Our previous work showed taxonomic differences between the gut microbiomes of healthy children compared to autoimmune children in a cohort of samples from Finland 
. In this work, the objective was to learn the metabolic potential of these bacterial communities by shotgun sequencing DNA extracted from the stool collected from children at approximately the time when they were diagnosed as autoimmune. A new approach to the analysis of metagenomic DNA is described herein which allows the statistical quantitative analysis of the functional differences between cases and controls.
The analysis methodology described herein revealed striking functional differences between cases and controls. These were seen at the levels of community processes, whole pathways as well as for individual genes. At the community level, the microbiome of the healthy children are far more functionally diverse than are the autoimmune microbiomes. For example, for nearly every major function category, as defined by the SEEDS subsystems, the relative abundance of reads was statistically higher in controls rather than cases. These major categories include amino acid metabolism, carbohydrate metabolism, RNA metabolism, DNA metabolism, cell wall and capsule proteins, nucleotides and nucleosides, cofactors and vitamins, motility and chemotaxis, nitrogen metabolism, membrane transport, phosphorous metabolism, virulence, and respiration.
The lower functional diversity in cases suggests that the case microbiomes possess more bacteria that are fastidious, requiring more nutrients in the external environment for survival and growth. If the gut epithelial layer in autoimmune children is leaky as suggested by Vaarala et al. 
, the host may be leaking more substrates into the gut than is typically seen in healthy large intestines. In contrast, the abundance of reads that map to ORFs of unknown function is statistically higher in cases than controls. So although, the vast majority of bacteria found in both control and case samples can be identified to the genus level, the case genomes are much less well characterized at the functional level. One reason for this is that there appears to be a much higher abundance of anaerobes in cases than controls. One indicator of higher anaerobicity in cases is the higher number of reads mapping to sulfur metabolism in cases. Control communities have far more genes involved in aerobic respiration while cases have far more anaerobic respiratory reductases. As anaerobic bacteria are more difficult to characterize genetically than are aerobic bacteria, it is not surprising that less is known about their biochemistry than the aerobes.
Although more reads controls are found that match known virulence determinants than in cases, there are significantly more reads in cases for some very specific virulence factors such as adhesions, Staphylococcus pathogenicity island genes, and antibiotic resistances. The forty most abundant differences between cases and controls show that stress responses, virulence factors, phages, prophages, quorum sensing, and motility genes are much more abundant in terms of the % of total reads in cases than controls (). In contrast, controls are more abundant in functions related to central metabolism such as DNA, RNA, and protein metabolism and respiration.
Forty known functions that differ significantly between cases and controls (p value≤0.01) as determined by the log of the ration between cases and controls.
One of the intriguing findings with these data relates to a possible role of butyrate production in the maintenance of gut health. Butyrate is known as an anti-inflammatory short chain fatty acid that contributes to colon health 
. The 16S rRNA mining of these data shows that many of the bacterial genera significantly more abundant in controls compared to cases are butyrate producers. In addition, butyrate induces mucin synthesis 
, decreases bacterial transport across metabolically stressed epithelia 
, and improves the intestinal barrier by increasing tight junction assembly 
. Mucin is a glycoprotein made by the host that is believed to maintain the integrity of the gut epithelium. Perspective signatures of increased mucin synthesis in the gut may be the presence of Prevotella
as both genera are known to degrade mucin 
are significantly more abundant in the controls than in the cases. Thus, a working hypothesis for a role for bacteria in preventing autoimmunity is that the presence of butyrate producing bacteria in healthy individuals may be inducing mucin synthesis in the gut, which maintains gut integrity. The presence of Akkermansia
in the gut may provide a useful, simple prediction of mucin content in the gut.
All of these data, as well as work from others in the literature, suggest a model for the role of bacteria in a healthy gut (). The total number of lactic acid producing and butyrate producing bacteria is higher in controls than in cases 
. Butyrate induces mucin synthesis 
. The higher number of butyrate producers in controls is confirmed by 16S rRNA analysis and by a higher abundance in controls of a key enzyme that catalyzes a rate-limiting step in butyrate synthesis. Mucin is well known as a glycoprotein produced by the host that contributes to gut integrity 
are much more abundant in controls compared to cases (, supporting dataset S4
). These bacteria are mucin degraders 
, often found in the human gut 
, and this work suggests that they may be useful indicators of gut integrity.
Model for a bacterial role in gut integrity leading to either a healthy state or autoimmunity for type 1 diabetes.
In contrast to controls, the relative absence of Prevotella
in cases suggests a lack of mucin on the epithelial layer of intestines of cases, and may be a diagnostic of future or current gut permeability. In addition, cases have a much larger population of bacteria such as Bacteroides
compared to controls. These bacteria ferment glucose and lactate to propionate, acetate, and succinate. Unlike butyrate, these short chain fatty acids do not induce mucin synthesis 
. Other factors may also be increasing mucin synthesis. Addition of a mixture of amino acids to the diets of dextran sulfate sodium-treated rats also increases mucin synthesis 
. The control microbiomes have a significantly higher abundance of amino acid synthesis genes than do the case microbiomes.
Although lactic acid producers may be very important in the maintenance of gut health, the fate of the lactic acid produced by these bacteria may be equally important. If that fate is butyrate production, a healthy gut seems more likely. If the microbiome of the gut encourages the production of other short chain fatty acids, gut permeability may occur. Other factors also are likely to play a role in gut inflammation such as the large amount of adhesion genes found in cases. In addition, Lactobacillus
strains can induce specific changes in the immune system of NOD mice that can increase or decrease diabetes 
. Intestinal microbes and the innate immune system also interact to affect the development of diabetes in NOD mice 
. Data presented here and published elsewhere suggest that microbial-induced butyrate production, and subsequent mucin synthesis, with a corresponding enhancement of tight junctions may contribute to the development of autoimmunity for type 1 diabetes in humans.