3.1 Influence of mouse genotype and dietary zinc supplementation on bacterial and fungal communities
For bacterial analyses, 70 faecal and 111 gastrointestinal samples (representing 37 ileum, caecum and colon samples per mouse) were sequenced, resulting in 24,619,426 16S rRNA gene sequence reads. After rarefying to 8,000 reads/sample, 178 (70 faecal, 35 ileum, 37 caecum, 36 colon) samples were retained and we observed 375 ASVs classified into seven bacterial phyla and 45 genera. For fungi, we sequenced 139 (33 ileum, 33 colon, 33 caecum and 40 faecal) samples, resulting in 3,909,164 ITS2 reads. After rarefying to 500 reads/sample, 57 (13 ileum, 12 caecum, 17 colon and 15 faecal) samples were retained and we observed 65 ZOTUs classified into three phyla and 11 genera.
Overall, there was considerable overlap in bacterial (Figure 1) and fungal (Suppl. Figure 1) community compositions among the four experimental groups (WT mice fed 30 ppm zinc, WT fed 150 ppm zinc, KO Shank3B-/- mice fed 30 ppm, KO Shank3-/- mice fed 150 ppm). Of bacterial 16S rRNA gene sequences that could be assigned to phylum level, the dominant phyla across the entire dataset were Verrucomicrobia (35.3 mean ± 7.8% S.D.), Bacteroidetes (31.9 ± 6.2%) and Firmicutes (22.2 ± 8.4%) (Figure 1A). The genus Akkermansia dominated across the entire dataset regardless of genotype or dietary zinc level (35.0 mean ± 8.1% S.D.), with Faecalibaculum (6.8 ± 6.5%), Muribaculum (4.4 ± 3.6%), Escherichia/Shigella (4.3 ± 2.6%) and Dubosiella (4.1 ± 3.8%) also abundant (Figure 1B). Although the majority of the detected fungal ZOTUs (67.3 mean per sample ± 39.9% S.D.) could not be classified beyond kingdom level, the phylum Ascomycota was dominant among those taxa that could be assigned (20.1 ± 31.9%), with other ZOTUs classified to the phyla Basidiomycota (9.6 ± 25.7%) and Zygomycota (3.0 ± 8.1%) (see Additional file 2: Supplementary Figure 1). At genus level, the fungi Rhodotorula (6.4 ± 22.4%) and Cyberlindnera (6.2 ± 20.0%) were most prevalent among those identified, with nine other genera ranging from 0.01-2.70% relative abundance across the data set (Suppl. Figure 1B). The lack of distinct microbial signatures between the four experimental groups, as shown by the taxonomic bar charts, was also evident in the non-metric multidimensional scaling plots for both bacteria and fungi (Figures 2A and 2B). Nevertheless, there was marked similarity among gut section communities, with the fungal faecal communities specifically notably differing along the gastrointestinal tract (Figure 2B).
Bacterial community richness and diversity varied among samples, with 74 to 148 distinct bacterial ASVs observed per sample (Figure 3A). The caecum samples (based on Jost1, Shannon and Inverse Simpson) and faecal samples (based on richness) had the highest alpha-diversity, depending on the diversity measure applied. When aggregating samples from all gut locations for a given experimental group, there was significantly higher (p<0.05) bacterial diversity in the WT compared with the Shank3B-/- KO mice (Figure 3A), irrespective of alpha-diversity metric and zinc diet. For the fungal ITS2 data, significantly higher diversity – supported by all four alpha-diversity metrics – was observed between the WT30 group (control mice fed 30 ppm zinc), compared with KO30 (Shank3B-/- KO mice also fed 30 ppm zinc) group (Figure 3B), showing that gut fungal diversity is significantly affected in Shank3B-/- KO mice. In contrast, there were no significant differences between the Shank3B-/- KO mice fed the supplemented zinc diet (KO150) and the WT30 controls, suggesting that dietary zinc may revert fungal diversity to that observed in WT mice.
To identify potential factors contributing to the observed statistically significant differences in beta diversity, we performed PERMANOVA analyses to determine whether caging, genotype, dietary zinc level, or gut sample location also play a role. For bacteria, individual cage contributed the greatest amount to variation within the microbiota data, accounting for 12.8% (p=0.001) of observed variation (Table 2). Gut sample location, mouse genotype, dietary zinc level and the genotype:zinc diet interaction were also significant (p≤0.008), explaining 11.4%, 3.9%, 2.3% and 1.1% of microbiota variation, respectively.
Table 2. PERMANOVA of bacterial ASV data table modelling cage number and the interaction between genotype and zinc diet, 999 permutations. Sample type refers to sample location along the gastrointestinal tract.
|
Df
|
|
SumsOfSqs
|
MeanSqs
|
F.Model
|
R2
|
Pr(>F)
|
Cage
|
1
|
|
1.5843
|
1.5844
|
31.8620
|
0.1284
|
0.0010
|
Sample type
|
3
|
|
1.4028
|
0.4676
|
9.4040
|
0.1137
|
0.0010
|
Genotype
|
1
|
|
0.4803
|
0.4803
|
9.6580
|
0.0389
|
0.0010
|
Zinc diet
|
1
|
|
0.2883
|
0.2883
|
5.7970
|
0.0234
|
0.0010
|
Genotype:Zinc diet
|
1
|
|
0.1324
|
0.1324
|
2.6630
|
0.0107
|
0.0080
|
Residuals
|
170
|
|
8.4533
|
0.0497
|
0.6850
|
|
|
Total
|
177
|
|
12.3414
|
1.0000
|
|
|
|
Bray-Curtis dissimilarity non-metric ordination plots showed that bacterial communities associated with ileum samples were overall less similar compared to the caecum, colon and faecal samples within a mouse individual. Despite the high overall similarity of bacterial faecal communities to caecum and colon communities in the dataset as a whole, when considering each individual mouse, caecum and colon communities were the most similar (Figure 2A). A broadly similar trend was observed for the fungal data, whereby dietary zinc levels explained 3.7% of observed variation overall (p=0.007), while gut sample location accounted for almost one-quarter (23.4%, p=0.001) of variation (Table 3; Figure 2B); these results reflect substantially greater discrimination of faecal fungal communities based on gut location than observed for bacteria.
Table 3 PERMANOVA of fungal ZOTU data table modelling zinc diet and sample type, 999 permutations. Sample type refers to sample location along the gastrointestinal tract.
|
Df
|
SumsOfSqs
|
MeanSqs
|
F.Model
|
R2
|
Pr(>F)
|
Zinc diet
|
1
|
0.6988
|
0.6988
|
2.6457
|
0.03709
|
0.007
|
Sample type
|
3
|
4.4062
|
1.4687
|
5.5609
|
0.23389
|
0.001
|
Residuals
|
52
|
13.7340
|
0.2641
|
|
0.72902
|
|
Total
|
56
|
18.8390
|
|
|
1.0000
|
|
3.2 Specific microbial taxa as biomarkers for experimental groups
We identified specific microbial taxa that were differentially abundant among the four experimental groups, using Kruskal-Wallis rank-sum and Dunn testing with adjustments for multiple pairwise comparisons. A total of 74 bacterial ASVs differed significantly in abundance between the WT30 controls and the KO30 groups (See Additional file 1: Supplementary Table 2), with the majority classified as members of the phyla Bacteroidetes and Firmicutes, and the families Muribaculaceae and Lachnospiraceae. Individual ASVs with the lowest p-values (p<3.78 x 10-9) were ASVs 182/155/135 (all Akkermansia), ASV 90 (Lachnoclostridium) and ASV 69 (Muribaculaceae). Sixty-five ASVs differed significantly (p<0.05) in relative abundance between the KO30 and KO150 groups, with 65 ASVs also differing between the WT30 and KO150 groups, including the bacterial taxa described earlier in this paragraph (See Additional file 1: Supplementary Table 2). Across all experimental groups, 24 fungal ZOTUs were identified as being differentially abundant (See Additional file 1: Supplementary Table 3). Only five of these ZOTUs could be assigned to below kingdom level: Rhizopus ZOTUs 5, 7 and 502, and Aspergillus ZOTUs 11 and 19 (See Additional file 1: Supplementary Table 3).
3.3 Effect of mouse genotype and dietary zinc supplementation on host gene expression
In total, the transcriptomes of 24 (12 ileum, 12 colon) samples were sequenced, resulting in 787,500,000 reads. Following quality control, trimming and removal of ribosomal RNA, 787,447,474 reads remained. For each sample, 60-76% (mean = 7,394,185) of reads mapped to the mouse genome. A single colon sample failed QC and was removed from subsequent analyses.
Gene expression patterns correlated strongly with sample location. Compared to the colon, ileal samples exhibited lower expression of functional gene groups related to KEGG categories Digestive system, Cell motility and Metabolism of other amino acids, while genes relating to energy metabolism were typically more highly expressed in the colon (Figure 4). KEGG functional groups of genes related to Signal transduction, Endocrine and metabolic diseases, and Neurodegenerative diseases were particularly highly expressed in all colon and ileum samples, encompassing ~ 86 - 89% of all transcripts (KEGG Level 2, Figure 4). Overall, there were KEGG group expression differences between the WT and Shank3B-/- KO genotypes as well as between the 30 ppm and 150 ppm zinc diets within a genotype, with differences most notable among colon samples. Genes involved in energy metabolism were enriched in the colon of Shank3B-/- KO mice (on both zinc diets) compared to wild type mice on the control zinc diet. Additionally, immune disease genes had decreased expression on the supplemented zinc diet (150 ppm) compared to the control zinc diet (30 ppm) in the colon (Figure 4).
Across the entire data set, a total of 1,387 genes were significantly differentially expressed based on differential expression analysis. Consistent with the clustering shown in Figure 4, ileum and colon samples were separated in two-dimensional space in MDS plots (Figure 5A), with greater diversity in the expression of these genes in the colon. By contrast, distribution of samples in the ordination did not correlate with experimental group, which encompasses zinc diet and mouse genotype (Figure 5B). When differential expression analyses were carried out on each of the two gut sample locations (with data from the four experimental groups combined per location), 1,271 genes among the ileum samples and 741 genes among the colon samples were significantly differentially expressed (data not shown). Many of these are related to antimicrobial interactions (alpha-defensins, mucin 2, carbonic anhydrase 1, intelectin) and host metabolism genes that may be regulated by gut microbiota (e.g. NADH dehydrogenases).
To investigate non-parametric differential feature (gene expression) abundance between treatments, linear discriminant analyses of effect size (LEfSe) was used. This identified 54 genes (biomarkers) that were significantly differentially expressed between the KO30 and KO150 groups (Figure 6A). For this analysis, ileum and colon samples for each experimental group were combined to find key differences regardless of sample type. NADH dehydrogenases, DNA/RNA binding proteins, zinc finger and transmembrane proteins were among the most differentially expressed gene categories. We focused our subsequent analysis on two main categories: alpha defensins, due to their importance in the control of bacteria at the mucosal barrier, and genes related to tight junction proteins, due to their integral role in mediating gastrointestinal permeability [71]. Overall, alpha defensin genes were more highly expressed in the ilea of Shank3B-/- (KO) mice - with slightly increased expression in the supplemented zinc diet KO mice - compared to the ilea of wild-type mice (Figure 6B). Expression of tight junction genes was significantly lower in the colon samples from Shank3B-/- KO mice receiving zinc supplementation compared to their wild-type counterparts on the same diet (p =0.0139), while the ileum samples showed the opposite trend with greater expression in Shank3B-/- KO mice than wild type mice (Figure 6C).