One hundred and sixteen adult C57BL/6N mice (8 weeks old, female = 58, male = 58) were used in this study of oxytocin. All mice were obtained from and housed in the Laboratory Animal Unit (LAU) at the University of Hong Kong. The experimental protocol had been approved by the Committee on the Use of Live Animals for Teaching and Research, the University of Hong Kong (CULATR case no: 2189–10, 2624–12). The behaviour holding room was maintained at 21°C but, unlike the general areas of the LAU, had a reversed day-night cycle (light on: 7PM-7AM). Mice were therefore acclimatized for a minimum of one week before testing. All behavioural tests were conducted in the dark phase of the light-dark cycle.

The social interaction test was carried out in a neutral cage with standard bedding (30 cm×30 cm×30 cm) following methods previously published [33]. On the first three days before the test, each mouse, including mice of the same strain, age and sex to be used as ‘strangers’, were habituated to the cage in one 10min session each day. On day 4, the test day, either oxytocin or saline was administered 30min before the test. Pairs of an unfamiliar, ‘stranger’ mouse and experimental mouse were placed into the neutral cage for 10 min. The time spent in social interaction (active contact such as sniffing, following and grooming the partner) was measured over 10 min.

The effects of oxytocin dose and exposure were analysed using a Univariate General Linear Model (GLM, SPSS 22), in which the drug dose and exposure (single or repeated) were the between-subject factors.

The open field exploration test was tested following methods previously published [34]. It was performed in 40×40 cm² rectangular white arenas. At the beginning of the session, the mouse was gently placed in the centre of the arena, and allowed to explore for 1hour while drug free. The mouse was then briefly removed from the arena, and drug administered quickly via the subcutaneous (s.c.) route. The arena was cleaned with 70% ethanol; the mouse was placed back in the arena and allowed to explore for another 30min. The mouse was again briefly removed from the arena and injected with 2.5 mg/kg d-amphetamine (calculated as the salt, sigma-Aldrich, Switzerland) dissolved in a 0.9% NaCl solution via the intraperitoneal route. The volume of injection was 5 ml/kg. The arena was cleaned again. The mouse was placed back in the arena and allowed to explore for another 1hour (Fig 2).

Equal numbers of male and female mice received a single subcutaneous dose of 10 μg/kg, 100 μg/kg or 1000 μg/kg of oxytocin (Sigma Chemicals, St Louis, Mo., USA) in an injection volume of 5 ml/kg prior to testing. All mice were given one week of ‘rest’ between each test. Each mouse received the same dose of oxytocin or saline for all tests. To control for the potential effect of previous oxytocin exposure and/or behavioural test order, half the mice followed test sequence I; half followed test sequence II (Fig 3). Twenty mice (10 females; 10 males) received saline injections only. There were 32 mice (16 females; 16 males) in each oxytocin dose group: 10 μg/kg, 100 μg/kg or 1000 μg/kg. Where appropriate, whether this was the first exposure to oxytocin or a repeat exposure was entered as a fixed factor in the analyses.

The open-field and amphetamine challenge test were carried out last; therefore, all animals had previously received 3 doses of oxytocin or saline. In this test, to clarify whether pre-exposure to oxytocin was sufficient to modify low-dose (2.5 mg/kg) amphetamine response, or whether an additional dose of oxytocin was necessary, half the cohort received saline immediately before the amphetamine challenge; half received a single dose of oxytocin (10 μg/kg, 100 μg/kg or 1000 μg/kg)

Twelve mice (female = 6, male = 6) were naïve to behavioural testing; half were exposed to 1000 μg/kg oxytocin, half to saline vehicle injection, once a week for 3 weeks were selected for proteomics and western blot analysis. The mice were sacrificed by cervical dislocation. The brain was removed and placed immediately on ice, and the striatum dissected free of the remainder of the brain tissue. The striatal tissue from each mouse was homogenized in 0.5ml of ice-cold 0.32 M sucrose buffered with 5 mM HEPES, pH7.5.

Striatum was homogenized in 2D lysis buffer (7M Urea, 2M Thiourea, 2% CHAPS, 10mmol DTT, 0.5% IPG buffer) for 30 seconds, then centrifuged at 15,000×g for 15 minutes to remove cell debris. Supernatant for 2-DE was collected and quantified with protein assay kit (Pierce^® 660nm Protein Assay, Thermo Scientific).

The 2D procedure followed the protocol from Bio-Rad Laboratories (Bio-Rad Laboratories, Inc.), with minor modifications. Rehydration and first dimension were carried out using Bio-Rad isoelectric focusing (IEF) separation (Bio-Rad IEF system). The 350ul rehydration buffer containing 180ug protein was loaded on 18cm, pH 4–7, IPG DryStrips (Bio-rad). IEF conditions were: 50V, 18h, 1000V, 1h; 3000V, 3h, 8000V, 6h. After the IEF run was complete, the IPG strips were equilibrated for 2×20min in the equilibration buffer containing 50mM Tris, 6M urea, 30% glycerol, 2% SDS. The first equilibration was performed in the equilibration buffer with 1.0%w/v DTT followed by a second equilibration with 2.5% w/v iodoacetamide. The strips were subsequently subjected to a secondary dimensional separation by 12% SDS-PAGE (Bio-Rad vertical system). The SDS-PAGE was performed at 10mA/gel for 30min, and then 45mA/gel, until the dye front reached the bottom of gels. After fixing in 50% ethanol, 12% acetate solution and staining with silver nitrate, gels were scanned at 300dpi resolution and the images were analysed with Image Master Platinum^™ software (GE Health care).

Protein spots with p < 0.05 and > 1.3-fold change were considered significant. Spots of interest were picked from sliver-stained gels, followed by in-gel trypsin digestion and peptide extraction [35] for protein identification by tandem mass spectrometry analysis using ABI 4800 MALDI TOF/TOF^™ MS Analyzer (Applied Biosystems, Foster City, CA, USA). The combined peptide mass fingerprinting (PMF) and MS/MS peptide fragmention data were submitted to the NCBInr database and SwissProt database using the software MASCOT version 2.2 (Matrix Science). For all significant protein identifications, both protein and total ion scores were above or equal to C.I. 99%.

The striatum was homogenized at 4°C in lysis buffer (1:5,wt/vol) containing 1mM EDTA and 20mM phenylmethylsulphonyl fluoride. The proteins from the resultant supernatant were determined (Bio-Rad Protein Assay) for Sodium Dodecyle-Sulphate Polyacrylamide gel electrophoresis (SDS-PAGE). Protein (20ug/lane) was subjected to electrophoresis on a 10% (wt/vol) polyacrylamide gel in SDS, and gels subsequently processed for electroblotting to polyvinylidene difluoride (PVDF) membranes. The blotted PVDF membranes were saturated with 5% (wt/vol) of skimmed milk in Tris Buffer Saline, PH 7.4 and 0.1% (vol/vol) of Tween 20 for 1h at room temperature. The membranes were sequentially incubated with primary antibodies to anti-Calcineurin B (rabbit polyclonal IgG antibody, 1:500 dilution, Cat# ab136526, Abcam); anti-GAD 67 (rabbit polyclonal IgG antibody, 1:1000 dilution, Cat# ab52249, Abcam) and anti-GAD 65 (rabbit polyclonal IgG antibody, 1:500 dilution, Cat# ab75750, Abcam) overnight at 4°C following by a peroxidase-labelled anti-mouse/rabbit IgG (1:2000 dilution, Boehringer Mannheim, Germany) for 1h at room temperature. After thorough washing, positive bands were revealed using ECL western blotting detection reagents and autoradiography film (Amersham, Biosciences, UK). The intensities of the bands were quantified using IMAGE QUANT software (Molecular Dynamics, USA).

Behavioural results from each task were analysed using a GLM running in SPSS22 with ‘Dose’ of drug (saline, 10 μg/kg, 100 μg/kg, 1000 μg/kg oxytocin), ‘Sex’ and ‘Repeat exposure’ to oxytocin/saline (where appropriate) as between subject factors. Where there was a significant main effect of Sex, data were analysed in each Sex separately. If there was no effect of repeat exposure, data were combined for further analyses. Significant findings observed using GLM were explored using post-hoc Bonferroni tests were appropriate.

In PPI, drug treatment may alter the baseline startle response to pulse stimuli. This impacts upon subsequent analysis of %PPI [30]. Therefore, our PPI design included 3 different pulse conditions so that, should baseline startle reactivity to any pulse condition be significantly altered by oxytocin, that pulse condition could be excluded from further analysis of %PPI.

Multivariate analysis using partial least squares-discriminant analysis (PLS-DA, SIMCA 12.0, Umetrics, Umea, Sweden) was performed to examine whether C57BL/6N mice with oxytocin exposure could be discriminated from those with saline exposure on the basis of full proteome profile [36]. Subsequently Analysis of Variance running in the standard in multivariate data analysis (SIMCA-P+ 12.0, Umetrics) was used to identify proteins with significant up or down regulation in the oxytocin treated group compared to saline treated group. Protein expression differences in Western Blot were analysed using t-tests.

In saline treated animals, the order of testing had no effect on any test.

Basal locomotor activity: All mice had been given three doses of oxytocin or saline before the open field and amphetamine challenge (Fig 3); all had completed PPI, novel object and social interaction paradigms. The total distance moved in a one-hour open field test provided a measure of basal exploratory locomotor activity. Following an injection of oxytocin or saline, locomotion was measured for 30 minutes. Finally, each animal received 2.5 mg/kg amphetamine and locomotion measured for 60 minutes.

Previous exposure to oxytocin lowered baseline activity in both sexes (F _{(3, 100)} = 3.649, p = 0.015), especially at 100 μg/kg (p = 0.009). There was no effect of an additional dose of oxytocin before testing (Fig 6A and 6B).

Amphetamine-induced locomotor activity: Oxytocin significantly attenuated the response to amphetamine (F _{(3, 100)} = 11.743, p < = 0.001). There was no effect of an additional dose of oxytocin before testing. As previous oxytocin exposure altered baseline activity, the data were reanalysed with baseline distance entered as a covariate. This did not alter the pattern of results; thus differences in baseline activity could not explain attenuation of amphetamine-induced hyperlocomotion by oxytocin. There was however a significant time × dose × sex interaction in the amphetamine response (F _{(33, 1167)} = 2.115, p = 0.000), therefore, post-hoc analysis examined each sex separately.

Oxytocin attenuated the response to amphetamine in females (F _{(3, 54)} = 10.477, p = 0.000), especially at 100 μg/kg (p = 0.033) and 1000 μg/kg doses (p = 0.000, Fig 6C). It also attenuated the response to amphetamine in males (F _{(3, 54)} = 3.846, p = 0.015), particularly at the highest dose (1000μg/kg) (p = 0.008, Fig 6D). As 3 repeated weekly doses of 1000ug/kg oxytocin caused a clear suppression of amphetamine-induced locomotion in both male and female mice, this dose regime was chosen for the 2D proteomics analysis.

A total of 811 individual protein spots were visualized across 12 gels (saline: n = 3 males, n = 3 females; oxytocin: n = 3 males, n = 3 females). All spots were included in a partial least squares-discriminant analysis (PLS-DA), which clearly separated protein expression in mice exposed to oxytocin from saline controls (Fig 7A).

Analysis of covariance confirmed significant drug-related differences in the expression of 15 proteins (Table 1). Western blot analysis of striatal tissue from animals used for behavioural testing confirmed that a protein identified in 2D proteomics, calcineurin, was significantly lowered by oxytocin.

2D-gel results were confirmed by Western blot analysis in a selected target protein, calcineurin. In addition, because 2D-gel analysis suggested that oxytocin may act upon GABA-glutamate pathways, we examined the effects of oxytocin on glutamic acid decarboxylase (GAD65 and GAD67) post-hoc. GAD catalyses the conversion of glutamate to GABA, and is a marker for GABAergic neurons. Both isoforms have been reported to be lower in schizophrenia and related neurodevelopmental disorders [37] and dopamine inhibits striatal GAD67 [38]. We expected that an ‘anti-psychotic’ effect of oxytocin would include up-regulation of GAD67 in striatum of C57BL/6N mice.

In addition, GAD67 expression (but not GAD65) was significantly elevated by oxytocin (Fig 7B and 7C).

Deficits in sensorimotor gating measured using PPI are associated with neurodevelopmental disorders. We selected the C57BL/6N mouse strain for this study because it has lower PPI than many other in-bred strains [39] and has been suggested to model aspects of schizophrenia [40]. Thus, any improvement elicited by oxytocin should be evident above the low baseline PPI expected in these animals.

Oxytocin increased PPI in male mice—a direction of effect similar to that reported using conventional anti-psychotics [41]. This result is partly consistent with previous reports that high doses of oxytocin (1000μg/kg) improve PPI in male Brown-Norway rats [42] and reverse PPI deficits induced by NMDA antagonists in rats [28]. However, the sensitivity of rodents to oxytocin is likely to be species and strain specific [43], and to be dependent upon the parameters of the PPI paradigm [44].

Although oxytocin attenuated PPI in female mice, as can be seen from Fig 4C and 4D, the levels of PPI in female mice were higher than even males treated with oxytocin. Thus oxytocin improved a behaviour ‘impaired’ in males, but had limited impact upon behaviour essentially intact in females. The latter observation fits with clinical evidence that PPI is intact in females with schizophrenia [45].

Our results contrast with the previous observation that neither oxytocin knockout [3] nor oxytocin receptor knockout [46] disrupts PPI. Thus, oxytocin is not ‘necessary’ for PPI, but does alter it. One mechanism for this action may be through the glutamate-GABA system which contributes to PPI [47] and is thought to be modulated by oxytocin [48]. Consistent with this, we found that oxytocin down-regulated ornithine aminotransferase, which catalyses the transamination of ornithine to glutamate [49]; and, up-regulated 14-3-3 protein zeta/delta (or KCIP-1), involved in GABA_B receptor coupling and dopamine D2 receptor inhibition [50]. In addition, oxytocin increased the level of GAD67 which reflects intraneuronal gamma-aminobutyric acid levels [51]. Thus, we speculate that the net effect of oxytocin may be to promote GABAergic signalling and restrain glutamate and dopamine activity, improving PPI.

In the first step of the novel object paradigm, oxytocin increased time spent exploring the objects, possibly representing an anxiolytic action [52]. However, the novel object component of this paradigm indexes memory for a familiar object and oxytocin disrupted recognition memory in males (Fig 3A) without altering this measure in females.

Again a pro-GABA action may contribute through calcineurin and KCIP-1 regulation of GABA_A and GABA_B receptors respectively. In line with this, GABA_A [53] and GABA_B [54] agonists impair recognition memory and calcineurin knockout has also been reported to impair memory in mice [55].

Oxytocin increased social interaction time in both females and males, but the dose response was sex-dependent (Fig 5B–5D). Oxytocin had a cumulative effect on social interaction in female mice; whereas previous exposure to oxytocin in males was no different to a single exposure. These facilitatory effects of oxytocin on social interaction are consistent with studies in oxytocin knock-out mice [56]. As discussed above, it is possible that oxytocin triggered changes in GABA signalling contribute. For example, GABA_A agonists increase social interaction time in rats, and antagonists do the opposite [57]. No dose of oxytocin used reduced social interaction, ruling out sedative effects.

Amphetamine is an indirect-acting dopamine agonist that increases synaptic dopamine; patients with schizophrenia are particularly vulnerable to amphetamine. Here, oxytocin suppressed amphetamine-induced locomotion in both sexes (even controlling for baseline activity differences). A possible explanation is that oxytocin promotes GABA-medicated inhibition of striatal dopamine release [58], and the increased level of GAD67 following oxytocin exposure points to increased GABAergic inhibitory activity. However, our proteomics results suggest additional pathways for oxytocin to limit dopamine mechanisms. For example, the 14-3-3 protein zeta/delta, up-regulated by oxytocin, is known to inhibit PKC [59] and PKC inhibitors have been shown to block amphetamine-induced dopamine release from rat striatal synaptoneurosome [60].

The attenuation of the amphetamine response occurred in mice previously exposed to oxytocin, with or without an additional acute dose of oxytocin (Fig 6). This indication that intermittent exposure to oxytocin can have ‘lasting’ effects will be important to explore further in light of recent evidence that chronic administration of oxytocin may have undesirable effects, disrupting social bonding in male voles [61].

In addition to actions on the glutamate-GABA systems, oxytocin had other wide-ranging effects including on proteins involved in cell structure and plasticity—tubulin-α dihydropyrimidinase-related protein 2 (DRP-2) and D-3-phosphoglycerate dehydrogenase (3-PGDH) [62–64]; and mitochondrial function—upregulating ATP synthase subunit alpha [65]; DJ-1 [66], and enolase (linked to recovery of cerebral metabolism in patients with schizophrenia treated with anti-psychotic medication [67]).

There are some limitations to our current study. To reduce the number of animals used, we tested each mouse in every behavioural paradigm design. The order of testing did not impact upon findings in the animals exposed to saline, and previous exposure to oxytocin did not influence results of PPI in either sex or social interaction in males. However, previous exposure was sufficient to suppress the amphetamine response in both sexes and improve social interaction in females. Unfortunately, our study was not designed to test the effect of an acute exposure to oxytocin in naïve animals in the amphetamine challenge (without greatly increasing the number of animals used).

The half-life of oxytocin is 10-15min [68], but only about 0.1% of oxytocin can cross the blood brain barrier [69]. Therefore the effects studied here are most liekly indirect. The timing of oxytocin administration was challenging to match across tasks. For example, in PPI, as in other drug studies using this paradigm, oxytocin was administered 10 mins before the task, but data collected in the task for a further 45 mins. In the social interaction and amphetamine challenge tests the effects of oxytocin were captured after 30mins. Thus for these tasks the effects of oxytocin from 30mins would have been captured. However, oxytocin was given after the ‘learning phase’ of the object recognition task to establish if it had any effect on the consolidation stage.

We observed subtle sex differences in the oxytocin dose response. However, the study was not primarily designed to compare sex differences in response to oxytocin. Rather the study was designed to look at the effects of oxytocin and both male and female animals were included. This is increasingly recognised as important and the NIH now requires the reporting of plans for a balance of male and female animals in preclinical studies unless single sex inclusion is specifically warranted [70].

Unfortunately, the underlying differences in baseline behaviours make interpretation difficult. For example, as previously reported, female C57BL/6N mice had higher levels of PPI than males [71]. This is not thought to be a consequence of the oestrous cycle which does not alter PPI and anxiety-related behaviours in female C57BL mice [72]. Male C57BL/6N mice also had longer social interaction times than female mice (Fig 5B, 5C&5D); possibly meaning they were less anxious than females. However, in addition to these baseline behavioural differences between the sexes, the differential response to oxytocin may also be a consequence of sex differences in expression of oxytocin and its receptors [7] and further work is therefore needed to clarify sex effects of oxytocin.

We only examined 2D protein differences in the striatum in response to repeated exposure to 1000 μg/kg oxytocin in naïve animals, and we cannot directly link differences to phenotypic differences. However, confirmatory Western blot studies carried out using tissue from animals involved in the behavioural experiments adds weight to the likelihood that the proteins implicated contribute to the behavioural differences observed. Protein expression differs across brain regions [73] and we do not know if our results generalize across the brain. Nor did we examine protein expression in each sex separately. However, the PLS-DA suggests the distinct protein profile in animals exposed to oxytocin and saline involved both sexes and the effects of oxytocin on social interaction and amphetamine challenge tasks were broadly similar in both sexes. Lastly, the proteomics approach used here was a standard analysis that does not fully sample intracellular organelles. The future separation of subcellular fractions would be useful.