Animals
Twenty adult (6 months of age) male ferrets (Mustela putorius furo) were used in this study. These animals were housed separately on a 12-h light-dark cycle. Animals had ad libitum access to food and water. All animals were fed with Teklad 2072 ferret diet (Envigo, Indianapolis, IN). The manufacturer lists the following average nutrient profile for this diet: 39% protein, 19% fat, 1.2% crude fiber. All animal procedures were performed in compliance with the National Institutes of Health guide for the care and use of laboratory animals (NIH Publications no. 8023, revised 1978), and approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill and the United States Department of Agriculture (USDA Animal Welfare).
The adult ferrets used in this study were born to 6 pregnant ferrets in house. As previously described [7], at G30, i.e., day 30 of gestation, the pregnant mother animals were randomly assigned to receive either 10 mg/kg PolyIC (polyinosinic:polycytidylic acid, potassium salt, Sigma-Aldrich, St. Louis, MO, dissolved 10 mg/ml in PBS) or 1 ml/kg PBS via intraperitoneal (i.p.) injection under anesthesia introduced by 4–5% then sustained by 1–2% isoflurane. Immune activation was confirmed by elevated rectal temperature measurements taken immediately before and 3 hours after the PBS or PolyIC injection. Blood samples (0.5 ml) were also drawn from the jugular vein at the same time points. Serum cytokine level (IL1β, IL-2, IL-6, INF-γ and TNFα) was then detected by the UNC Animal Clinical Chemistry and Gene Expression Laboratory using multiplexed biomarker immunoassays (Luminex MAGPIX system, Luminex Inc., Austin, TX, using the canine cytokine kit). Ferret offspring (kits) were born at around G40-G41. The kits were kept with their mother until weaning at postnatal day 42. At this time, the male ferrets were separated for single housing [7].
Probiotic intervention
The probiotics used in this study was VSL #3 (Alpha Sigma, Covington, LA), a mixture of 8 strains of lactic acid–producing bacteria (L. plantarum, L. delbrueckii subsp. Bulgaricus, L. casei, L. acidophilus, Bifidobacterium breve, B. longum, B. infantis, and Streptococcus salivarius subsp. thermophilus) [11]. We chose to use VSL#3 since it includes several different gram-positive strains as we had no a priori hypothesis for which strain would be the most beneficial in the context of our study. VSL#3 has been investigated in several other animal models such as dogs [12] and rodents [13,14,15,16,17]. Thus, further motivation for the choice of probiotic brand was to build on these previous studies. The VSL#3 was shipped in a cooler with cool packs and was immediately refrigerated at arrival until preparation for administration. The VSL#3 suspension was prepared daily by diluting 4.5 × 109 colony-forming units (CFU) of freeze-dried VSL#3 in 10 ml Kitty Milk Replacement (KMR).
Study schedule
All ferrets received 10 ml of KMR with probiotics or KMR alone every day in the evening for 6 consecutive days. All animals were also placed on 12 h of water restriction per day during this time. Weight variations and changes in health condition were monitored and none occurred. Water restriction before administration of the probiotic mixture helped the ferrets more easily accept 10 ml of liquid without the discomfort. The 10 ml of fluid were fed to the animal with a syringe without needle. Care was taken that the animal swallowed the fluid. All adult ferrets (10 each from the PBS or PolyI:C condition) were randomly chosen to receive either the probiotic mixture or KMR alone, creating 4 groups (Probiotic:PolyI:C, Probiotic:PBS, KMR;PolyI:C, KMR:PBS).
The researchers involved with the administration of the interventions and behavioral assays were blinded to the experimental condition of all animals. This included both the identity of the animal pertaining to their assignment to the MIA or control (PBS) group as well as assignment to probiotics or control (KMR).
Microbiome sample collection and sequencing
Stool collections took place in the housing facilities of the animals. Stool samples were taken 3 times. Collection time points are labeled with letters A, B, and C throughout the manuscript. Collection A was performed 22 days before the first day of probiotic treatment as a baseline. Collection B was performed 2 days after the probiotic treatment concluded in order to avoid recording a predictably high number of bacteria sourced from the probiotic itself. Collection C occurred 9 days after the probiotic treatment concluded.
On each collection day, each single house cage was cleared of enrichment materials and litter boxes and disinfected with Vimoba 128 (Quip Laboratories, Wilmington, DE). Each sample was collected within 5 min of deposit and immediately frozen in a dry ice bath (−80o C) before further processing. The 16S rDNA amplicon sequencing was performed at the University of North Carolina Microbiome Core Facility (Chapel Hill, NC) as previously described [7]. Briefly, DNA was extracted from stool samples using MagMAX Total Nucleic Acid Isolation kit (Thermo Fisher Scientific, Waltham, MA). From this DNA the V3-V4 region of the bacterial 16S rRNA gene was amplified. This PCR product was purified and then used to generate a sequencing library for sequencing on an Illumina MiSeq instrument.
Behavioral assessments
Behavioral tests were performed around the time-point of the first sample collection (Time-point A) and after sample collection at Time Point C. No behavioral assessments were performed at Timepoint B since the goal of the study was to investigate behavioral changes after completion of a course of probiotic administration. All behavioral tests were performed in a well-illuminated white 1.5 X 1.5 m2 arena as previously described [7]. The arena was cleaned with 70% ethanol between tests on different animals, and the tests were conducted by experimenters blind to the group membership of the animals. The activity within the box was captured by a top-mounted camera as previously described (Microsoft LifeCam Cinema 720p HD Webcam, 30 Hz frame rate) [7].
The novel object recognition test is widely used to investigate recognition memory without the need for behavioral training or conditioning [18]. It has been used to demonstrate behavioral changes in rodents that are akin to ASD symptoms [5, 19, 20]. In summary, the animal was first placed in the arena for acclimation for 3 min. In the next Learning Phase, the animal was removed while 2 identical objects were placed in opposite corners of the arena allowing for 15 cm of space between the object and the walls. The animal was replaced into the arena and allowed to interact with the objects for 5 min. Two and a half hours passed before the recall phase began. In this phase, one of the identical objects was swapped with a similarly sized and weighted novel object. The animal was placed back into the arena and interactions with either object were recorded for 5 min. Recognition of the novel object is characterized as the time the animal spends interacting with the novel object minus the time spent interacting with the old one. The two types of objects used were green and black 10 lb. kettle bells with ferret toys attached to them. The identity of the novel object and the novel object location were randomized and balanced across animals of the two treatment groups.
This social interaction test has historically been used to investigate how animals interact with strange and familiar conspecifics. A three chamber paradigm test has been successfully utilized in rodent studies to observe social affiliation and social memory [21,22,23]. This test is also used to measure ASD related behavioral changes [5, 24]. In the ferret version of this task, two identical cages were placed within the arena 15 cm away from the walls. The cages are 49 × 33 × 26 cm3 (L x W x H) and have windows at all four sides and were weighed down with 10 lb. weights (Fig. 1). The animal was placed into the arena with the empty cages to acclimate for 10 min before the next phase. Once the test animal was removed from the arena, a “stranger” animal of identical sex that was not utilized in this study was placed in one of the cages. The test animal was replaced in the arena for 10 min. The sociability of the animal is determined by subtracting the amount of time the animal spends interacting with the empty cage from the amount of time it spends interacting with the cage containing the stranger. Following this phase, another stranger was placed into the other cage. The familiar and stranger animal locations on either side of the arena were randomized across the study. The interactions of the test animal were recorded again for 10 min. The social preference of the animal is characterized by subtracting the time spent interacting with the familiar animal from the time spent interacting with the stranger. All cages and the surrounding arena were cleaned with 70% ethanol between tests.
Behavioral data analysis
BORIS software [25] was used to manually log behavioral events in novel object recognition and social interaction test as previously described [7]. BORIS enables frame-by-frame analysis of a video and allows recording of the duration of interaction in each assay. In analyzing the novel object recognition video, we logged the time periods during which the animal actively explored either of the objects by pointing their noise towards the object. For the social interaction videos, we logged the timing when the animal explored either of the cages in a similar way.
Microbiome analysis
Initial analysis of the microbiome sequencing data, including quality control of the fastq files, quantification of operational taxonomical units (OTUs), and calculation of alpha and beta diversity, was performed as previously described [7]. Briefly, fastq files were quality filtered for reads with > 70% of bases having a quality score > 24 using FastQC. QIIME [26] was used to join quality filtered fastqs and align them to the GreenGenes 13 05 reference of 16S rRNA sequences, to pick and quantify OTUs (UCLUST), and calculate alpha and beta diversity (samples rarefied at depth of 1000 for alpha diversity). An empty sample was included as a control for reagent contamination. This sample generated only 136 reads and did not have any identifiable OTUs, ruling out any significant reagent contamination. Differences in groups by beta diversity were tested using QIIME’s implementation of the PERMANOVA test (p < 0.05 considered significant).
To determine whether OTU relative frequency related to animal behavior, we performed spearman correlations of OTU relative frequency at time point C with percent time novel object minus percent time at familiar object and with percent time at novel anima minus percent time at familiar animal. To determine whether the relative frequencies of OTUs and of OTUs representing probiotic organisms (genera Bifidobacterium, Lactobacillus and Streptococcus, annotated as k__Bacteria;p__Actinobacteria;c__Actinobacteria;o__Bifidobacteriales;f__Bifidobacteriaceae;g__Bifidobacterium, k__Bacteria;p__Firmicutes;c__Bacilli;o__Lactobacillales;f__Lactobacillaceae;g__Lactobacillus, and k__Bacteria;p__Firmicutes;c__Bacilli;o__Lactobacillales;f__Streptococcaceae;g__Streptococcus, respectively) were related, we performed Spearman correlations between the relative frequency of OTUs at time point C and the relative frequencies of probiotic OTUs at time point B. We used time point B for probiotic OTUs, as we found that probiotic OTU frequency in treated animals tended to peak at time point B. OTUs with a frequency of zero in more than 10 animals (half of all animals) were excluded from correlation analyses to avoid spurious correlation results.
To identify OTUs that may connect probiotic treatment with changes in behavior, we filtered for OTUs that were correlated with either time at novel animal or time at novel object and with at least one of the three probiotic organism OTUs. For all correlation tests, we filtered for a correlation coefficient (Spearman’s rho) greater than 0.3 and a p value less than 0.1. Although a p value less than 0.05 is generally considered to be statistically significant, we chose the higher p value threshold to capture correlations that were trending significance. All correlation tests and plots were generated in R. Correlation tests of significance were performed using Algorithm AS 89 as implemented in cor.test(). Plots of probiotic quantities at the three time points were generated using the ggplot2 package using the loess method to smooth the curves.