of behaviors relevant to schizophrenia.
An important question about the pathophysiology of schizophrenia is why the psychotic symptoms emerge in the third decade of life, despite high heritability and the presence of premorbid abnormalities in early childhood [35]. Longitudinal magnetic resonance imaging (MRI) studies reveal progressive structural brain changes prior to the first episode of psychosis [36]. However, it remains unknown whether these premorbid structural abnormalities can be prevented by prophylactic early intervention. genotype-drug interaction term or the genotype term in ANOVA analyses. For microarrays, this analysis was performed using the fold-change magnitude in log 2-space, for RT-qPCR, RNE values (log 2-based fold changes in mRNA quantity, normalized to Gapdh and b-actin levels). We obtained an overall validation rate of 0.60 (P = 0.037). r, Pearson correlation factor.  Many schizophrenia patients have a prodromal period [1,37] of social and functional decline [38], that provides an opportunity for early intervention. pWe have chosen to investigate the potential for valproic acid to prevent the emergence of behavioral abnormalities in the Disc1-L100P mouse model for several reasons. First, valproic acid is already approved for use in teenagers as a mood stabilizer and anticonvulsant [16,39]. 15 Second, schizophrenia is caused by multiple factors that interact at many biological levels [40], so improved treatment may have to target multiple mechanisms [41]. Valproic acid is a drug with many pharmacological actions [16,42,43,44,45] and targets, including GABA transaminase, voltage-gated sodium channels, glycogen synthase kinase (GSK)-3, and histone deacetylases (HDACs) [16]. Thus, it was used as a screen to test if early intervention in an animal model for schizophrenia is feasible. Finally, we chose not to test antipsychotic medications because clinical trials during the prodrome of schizophrenia showed limited benefit [46].
To our knowledge, this is the first report of delayed onset of schizophrenia-related behaviors in a genetic animal model of schizophrenia. However, adolescent onset of schizophrenia-related behaviors has also been observed with prenatal immune activation by polyinosinic-polycytidilic acid (poly I:C) [47], and risperidone pre-treatment can prevent brain and behavioral abnormalities if administered in adolescence [48]. The beneficial effects of valproic acid on Disc1-L100P mice are similar to its efficacy in other mouse models of schizophrenia [19,49,50]. Importantly, we found that chronic valproic acid treatment prevented schizophrenia-related behaviors for three weeks after the last dose. Thus, we argue that valproic acid treatment may have altered brain development, rather than just having acute effects, which represents a new paradigm for preventive treatments [51].
Genome-wide expression profiling identified large effects of the Disc1-L100P mutation in the hippocampus, an area with high DISC1 expression [52]; smaller but still significant effects were also seen in the brain stem, consistent with a wide-ranging DISC1  functional network [53]. Valproic acid alone had minimal effects on gene expression in hippocampus and striatum, consistent with the limited behavioral effect on WT animals. As expected, valproic acid affected expression of transcription factors (Egr2 and Fosb), genes involved in epigenetic regulation (Hist1h1c and Hist1h2be) and neuronal excitability (Clcn2) [16]. Valproate had broad transcriptional effects in DISC1 mutants, correcting 21.3% of altered gene expression in the hippocampus. Moreover, we detected convergent effects of both Disc1-L100P and valproate on expression of 14 genes, of which five are independently associated with schizophrenia (Tables S2-6).
Glial cells are important in synaptic function, maturation and elimination, and are thus a potential therapeutic target for brain disorders [54]. Here, we describe for the first time, an association of schizophrenia-related behavior in Disc1 mutant mice with increased LCN2 and GFAP + glial cells in the SVZ. We found similar results with pharmacological inhibition (Figure 4a-d) and genetic inactivation of Lcn2 (Figure 5c-d; Table S8). LCN2 is a glycoprotein initially purified from neutrophils [55], that transports fatty acids and iron, induces apoptosis, regulates innate immunity [30], and is a pro-neoplastic factor [30]. LCN2 is secreted by astrocytes [56] and over-expression in zebrafish increased the number and activity of GFAP + glial cells [34]. Furthermore, mild cognitive impairment is associated with higher plasma LCN2 levels [57]. One possible connection between Disc1 and Lcn2 is the Rho-ROCK pathway, through which LCN2 modifies reactive astrocytes [34]. The Disc1-L100P mutation decreased expression of three Rho-related genes (Cdc42ep2; Arhgap24 and Rock2; Table S3).
There are several limitations of the work described here. A recent study of the Disc1-L100P mice backcrossed onto the C57BL/6JJcl strain did not show the same phenotype as the original paper characterizing this mutant [11,58]. C57BL/6JJcl was the original strain subject to ENU mutagenesis, but C57BL/6J was the strain backcrossed before phenotyping in the original report. Thus it is possible that genetic background could affect the phenotype, and that the results of the cross to the Lcn2 knock-out mice are not specific to inactivation of the Lcn2 gene. Previous studies have found that the antipsychotics haloperidol [13] and clozapine [11] corrected behavioural abnormalities in adult (week 12-16) Disc1-L100P mice, but did not assess effects on behaviour when treatment was delivered earlier, nor whether the drug effects persisted after treatment stopped, as was done in the current study. Finally, the L100P (T334C) mutation is not a polymorphism in humans associated with schizophrenia, so direct comparison across species is not possible. However, the Disc1-L100P mutation does provide general insights about the possible effects of DISC1 variants in humans.
In conclusion, the current study demonstrates that Disc1-L100P mutant mice have a developmental onset of behavioral and **p,0.01; ***p,0.001 in comparison with WT mice. (f) PPI deficit in Disc1-L100P mutants was normalized at all three pre-pulses by genetic ablation of Lcn2. ANOVA with repeated measures found a main effect of genotype [F 3,31 = 14.1, p,0.001], and pre-pulse intensities [F 2,62 = 6.6, p,0.05]. n = 6211 mice per genotype. ***p,0.001 in comparison with WT mice (g) Deletion of Lcn2 in Disc1-L100P mutants also restored LI. ANOVA detected a significant effect of genotype [F 3,47 = 13.5; p,0.001], pre-exposure [F 1,47 = 62.4; p,0.001] and their interaction [F 3,47 = 9.9, p,0.001]. n = 628 per experimental group. ***p,0.001 non-pre-exposed (NPE) in comparison with pre-exposed (PE) animals to the conditioned stimulus (CS) within each genotype. doi:10.1371/journal.pone.0051562.g005 Figure 6. Summary of results. Lcn2 levels correlate with GFAP in subventricular zone (SVZ) and these are in turn associated with abnormal PPI, an endophenotype for schizophrenia in Disc1-L100P mice. The 3D quadratic surface graph illustrates surfaces fitted by a smoothing technique to the average PPI, GFAP and Lcn2 intensity in SVZ data. The color spectrum (from green to brown) represents the GFAP and Lcn2 intensity data (from low to high, respectively). Pearson correlation coefficients are: r = 20.73; p,0.001 -for PPI and Lcn2; r = 20.56; p,0.05 -for PPI and GFAP and r = 0.77, p,0.001-for Lcn2 and GFAP (N = 16). doi:10.1371/journal.pone.0051562.g006 cellular abnormalities that parallels some clinical features of schizophrenia. We have shown that glial proliferation is increased by Disc1-L100P mutation and that early treatment with valproic acid or by genetic inactivation of Lcn2 can rectify this abnormal proliferation in conjunction with normalizing behavior. Future experiments will investigate molecular mechanisms underlying the functional association of DISC1 and LCN2 in the brain and the potential for valproic acid as a treatment in prodromal patients. LCN2 might also be useful as new biomarker for the prodromal stage of schizophrenia and as a potential new therapeutic target.

Supporting Information
Table S1 Acoustic Startle Response in Vehicle2/Valproic acid-treated Disc1-L100P and WT mice.