Introduction

Epidemiological evidence for a link between maternal bacterial and viral infections during pregnancy and increased risk in the offspring for schizophrenia and related disorders (Mednick et al. 1988; O’Callaghan et al. 1994; Hultman et al. 1999; Brown et al. 2000, 2004a) has led to recent attempts to model the mechanisms involved in animals. Support for a causal relationship between in utero immunological challenge and enhanced prevalence of psychosis-related behavioural, cognitive and pharmacological abnormalities in later life has been established in rats and mice using bacterial endotoxin (Borrell et al. 2002; Fortier et al. 2004; Golan et al. 2005), human influenza virus (Shi et al. 2003) or a synthetic viral mimic (Zuckerman et al. 2003a; Zuckerman and Weiner 2003, 2005; Meyer et al. 2005; 2006; Ozawa et al. 2006). Prenatal immunological stimulation has also been shown to result in long-term structural alterations in the hippocampus and cortical mantle in neonatal and adult offspring (Fatemi et al. 1999, 2002; Golan et al. 2005) and in altered expressions of genes implicated in schizophrenia (Fatemi et al. 2005; Harrison and Weinberger 2005). The functional consequences of prenatal immune challenge typically remain behaviourally latent until adulthood when the impact of the prenatal immunological manipulation on behaviour and cognition eventually unfolds (Zuckerman et al. 2003a; Zuckerman and Weiner 2003; Ozawa et al. 2006). This maturational delay between the immunological manipulation in utero and the onset of psychopathology in adulthood agrees well with the developmental aetiology and disease process of schizophrenia and related disorders (Weinberger 1987; Rapoport et al. 2005).

The precise mechanisms underlying the epidemiological link between maternal infection during pregnancy and increased risk for schizophrenia in the offspring remain largely unknown. One hypothesis asserts that the maternal immune response to the infectious process may be the critical link in this association (Buka et al. 2001; Pearce 2001; Patterson 2002; Brown et al. 2004b). In particular, it has been suggested that the elevation of pro-inflammatory cytokines in the maternal host and thereafter in the foetal environment may be one of the key events in triggering mal-neurodevelopment in the developing foetus (Gilmore and Jarskog 1997; Urakubo et al. 2001; Gilmore et al. 2004, 2005; Ashdown et al. 2006). Hence, inflammation-mediated disruption of foetal brain development may predispose the early organism to the emergence of psychopathological traits in later life, suggesting furthermore that neurodevelopmental perturbations implicated in the ethiopathogenesis of schizophrenia already start in utero (Jakob and Beckman 1986; Bracha et al. 1991, 1992; Murray et al. 1991; Gilmore et al. 1996).

However, the rodent and primate brain undergoes significant development and maturation in early postnatal life, which is characterized by an intricate mother–infant relationship (Fleming et al. 1999). It is therefore not surprising that early life adversity in the form of child neglect and/or abuse is a critical factor in the development of affective psychopathology in humans (Parker et al. 1995; Lehmann and Feldon 2000; Kendler et al. 2002, 2006; Cirulli et al. 2003; Pryce and Feldon 2003; Pryce et al. 2005). Moreover, human adoption studies have demonstrated that adoptees with a high genetic risk for schizophrenia are significantly more sensitive to adverse rearing patterns in adoptive families than are adoptees at low risk (Wahlberg et al. 1997; Tienari et al. 2004; Mäki et al. 2005), and therefore are more prone to the emergence of psychosis in later life. This is in full agreement with the hypothesis that the aetiology of schizophrenia involves the combined effects of environmental and genetic factors (Franzek and Beckmann 1996; Cardno et al. 1999; Tsuang 2000; Petronis 2004). Experimentation in animals has further confirmed that interference to normal postnatal mother–infant interactions can induce psychosis-related behavioural and pharmacological abnormalities in the offspring (Weiss and Feldon 2001; Weiss et al. 2001; Ellenbroek and Cools 2002; Brake et al. 2004).

Maternal stress during pregnancy represents a psychophysiological event that can alter postpartum maternal behaviour and thus expose the developing offspring to adverse postnatal rearing environment (Brayden et al. 1992; Maughan and McCarthy 1997; Meek et al. 2001; Patin et al. 2002). However, although immunological stimulation is strongly associated with activation of the stress response axes (Goebel et al. 2002; Haddad et al. 2002; Webster and Sternberg 2004; Silverman et al. 2005), it is virtually unknown whether gestational immune challenge is associated with alterations in postpartum maternal behaviour. Hence, the role of postnatal maternal factors in the precipitation of psychosis-related abnormalities following prenatal immune challenge still remains uncharted.

We therefore sought to dissect the relative contributions of prenatal inflammatory events and postnatal maternal factors in triggering and precipitating psychopathology in the offspring. Here, pregnant mice were exposed to the inflammatory agent polyriboinosinic-polyribocytidilic acid (PolyI:C) or vehicle treatment on gestation day (GD) 9, and the offspring derived were then cross-fostered to surrogate rearing mothers, which previously had either experienced inflammatory or vehicle treatment during pregnancy. The effectiveness of GD 9 PolyI:C treatment to induce psychosis-related behavioural, cognitive and neurochemical abnormalities in the offspring has been firmly established before (Shi et al. 2003; Meyer et al. 2005, 2006). In the present study, we aimed to elucidate the relative effects of the prenatal and postnatal factors specifically on the development and expression of latent inhibition (LI). LI is a selective learning paradigm, in which prior repeated pre-exposures to the to-be-conditioned stimulus retards subsequent development of the conditioned response (CR) following explicit parings between the same conditioned stimulus (CS) and an unconditioned stimulus (US). Deficiency in LI has been related to the attentional deficits commonly observed in schizophrenic patients (Feldon and Weiner 1992; Gray et al. 1991; Gray 1998; Weiner 2003), which may be central to the patients’ difficulties in forming, maintaining and/or retrieving selective associations between events (Hemsley 1976; Frith 1979; Nuechterlein and Dawson 1984; Perry and Braff 1994).

Here, three different conditioning procedures were used to asses juvenile and adult LI, namely conditioned freezing, conditioned active avoidance and conditioned taste aversion. This allowed us to study the relative effects of the prenatal and postnatal manipulations on the development of adult LI without any potential confounds resulting from repeated testing with the same behavioural paradigm from pre-adolescence to adulthood, and it further served to verify the generality of our findings across different conditioning procedures in adulthood.

Materials and methods

Animals

Female and male C57BL6/J breeders were obtained from our in-house specific pathogen-free (SPF) colony at the age of 10–14 weeks. Littermates of the same sex were kept in groups of 3–5 mice. Breeding began after 2 weeks of acclimatization to the new animal holding room. The breeding procedure and the verification of pregnancy have been fully described before (Meyer et al. 2005). All procedures described in the present study had been previously approved by the Swiss Cantonal Veterinary Office, and are in agreement with the Principles of Laboratory Animal Care (NIH publication No. 86-23, revised 1985).

Prenatal treatment

Pregnant dams on GD 9 received either a single injection of PolyI:C or vehicle solution (saline) via the intravenous route at the tail vein under mild physical constraint. PolyI:C (potassium salt) was obtained from Sigma Aldrich (Switzerland) and dissolved in saline to obtain the desired dosage (5 mg/kg, calculated based on the pure form PolyI:C). The volume of injection was 5 ml/kg. All animals were immediately returned to the home cage after the injection procedure.

Postnatal cross-fostering

On the day of birth [postnatal day (PND) 0], the offspring born to PolyI:C- and vehicle-treated dams were culled to litters of 6–8 animals each (with 3–4 animals per sex) and cross-fostered to surrogate rearing mothers. Male and female pups were removed from the original mother and gently placed in a cage containing sawdust bedding for a maximal total of 10 min. Half of a given litter was then placed with a PolyI:C-treated surrogate rearing mother and the other half with a vehicle-treated rearing mother. Each surrogate mother thus simultaneously fostered pups originating from both prenatal treatment conditions, but not its own offspring. PolyI:C and vehicle offsprings were marked on the left and right hind paw, respectively. A total of 24 litters (12 PolyI:C and 12 vehicle) were cross-fostered to 24 rearing mothers, half of which had been subjected to PolyI:C exposure during pregnancy and the other half to vehicle treatment only.

The cross-fostering resulted in four experimental treatment groups: (1) offspring subjected to prenatal vehicle exposure and raised by a vehicle-treated surrogate mother (SAL–SAL), (2) offspring subjected to prenatal vehicle exposure and raised by a PolyI:C-treated surrogate mother (SAL–POL), (3) offspring subjected to prenatal PolyI:C exposure and raised by a vehicle-treated surrogate mother (POL–SAL), and (4) offspring subjected to prenatal PolyI:C exposure and raised by a PolyI:C-treated surrogate mother (POL–POL).

Behavioural testing

The offspring were weaned and sexed at PND 24. The pups were weighed and littermates of the same sex were caged separately, thereby separating the animals according to the four treatment conditions. All animals were maintained under ad libitum food (Kliba, 3430, Klibamühlen, Kaiseraugst, Switzerland) and water diet, and kept in a temperature and humidity-controlled (21±1°C, 55±5%) animal vivarium under a 12:12 h reversed light–dark cycle (lights off at 0700 h).

Experiment 1 assessed LI in pre-adolescence, i.e. when the offspring reached 35–45 days of age, and only included male subjects. Each of the four treatment groups consisted of subjects derived from multiple independent litters formed by the cross-fostering procedure. Animals belonging to the same litter were, as far as possible, allocated evenly across levels of the between-subject factors of pre-exposure in each experiment, namely the non-pre-exposed (NPE) and pre-exposed (PE) conditions, in order to minimize potential confounds resulting from litter effects (Zorrilla 1997).

Experiments 2 and 3 tested LI in adulthood, i.e. when the offspring reached PND 95–135, and were conducted in both experimentally naïve animals and animals that had been previously tested in pre-adolescence (Experiment 1). Again, only male subjects were included. They were evenly distributed across levels of the between-subject factors of pre-exposure (NPE and PE), with the previous pre-exposure experience in Experiment 1 counterbalanced across the two pre-exposure conditions. The number of subjects employed in each of the behavioural tests is listed in Table 1.

Table 1 Sample size of each treatment group and experimental conditions
Full size table

Experiment 1: assessment of LI using a conditioned freezing paradigm in pre-adolescence

The apparatus comprised two sets of four test chambers. One set consisted of four metal operant boxes (30×25×29 cm; Model E10-10, Coulbourn Instruments, Allentown, PA, USA), each installed in a ventilated and sound-insulated Coulbourn Instruments chest. The animal was confined to a rectangular enclosure (17.5×13 cm) in the middle of the operant box. The other set of four chambers comprised four semi-circular (19 cm in diameter) Plexiglas enclosures, each installed in a ventilated, sound-insulated, wooden cabinet. The construction of the test chambers as well as the image analysis software used to evaluate freezing behaviour has been fully described previously (Meyer et al. 2005).

The test procedures consisted of four phases: Pre-exposure, conditioning, context test and CS test. In the pre-exposure phase, the animals were placed in the appropriate test chamber: CS-pre-exposed (CS-PE) subjects received 40 presentations of a 30-s tone stimulus at a variable inter-stimulus interval of 40±30 s; NPE subjects were confined to the chamber for an equivalent period of time without any stimulus presentation. Conditioning commenced immediately at the end of the pre-exposure phase without removing the animals from the chambers. Conditioning comprised three discrete trials of CS–US pairings. Each trial began with the 30-s tone stimulus (identical to the one used during pre-exposure) followed immediately by the delivery of a 1-s foot-shock set at 0.3 mA. Each trial was preceded and followed by a 180-s interval.

The context test took place 24 h later when the subjects were returned to the same chambers for a period of 480 s in the absence of any discrete stimulus.

The test of conditioned responding to the tone CS was conducted another 24 h later, when the animals were placed in the same test chambers again. Following an initial period of 180 s acclimatization, the tone CS was delivered and remained on for 480 s, and the amount of time conditioned freezing occurred in the presence of the tone stimulus was recorded as described before (Meyer et al. 2005, 2006). The test phase was concluded with a final 180 s CS-free period.

The data on freezing behaviour collected in the three phases (conditioning, context-test, CS-test) were analysed separately.

Experiment 2: assessment of LI using a two-way active avoidance paradigm in adulthood

The apparatus consisted of four identical two-way shuttle boxes (model H10-11M-SC; Coulbourn Instruments) as described before (Meyer et al. 2006). The internal dimensions of each shuttle box, as measured from the raised grid floor, were 35.5×18×32.5 cm. The box was separated by an aluminium wall into two identical compartments, which were interconnected by an opening (6.5×8 cm) in the partition wall, thus allowing the animal to freely traverse from one compartment to the other. The grid floor was made of stainless steel rods, 4 mm in diameter and spaced 7 mm apart (centre to centre), and was connected to a constant current shock generator (model H10-1 M-XX-SF; Coulbourn Instruments). Through the grid floor, an electric shock (0.3 mA) could be delivered. The CS was an 80-dBA white noise delivered by a speaker placed on the floor of the cubicle and behind outside the shuttle box. Shuttle response was detected by a series of photocells (H20-95X, Coulbourn Instruments) mounted on the side of both shuttle compartments. The boxes were illuminated throughout the experimental session by two diffuse light sources (1.1 W), each mounted 19 cm above the grid floor on the sidewall on each of the two compartments.

The test procedures consisted of two phases: pre-exposure and conditioning, conducted 24 h apart. Animals from each of the four treatment groups were allocated to one of two conditions: CS-pre-exposure (PE) and non-pre-exposure (NPE).

In the pre-exposure phase, the PE subjects were placed in the appropriate test chambers and presented with 100 pre-exposures to a 5-s noise CS according to a random inter-stimulus interval schedule of 40±15 s. The NPE subjects were confined to the chamber for an equivalent period of time without any stimulus presentation. The total number of shuttle responses performed during the pre-exposure session was recorded.

On the conditioning day, the subjects were returned to the same shuttle boxes and received a total of 100 avoidance trials presented with an ITI of 40±15 s. Each avoidance trial began with the onset of the white noise CS. If an animal made a shuttle into the adjacent compartment within 5 s of the CS onset, the CS would be terminated without any delivery of the shock-US and an avoidance response was scored. Avoidance failure led immediately to an electric foot-shock presented coincident with the CS. This could last for a maximum of 2 s, but would be terminated by a shuttle response during this period, which would then be counted as an escape response. To index conditioned avoidance learning, the mean avoidance response per ten-trial block was submitted to statistical analysis.

Experiment 3: assessment of LI using a conditioned taste aversion paradigm in adulthood

The apparatus consisted of Macrolon type III cages with modified grid tops, which allowed the efficient placement or removal of the drinking tubes (two per cage). The apparatus has been fully described elsewhere (Meyer et al. 2004).

The animals were kept singly in the modified cages throughout the experiment. One day after the animals were switched to single caging, they were gradually introduced to a water deprivation regime over a 5-day period to achieve 23 h water deprivation on the fifth day. On all subsequent days, the animals were allowed daily access to fluid in two 30-min periods, separated by a 5-h interval. During the first drinking period, the animals had free access to two tubes, and the second period always consisted of water only, which was presented in regular water bottles. The experimental procedures consisted of four stages: baseline, pre-exposure, conditioning and test.

Baseline

From days 1 to 4, the animals were allowed access to water in both drinking periods on each day. This served to stabilize the volume of daily water intake. The allocation of the subjects to the two pre-exposure conditions was counterbalanced in each treatment group according to the animals’ baseline performance.

Pre-exposure

On day 5, animals allocated to the NPE groups had access to water as described above. Animals in the PE groups were given access to 10% sucrose solution in both drinking tubes during the first drinking period. The amount of liquid consumption was recorded (sucrose solution for PE subjects, and water for NPE subjects).

Conditioning

On day 6, all animals were given access to 10% sucrose solution presented in two drinking tubes for a total of 30 min, followed by an intraperitoneal (i.p.) administration of lithium chloride (LiCl) solution 5 min later. LiCl was dissolved in 0.9% NaCl to achieve a concentration of 0.25 M on the day of administration. It was injected in a volume of 2% v/w body weight. The amount of sucrose intake on the conditioning session was recorded.

Test

On day 7, conditioned aversion to the sucrose taste was measured in a two-choice test with one drinking tube filled with 10% sucrose solution, and the other with water, and indexed by sucrose consumption as a proportion (in percent) of total liquid consumed during the first drinking period when the animals were confronted with a choice between sucrose and water. The data collected on the four phases were analysed separately.

Statistical analysis

All data were analysed using parametric analysis of variance (ANOVA) followed by separate ANOVAs restricted to the prenatal and/or postnatal treatment conditions whenever a main effect or interaction involving the factor of treatment attained statistical significance. Statistical significance was set at P<0.05. All statistical analyses were carried out using the statistical software StatView (version 5.0) implemented on a PC running the Windows XP operating system. Preliminary analyses indicated that the test history never interacted with any of the between-subject factors of treatment and of pre-exposure in Experiments 2 and 3, so that experimentally naïve subjects and animals having been tested in pre-adolescence (Experiment 1) were combined in the final analysis and in the presentation of the behavioural results obtained in Experiments 2 and 3.

Results

Body weights

Body weights of experimentally naïve female and male subjects measured on PND 35 and PND 95 were analysed separately using 2×2×2 (sex × prenatal treatment × postnatal rearing) randomized block ANOVA.

The analysis conducted on the body weights at PND 35 revealed no significant main effects or interactions involving any of the three between-subject factors. The overall mean body weight (±SEM) of subjects at PND 35 was 19.02±0.35 g.

At PND 95, however, the body weight of female subjects was generally lower compared to male animals, as indicated by the presence of a significant main effect of sex [F(1,64)=61.13, P<0.001]. In addition, the adoption of neonates by PolyI:C-exposed surrogate rearing mothers resulted in a decrease in their body weight, and this effect was independent of sex and prenatal treatment history. This impression was supported by the presence of a significant main effect of postnatal rearing [F(1,64)=4.34, P<0.05]. The mean body weight (±SEM) of offspring raised by vehicle-treated surrogate mothers was 27.68±0.58 g, and that of animals adopted by PolyI:C-exposed rearing mothers was 23.72±0.63 g. No other main effects or interactions attained statistical significance.

Experiment 1: assessment of LI using a conditioned freezing paradigm in pre-adolescence

Conditioning

The development of the conditioned freezing response was assessed by comparing the levels of percent time freezing at the three successive presentations of the tone CS. The percentage of time freezing was submitted to a 2×2×2×3 (prenatal treatment × postnatal rearing × pre-exposure × trials) split-plot ANOVA.

The amount of freezing increased as a function of trials in all treatment groups as evident by the overall main effect of trials [F(2,92)=125.70, P<0.001]. The increase of percent time freezing across successive trials was retarded in PE animals compared to NPE subjects, leading to a significant main effect of pre-exposure [F(1,46)=38.15, P<0.001] and its interaction with trials [F(2,92)=17.96, P<0.001]. However, the efficacy of repeated tone-stimulus pre-exposures to impede the generation of conditioned freezing to the tone (i.e. the LI effect) was enhanced in animals exposed to prenatal PolyI:C treatment regardless of the neonatal rearing condition, leading to the emergence of a significant prenatal treatment × pre-exposure × trials interaction term [F(2,92)=4.27, P<0.05]. Separate ANOVAs restricted to each of the four experimental conditions were then conducted at each of the three conditioning trials. These analyses showed that prior stimulus pre-exposures was effective in reducing the conditioned freezing response to the tone stimulus in the second trial in all experimental groups, as evident by the presence of a significant main effect of pre-exposure in each of the four groups [SAL–SAL: F(1,13)=12.46, P<0.01; SAL–POL: F(1,12)=25.75, P<0.001; POL–SAL: F(1,12)=17.60, P<0.001; POL–POL: F(1,9)=16.25, P<0.001]. However, this CS-pre-exposure effect disappeared by the third conditioning trial in animals exposed to prenatal vehicle treatment [SAL–SAL and SAL–POL animals: Fs<1.0, ns], but persisted into the final trial in prenatal PolyI:C-offspring regardless of the postnatal rearing condition [POL–SAL: F(1,12)=4.79, P<0.05; POL–POL: F(1,9)=24.24, P<0.001] (Fig. 1).

Fig. 1
figure 1

The effects of the prenatal and postnatal manipulations on the acquisition of the freezing response on the conditioning day in pre-adolescence. The development of the conditioned freezing response was assessed by comparing the levels of percent time freezing at successive presentations of the tone conditioned stimulus (CS), which was immediately followed by a 1-s foot shock. Repeated pre-exposures to the tone stimulus was effective in retarding the expression of freezing to the tone in all experimental groups, and this latent inhibition (LI) effect was particularly evident during the second conditioning trial. However, in contrast to juvenile animals that were subjected to prenatal vehicle treatment (i.e. SAL–SAL and SAL–POL offspring), the LI effect persisted into the third conditioning trial in prenatally PolyI:C-exposed subjects regardless of the postnatal rearing condition (i.e. in POL–SAL and POL–POL offspring). All values refer to mean ± SEM. Symbols (**) and (***) refer to a statistical significance of P<0.01 and P<0.001, respectively, based on separate ANOVAs restricted to each conditioning trial and experimental treatment group

Full size image

Context-freezing

The expression of conditioned freezing towards the context in which conditioning took place was assessed when the animals were returned to the conditioning chamber 24 h following conditioning. Data were expressed as percent time freezing per 1-min bin and analysed using a 2×2×2×8 (prenatal treatment × postnatal rearing × pre-exposure × trials) split-plot ANOVA.

There was a general increase in the level of freezing as a function of time, leading to a significant main effect of bins [F(7,322)=12.60, P<0.001]. However, the expression of freezing towards the context was generally low, and no other main effects or interactions involving the factors of treatment or pre-exposure attained statistical significance. The overall mean ± SEM of context-freezing per pre-exposure condition was 10.83±1.65 (NPE) and 10.90±2.19 (PE).

Tone-freezing

The level of freezing during the 3-min pre-tone period was generally low, and no significant group differences were detected in the analysis of percent time freezing over this period.

Conditioned freezing towards the discrete tone CS was assessed 48 h following conditioning. For the purpose of analysis, the percentage of time freezing during the first 3-min period of the tone presentation was expressed as a function of 1-min bins and subjected to a 2×2×2×3 (prenatal treatment × postnatal rearing × pre-exposure × 1-min bins) split-plot ANOVA.

Non-pre-exposed subjects generally showed a higher level of freezing towards the tone than PE animals, leading to a significant main effect of pre-exposure [F(1,46)=12.56, P<0.001]. However, the presence of a significant prenatal treatment × postnatal rearing × pre-exposure [F(2,92)=4.49, P<0.05] interaction term indicated that the expression of LI in the tone test was differentially affected by both the prenatal and postnatal treatment. Subsequent analyses restricted to each of the four treatment conditions revealed a significant pre-exposure effect in animals that were subjected to prenatal vehicle treatment and raised by a vehicle-treated surrogate mother (SAL–SAL offspring) [F(1,13)=6.12, P<0.05] and in animals that were exposed to prenatal PolyI:C treatment, independent of the postnatal rearing condition [POL–SAL offspring: [F(1,12)=13.12, P<0.01; POL–POL offspring: F(1,9)=12.75, P<0.01] (Fig. 2). In contrast, no equivalent pre-exposure effect was detected in animals that were subjected to prenatal vehicle treatment and adopted by PolyI:C-exposed rearing mothers due to the reduction in the freezing response displayed by NPE subjects [F<0.01, ns] (Fig. 2).

Fig. 2
figure 2

The effects of the prenatal and postnatal manipulations on the expression of LI during the test of conditioned tone-freezing in pre-adolescence. Conditioned freezing was assessed over a period of 180 s in the presence of the tone CS. For every experimental group, the freezing response is plotted as a function of 1-min bins for the initial 3-min pre-tone period, in which no explicit stimulus was presented, and for the subsequent tone-presentation period. Following the onset of the tone-CS, a significant presence of LI was observed in juvenile offspring that were subjected to prenatal vehicle (saline) treatment and adopted by vehicle-treated surrogate mothers (SAL–SAL), as well as in animals that were exposed to prenatal PolyI:C treatment, independent of the postnatal rearing condition (i.e. in POL–SAL and POL–POL animals). In contrast, a loss of the LI effect was detected in prenatally vehicle-treated offspring that were raised by PolyI:C-exposed surrogate mothers. All values refer to mean ± SEM. Symbols (*) and (**) refer to a statistical significance of P<0.05 and P<0.01, respectively, based on separate ANOVAs restricted to each experimental condition

Full size image

In order to further assess the effects of the prenatal and postnatal manipulations on the expression of the tone-freezing response per se, an additional 2×2×3 (prenatal treatment × postnatal rearing × 1-min bins) split-plot ANOVA restricted to NPE subjects was conducted. This analysis showed that the prenatal administration of PolyI:C enhanced the conditioned responding regardless of the postnatal rearing condition (Fig. 3), as supported by the significant main effect of prenatal treatment [F(1,23)=11.27, P<0.01]. In addition, the adoption of neonates by PolyI:C-treated surrogate mothers led to a marginal reduction in the expression of the tone-freezing response [main effect of postnatal rearing: F(1,23)=3.04, P=0.095] (Fig. 3). No interactions involving the two between-subject factors attained statistical significance.

Fig. 3
figure 3

The effects of the prenatal and postnatal manipulations on the expression of the conditioned freezing response in the non-pre-exposed (NPE) condition during the tone-test day in pre-adolescence. The graph illustrates the overall mean percent time freezing displayed by NPE subjects during the 3-min period of the tone presentation. Animals subjected to prenatal PolyI:C (POL) exposure displayed an enhanced conditioned freezing response compared to offspring of dams treated with saline (SAL) during mid-pregnancy, regardless of the postnatal rearing condition. In contrast, the adoption of neonates by PolyI:C-exposed surrogate mothers led to a marginal decrease in the tone-freezing response in juvenile animals. All values refer to mean ± SEM. Symbols (**) and (#) refer to a statistical significance of P<0.01 and P<0.1, respectively, associated with the main effects of prenatal treatment and postnatal rearing taken from the corresponding ANOVA restricted to NPE animals

Full size image

Experiment 2: assessment of LI using a two-way active avoidance paradigm in adulthood

Pre-exposure

The total number of shuttle responses performed during the pre-exposure session was subjected to a 2×2×2 (prenatal treatment × postnatal rearing × pre-exposure) randomized block ANOVA. This analysis yielded no significant main effects or interactions, and notably the spontaneous locomotor activity did not differ between treatment and pre-exposure conditions. The mean total number of spontaneous shuttles (±SEM) in the two pre-exposure conditions was: NPE=75±13 and PE=66±9.

Conditioning

Acquisition of conditioned avoidance responding was evident by an increase in the number of avoidance responses as a function of blocks, leading to a significant main effect of blocks [F(9,342)=145.40, P<0.001] in a 2×2×2×10 (prenatal treatment × postnatal rearing × pre-exposure × ten-trial blocks) split-plot ANOVA of this measure. Both the prenatal treatment and the postnatal rearing condition affected the development of LI in conditioned avoidance learning. This was supported by the presence of the significant pre-exposure × prenatal treatment [F(1,38)=5.11, P<0.05] and pre-exposure × postnatal rearing × blocks [F(9,342)=2.94, P<0.01] interaction terms. Subsequent analyses restricted to each of the four treatment conditions revealed a clear LI effect in animals that were exposed to prenatal vehicle treatment and raised by vehicle-treated surrogate mothers [F(1,10)=40.83, P<0.001], i.e. with the PE subjects acquiring the avoidance response slower than the NPE subjects (Fig. 4). In contrast, this effect was equally absent in the other three treatment conditions, leading to the abolition of LI in these groups [all Fs<1.0, ns].

Fig. 4
figure 4

The effects of the prenatal and postnatal manipulations on the development of LI in a two-way active avoidance paradigm in adulthood. Conditioned avoidance learning was indexed by the number of avoidance responses on the day of conditioning and expressed in successive ten-trial blocks. The embedded histograms illustrate the overall mean avoidance per block across the 100 conditioning trials. A clear LI effect was detected in adult offspring that were subjected to prenatal vehicle (saline) treatment and adopted by vehicle-treated surrogate mothers (SAL–SAL). The prenatal exposure to PolyI:C resulted in the abolition of LI in adulthood, regardless of the postnatal rearing condition (i.e. in adult POL–SAL and POL–POL animals). In addition, a loss of the LI effect was also observed in prenatally vehicle-treated offspring that were raised by PolyI:C-exposed surrogate mothers (SAL-POL). All values refer to mean ± SEM. Symbol (**) refers to a statistical significance of P<0.01 based on separate ANOVAs restricted to each experimental condition

Full size image

An additional 2×2×10 (prenatal treatment × postnatal rearing × ten-trial blocks) split-plot ANOVA restricted to the NPE condition was conducted in order to assess the effects of the prenatal and postnatal manipulations on conditioned avoidance learning per se. This analysis indicated that conditioned avoidance learning was impaired in offspring adopted by PolyI:C-treated surrogate mothers regardless of the offspring’s prenatal treatment history (Fig. 5), as supported by the significant main effect of postnatal rearing [F(1,19)=3.12, P<0.05] and its interaction with ten-trial blocks [F(9,171)=4.56, P<0.001]. In contrast, the prenatal history exerted no effects on avoidance learning in the NPE condition, as the main effect of prenatal treatment was far from significance [F<1.0].

Fig. 5
figure 5

The effects of the prenatal and postnatal manipulations on the acquisition of conditioned avoidance learning in the NPE condition in adulthood. The graph illustrates the overall mean avoidance response per ten-trial block displayed by NPE subjects across a total of 100 conditioning trials. The adoption of neonates by PolyI:C-challenged surrogate mothers (POL) retarded the acquisition of conditioned avoidance learning in adulthood compared to offspring raised by SAL-treated surrogate mothers. On the other hand, the prenatal treatment history exerted virtually no effect on conditioned avoidance learning per se. All values refer to mean ± SEM. Symbol (*) refers to a statistical significance of P<0.05 associated with the main effect of postnatal rearing taken from the corresponding ANOVA restricted to NPE animals

Full size image

Experiment 3: assessment of LI using a conditioned taste aversion paradigm in adulthood

Pre-exposure

The amount of liquid consumption (sucrose solution for PE subjects, and water for NPE subjects) was subjected to a 2×2×2 (prenatal treatment × postnatal rearing × pre-exposure) randomized block ANOVA. This analysis yielded no significant main effects or interactions, and notably the amount of liquid consumption was not affected by the pre-exposure or treatment conditions. The mean (±SEM) value of water consumption in the NPE subjects was 1.64±0.12 ml, and the mean (±SEM) value of sucrose solution intake in the PE subjects was 1.48±0.05 ml.

Conditioning

The consumption of sucrose solution was affected by the sucrose pre-exposure history regardless of the prenatal and postnatal treatment conditions. PE subjects consumed more sucrose solution on the conditioning session compared to NPE subjects, leading to a significant main effect of pre-exposure in the ANOVA [F(1,27)=7.13, P<0.01]. The mean (SEM) value of sucrose solution consumption was: NPE=1.57±0.11 ml, PE=1.98±0.06 ml. No other main effects or interactions attained statistical significance.

Test

Sucrose solution consumption as a proportion (in percent) of total liquid consumed was used to index conditioned taste aversion and the data were subjected to a three-way ANOVA as described before. Pre-exposure to the sucrose solution led to reduced conditioned taste aversion, i.e. PE subjects consumed a higher proportion of sucrose solution than NPE subjects. However, this LI effect was present only in animals exposed to prenatal vehicle treatment but was absent in animals born to dams treated with PolyI:C during pregnancy, regardless of the postnatal rearing condition (Fig. 6a). These impressions were supported by the significant main effect of pre-exposure [F(1,27)=4.56, P<0.05] and its interaction with prenatal treatment [F(1,27)=8.40, P<0.01]. Subsequent analyses restricted to both prenatal treatment conditions confirmed a significant LI effect in animals born to dams treated with vehicle during pregnancy [F(1,15)=16.24, P<0.01] but not in prenatally PolyI:C-treated offspring [F<1.0, ns]. No other main effects or interactions attained statistical significance.

Fig. 6
figure 6

The effects of the prenatal and postnatal manipulations on LI in a conditioned taste aversion paradigm in adulthood. Conditioned taste aversion was demonstrated using a two-bottle test procedure and was indexed by the preference for the sucrose solution expressed in percentage of total liquid (sucrose solution and water) consumed on the test session. a Pre-exposure to the sucrose solution led to reduced conditioned taste aversion in adult animals subjected to prenatal (saline) treatment regardless of the postnatal rearing condition (i.e. in adult SAL–SAL and SAL–POL offspring). In contrast, this LI effect was lost in adult offspring exposed to prenatal PolyI:C treatment independent of the neonatal adoption procedure (i.e. in adult POL–SAL and POL–POL animals). b The prenatal treatment did not exert any effects on drinking behaviour per se, as the total amount of liquid consumption on the test day did not differ across prenatal and pre-exposure conditions. All values refer to mean ± SEM. Symbol (**) refers to a statistically significant difference (P<0.01) between the PE animals exposed to prenatal vehicle treatment (SAL–SAL and SAL–POL) and the combined NPE animals subjected to prenatal vehicle exposure (SAL–SAL and SAL–POL), as indicated by the main effect of pre-exposure taken from the corresponding ANOVA restricted to the prenatal vehicle treatment [F(1,15)=16.24, (P<0.01)]

Full size image

Additional analyses showed that the significant pre-exposure × prenatal treatment interaction term stemmed primarily from a prenatal treatment effect in the PE condition. An ANOVA restricted to the PE subjects yielded a significant main effect of prenatal treatment [F(1,16)=8.29, P<0.05], whilst this main effect in the analysis restricted to the NPE subjects was far from statistical significance [F(1,15)=1.18, P=0.20]. This suggested that prenatal PolyI:C treatment attenuated the effects of CS pre-exposure on subsequent conditioned taste aversion learning.

An analysis of the total amount of liquid consumed (sucrose solution and water) was also carried out to examine any possible effects of the prenatal and postnatal treatment conditions on drinking behaviour per se. A 2×2×2 (prenatal treatment × postnatal rearing × pre-exposure) randomized block ANOVA of this measure yielded no significant main effects or interactions, and notably the liquid consumption on the test day did not differ across the four treatment groups or between the two pre-exposure conditions (Fig. 6b).

Discussion

The present study replicated and extended the findings from our previous reports that the maternal exposure to PolyI:C at gestation day 9 in mice led to the attenuation of LI in a conditioned freezing paradigm in the adult offspring (Meyer et al. 2005, 2006). Here, a post-pubertal loss of LI following prenatal immunological stimulation was demonstrated in both active avoidance learning (Experiment 2) and conditioned taste aversion (Experiment 3). This readily suggests that the disruptive effects of prenatal PolyI:C treatment on LI expression in adulthood can be demonstrated independently of the conditioning procedure used. Moreover, our data support the findings from previous studies in rats showing that LI deficiency following prenatal immune activation exhibits a post-pubertal maturational delay (Zuckerman et al. 2003a; Zuckerman and Weiner 2003), i.e. prenatal PolyI:C exposure results in the disruption of LI only in adulthood but not at a juvenile stage of development. Indeed, the acquisition and expression of LI during pre-adolescence was, if anything, enhanced in animals subjected to prenatal PolyI:C exposure compared to prenatally vehicle-treated offspring (see Figs. 1, 2). Thus, a maturational delay between the immunological manipulation in utero and the onset of psychopathology in adulthood exists, and this is consistent with the developmental aetiology and disease process of schizophrenia and related disorders (Jakob and Beckman 1986; Weinberger 1987; Rapoport et al. 2005). However, the present findings of juvenile and adult LI deficiency in prenatal control animals that were raised by PolyI:C-treated surrogate mothers (Figs. 2, 4) highlight the possibility that immunological stress during pregnancy may affect maternal postpartum factors in such a way that the adoption of prenatal control animals by immune-challenged surrogate mothers is sufficient to precipitate learning deficits in the offspring. Specifically, these learning disabilities do not seem to be dependent on post-pubertal maturational processes, as they are equally present in pre-adolescence and adulthood.

Our attempts here to dissect the relative contributions of prenatal inflammatory events and postnatal maternal factors in the prenatal immune activation model have demonstrated that the gestational treatment histories of the rearing mother does not modulate the post-pubertal emergence of LI deficiency in offspring subjected to prenatal immune challenge. This is because the behavioural abnormalities emerging in animals subjected to prenatal PolyI:C exposure were independent of the postnatal adoption procedure, suggesting furthermore that the prenatal immunological manipulation induces robust maturation-dependent deficits in LI. It thus appears that the adoption of prenatally immune challenged neonates by control surrogate rearing mothers does not have any beneficial effects on the later appearance of selective learning deficits. Rather, the present results provide further support for the hypothesis that inflammation-mediated disruption of foetal brain development may predispose the early-life organism to the emergence of psychopathology in later life (Gilmore and Jarskog 1997; Pearce 2001; Patterson 2002; Brown et al. 2004a, b). In addition, our results provide no experimental evidence for the possibility that additive effects may exist between prenatal immunological stress and the postnatal adoption by immune-challenged surrogate rearing mothers, at least in the precipitation of deficiency in selective learning.

However, we also clearly demonstrated a pre- and post-pubertal loss of the LI effect in prenatally vehicle-treated animals that were adopted by a PolyI:C-treated surrogate mother (Figs. 2, 4). At both stages of development, this LI deficiency was largely mediated by reduced learning in the NPE subjects, indicating that the pre- and post-pubertal loss of LI in prenatal control offspring that were raised by PolyI:C-treated surrogate mothers may readily be attributable to a general deterioration in cognitive functions as such rather than to impairments in selective attention. Hence, our findings highlight for the first time that putative changes in postnatal maternal factors as a result of immune challenge during pregnancy are sufficient to lead to the emergence of learning disabilities in the juvenile and adult offspring. Since postpartum maternal behaviour and milk quality are the two crucial factors that are transmitted from the mother to the neonates (Fleming et al. 1999; Leckman and Herman 2002; Walker et al. 2004), the experience of immunological stress during pregnancy may adversely affect the quality (and/or quantity) of these factors and thus subject the organism with an unfavourable rearing environment in early life. Previous studies have already indicated that early life adversity can affect postnatal growth and brain development, and the disruption of which is linked to the eventual appearance of psychopathological behaviour later in life (reviewed in Weiss and Feldon 2001; Cirulli et al. 2003; Pryce and Feldon 2003; Pryce et al. 2005). The emergence of a post-pubertal reduction in body weight as a function of the adoption by immune-challenged rearing mothers further supports the possibility that PolyI:C-induced immunological stress during pregnancy may have critical impact on postpartum maternal factors.

It needs to be emphasized, however, that such putative alterations in maternal postpartum factors appeared to exert distinct effects in the offspring depending on their prenatal treatment histories. This is supported by the observation that only prenatally vehicle-treated offspring that were raised by PolyI:C-exposed surrogate mothers displayed LI deficits in pre-adolescence. In contrast, prenatal PolyI:C-offspring that were adopted by PolyI:C-exposed surrogate mothers showed normal or even enhanced LI at this stage of development. Since each surrogate mother concomitantly raised pups of both prenatal treatment conditions, it is unlikely that these distinct effects on the development of LI could be accounted for by within-litter differences in maternal behaviour towards the pups, because maternal behaviour is most often directed by and transmitted to the whole litter (Fleming et al. 1999; Walker et al. 2004). Rather, the ultimate impact of putative changes in postnatal maternal factors on postnatal brain development and associated behaviour in the offspring may be influenced by the prenatal treatment history. The data here readily suggest that offspring may be less sensitive to changes in maternal postnatal factors when they had been exposed to prenatal immunological stimulation compared to offspring that had received prenatal vehicle treatment only.

Previous studies in rats have indicated that the sensitivity of juvenile LI (at PND 35) to acute dopaminergic stimulation markedly differs from its capacity to disrupt LI in adulthood (Zuckerman et al. 2003b). In contrast to adult rodents, where the efficacy of low doses of amphetamine to abolish LI is well established (e.g. Weiner et al. 1984, 1988; Russig et al. 2003; Meyer et al. 2004, reviewed in Moser et al. 2000; Weiner 2003), it has been demonstrated by Zuckerman et al. (2003b) that LI in juvenile rats is resistant to the disruptive effects of low doses of amphetamine. The disruption of LI in adult rats by amphetamine is likely to be dopamine-mediated, because the amphetamine-induced attenuation of LI can be reversed by the administration of the dopamine-receptor antagonist, haloperidol (e.g. Warburton et al. 1994; Russig et al. 2003). The failure of amphetamine to disrupt LI in pre-adolescence is in agreement with the findings that the mesolimbic dopaminergic system is not fully developed at a juvenile stage of development (Infurna and Spear 1979; Bolanos et al. 1998; Andersen et al. 2001; Spear 2000). Zuckerman et al. (2003b) went on to show, however, that LI of juvenile rats can be attenuated by serotonergic manipulations, suggesting that the serotonergic modulation of LI is already operative in pre-adolescence and is, in contrast to the dopamine-related modulatory systems, not dependent on post-pubertal maturational processes. Hence, the loss of LI in juvenile offspring that were subjected to prenatal vehicle treatment and raised by PolyI:C-exposed surrogate mothers here is likely attributable to disturbances in the central serotonergic system. It will therefore be of great interest to further elucidate the pharmacological sensitivity of juvenile (and adult) offspring raised by immune-challenged surrogate mothers in order to characterize more specifically the neurochemical basis of the loss of LI in these animals.

It follows that the neural basis of the post-pubertal loss of LI due to prenatal PolyI:C exposure (regardless of the postnatal rearing condition) may differ from that one precipitated by the adoption of prenatal control animals by PolyI:C-exposed rearing mothers. Not only has the former emerged after a post-pubertal maturational delay, whilst the latter was already manifested in pre-adolescence, but the prenatal and postnatal manipulations exerted distinct long-term effects on avoidance learning as such. Whereas the adoption of offspring by PolyI:C-treated mothers impaired adult avoidance learning in the NPE condition (Fig. 5), prenatal PolyI:C administration facilitated avoidance responding in adult CS-PE subjects. Hence, similar to the findings obtained in the conditioned freezing paradigm in pre-adolescence (Figs. 2, 3), impaired conditioning in the NPE condition has contributed to the attenuation of adult LI in the active avoidance paradigm in prenatal control animals that were adopted by PolyI:C-treaded surrogate rearing mothers. In contrast, the efficacy of prenatal PolyI:C exposure to disrupt adult LI in active avoidance learning stemmed solely from its effects on improving conditioned responding in the PE condition. The latter suggestion is in full agreement with the findings that the prenatal PolyI:C-induced abolition of adult LI in the conditioned taste aversion paradigm was also clearly mediated by a facilitation in the learning performance of prenatally PolyI:C-exposed PE animals relative to the NPE subjects (Fig. 6a).

Further support for the possibility that the nature of adult LI deficiency is dissimilar between animals subjected to prenatal immune challenge and prenatal control animals that were raised by PolyI:C-treated surrogate mothers is provided by the observation that the long-term behavioural consequences of prenatal immune activation were similarly detectable in both aversive instrumental (i.e. conditioned active avoidance) and classical Pavlovian (i.e. conditioned taste aversion) conditioning. Conversely, the efficacy of the postnatal PolyI:C-rearing condition to abolish LI in adulthood was only apparent in active avoidance learning, but not in the conditioned taste aversion paradigm. As reviewed elsewhere (Moser et al. 2000; Weiner 2003), the LI effect is demonstrable across a variety of Pavlovian and instrumental conditioning procedures, but recent experimentation in animals has indicated that the efficacy of manipulations affecting its development and/or expression is not always generalizable one conditioning paradigm to the other (see e.g. Reilly et al. 1993; Purves et al. 1995; Gallo and Candido 1995; Buhusi et al. 1998; Schmajuk et al. 2000, 2001; Oswald et al. 2002; Schmajuk 2005; Pothuizen et al. 2006). This may further indicate the possibility that distinct neural substrates are involved in the regulation and modulation of LI depending on the specific conditioning procedure used (Schmajuk et al. 2000; Schmajuk 2005). Here, we found that the adoption of prenatally vehicle-treated neonates by PolyI:C-exposed surrogate rearing mothers produced a paradigm-specific disruption of adult LI, whilst prenatal immune challenge attenuated adult LI independent of the conditioning procedure used. Hence, the use of multiple associative learning procedures has allowed us to detect a potentially important distinction between the mechanism underlying the disruptive effects of the prenatal and postnatal manipulations on adult LI, respectively.

The present study provided the first experimental evidence for the emergence of maturation-independent learning disabilities following neonatal rearing by immune-challenged surrogate mothers. The mechanisms underlying this association are virtually unknown, because the role of postnatal maternal factors have received scant attention in the link between maternal immune activation during pregnancy and the higher risk for psychosis-related abnormalities in the resulting offspring. The induction of maternal stress responses, however, might represent one way as to how maternal postnatal factors could be affected by immunological stimulation during pregnancy. Indeed, stimulation of the peripheral immune system is immediately accompanied by the activation of the stress response axes (Goebel et al. 2002; Haddad et al. 2002; Webster and Sternberg 2004; Silverman et al. 2005), and maternal physiological and/or psychological stress during pregnancy can alter postpartum maternal behaviour (Brayden et al. 1992; Maughan and McCarthy 1997; Meek et al. 2001; Patin et al. 2002). The future investigations of the potential effects of gestational immune challenge on postpartum maternal behaviour and physiology will therefore be highly relevant for the elucidation of mechanisms underlying the link between the adoption of neonates by immune-challenged surrogate mothers and the precipitation of juvenile and adult learning deficits.