Possible role of the gut microbiota in the pathogenesis of anorexia nervosa

The fact that gut bacteria are involved in weight control and growth is not a novel finding, but is a long known basic theory in livestock industries. For example, antibiotics have often been used to promote livestock growth [19, 20], called antibiotic-induced growth promotion (AIGP). The use of antibiotics for AIGP is still being exploited in several countries, including the US and Japan. The precise mechanism for AIGP is unclear, but the fact that growth promotion is absent when administering antibiotics to germ-free animals deficient in bacteria [21, 22], indicates that the gut microbiota plays a critical role in this promoting effect [20]. Gordon et al. [23‐25] used an elegant method of fecal transplantation to demonstrate that the gut microbiota is critically important for regulating body weight. Furthermore, another group [26] showed that transplantation of Christensenella minuta into mice with an obese phenotype suppressed weight gain. Thus, these findings indicate that the gut microbiota can affect body weight regulation by modulating the gut microbial ecology.

When exposed to stress stimuli, the hypothalamic-pituitary-adrenal (HPA) axis is activated to maintain homeostasis of the body [27]. Interestingly, the HPA axis is known to be affected by both genetic determinants and postnatal environmental factors during infancy [28, 29]. Because gut microbes are an important environmental factor, we hypothesized that the gut microbiota plays a role in host stress responses. In 2004, we first demonstrated that gut microbes determine the HPA response to stressors, using germ-free (GF) and gnotobiotic mice [30]. The degree of plasma ACTH and corticosterone elevation in response to 1-h restraint stress was higher in GF mice than in specific pathogen free (SPF) mice. Furthermore, as summarized in Table 1, mono-association with B. infantis, a representative inhabitant of the neonate gut, decreased the HPA stress response to SPF, while mono-association with Bacteroides vulgatus had no effect. The hormonal stress response of rabbit-derived enteropathogenic Escherichia coli (EPEC)-monoassociated mice was substantially higher than that of GF mice, although such an exaggerated response was not found in mice with an EPEC mutant strain, ΔTir [31], which was uninternalized due to defects in the translocated intimin receptor.

The finding that commensal bacteria can substantially affect the HPA stress response later in life raises an interesting question about whether gut bacteria can change host behavior. This question has been addressed in the past decade by several independent groups, including our own [32‐37]. The commensal microbiota is a crucial factor in modulating the host’s behavioral profile. For example, we developed a reliable method to accurately analyze the behavior of GF mice maintained in an isolator [35]. This method enabled us to evaluate GF animal behavior without the risk of exposure to contamination. Using this system, GF mice were found to be more active and anxious than EX-GF mice that were transplanted with a normal SPF microbiota, based on open-field and marble-burying tests (Table 2). Colonization with B. infantis decreased the locomotor activity to the EX-GF level but had little effect on the anxiety levels. In contrast, mono-association with Clostridium coccoides reduced anxiety levels; however, it did not affect locomotor activity [35].

Collectively, these results indicate that gut microbiota can exert a substantial effect on the behavioral phenotype as well as the ability of the host to respond to stressors.

Patients with AN who have extremely low body weight often show marked resistance to weight gain in response to calorie intake; therefore, they usually require more calories to increase body weight than healthy subjects with normal weight [38‐40]. Several factors such as increased physical activity [41] and diet-induced thermogenesis [42] are suggested to be involved in this phenomenon; however, the precise mechanisms for this are largely unknown. In addition, other important complications of AN include comorbid anxiety and depressive disorders. A well-known study, “the Minnesota Starvation Experiment” [43], which was performed during World War II, clearly showed that starvation significantly affected both physical and psychological conditions [43, 44]. Thirty-six volunteers who refused military service for religious reasons participated in this study. They were psychologically resilient before the experiment; however, most experienced periods of severe emotional distress during the study. This indicates that some psychiatric symptoms found in patients with AN could be explained by the effects of starvation. However, how and to what extent starvation affects the psychological and social functioning of patients with AN remain largely unknown.

Considering the emerging effects of gut microbes on body weight control or behavioral phenotypes, we speculated that commensal bacteria may play an important role in the onset and exacerbation of AN. Undoubtedly, patients’ premorbid traits and psychosocial stressors are critical to the onset of AN. Prolonged physiological and psychological stress, including severe weight loss and extreme dietary habits, may influence the composition of the gut microbiota and subsequently induce “gut dysbiosis.” This disturbed bacterial community may contribute to a decrease in food efficiency or behavioral abnormalities. For example, hyperactivity, one such abnormal behavior, may further strengthen the resistance to weight gain.

Several researchers have investigated whether gut dysbiosis exists in the gut of patients with AN and have shown that patients with AN show “dysbiosis”, abnormal features of gut microbiota [45‐48]. Armougom et al. [45] reported that Methanobrevibacter smithii was more frequently detected in patients with AN than those with low body weight due to diseases other than AN. Interestingly, these methanogenic archaea are often found in patients with constipation-predominant irritable bowel syndrome [49, 50]. Mack et al. [47] found that the ratio of mucin-degrading bacteria to Clostridium cluster I, XI, and XVIII in an AN group was higher and the butyric acid-producing genus Roseburia was lower than that of the intestinal flora of normal body weight subjects. In addition, the proportion of Bacteroidetes phylum in patients with AN was significantly reduced compared to normal-weight subjects and did not recover after weight gain. Our group [48] used the intestinal bacterial analysis system YIF-SCAN® that was developed by Yakult Central Laboratory to examine the intestinal bacteria of female subjects with AN and age-matched healthy women. The total bacterial, Clostridium coccoides, Clostridium leptum, Bacteroides fragilis, and Streptococcus counts were significantly lower in the AN group than the control group. The detection rate of the Lactobacillus plantarum subgroup was significantly lower in the AN group than in the control group. Based on 16S rRNA sequencing methods, we also reported a lower relative abundance of the phylum Bacteroidetes in patients with AN than healthy age-matched controls [51]. As summarized in Table 3, several recently published papers [52‐57] have also shown changes in gut microbial ecology at the phylum or genus levels; however, no specific bacteria were identified among these reports. Thus, although the results differ depending on the patient’s background or analytical methods, “gut dysbiosis” is consistently observed in patients with AN.

The existence of gut dysbiosis does not always relate to impaired functional outcomes; namely, such disturbed microbiota may not be a causal factor contributing to AN pathologies but an epiphenomenon resulting from lengthy starvation. To test this, we transplanted the gut microbiota of patients with AN into GF mice to establish gnotobiotic mice (gAN) whose microbiota consisted of the intestinal microbiota of patients with AN [51]. The gAN mice showed a significantly decreased weight gain compared with the gnotobiotic mice (gHC) transplanted with the gut microbiota of healthy women. Similarly, food efficiency (weight gain/food intake) was also lower in the gAN mice than in the gHC mice. Moreover, the gAN mice also exhibited increased anxiety-like behaviors relative to the gHC mice, when evaluated by open field and marble-burying tests. Interestingly, marble-burying behaviors displayed the highest correlation with the relative abundance of the genus Bacteroides. Moreover, the administration of Bacteroides vulgatus, which belongs to the genus Bacteroides and is a predominant species of the Bacteroides fragilis group in adult humans [58], reversed the behavioral abnormalities in the gAN mice.

These results indicate that some characteristic features of patients with AN can be reproduced by transplanting the AN gut microbiota. Bacteria, such as Bacteroides vulgatus, may play a protective role against the development of pathologies specific to patients with AN. Recently, Glenny et al. [59] reported that fecal transplantation from patients with AN exerted no significant effects on body weight in GF mice. A precise reason for this discrepant result is unclear; however, this may be related to the methodology of fecal transplantation, which involved the use of frozen feces, as they suggested. Therefore, the results of our study, which utilized fresh feces, should be validated by future experiments using a larger number of mouse colonies and different human donors.

The above results based on animal experiments provide valuable information about some pathological features of AN; nonetheless, they are inapplicable to human conditions. Therefore, we performed metabolome analyses using human materials.

Increasing attention has been paid to the role of gut microbiota in AN. However, in terms of gut microbes, the following questions remain to be addressed.

Figure 1 shows our working hypothesis concerning the possible role of the gut microbiota in the development and perpetuation of AN.