In vitro fermentation of near-isogenic lines of wild type and waxy sorghum grain with human stool microbiomes yielded distinct effects on microbiome diversity
In the first set of experiments, we examined the outcomes of in vitro microbiome fermentation reactions across a set of near isogenic lines(NILs) of sorghum derived from six different elite genetic backgrounds. Differences in amylose in the grain from each pair of NILs was estimated by measuring residual starch content after in vitro digestion (Table 1). Across the six pairs of lines used for our studies, the wild type lines yielded 3-5 times more residual starch (0.21% to 0.39%) than their waxy derivative (0.07% to 0.1%). Grain from each of these six pairs of wild type lines and isogenic waxy derivatives was subsequently used in individual in vitro microbiome fermentation reactions with human stool microbiomes from 12 donors (3 females and 9 males) with distinct baseline microbiome compositional features (Figure S1).
Compositional features of the microbiomes from multiple subjects showed highly significant treatment effects in fermentations of waxy compared to parental lines across microbiomes from multiple human subjects. These treatment effects (waxy versus wild type) manifested as differences in ecological metrics (a-and b-diversity) of the microbiomes as well as significant differences in the relative abundances of individual and groups of taxa. With respect to ecological metrics, the Shannon index (α-diversity) was significantly lower in fermentations with waxy sorghum when compared to wild type lines (Figure 1A). PERMANOVA of Bray-Curtis distances also showed significant differences in b-diversity (p < 0.001). Subsequent analysis of β-diversity in samples from each individual microbiome by canonical analysis of principal coordinates (CAP) based on Bray-Curtis distance illustrated the strength of associations between b-diversity of the microbiomes with wild type or waxy grain lines (Figures 1B and S2), and PERMANOVA of Bray Curtis distance further highlighted the statistical differences in overall microbiome composition in the microbiomes from all 12 donors (p < 0.05, Figure S2). When compared to fermentations of the wildtype parental lines, our results collectively show that the waxy lines had major effects on the overall a- and b-diversity of the microbiomes, and that microbiomes from each subject were able to differentiate substrates from wild type versus waxy sorghum regardless of the sorghum genetic background in which the waxy mutations were introduced.
Taxonomic features of the human gut microbiome from in vitro fermentation of wild type and waxy sorghum revealed shared and individualized patterns of responsiveness among different human donors
Given the highly significant effects of the waxy versus wild type parental grain phenotypes on both α and β-diversity metrics of microbial communities, we next identified specific microbial taxa associated with the treatment effects (waxy versus wild type) from each donor microbiome. Statistical significance was tested at multiple taxonomic levels. At the phylum level, significantly higher abundances of Proteobacteria and Bacteroidetes were found in fermentation of waxy lines, whereas significantly higher abundances of Firmicutes and Actinobacteria were found in fermentations from wild type lines (Paired Wilcoxon test followed by FDR correction, Figure 2A). The trends stayed much the same at increasing levels of taxonomic resolution. At the genus level, we detected three or more genera that accounted for many of the differences at the phylum level (Paired Wilcoxon test followed by FDR correction, Figure 2B). For example, within the phylum Bacteroidetes, the genera Allistipes, Bacteriodes, Parabacteriodes and an unclassified taxon in Tannerellaceae each showed the same trend (greater abundances in fermentations of waxy lines) with statistical significance in at least three of the subject microbiomes. Similarly, Sutterella, Escherichia-Shigella, and an unclassified taxon from Enterobacteriaceae were each present at significantly higher abundances in fermentations from waxy lines from microbiomes of five or more donors. Accounting for most of the significant increase in Firmicutes from fermentations of parental lines were members of the family Lachnospiraceae, and genera from this family also showed the most consistent behavior across donor microbiomes. For example, Roseburia was significantly higher in fermentations of wild type sorghum across all 12 donor microbiomes while Blautia and Coprococcus were significantly higher in nine and ten microbiomes (Paired Wilcoxon test followed by FDR correction, Figure 2B). Independent analysis of the data by linear discriminant analysis effect size analysis (LEfSe) also identified similar bacterial taxa with the greatest contribution to treatment effects (waxy versus wild type; Figure 2B). LEfSe identified Escherichia-Shigella and Alistipes as the major taxonomic groups driving fermentations of waxy lines whereas the genera driving fermentations of wild type sorghum lines corresponded to members of the Lachnospiraceae, namely Roseburia, Coprococcus, and Blautia.
Individual species of Roseburia are highly responsive to amylose content across different human donors
The most consistent microbiome-wide treatment effect (waxy versus wild type) across hosts corresponded to increased abundances of the amylolytic genus Roseburia in fermentations of grain from parental versus waxy lines. Using species-specific qPCR reactions, we confirmed the observations from the 16S rRNA sequencing data and found that this behavior was shared by three of the major Roseburia species from human microbiomes (R. intestinalis, R. hominis, and R. inulinivorans) as each of these species were significantly enriched in fermentations from wild type parental sorghum lines across microbiomes from ten or more subjects compared to fermentations from near-isogenic waxy derivative lines (Figure 3A and B). R. faecis was much less enriched, showing significant differences in only three out of twelve subjects (Figure 3A and B).
Comparisons of CAZyme glycohydrolase families (GH) found in the genomes of representative strains of these four human Roseburia species offers an explanation as to why R. intestinalis, R. hominis, and R. inulinovorans were more responsive to differences in amylose content compared to R. faecis. Genomes of R. intestinalis, R. hominis and R. inulinivorans had five to fourteen different genes encoding GH3-family enzymes associated with starch degradation whereas R. faecis carried only two genes encoding GH3 enzymes (Figure 3C). Thus, enrichment of the GH3 enzyme family in R. intestinalis, R. hominis, and R. inulinivoransmay provide a selective advantage for growth on starch-rich substrates and may explain why these three species were more responsive to treatment effects (waxy versus wild type) across individual microbiomes.
Waxy sorghum leads to reduced butyrate production in in vitro fermentations with microbiomes from multiple human donors
The significant decreases in the abundances of amylolytic, butyrate-producing members of the Lachnospiraceae (e.g., Roseburia) in fermentations of waxy lines across multiple microbiomes would also be expected to be accompanied by decreased production of butyrate, the major end-product of starch fermentation by these organisms. Measurement of the major SCFAs by gas chromatography of supernatants from the fermentations (Figure 4A) confirmed this hypothesis, with significantly lower concentrations of butyrate in fermentations from waxy lines compared to wild type (average of 24% decrease, p < 0.001, rANOVA followed by FDR correction). Concentrations of the other major SCFA (acetate, propionate, isobutyrate and isovalerate) were not significantly affected by treatment (waxy versus wild type lines).
Correlation analysis (Spearman’s correlation) further supports substantial roles for members of the Lachnospiraceae in butyrate production as the relative abundances of Roseburia(9 out of 12 microbiomes) and Coprococcus (8 out of 12 microbiomes) had some of the strongest correlations with butyrate production (Figure 4B, Spearman’s correlation with FDR correction). Notably, Butyricicoccus, Blautia,and Faecalibacterium also showed significant correlations with butyrate production in at least 5 different microbiomes (Figure 4B). Thus, while butyrate production in these fermentations is polymicrobial, members of the Lachnospiraceae family, particularly Roseburia and Coprococcus, seem to have the most significant microbiome-wide contributions to butyrate production and these taxa are known to possess pathways for fermentation of glucose to butyrate , thus suggesting that they are efficient at utilizing amylose present in the wildtype parental sorghum lines.
Resistant starch extracted from wild type sorghum lines restored amylose deficiency in waxy sorghum
Mutations affecting the synthesis of major seed components (i.e., starches in the endosperm) can also have pleiotropic effects on other major components of the seed (i.e., protein content) . We therefore used “molecular complementation” experiments to confirm that the differential microbiome effects observed in fermentation of whole grain from wild type and waxy lines were due to amylose content alone and not an unknown pleiotropic effect of the GBSS mutations on other seed components. Molecular complementation was achieved by introducing residual, digestion-resistant starch purified from wild type sorghum lines into fermentations with grain from amylose-deficient waxy lines and examining effects on microbiome phenotypes. Microbiome data from the fermentations were analyzed by comparing Bray-Curtis distance of microbiomes from i) fermentation of waxy sorghum lines alone, ii) fermentations of waxy lines supplemented with amylose-enriched (digestion-resistant) starch from wildtype lines, and iii) fermentations of wild type lines alone. The addition of the residual, digestion-resistant starch from wild type lines indeed caused significant shifts in b-diversity in the microbiomes across multiple human subjects, with microbiomes from most subjects responding to complementation with profiles that were intermediate to profiles from fermentation of parental or waxy lines (Figures 5 and S3) but demonstrating statistically significant responses to complementation (Kruskal-Wallis test followed by post hoc pairwise multiple comparisons using Dunn’s Test; Figure 5A).
We further investigated molecular complementation at higher taxonomic resolution by qPCR quantification of Roseburia species in the fermentation reactions (Figure 5C). These reactions showed significant increases in the abundance of R. intestinalis, R. hominis, and R. inulinivorans but not R. faecis in the fermentations from waxy lines complemented with resistant starch from wild type lines compared to fermentations from waxy sorghum lines (Kruskal-Wallis test followed by post hoc pairwise multiple comparisons using Dunn’s Test, Figure 5C). Molecular complementation of residual starch from digestion into fermentations with waxy sorghum lines also produced an expected stimulation in butyrate production, similar to the levels observed during fermentations with wild type lines (rANOVA followed by FDR correction, Figure 5D). Thus, introduction of amylose-enriched, digestion-resistant starch from parental lines into fermentation reactions of waxy lines promotes restoration of microbiome profiles in fermentations from waxy lines to profiles that are observed from fermentations of wild type parental sorghum lines, including stimulation of some of the most responsive amylolytic taxa (Roseburia) and concomitant changes in butyrate production. Consequently, the significant microbiome phenotypes caused by the waxy mutation in our in vitro fermentation reactions appear to be primarily dependent on the effects of waxy mutations on amylose content of the grain.
Waxy behaves as a Microbiome-Active Trait in many small grain commodities
The waxy starch trait has been developed in many small grain plant species due to its unique physicochemical properties. As with our sorghum lines, waxy lines of these other small grain species have a lower concentration of digestion-resistant starch compared to corresponding wild type lines after digestion (Table 1). To determine if the waxy trait in a whole-grain context from other species of small grains shows similar effects on the microbiome as we observed in sorghum, we compared in vitro fecal fermentations on grain derived from waxy and wild type lines of sorghum, maize, millet, rice, and wheat using the same 12 donor microbiomes for all grains tested. While the sorghum and maize grain for this experiment were derived from NILs of parental and waxy derivatives, grain from wheat, rice, and millet were not derived from isogenic pairs.
Combined data from all 12 human microbiomes showed significant reduction of butyrate production in fermentations of waxy grain from rice, sorghum, and maize with the most significant reduction occurring between waxy and wild type lines of sorghum (Paired Wilcoxon test, Figure 6A). Microbiome analysis revealed significant abundance differences in many genera (Figure 6B) and, like the butyrate production data, microbiome responsiveness to waxy and wild type lines of sorghum showed the most significant taxonomic responses based on the number of taxa showing statistically significant responses. A small number of microbial taxa showed shared responses to wild type versus waxy fermentations across multiple crop species (e.g., waxy grain from sorghum, maize, and millet all yielded significant reductions in abundances of Roseburia and elevated abundances of Escherichia based on two-way rANOVA with FDR correction). However, the taxonomic responses were largely unique to each of the crop species (Figure 6B). The unique effects of waxy wheat and rice may be due in part to the lines not being isogenic (e.g., contribution from variation at other genetic loci) and/or differences in the penetrance of the waxy mutations or differences in the physiochemical characteristics of amylose from those species (e.g., different degrees of polymerization). The important finding, however, is that human microbiomes appear to display significantly different fermentation patterns of grain from waxy versus wild type lines from each of the crop species and further experimentation is clearly warranted to understand relationships between waxy phenotypes in these crop species, physiochemical characteristics of the starches, and their impacts on fermentation by gut microbes.
The waxy phenotype in whole grain sorghum alters the gut microbiome in human microbiome-associated mice
While in vitro fermentation with human microbiomes is an excellent model for estimating the capacity of different substrates to influence the microbiome, little is known about how substrate-driven microbiome changes under in vitro growth conditions relate to the capacity of the same substrates to drive changes in the microbiome in humans or animal model systems. To address this knowledge gap, we used a human microbiota-associated (HMA) mouse model to determine if significant microbiome changes could be detected in HMA lines fed diets supplemented with 20% whole grain flour from isogenic wild type or waxy sorghum lines. Four unique HMA mouse lines were created by colonizing germ-free C57BL/6 mice with one of four human microbiomes (S766, S772, S776, and S778) demonstrating the most significant differential responses to the sorghum substrates during in vitro fermentation studies (Figure 7A). After introduction of the microbiomes, the HMA lines were divided into three treatment groups of six animals per treatment per HMA line. Among the treatments within an HMA line, one was fed a low-fiber diet while the others were fed diets supplemented with 20% sorghum from either the wild type line or the isogenic waxy derivative.
Characterization of fecal (eight time points) and cecal (terminal time point) bacterial communities of HMA mice by 16S rRNA gene sequencing over time revealed that both wild type and waxy sorghum diets increased the α-diversity (Shannon index, total ASVs, and Pielou’s evenness index) of the bacterial community compared to a low fiber control diet. There were no significant differences in the Shannon index or number of ASVs in the microbiomes of mice fed waxy or wild type sorghum diets, but a lower Pielou’s evenness index was observed in mice fed waxy sorghum (Kruskal-Wallis test followed by post hoc pairwise multiple comparisons using Dunn’s Test, Figure 7B). In contrast, constrained ordination analysis of Bray-Curtis distances showed significant differences (p < 0.05, PERMANOVA) in the b-diversity of HMA mouse microbiomes from all four human donors when fed a waxy sorghum diet compared to a wild type sorghum diet (Figure 7C). Thus, even with sorghum representing only 20% of the diet, significant treatment effects (wild type versus waxy sorghum) were detected in the microbiomes of all four HMA mouse lines.
Consistent with the changes in β-diversity, significant differences in the relative abundances of bacterial genera were also detected in each HMA mouse line fed diets containing waxy versus wild type sorghum (Paired Wilcoxon test followed by FDR correction, Figure 7E). The patterns of taxa showing statistically-significant treatment effects (waxy versus parental) were unique to each individual microbiome, a result commonly observed in experiments using HMA mice that harbor different donor microbiomes [39–41]. In HMA mice carrying the stool microbiome of subject S772, taxa that were significantly less abundant in animals fed the waxy sorghum diet included members of the Bacteriodales (Muribaculaceae), Erysipelotrichiaceae (Allobaculum), Lachnospiraceae (Agathobacter and Anaerostipes), Ruminococcaceae (Faecalibacterium), Clostridia (Christensensellaceae) and Enterococcaceae (Enterococcus) whereas abundances of Erysipeloatoclostridium (Erysipelotrichaceae) and Phascolarctobacterium (Acidaminococcaceae)were higher in mice fed waxy sorghum. Microbiomes from HMA mice harboring stool from subject S778 shared some overlapping taxonomic responses with those carrying the S772 microbiota when receiving waxy sorghum compared to wild type sorghum, including decreased abundances of Christensenellaceae and increased abundances of Erysipeloatoclostridium and Phascolarctobacterium. Another intriguing aspect of the microbiome-dependent responses was the observation that treatment effects on several of the significant taxa were not necessarily in the same direction for each microbiome, implying that the microbiome context is a major determinant of how individual microbial taxa may respond. However, we did note that three taxa, Faecalibacterium, Christensenellaceae, and Enterococcus, exhibited significant diet-driven responses in the same direction (e.g., decreased in waxy sorghum diet) across HMA mouse lines from two or more donor microbiomes.
Although dietary treatments (wild type versus waxy) had significant effects on compositional features of the microbiomes in HMA mouse lines, a notable difference between responses of the microbiomes in the in vitro fermentations compared to the HMA feeding experiment was the absence of treatment effects on Roseburia in the HMA mice.Given the strong responsiveness of this organism to wild type versus waxy grain in the in vitro experiments and the consistency of this response across multiple human microbiomes (Figure 2), absence of significant treatment effects in the HMA mice was unexpected. These disparate results are explained, however, by inefficient colonization of Roseburia species, as qPCR assays for Roseburia species (Figure S4) showed that populations of Roseburia species declined rapidly after microbiome introduction in all four HMA lines. Indeed, by day 7, populations of all four Roseburia species declined to levels near the threshold for detection, suggesting that Roseburia did not efficiently engraft and persist in ex-germ-free mice. Further study is needed to determine if poor colonization of the mouse host is a general feature of Roseburia.
In addition to the microbiome phenotypes, we also tested for treatment effects (wild type versus waxy) on feed intake and weight gain throughout the study. Remarkably, we found that animals from all four HMA mouse lines fed diets supplemented with waxy sorghum gained significantly more weight compared to their counterparts receiving either the wild type sorghum or low fiber control diets (rANOVA followed by FDR correction, Figure 7D). These weight gain phenotypes were observed in the absence of any significant treatment effects on feed intake, and are thus driven by compositional differences (e.g., amylose content) between the diets alone, most likely by more efficient digestion of the higher levels of amylopectin in feed derived from waxy lines. Studies comparing weight gain from consumption of wild type and waxy grains are limited, but one study in broilers demonstrated significant increases in weight gain in animals fed grain from waxy versus non-waxy hybrids of maize .