Gut microbial taxa elevated by dietary sugar disrupt memory function

Juvenile male Sprague Dawley rats (Envigo; arrival postnatal day (PN) 26–28; 50–70 g) were housed individually in standard conditions with a 12:12 light/dark cycle. All rats had ad libitum access to water and Lab Diet 5001 (PMI Nutrition International, Brentwood, MO; 29.8 % kcal from protein, 13.4% kcal from fat, 56.7% kcal from carbohydrate), with modifications where noted. Treatment group sizes for Aim 1 experiments are derived from power analyses conducted in Statistica Software (V7) based on our published data, pilot data, and relevant publications in the literature. All experiments were performed in accordance with the approval of the Animal Care and Use Committee at the University of Southern California.

Twenty-one juvenile male rats (PN 26–28) were divided into two groups with equal bodyweight and given ad libitum access to (1) 11% weight-by-volume (w/v) solution containing monosaccharide ratio of 65% fructose and 35% glucose in reverse osmosis-filtered water (SUG; n = 11) or 2) or an extra bottle of reverse osmosis-filtered water (CTL; n = 10). This solution was chosen to model commonly consumed sugar-sweetened beverages (SSBs) in humans in terms of both caloric content and monosaccharide ratio27. In addition, all rats were given ad libitum access to water and standard rat chow. Food intake, solution intake, and body weights were monitored thrice-weekly except were prohibited due to behavioral testing. At PN 60, rats underwent Novel Object in Context (NOIC) testing, to measure hippocampal-dependent episodic contextual memory. At PN 67 rats underwent anxiety-like behavior testing in the Zero Maze, followed by body composition testing at PN 70 and an intraperitoneal glucose tolerance test (IP GTT) at PN 84. All behavioral procedures were run at the same time each day (4–6 h into the light cycle). Investigators were blind to animal groups when scoring the behavioral tasks such that the scorers did not know which animal was in which group. Fecal and cecal samples were collected prior to sacrifice at PN 104.

In a separate cohort of juvenile male rats (n = 6/group) animals were treated as above, but on PN day 60 rats were tested in the Novel Object Recognition (NOR) and Open Field (OF) tasks, with two days in between tasks. Animals were sacrificed and tissue punches were collected from the dorsal hippocampus on PN day 65. Tissue punches were flash-frozen in a beaker filled with isopentane and surrounded dry ice and then stored at −80 °C until further analyses.

Twenty-three juvenile male rats (PN 26–28) were divided into two groups of equal bodyweight and received a gavage twice daily (12 h apart) for 7 days (only one treatment was given on day 7) of either (1) saline (SAL; n = 8), or (2) a cocktail of antibiotics consisting of Vancomycin (50 mg/kg), Neomycin (100 mg/kg), and Metronidazole (100 mg/kg) along with supplementation with 1 mg/mL of ampicillin in their drinking water (ABX; n = 15), which is a protocol modified from28. Animals were housed in fresh, sterile cages on Day 3 of the antibiotic or saline treatment, and again switched to fresh sterile cages on Day 7 after the final gavage. All animals were maintained on sterile, autoclaved water and chow for the remainder of the experiment. Rats in the ABX group were given water instead of ampicillin solution on Day 7. Animals in the ABX group were further subdivided to receive either gavage of a 1:1 ratio of Parabacteroides distasonis and Parabacteroides johnsonii (PARA; n = 8) or saline (SAL; n = 7) thirty-six hours after the last ABX treatment. To minimize potential contamination, rats were handled minimally for 14 days. Cage changes occurred once weekly at which time animals and food were weighed. Experimenters wore fresh, sterile PPE, and weigh boxes were cleaned with sterilizing solution in between each cage change. On PN 50 rats were tested in NOIC, on PN 60 rats were tested in NOR, on PN 62 rats were tested in the Zero Maze, followed by OF on PN 64. Investigators were blind to animal groups when scoring the behavioral tasks such that the scorers did not know which animal was in which group when timing the behavior (NOIC, NOR, Zero Maze, OF). On PN 73 rats were given an IP GTT, and on PN 76 body composition was tested. Rats were sacrificed at PN 83 and dorsal hippocampus tissue punches and cecal samples were collected. Tissue punches were flash-frozen in a beaker filled with isopentane and surrounded by dry ice and cecal samples were placed in microcentrifuge tubes embedded in dry ice. Samples were subsequently stored at −80 °C until further analyses.

IP glucose tolerance test (IP GTT)

Animals were food-restricted 24 h prior to IP GTT. Immediately prior to the test, baseline blood glucose readings were obtained from the tail tip and recorded by a blood glucose meter (One-touch Ultra2, LifeScan Inc., Milpitas, CA). Each animal was then intraperitoneally (IP) injected with dextrose solution (0.923 g/ml by body weight) and tail tip blood glucose readings were obtained at 30, 60, 90, and 120 min after IP injections, as previously described23.

The Zero Maze is an elevated circular track (63.5 cm fall height, 116.8 cm outside diameter), divided into four equal-length sections. Two sections were open with 3 cm high curbs, whereas the 2 other closed sections contained 17.5 cm high walls. Animals are placed in the maze facing the open section of the track in a room with ambient lighting for 5 min while the experimenter watches the animal from a monitor outside of the room. The experimenter records the total time spent in the open sections (defined as the head and front two paws in open arms), and the number of crosses into the open sections from the closed sections.

The novel object in context task

NOIC measures episodic contextual memory based on the capacity for an animal to identify which of two familiar objects it has never seen before in a specific context. Procedures were adapted from prior reports29,30. Briefly, rats are habituated to two distinct contexts on subsequent days (with the habituation order counterbalanced by the group) for 5-min sessions: Context 1 is a semi-transparent box (15 in. W × 24 in. L × 12 in. H) with orange stripes and Context 2 is a grey opaque box (17 in. W × 17 in. L × 16 in. H) (Context identify assignments counterbalanced by the group), each context is in a separate dimly lit room, which is obtained using two desk lamps pointed toward the floor. Day 1 of NOIC begins with each animal being placed in Context 1 containing two distinct similarly sized objects placed in opposite corners: a 500 ml jar filled with blue water (Object A) and a square glass container (Object B) (Object assignments and placement counterbalanced by the group). On day 2 of NOIC, animals are placed in Context 2 with duplicates of one of the objects. On NOIC day 3, rats are placed in Context 2 with Objects A and Object B. One of these objects is not novel to the rat, but its placement in Context 2 is novel. All sessions are 5 min long and are video recorded. Each time the rat is placed in one of the contexts, it is placed with its head facing away from both objects. The time spent investigating each object is recorded from the video recordings by an experimenter who is blinded to the treatment groups. Exploration is defined as sniffing or touching the object with the nose or forepaws. The task is scored by calculating the time spent exploring the Novel Object to the context divided by the time spent exploring both Objects A and B combined, which is the novelty or “discrimination index”. Rats with an intact hippocampus will preferentially investigate the object that is novel to Context 2, given that this object is a familiar object yet is now presented in a novel context, whereas hippocampal inactivation impairs the preferential investigation of the object novel to Context 229.

The apparatus used for NOR is a grey opaque box (17 in. W × 17 in. L × 16 in. H) placed in a dimly lit room, which is obtained using two desk lamps pointed toward the floor. Procedures are adapted from ref. 31. Rats are habituated to the empty arena and conditions for 10 min on the day prior to testing. The novel object and the side on which the novel object is placed are counterbalanced by the group. The test begins with a 5-min familiarization phase, where rats are placed in the center of the arena, facing away from the objects, with two identical copies of the same object to explore. The objects were either two identical cans or two identical bottles, counterbalanced by the treatment group. The objects were chosen based on preliminary studies which determined that they are equally preferred by Sprague Dawley rats. Animals are then removed from the arena and placed in the home cage for 5 min. The arena and objects are cleaned with 10% ethanol solution, and one of the objects in the arena is replaced with a different one (either the can or bottle, whichever the animal has not previously seen, i.e., the “novel object”). Animals are again placed in the center of the arena and allowed to explore for 3 min. Time spent exploring the objects is recorded via video recording and analyzed using Any-maze activity tracking software (Stoelting Co., Wood Dale, IL).

OF measures general activity level and also anxiety-like behavior in the rat. A large gray bin, 60 cm (L) × 56 cm (W) is placed under diffuse even lighting (30 lux). A center zone is identified and marked in the bin (19 cm L × 17.5 cm W). A video camera is placed directly overhead and animals are tracked using AnyMaze Software (Stoelting Co., Wood Dale, IL). Animals are placed in the center of the box facing the back wall and allowed to explore the arena for 10 min while the experimenter watches from a monitor in an adjacent room. The apparatus is cleaned with 10% ethanol after each rat is tested.

Body composition (body fat, lean mass) was measured using LF90 time-domain nuclear magnetic resonance (Bruker NMR minispec LF 90II, Bruker Daltonics, Inc.).

P. distasonis (ATCC 8503) was cultured under anaerobic conditions at 37 °C in Reinforced Clostridial Medium (RCM, BD Biosciences). P. johnsonii (DSM 18315) was grown in anaerobic conditions in PYG medium (modified, DSM medium 104). Cultures were authenticated by full-length 16S rRNA gene sequencing. For bacterial enrichment, 109 colony-forming units of both P. distasonis and P. johnsonii were suspended in 500 µL pre-reduced PBS and orally gavaged into antibiotic-treated rats. When co-administered, a ratio of 1:1 was used for P. distasonis and P. johnsonii.

Gut microbiota DNA extraction and 16s rRNA gene sequencing in sugar-fed and control rats

All samples were extracted and sequenced according to the guidelines and procedures established by the Earth Microbiome Project32. DNA was extracted from fecal and cecal samples using the MO BIO PowerSoil DNA extraction kit. Polymerase chain reaction (PCR) targeting the V4 region of the 16S rRNA bacterial gene was performed with the 515F/806R primers, utilizing the protocol described in Caporaso et al.33. Amplicons were barcoded and pooled in equal concentrations for sequencing. The amplicon pool was purified with the MO BIO UltraClean PCR Clean-up kit and sequenced by the 2 × 150 bp MiSeq platform at the Institute for Genomic Medicine at UCSD. All sequences were deposited in Qiita Study 11255 as raw FASTQ files. Sequences were demultiplexed using Qiime-1 based “split libraries” with the forward reads only dropping. Demultiplexed sequences were then trimmed evenly to 100 bp and 150 bp to enable comparison to other studies for meta-analyses. Trimmed sequences were matched to known OTUs at 97% identity.

Gut microbiota DNA extraction and 16S rRNA gene sequencing for Parabacteroides-enriched and control rats

Total bacterial genomic DNA was extracted from rat fecal samples (0.25 g) using the Qiagen DNeasy PowerSoil Kit. The library was prepared following methods from (Caporaso et al.33). The V4 region (515F–806R) of the 16S rDNA gene was PCR amplified using individually barcoded universal primers and 30 ng of the extracted genomic DNA. The conditions for PCR were as follows: 94 °C for 3 min to denature the DNA, with 35 cycles at 94 °C for 45 s, 50 °C for 60 s, and 72 °C for 90 s, with a final extension of 10 min at 72 °C. The PCR reaction was set up in triplicate, and the PCR products were purified using the Qiaquick PCR purification kit (QIAGEN). The purified PCR product was pooled in equal molar concentrations quantified by nanodrop and sequenced by Laragen, Inc. using the Illumina MiSeq platform and 2 × 250 bp reagent kit for paired-end sequencing. Amplicon sequence variants (ASVs) were chosen after denoising with the Deblur pipeline. Taxonomy assignment and rarefaction were performed using QIIME2-2019.10.

Hippocampi from rats treated with or without sugar or Parabacteroides were subject to RNA-seq analysis. Total RNA was extracted according to the manufacturer’s instructions using RNeasy Lipid Tissue Mini Kit (Qiagen, Hilden, Germany). Total RNA was checked for degradation in a Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA). Quality was very high for all samples, and libraries were prepared from 1 µg of total RNA using a NuGen Universal Plus mRNA-seq Library Prep Kit (Tecan Genomics Inc., Redwood City, CA). Final library products were quantified using the Qubit 2.0 Fluorometer (Thermo Fisher Scientific Inc., Waltham, MA, USA), and the fragment size distribution was determined with the Bioanalyzer 2100. The libraries were then pooled equimolarly, and the final pool was quantified via qPCR using the Kapa Biosystems Library Quantification Kit, according to the manufacturer’s instructions. The pool was sequenced in an Illumina NextSeq 550 platform (Illumina, San Diego, CA, USA), in Single-Read 75 cycles format, obtaining about 25 million reads per sample. The preparation of the libraries and the sequencing were performed at the USC Genome Core (http://uscgenomecore.usc.edu/).

Data quality checks were performed using the FastQC tool (http://www.bioinformatics.babraham.ac.uk/projects/fastqc) and low-quality reads were trimmed with Trim_Galore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). RNA-seq reads passing quality control were mapped to Rattus novegicus transcriptome (Rnor6) and quantified with Salmon34. Salmon directly mapped RNA-seq reads to Rat transcriptome and quantified transcript counts. Txiimport35 was used to convert transcript counts into gene counts. Potential sample outliers were detected by principal component analysis (PCA) and one control and one treatment sample from the Parabacteroides experiment were deemed outliers (Fig. S1) and removed.

Identification of differentially expressed genes (DEGs)

DESeq236 were used to conduct differential gene expression analysis between sugar treatment and the corresponding controls or between Parabacteroides treatment and the corresponding controls. Low-abundance genes were filtered out and only those having a mean raw count > 1 in more than 50% of the samples were included. Differentially expressed genes were detected by DESeq2 with default settings. Significant DEGs were defined as Benjamini–Hochberg (BH) adjusted false-discovery rate (FDR) < 0.05. For heatmap visualization, genes were normalized with variance stabilization transformation implemented in DESeq2, followed by calculating a z-score for each gene.

For the pathway analyses, DEGs at an unadjusted P value < 0.01 were used. Pathway enrichment analyses were conducted using enrichr37 by intersecting each signature with pathways or gene sets from KEGG38, gene ontology biological pathways, cellular component, molecular function39, and Wikipathways40. Pathways at FDR < 0.05 were considered significant. Unless otherwise specified, R 3.5.2 was used for the analysis mentioned in the RNA sequencing section.

Data are presented as means ± SEM. For analytic comparisons of body weight, total food intake, and chow intake, groups were compared using repeated-measures ANOVA in Prism software (GraphPad Inc., version 8.0). Taxonomic comparisons from 16S rRNA sequencing analysis were analyzed by analysis of the composition of microbiomes (ANCOM). When significant differences were detected, Sidak post-hoc test for multiple comparisons was used. The area under the curve for the IP GTT testing was also calculated using Prism. All other statistical analyses were performed using Student’s two-tailed unpaired t tests in excel software (Microsoft Inc., version 15.26). Normality was confirmed prior to the utilization of parametric testing. For all analyses, statistical significance was set at P < 0.05. A predetermined criterion for exclusion was utilized and was based on the Grubbs Outlier Test (Prism, Graphpad Inc.) using alpha = 0.05.