Ever since the publication of the ecologic study linking insufficient breast-feeding to T1D incidence over time (43), numerous studies have been conducted on breast-feeding and T1D, and still the role of breastfeeding initiation, and whether an optimal duration of breastfeeding exists, remains unclear. Prospective studies of children at increased risk for T1D have reported inconsistent results regarding the association between breast-feeding and T1D (44–47). A large, population-based study of two birth cohorts in Norway suggests that while initiation of breastfeeding may reduce the risk of T1D, among those who were breastfed there does not appear to be a benefit of prolonging breastfeeding on the risk of T1D (48).

While the findings regarding breastfeeding and T1D may be equivocal, breast milk has a profound impact on the gut microbiota of the infant, as reviewed in Li et al. (49). Breastfeeding could lead to a healthier infant gut microbiota at an early stage of life compared with infants who are never breastfed (50, 51). As mentioned previously, the TEDDY study (21, 22) found that breast milk was the most significant factor associated with the structure of the developing microbiota. Current breast-feeding was associated with higher levels of Bifidobacterium species (B. breve, B. bifidum, and B. longum). B. longum subsp. infantis is a particularly important degrader of human milk oligosaccharide (HMO), and a genetic strain that is specific for the utilization of HMO within a subset of B. longum was present only in samples of breast-fed infants in TEDDY (21). Similarly, in a small Korean study, an increase in Bifidobacterium was observed in newborn infants that were breast-fed, whereas an increase in Klebsiella and Serratia was detected in newborn infants fed infant formula, suggesting a beneficial impact of breast milk (52).

The cessation of breast-feeding results in a functional shift in the gut microbiota (49) and a faster maturation of the gut (22). However, even with this shift, the impact of breastfeeding on the microbiota is still seen in 6-year old children, suggesting a long-term impact of exposure to breastmilk. The relative abundance of genera belonging to Firmicutes and Actinobacteria (and in particular Bifidobacterium, the predominant genus of Actinobacteria), was significantly different at 6 years of age between children that had been exclusively breast-fed for 4 months and those that had not (53). Bifidobacterium are important species in the healthy gut microbiota that can partially breakdown complex starches and oligosaccharides, making them available via “cross-feeding” to support butyrate producing bacterial populations such as Roseburia sp., Eubacterium hallii, and Anaerostipes caccae (54).

Age at introduction of complementary foods in infants has been examined regarding IA and T1D risk. Prospective observational studies suggest that late introduction of cereals (46) or gluten (47) is associated with increased risk of T1D and IA, respectively. However, an intervention in which gluten introduction was delayed to 12 months, vs. 6 months, showed no effect on IA or T1D risk (55, 56). It has been hypothesized that introduction of complementary foods impacts health by altering the microbiota. Introduction of any complementary foods at 3 months of age (vs. later) was associated with increased Bilophila wadsworthia and Roseburia in the infant's gut microbiota after adjustment for breast-feeding (57). However, others suggest that that the maturation of the gut microbiota is driven by the cessation of breast milk rather than the introduction of complementary foods (22). Using samples collected from 44 children from the German BABYDIET study, Endesfelder et al. showed that children with the most complex diet at 6-months of age were dominated by Bacteroides and this Bacteroides dominated community was associated with later development of IA. This suggested that fast-tracking of a more complex and adult-like microbial community may be related to establishment of taxa associated with T1D risk (58).

Researchers have long been interested in gluten intake as a risk factor for T1D due to the similarities between and co-occurrence of T1D and celiac disease in individuals and families. The Diabetes Autoimmunity Study in the Young (DAISY) examined early childhood intake of gluten and did not find an association with either development of IA or progression from IA to T1D (59). Related to gluten intake is the intake of dietary fiber. TEDDY reported that intake of soluble fiber was not associated with the development of IA or T1D (60). However, Hakola et al. (61) found that in Finland increased early childhood gluten and dietary fiber intake were both associated with increased risk of IA, though they were not associated with T1D risk. A large nationwide cohort study in Norway confirmed the association between early childhood intake of gluten and a higher risk of T1D (62), but was not able to address the dietary fiber association.

Studies in normal volunteers show that introduction of gluten-free diets changes the composition of the microbiota (63). This gluten-induced dysbiosis may activate an inflammatory pathway that could lead to autoimmune diseases such as T1D (64, 65). There are no data in humans, but NOD mice on a gluten-free diet showed decreased Bifidobacterium, Tanerella, and Barnesiella species and increased Akkermansia species in feces, as well as a reduced diabetes incidence compared with mice not on a gluten-free diet (66).

Two prospective cohorts of at-risk children (DAISY and DIPP) have shown that increased intake of omega-3 fatty acids (67, 68) or increased biomarker of omega-3 fatty acid status (67, 68) were linked to a reduced risk of IA. Omega-3 fatty acids are associated with alteration of the gut microbiota (69). A trial involving omega-3 fatty supplements for 8 weeks resulted in increases in the Clostridiaceae, Sutterellaceae, and Akkermansiaceae families, which then declined when supplementation was stopped (70). Trials in individuals at risk for metabolic syndrome (71) and with new onset type 2 diabetes (72) showed an impact of fatty acid supplementation on the composition of the microbiota. While it is not clear whether these changes would have an appreciable impact on IA or T1D risk, one can hypothesize that they may lead an increase in the production of anti-inflammatory compounds (69) that may affect the development of IA and T1D.

Vitamin D is obtained both from the diet and synthesis in the skin, and then metabolized to 25-hydroxyvitamin D (25OHD) in the liver. Higher plasma 25OHD levels were associated with lower risk of IA in both TEDDY (73) and in the TRIGR Ancillary Study (74), but not in other cohort studies of at-risk children (75–77). In children with IA, lower 25OHD levels were not associated with increased risk of progression to T1D (75, 78), suggesting that the protective role of vitamin D occurs early in the disease process. Vitamin D deficiency may contribute to autoimmunity via effects on the composition of the microbiota, although existing evidence is limited to animal studies or small human studies (79). Alternatively, vitamin D may act via direct effects on immune function rather than on the microbiota (80). Studies examining vitamin D intake and the gut microbiota have been inconsistent (81–83). A genome wide association study (GWAS) showed that the vitamin D receptor (VDR) gene was associated with beta diversity of the gut microbiota (84). How vitamin D affects the microbiota in this instance is not clear, although it is likely through an indirect mechanism as bacteria do not express the VDR.

Increased intake of total sugars, and higher glycemic index of the diet are associated with increased risk of progression to T1D in autoantibody positive individuals (85, 86). A standard deviation (i.e., 49.7 g) increase in total sugar intake per day was associated with a 1.75-fold increased risk of progression to T1D after an average of 10.5 years (86). Sugar intake and glycemic index were not associated with initial development of IA, suggesting that these factors are only related to acceleration of the later stages of T1D. Potential mechanisms by which dietary carbohydrates and sugars may have an effect on T1D include metabolic (87, 88), and immunological pathways (89, 90), as well as pathways involving the microbiota, as reviewed in (91). Studies suggest that the high amount of simple sugars in our diet has lasting and detrimental effects on our microbiota (92–94). It is hypothesized that the replacement of microbiota-accessible carbohydrates (i.e., fiber) with the fats and simple carbohydrates that are characteristic of the Western diet may be negatively shaping the microbiota thereby leading to poor health outcomes.

Several strategies are being considered to attempt to remodel the gut microbiota in T1D with an aim to improving immunological tolerance and beta cell function. Live bacteria (termed probiotics) can be introduced with a goal to either colonize the gut with a species that is absent, to increase the abundance of an existing species, or to provide beneficial metabolites without colonization. Alternatively, the goal may be to increase the overall community diversity by introducing a large bolus of diverse microorganisms with or without depletion of the existing microbiota as is the case with fecal microbiota transplantation (FMT). Another major approach under investigation is the use of “prebiotics” and diet-based therapy where the goal is to provide a food source to encourage the growth of beneficial bacteria. Alternatively, dietary supplements can be designed to directly deliver these bacterial metabolites themselves (95). The advantage of prebiotic approaches is that they do not rely on bacterial colonization, which may be ineffective if the T1D-associated intestinal environment is resistant to the introduction of beneficial bacteria. This has been shown to be the case in NOD mice, which were resistant to passive introduction of new species (96). T1D-associated genetic risk factors are thought to contribute to altering the intestinal immune environment, making it less hospitable to certain bacteria (38, 40, 97). Another approach is to identify the major modifiable lifestyle factors that lead to the establishment of a dysbiotic gut microbiota, but even this may need to be supported by one of the other strategies above, and lifestyle changes may be difficult to implement.

The notion of modulating gut homeostasis with supplementation of tolerogenic commensal microbes offers the potential for a safe method of intervention in the disease prevention setting. TEDDY reported an association between decreased risk of IA and early supplementation of probiotics (between the age of 0–27 days) in children with the highest risk HLA genotype (DR3/4), whereas the responses were highly variable in other age groups (98). These data suggest that early intervention with probiotics in infancy in T1D-susceptible populations, when the microbiota is being established, may be protective of future disease through probiotic colonization. However, this study did not look at dose or type of probiotic consumed, so more rational approach should be taken to identify disease protective probiotic strains. A number of Lactobacillus and Bifidobacterium species considered “Generally Regarded as Safe” (GRAS) microorganisms are commonly used in dietary supplements as probiotics worldwide and could potentially be trialed in T1D-susceptible newborns.

The use of probiotics for the prevention of T1D was initially tested with heat killed Lactobacillus casei BL21 in the NOD rodent model (99). It was found that the onset of diabetes was significantly higher in the control group than in the group that received L. casei (99). Similar results were obtained by weekly feeding the VSL#3 formulation to NOD rodents (100). VSL#3 is a formulation containing three species of Bifidobacterium (B. longum, B. infantis, and B. breve), four species of Lactobacillus (L. acidophilus, L. paracasei, L. delbrueckii subsp. bulgaricus, and L. plantarum), and Streptococcus thermophilus. The decrease in T1D onset was positively correlated with lower T cell infiltration in the pancreas in the VSL#3 treated group, suggesting a direct mechanistic role of the microbiota in T1D. It was also observed that VSL#3-treated mice was associated with increased IL-10 mRNA expression and with the presence of IL-10-positive infiltrating mononuclear cells (MNCs) in the pancreas. Clinical trials are underway to test the role of the VSL#3 probiotic formulation (Visbiome probiotic, ClinicalTrials.gov identifier NCT04141761, Table 1) on children and adolescents with T1D. In a randomized clinical trial designed to test for the prevention of allergy in a high-risk cohort, VSL#3 provided during the first 6 months of life was not related to the development of T1D (amongst n = 14 cases) or development of islet autoantibodies (amongst n = 25 cases) during the 13-year follow-up (101). However, as this study was not powered to investigate T1D these results should be viewed with caution. Furthermore, a drawback of this formulation is the inability to determine which of the strains within the formulation is the one exerting a beneficial effect.

Through interventions with different probiotics in different rodent models of disease, it has been observed that the mechanisms of action of probiotics are diverse and strain specific. The effects range from immunomodulatory to immunostimulatory (102). The strain specificity is usually overlooked; however, it is probably the biggest factor to consider in most animal and human studies looking to use probiotics for the prevention or treatment of a disease. In addition, many commercially available probiotic formulations are selected based on their robustness in industrial production and shelf life, instead of their beneficial effects on human subjects. Currently, many clinical trials are underway in order to be able to add health claims into their labels. It is possible that the variability observed on the TEDDY cohort may be due to the use of a diverse array of probiotic products not specifically selected for T1D prevention. Consequently, there is a significant need to identify bacterial strains that are associated with a beneficial health outcome. The availability of large microbiota datasets offers the possibility of (i) testing if GRAS microorganisms are negatively correlated with disease onset and (ii) isolate and test unconventional microbial species that may be used to prevent T1D onset.

Most traditional probiotic bacteria originate from food, typically fermented dairy products or are previously isolated strains now currently used as commercial supplements. For example, Groele et al. plan to trial two common probiotic strains Lactobacillus rhamnosus GG and Bifidobacterium lactis Bb12 in new-onset T1D (111). Probiotic bacteria are not necessarily well-adapted to colonization and modulation of the human gut and other strains, termed “next-generation probiotics,” that are indigenous to the human microbiota may be more successful at inducing long-term benefits in T1D. When looking to identify which species should be introduced, it is important to consider the complex inter-relationships exist between species with different functional roles that support the growth of other community members. For example, one species may produce metabolites or substrates required to support the growth of another species, known as cross-feeding. One such interaction occurs when butyrate producers such as Faecalibacterium prausnitzii stimulate production of mucin by epithelial cells which in turn promotes the growth of mucinophiles such as Akkermansia. Mucin degradation and fermentation by Akkermansia releases oligosaccharides and acetate which in turn further supports the growth of butyrate producers in a feedback loop (112). This has led to the concept that increasing the abundance of a single species such as Akkermansia could then enhance the abundance of other beneficial species resulting in a broader beneficial impact. Indeed, some studies have suggested that a reduction in Akkermansia may be associated with IA or T1D (22, 58). This concept was tested by introduction of Akkermansia muciniphila by oral gavage into NOD mice resulting in an increase in mucin production by goblet cells in the colon, while reduced expression of the inflammatory marker Emr1 and increased expression of the anti-inflammatory marker Ym1 was also observed (113). The combined effect of these changes included increasing immune regulatory factors (increased IL-10 and TGF-beta in the pancreatic LN and Tregs in the islets) and delayed diabetes onset. It is likely that many more species with similar functions may also protect from T1D, at least in animal models. Recent progress in the ability to monoculture many more members of the gut microbiota community means that efforts are now ongoing aiming to systematically screen for taxa that can prevent T1D. These efforts may then identify candidates which can be tested for beneficial effect in human T1D.

FMT is the practice of transferring the total bacterial population from a healthy donor stool sample (either by rectal or oral routes into the gastrointestinal tract) with a goal of remodeling the recipient's microbiota for therapeutic benefit. This approach has gained traction due to dramatic benefits found using FMT for treatment of recurrent Clostridium difficile infection (114). In C. difficile, FMT was shown to significantly increase the diversity of the recipient's microbiota 2-weeks after transfer. Using genetic approaches to track donor derived microbial strains, Li et al. showed that donor derived bacteria persisted for 3-months after FMT (115). However, colonization was more successful for species that were already present in the recipient's own microbiota, supporting that host compatibility is a barrier to introduction of new taxa. In a study of individuals with metabolic syndrome, FMT from lean donors induced short-term changes in the recipients gut microbiota composition and improved insulin sensitivity and decreased glycated hemoglobin levels at 6 weeks (116). Taxa that were increased at 6 weeks included acetate producer Bifidobacterium pseudolongum, lactate-producing Lactobacillus salivarius, butyrate-producing Butyrivibrio, Clostridium symbiosum, and Eubacterium species. This suggests that modest benefits to glycemic control can be gained by FMT induced remodeling of the gut microbiota, but that strategies to increase the durability of the effects are needed.

The first study of FMT in T1D was recently published (41). Nieuwdorp and colleagues completed a trial of allogeneic vs. autologous FMT in individuals with new-onset T1D (Netherlands Trial Register identifier NL3542, Table 1). They found that the participants that received back their own autologous stool sample had preserved beta cell function (stimulated C-peptide) compared with those that received FMT from an allogeneic healthy donor (41), and this was correlated with certain microbial metabolites. This may indicate that the microbiota from the subject's own sample were better adapted to re-colonize than species from a non-diabetic donor. Results from this study demonstrate that FMT and microbiota-targeted therapy in general holds promise for promoting immune tolerance and preserving beta cell function in T1D. In the long-term, it is unlikely that FMT in its current form will progress to a wide-spread therapy for T1D due to both safety and palatability considerations. Recently, two cases of drug-resistant E. coli bacteraemia occurred following FMT in two separate clinical trials resulting in the death of one immunocompromised patient (117). It is more likely that if FMT demonstrates therapeutic benefit, knowledge from these studies will lead to the development of defined microbial consortium-transfer based approaches. The challenge will be in how to tailor such products to account for compatibility with core microbiota present in each recipient that together with diet and genetic background strongly influence donor engraftment (115).

The use of diet-based therapy to target the gut microbiota can circumvent the need for engraftment of donor microbes into the system. These include prebiotics, which are defined as “a substrate that is selectively utilized by host microorganisms conferring a health benefit” (118). These include certain oligosaccharides (including human milk oligosaccharides) and fermentable fibers. Using such an approach, Ho et al. delivered a prebiotic, oligofructose-enriched inulin or placebo, for 12 weeks in children that had T1D for at least 1 year (119). At the completion of the diet (3 months), the prebiotic group (n = 17) had significant preservation of c-peptide compared to the placebo group (n = 21), though this benefit was not sustained at 6 months after the diet was stopped. The prebiotic administration was associated with a significant increase in the Bifidobacterium genera. Other approaches introduce specific modifications to prebiotics to increase the delivery of beneficial metabolites such as SCFAs. For example, high amylose maize starch (HAMS) is a resistant starch that can be modified by acetylation (HAMSA) and/or butyrylation (HAMSB). When HAMSA or HAMSB are fermented in the colon, large quantities of acetate and butyrate are released. In NOD mice, a HAMSAB diet provided before disease onset almost completely prevented progression to autoimmune diabetes (95). Protection was associated with expansion of regulatory T cells and a decreased frequency of islet-specific effector T cells via a mechanism dependent on remodeling of the gut microbiota. HAMSAB is being tested in adults with established T1D in the “TOGeTHER” (Type One Gut Therapy) trial (Australia New Zealand Clinical Trials Registry identifier ACTRN12618001391268, Table 1) for its potential to increase systemic acetate and butyrate delivery and modulate the microbiota, immune tolerance and glycemic control. Similarly, HAMSAB is being tested in new-onset T1D in a randomized cross-over trial (NCT04114357). In a related approach, de Groot et al. gave oral sodium butyrate to adults with established T1D for 1 month (120). Unlike HAMSAB, which is utilized in the colon by the gut microbiota releasing acetate and butyrate, oral sodium butyrate capsules deliver butyrate mainly to the small intestine. In this study, oral butyrate delivery did not lead to increased butyrate availability in the colon, nor did it impact peripheral innate immunity, islet-specific T cell autoimmunity or beta cell function. However, this mode of butyrate delivery, dose, or duration of treatment may not have been appropriate to achieve the disease protective effects seen in animal studies. In studies in healthy subjects, three different resistant starches: maize, potato, and tapioca-based starches with different chemical cross-linking, were compared in a dose-escalation study with a non-resistant (digestible) control starch (121). The authors found that the crystalline maize resistant starch increased butyrate production together with Eubacterium rectale (a major butyrate producer), while the cross-linked tapioca resistant starch increased the SCFA propionate and Parabacteroides distasonis. Parabacteroides distasonis produces succinate, which is then converted to propionate by other bacteria. Neither the potato-based resistant starch nor the control starch increased SCFA concentrations. These data suggest that specialized dietary fibers can be used to target specific SCFA, and those that favor butyrate production could be specifically evaluated in T1D.

While much work is focusing on ways to increase SCFA production by the gut bacteria in T1D, many other microbial pathways and bioactive compounds produced by commensal bacteria may be beneficial for immune tolerance and glucose homeostasis. For example, the secondary bile acids isoDCA, isoalloLCA, and lithocholic/3-oxo-lithocholic acids, which are metabolized by bacteria in the colon from primary bile acid waste, have been shown to promote the generation of peripheral Tregs in the colon (122–124). Supplementation of these secondary bile acids was achieved in mice by either delivering them in drinking water, incorporating them into the mouse chow or by transferring a consortium of bacteria engineered with the metabolic capability to produce isoDCA. Thus, translation of such approaches to humans may have potential to reduce intestinal inflammation and potentially autoimmunity. Another example are microbe derived metabolites derived from breakdown of the amino acid tryptophan into indoles. These indoles are ligands for the aryl hydrocarbon receptor (AhR) which then promotes gut barrier integrity, Tregs and IL-22 production. In mouse models of T1D, oral administration of AhR ligands have been shown to expand Tregs, reduce insulitis and diabetes progression (125, 126). The beneficial gut bacteria Faecalibacterium prausnitzii is a high butyrate producer, but also produces an anti-inflammatory protein called MAM which inhibits NFκB activation and alleviates chemically-induced colitis (127). Identification of anti-inflammatory molecules produced by gut bacteria is currently a rich field, and it is likely that many candidates will be identified with promise for human disease modulation including T1D. Such molecules (now termed “postbiotics”) could be delivered as stand-alone drugs or the bacteria that produce these molecules could be targeted for expansion by prebiotics or probiotic based therapies.

While microbiota-based therapies hold much potential, it is quite likely that combination therapy approaches will be needed, especially at disease onset when autoimmunity and beta cell destruction is well-advanced. In cancer, it has recently been shown that the efficacy of check-point inhibitor immunotherapy relies on the gut microbiota (128, 129). At least in mouse models, co-delivery of a probiotic cocktail of Bifidobacterium strains improved the outcome of anti-PDL1 cancer therapy (130) and likewise Bacteroides thetaiotaomicron or Bacteroides fragilis transfer enhanced anti-CTLA-4 cancer therapy (128). There is good rationale that such combination approaches may also be of benefit in T1D. For example, low-dose IL-2 therapy, which protects NOD mice from diabetes (131) and is currently being trialed in human T1D, expanded Treg in the gut mucosa and led to an altered gut microbiota in NOD mice (40). Therefore, there may well be synergy between microbiota targeted and IL-2 based therapy. Similarly, in a rat model of T1D, two immunotherapies which prevent disease: the histone deacetylase inhibitor ITF-2357 and IL-1 receptor antagonist (Anakinra) also modify the gut microbiota. Using this rationale, Chan et al. showed that combining anti-CD3 immunotherapy with a prebiotic oligofructose increased the rate of disease reversal when treatment was started at the onset of hyperglycemia (132). In a more sophisticated approach, Takiishi et al. have used a probiotic strain of Lactococcus lactis modified to secrete the autoantigen proinsulin and IL-10 in combination with anti-CD3 therapy (133, 134). They showed that this combination therapy was more effective than either treatment alone in reversing disease at the onset of hyperglycemia in NOD mice. Hence, combination therapies targeting multiple disease pathways, including via the gut microbiota, could prove beneficial for treatment of T1D.