Brain Gene Expression in a Novel Mouse Model of Postpartum Mood Disorder

Abstract Background Steroid sulfatase (STS) cleaves sulfate groups from steroid hormones; its expression/activity increases in late pregnancy and into the postpartum period. STS-deficient human and mouse mothers display elevated psychopathology and abnormal behaviour respectively; in mice, these effects can be partially normalised by antipsychotic (ziprasidone) administration. Methodology We compared brain gene expression in new mouse mothers administered the STS inhibitor 667-Coumate, or vehicle; significant changes were followed-up with pathway analysis and quantitative polymerase chain reaction (qPCR). Finally, the effects of combined 667-Coumate and ziprasidone administration on expression of the most robustly differentially-expressed genes were examined. Results Surprisingly, no between-group gene expression changes were detected at a False Discovery Rate (FDR)-corrected p<0.1. 1,081 unique expression changes were detected at p<0.05, two top hits were verified by qPCR, and pathway analysis indicated enrichment of genes involved in olfactory transduction. The expression of Stoml3 and Cyp2g1 was unaffected by ziprasidone administration. Conclusions Postpartum behavioural abnormalities in STS-deficient mothers are likely to be the culmination of many small gene expression changes. Our data are consistent with the idea that olfactory function is key to maternal behaviour in mice, and suggest that aberrant expression of olfactory system genes may underlie abnormal maternal behaviour in STS-deficient women.


Introduction
Steroid sulfatase is an enzyme which cleaves sulfate groups from a variety of steroid hormones e.g. dehydroepiandrosterone sulfate (DHEAS), thereby altering their solubility and activity [1]. STS is expressed in numerous mammalian tissues, with highest expression in the placenta (www.ncbi.nlm.nih. gov/unigene/); in the developing and adult human brain, relatively high STS expression and activity is seen in the cortex, thalamus, cerebellum, basal ganglia, hippocampus and hypothalamus [2,3]. STS deficiency is associated with increased developmental and mood disorder risk and a number of behavioural differences including: inattention, increased impulsivity and altered mood and social function [4][5][6]; these behavioural differences may be mediated, in part, by underlying changes in serotonergic or cholinergic function [7][8][9]. Depression, prior and current history of medical disorders, and cognitive impairment, have previously been highlighted as important risk factors in suicidality [10].
In both humans and rodents, STS expression and activity increases in brain and peripheral tissues towards the end of pregnancy and into the postpartum period; hence, enzyme deficiency or dysfunction could potentially be associated with postpartum psychopathology [11,12]. Consistent with this, we have recently shown that women who are heterozygous for genetic mutations encompassing STS are at increased risk of postpartum mood disorders [4]. We have also demonstrated that female mice, in which STS activity is acutely, and systemically, inhibited with 667-Coumate shortly after giving birth, show altered maternal behaviour (specifically anxietyrelated and startle phenotypes) relative to vehicle-treated mice; these drug-induced behavioural abnormalities can be partially reversed by concurrent administration of the atypical antipsychotic drug ziprasidone [13].
To investigate the neurobiology underlying the postpartum behavioural phenotypes in STS deficient individuals, we compared whole brain gene expression in behaviourallydefined 667-Coumate and vehicle-treated new mouse mothers. Given the large between-group behavioural differences, we suspected that screening by microarray would be able to readily identify robust gene expression differences and candidate biological pathways, and we reasoned that analysis of whole brain tissue would capture activity changes across multiple interacting brain regions (no single brain region has yet been especially strongly implicated in postpartum mood disorders). We successfully identified a number of nominallysignificant gene expression differences, implicating a specific biological pathway.  [14,15]. There is some evidence that 667-Coumate also binds carbonic anhydrase in erythrocytes, an observation which may explain its in vivo stability and deliverability [16]. Behavioural testing comprised of sequential assessment on the elevated plus maze, in a locomotor activity chamber, and in a startle/prepulse inhibition paradigm.

Drug administration and behavioural analysis
A second batch of mice had 667-Coumate injections as described above, but also Tissue collection, RNA extraction and cDNA synthesis 3hr after behavioural testing, subjects were culled by cervical dislocation; whole brains were immediately removed, bisected sagitally, and frozen on dry ice. The 3hr post-behaviour timepoint was chosen in order to allow acute physiological changes induced by moderately aversive behavioural experiences to return to baseline. High-quality total RNA was extracted from the right hemisphere of the brain using custom-designed primers ( Table 1) and SensiMix (Bioline, London, UK). qPCR data were analysed using ΔC t methods as described previously [19] with normalisation to the mean of three 'housekeeping gene standards' (Hprt, Gapdh and Rn18s) whose expression was significantly (p<0.001) correlated within the samples. Groups were compared by either Mann-Whitney U test, unpaired t-test or One Way ANOVA.

Results
Behavioural and endocrine comparison in animals used for microarray analysis A subset of vehicle and 667-Coumate-treated mice from [13] which differed maximally on pertinent behavioural measures were selected Table 1. Primer sequences used for quantitative PCR analysis. Primers were designed to allow optimum amplification, to span intron-exon boundaries where possible, and to amplify key coding gene transcripts.

Gene
Forward primer 5'-3' Reverse primer 5'-3' for subsequent analysis, with the rationale being that these animals would exhibit the greatest difference in brain gene expression. Quantitative PCR (qPCR) qPCR analysis was performed for five of the most highly-differentially expressed genes  [21,22]. Whilst the direction of expression difference was generally consistent across the microarray and qPCR analyses for these genes, only two genes (Cyp2g1 and Stoml3) were significantly differentially-expressed according to qPCR  (Figure 2).

Discussion
Although our vehicle and 667-Coumate treated groups differed substantially both in terms of their behaviour and with respect to a peripheral endocrine marker of STS inhibition (DHEAS:DHEA ratio), and the microarray experiment was performed with standard quality control procedures, we identified surprisingly few gene expression differences between the groups; those changes which we did identify by microarray were relatively small in magnitude and several were not significantly different by quantitative PCR. These data indicate that, at least at the timepoint we assayed, large brain gene expression differences do not substantially contribute towards abnormal maternal behaviour in the STS inhibition mouse model, and suggest that the behavioural differences are associated with another underlying biological mechanism e.g. the aggregate effect of small expression changes across many genes. This idea is consistent with our previous observations of: a) few large, statistically-significant, gene expression differences between whole brain samples from male mice lacking the Sts gene and wildtype animals [23] despite considerable between-group behavioural differences [6][7][8][24][25][26] and b) evidence that small brain expression changes (<1.5-fold) detectable between 667-Coumate and vehicle-treated mice by qPCR, but not detectable by the present microarray study, might be associated with postpartum behavioural phenotypes [13].
We did not identify any overlap between genes significantly differentially expressed in the current study, and those whose expression was altered in whole brain from Sts-deficient male mice [23], and nor was there any overlap with genes whose skin expression was altered in male patients with steroid sulfatase deficiency [27]; hence, the genetic mechanisms associated with acute STS inhibition in females and constitutive STS deficiency in males may be largely dissociable.
We did observe robust upregulation of the Cyp2g1 and Stoml3 genes (~3-fold and 1.5fold respectively) in 667-Coumate treated whole brain. The expression of both genes is restricted to the olfactory system in mice [28,29]. CYP2G1 is a major P450 enzyme in the olfactory mucosa of rodents, and, whilst its absence in homozygous knockout mice does not appear to impair olfaction, it does result in altered steroid hormone metabolism and metabolic activation of coumarin [28].
Increased expression of Cyp2g1 following administration of 667-Coumate, a tricyclic coumarin sulfamate [30], is perhaps, therefore, unsurprising, and further evidence that the drug is influencing neurophysiology. In man, functional Cyp2g1 orthologues appear to be rare [31]. Whilst the biological roles of STOML3 remain to be fully clarified, there is some evidence that the protein mediates mechanosensory processes [32].  [33], and is consistent with an extensive literature on the importance of olfactory (and associated limbic) function in mammalian mothers [34], with evidence that several steroid sulfates act as ligands within the mouse accessory olfactory system [35], with the expression of multiple olfactory receptors within human brain tissue [36], and with recent findings that olfactory processes are perturbed in both a genetic mouse model of abnormal maternal behaviour [37] and multiple mood disorders [38]. With regard to specific candidate genes of interest, both [37] [38,39].
Additionally, it is important to recognise that olfactory abnormalities are likely to be only one of many biological mechanisms contributing to altered maternal behaviour.
Our study is limited in two main ways. First, we examined gene expression across the whole brain. Whilst this strategy is useful for capturing widespread expression changes and is sensible given our lack of knowledge about the underlying neuroanatomy of postpartum mood disorders, it is unlikely to detect gene expression changes between groups that are regionally-specific. Second, microarray technology, whilst cheap and easily implementable, is relatively insensitive to small gene expression changes, and has low resolution with regard to determining differentially expressed splice variants [40].
Hence, future between-group comparisons might use a more specific, and more sensitive, technique such as RNA sequencing to assay gene expression in selected brain regions of importance to maternal behaviour and mood regulation, and whose chemistry is known to be altered by STS deficiency e.g.
hippocampus. show interquartile range with median and mean (bold line) values, and whiskers represent 5% and 95% confidence intervals.