Ethical note
The experimental procedures were conducted in accordance with German animal protection laws. Animal facilities were approved (dated 18 April 2002) for keeping and breeding guinea pigs for research purposes by the local government authority responsible for health, veterinary and food monitoring (Gesundheits-, Veterinär- und Lebensmittelüberwachungsamt) under the licence number 530.42 16 30-1.
Subjects and housing conditions
The experiment was conducted in two batches, using descendants of outbred, multicoloured and shorthaired breeding stocks from the Universities of Bielefeld (batch I) and Münster (batch II). Due to frequent exchange of animals, the breeding stocks of both universities are rather similar genetically. Animals in batch II were moved shortly after independence from the University of Münster to Bielefeld. All adult animals and their offspring could clearly be individually distinguished by their natural fur colours and patterns which were recorded by taking pictures on the day of birth.
Animals were housed under natural light conditions with additional artificial light from 6 a.m. to 8 p.m. and a temperature of 20 ± 3°C. Wood chips were used for bedding, several plastic huts provided cover, and food and water was available at several feeding stations. Pellet food (guinea pig chow, Höveler, Langenfeld, Germany), hay, and water were provided ad libitum, and lettuce, carrots or bell pepper were given every other day. Additionally, drinking water was supplemented with vitamin C (ascorbic acid, approximately 1 g/l) once a week.
All experimental animals were housed shortly after weaning in a single large group (one group for each batch) comprising fourteen immature females and two immature males in a 15 m2 enclosure. At the time the groups were founded (spring 2011 and autumn 2011), neither females nor males had reached sexual maturity. Two months later, females were randomly assigned to one of the two different treatments (individual housing vs. group housing). In both batches, eight females were assigned to individual housing conditions while six females were assigned to group housing conditions. These females stayed together with the remaining females of their original group, resulting in one group of six females for each batch. All females were pregnant at this time. All animals were moved to new enclosures in a different room on the day the treatment groups were formed to ensure both group and individually housed females experienced a change of cages and rooms. Individually housed females were moved into standardised 0.8 m2 enclosures with wood chips for bedding, one hut for cover, a feeding dispenser and a water bottle. Females assigned to the group treatment remained together with the other females assigned to the group treatment but were moved into a new 8.5 m2 enclosure in the same room as the individually housed females, with several huts, feeding dispensers and water bottles. All animals had acoustic and olfactory contact, but individually housed females were prevented from having visual or social interaction with other animals. The six group females stayed together with the males until shortly before parturition. Two days before the estimated date of parturition (mean age of maturity + mean duration of pregnancy of 68 days), the two males were removed from the group-housed females to prevent post-partum pregnancies.
Each enclosure was subsequently checked daily in the morning and in the late afternoon for newborn pups. All females gave birth successfully, although one mother died shortly after parturition. Birth dates, litter size, mass and sexes of pups were noted within 12 hours after birth. In total, 91 pups were born: 43 in the first batch and 47 in the second batch. Six pups from two females were excluded from the experiment because they only conceived shortly before the treatment started and therefore differed strongly from the other females. Individually housed females were, on average, pregnant for 21.5 ± 4.5 days and group-housed females for 22.1 ± 5 days when placed into the treatments. There was no difference in pregnancy stage between treatments (F1,21=0.12; p=0.74) or between batches (F1,21=0.8; p=0.36). One son of a group-housed female was stillborn and only included in the analysis of birth sex-ratios, resulting in a final sample size of 84 offspring for the other measures on the day of birth. Four more offspring died before the end of the experiment (a daughter and two sons of two individually-housed females and a daughter of a group-housed female) and were included in the experiment until the day they died. These unexplained deaths were unlikely to be related to the treatments and the exclusion of these pups from the analysis slightly changed some estimates but not the overall significance of the effects.
After a female had given birth, the mother and her young were transferred into a 15 m2 enclosure with other mothers and their pups (see figure 5). Prior to this transfer, we placed one mother and her two pups not belonging to the experiment in the enclosure so that the first experimental mother would not be alone. For the consecutive mothers, the group consisted of all the other mothers and pups present in the enclosure at the time of transfer. The mean time from separation of males until parturition was 12 days, ranging between 2 and 18 days, and females gave birth on average 45.9 ± 4 days after they were assigned to the housing conditions.
There were slight differences in the initial formation of the groups between the two batches. In batch I, the fourteen nulliparous females and two males were housed together when all animals were approximately 30 days old, ranging between 25 and 35 days of age. The females were derived from eleven litters, the males from two additional litters. Female guinea pigs born in spring mature at around 45 days, as do their wild counterparts [61]. Male guinea pigs mature at between 50 and 70 days of age [62]. Thus, the group was given approximately one month to form stable bonds between individuals before animals started reproducing. In batch II, all females used for breeding came from different litters and the fourteen females were put together shortly after weaning, as in the first batch. The two males were added to the females two weeks later than in batch I because we first genotyped all individuals at 12 variable microsatellite loci [30, 63] to select the males least related to each other and the females. Both female and male guinea pigs mature later in autumn than in spring [61]. Consequently, the time the group remained together before onset of reproduction was approximately the same as in the first batch, as confirmed by the fact that the stage of pregnancy did not differ between treatments (see above). Otherwise, the experimental procedures, including housing conditions, were the same as in the first batch.
Behaviour
We assessed anxiety/emotionality of all pups born during the experiment using a struggle test, hand-escape test and novel environment test [46] (for details of the tests see below). We assessed boldness by means of a novel object test and sociability through a social separation test. Struggle duration, hand-escape latency and exploration behaviour in an open field test were correlated in the wild congener of the domesticate guinea pig, the wild cavy, indicating that all of these behaviours may reflect the same underlying behavioural category [46]. Hand-escape latency and open field behaviour were also correlated to cortisol baseline levels, hence potentially representing a measure of anxiety [45, 46]. Furthermore, struggle duration and hand escape latency showed temporal consistency even in two to three-day-old juveniles, suggesting that these tests are suited to measure behavioural phenotype in early life-stages. Approaching a novel object is often regarded as boldness or curiosity [64] and represents a different behavioural category than anxiety or emotionality in guinea pigs [65]. To measure how pups of the different treatments react to a brief social separation from their mother, we adopted a well established separation paradigm, in which guinea pigs of all ages react strongly with behavioural, physiological and immunological changes [66]. To test whether the behaviours that we measured represent different behavioural categories or the same underlying trait, we correlated the variables of each test with each other (see table 1). We correlated only the first measurements of each test with each other to exclude any possible influences of habituation or carry-over effects between repeated test situations. Most variables were not correlated with each other even when measured during the same test (e.g. social separation test). In cases where we found significant correlations (3 cases), the strength of correlation was only moderate (ranging between 0.3 and 0.43), so that we decided to analyse each behaviour separately rather than computing composite variables.
Measurement schedule
(see table 2 for overview)
Day of birth
Body mass, length and behaviour of all pups in a litter were measured within the first twelve hours after birth. If they were still wet when found, the first measurements were taken at the subsequent check when they were completely dry and able to walk, indicating that at least two hours had passed since birth. As mothers were accustomed to the observers, they did not flee or show signs of distress, and pups could be removed calmly from their side.
Struggle test
Each pup was removed singly from its mother, gently turned and held on its back in the hand of the observer and the total duration the animal's struggle during thirty seconds was recorded by starting and stopping a stopwatch each time the animal started and stopped moving (duration of struggling). Afterwards, each pup was weighed and its length (from snout to tail bone) was recorded. After the measurements, the pup was placed in a dark transport box. After all juveniles had been measured, the mother was also put into the transport box, and the family was transported to the room and enclosure that contained all other pups and their mothers.
Hand-escape test
In the second batch, an additional behavioural test was conducted immediately before the family was released into the new enclosure. For this test, the mother was placed beneath a hut in a small compartment (1 x 0.5 m2, see figure 5) separated from the main enclosure by an opaque plastic wall. The pups were then removed one by one from the transport box and placed into the compartment on the observer's outstretched hand. From this vantage point they could see their mother at a distance of approximately 60 cm. We then measured the time they took to leave the hand (latency to leave the hand). If the pup did not leave the hand within sixty seconds, it received the maximum score and was placed beside its mother. The dividing wall was removed after all juveniles were released so that the whole family could move into the main enclosure.
Day after birth
Between 9:30 and 11:00 in the morning, each pup was caught individually and two saliva samples were taken for later cortisol analysis using a Q-tip (cotton-swab). Sampling started less than a minute after disturbing animals for capture. During the sampling, the pup was placed on the lap of the observer. Both sides of the Q-tip were used to collect two consecutive samples within a few minutes that were analysed separately and then averaged. The Q-tip was cut in the middle and placed with the cotton part downwards into a 0.5 ml eppendorf tub that had been pierced at the bottom. The pierced eppendorf tube containing the Q-tip was placed into a 2.0 ml eppendorf tube and stored on ice for up to 1 hour before further processing (see below).
Social separation test
This test was performed one day after birth in the afternoon (between 2 pm and 4 pm). For testing, a smaller compartment was temporarily separated off within the main enclosure (see figure 5), and pups were caught one after the other for testing. As soon as a pup was placed in the test arena, we measured the cumulative duration of activity for thirty seconds (activity) with a stop watch. We scored as activity each movement that resulted in a change of body position (excluding head movements). Afterwards, we counted the number of calls the pup emitted within thirty seconds (number of calls). After that, we removed the wire mesh wall and measured the time it took the pup to reach its mother (latency to reach the mother). One pup made no attempt to reach its mother within three minutes, so the test was stopped and the maximum latency of 180 s was used for analysis.
One-week-old pups
Each juvenile was caught singly in the morning (9:30 a.m. to 11:00 a.m.) when it was seven days old. First the struggle test was repeated, then body mass and body length were measured. In batch II the hand test was also repeated. For practical reasons the social separation test was conducted again on the afternoon of day seven only in batch I.
Novel environment test and novel object test
Instead of the social separation test, in batch II we measured exploration (duration of activity) of a novel environment and boldness (approach to a novel object) on day eight. The novel environment test lasted for two minutes and was conducted in the same room where the animals were housed. The test animal was removed from the group and placed into a 0.5 m2 box with fresh wood chips covering the floor. The duration of activity within the first two minutes was measured, scoring each movement that resulted in a change of body position (excluding head movements). After two minutes, the animal was randomly placed in one of the corners of the box and a novel hut with a different shape and colour than the huts the animals were familiar with was placed in the opposite corner of the box. As a measure of boldness, we recorded how long it took the animals to touch that novel object using a maximum latency of two minutes. Unexpectedly, a large proportion of animals did not touch the novel object within the two minutes. We therefore assigned a score of 1 or 0 for the analysis, depending on whether or not the animal touched the novel object within the two minutes.
During the morning of day 8, each pup was again caught singly and another saliva sample was taken as described above. After the initial sample, the pup was introduced into a 30 x 40 cm plastic box from where it could hear and smell other animals but not see them. It was left there for 30 minutes before we took a second saliva sample to measure cortisol response to social separation.
Weaned juveniles (day 21 and day 35)
When pups were 21 days old (around weaning) and 35 days old (independence), body mass and body length were measured (between 9:30 a.m. and 11:00 a.m.). In the second batch the novel environment test was conducted again on day 21, except for two animals that were overlooked.
One of three trained observers took measurements and made direct observations of behaviour according to a pre-planned schedule that was unbiased with respect to the experimental treatment and observer. To ensure uniform recording of behaviour, we used unequivocal measures established during previous experiments. Observers were always visible to the animals during testing. Observers knew which pups belonged to which experimental treatment on the day of birth when they removed pups from their individually housed or group-housed mothers. Afterwards, all pups were housed in one large group and observers were not aware of the experimental treatment of individuals during testing.
Cortisol measurements
All samples on a given day were immediately stored on ice for up to 1 hour. They were then centrifuged for 10 minutes at 5000 rpm to spin the saliva from the cotton through the pierced small eppendorf tube into the larger eppendorf tube. All samples were frozen at -20° Celsius until further analysis by enzyme immunoassay, following the instructions of the manufacturer (Demeditec DES 6611, sensitivity 14 pg/ml and cross-reactivity with all tested endogenous steroids < 2%). For the assay, thawed saliva samples were centrifuged again and 5-20 µl samples were diluted in 220 µl phosphate buffer (pH 7.0, containing 1% bovine serum albumin). 100 µl of the dilution were assayed against a cortisol standard diluted in phosphate buffer in six steps of 1:3 dilutions ranging from 20 ng/ml to 82.3 pg/ml (intra-assay coefficient of replicate samples was 16.3% and inter-assay coefficient of variation was 20.4 %). All measurements were well above the detection limit of the assay (lowest amount measured was 7 ng/ml). No saliva sample could be obtained for eight pups on day 1 (four offspring each from individually-housed and group-housed mothers), for five pups on day 8 for the baseline (three from individually-housed mothers and two from group-housed mothers) and for eight pups for the cortisol response (five from individually-housed mothers and three from group-housed mothers).
Statistical analyses
Analysis of litter size, pup growth, and sex-ratio
Data were analysed using mixed-effect models in R 2.13.1 [67] with restricted maximum likelihood estimation and gaussian error distribution, unless stated otherwise. To control for common genetic or environmental effects unrelated to the treatment and avoid pseudoreplication, we included mother ID as a random effect in the models and individual ID nested within mother ID whenever there were repeated measures from the same mothers or the same offspring. Residuals of the models were checked visually for distribution and variance homogeneity by using Q–Q plots. Batch was included as a fixed factor but turned out to be significant only for body weight and cortisol levels (pups from the second batch weighed less, they had higher baseline levels of cortisol and a reduced cortisol response) and was therefore removed from all other models.
The effect of treatment during pregnancy on litter size was analysed by a t-test. Weight and size of pups were also analysed with a lme, including age, age2 and age3 (for offspring weight only) as covariates to model the changing slope of the growth curve. Additionally, treatment and offspring sex were included as fixed effects and offspring ID nested within mother ID as random effects. Offspring sex-ratios were analysed with a generalized linear mixed-effect model (package lme4), using a binomial error distribution and including treatment as a fixed effect and mother ID as a random effect.
Baseline cortisol was analysed for the day after birth in a model including treatment, sex, and batch as fixed factors, and offspring ID nested within mother ID as random factors. For day 8 we analysed the cortisol baseline level as described above and the cortisol response (the level of the second sample divided by the level of the first sample).
Behavioural measures
Six behavioural measures (struggling, latency to leave the hand, activity and number of calls emitted by pups during the social separation test, latency to mother, and exploration activity) were analysed using lmes with gaussian error distribution and restricted maximum likelihood estimation. Latency to mother and struggling were log-transformed to normalize the distributions (after adding 1 to all latency values to allow transformation of zeros). For analyses of the latency to leave the hand, separate variances were estimated for each group (using the varIdent function of the package nlme) since the pups from individually housed mothers were far more variable than pups from group-housed mothers. Approach to novel object in the novel environment test was analysed using a generalized linear mixed-effect model with binomial error structure. For all variables, we analyzed treatment, sex, and their interaction as main effects, while mother ID and offspring ID nested within mother ID were random effects. Data are shown as mean ± SEM, and a p < 0.05 was regarded as significant.