Introduction

The term intestinal microbiota refers to the set of microorganisms (bacteria, archaea, fungi, and viruses) that coexist in the gut [1]. In recent years, intestinal microbiota has become the subject of renewed interest not only in the field of science, but also across society, and industry. Thus, 96.8% of the 53,921 publications in PubMed under the label “gut microbiota” were published in the last decade. Moreover, the microbiota has become one of the scientific topics receiving broad coverage in the press [2, 3], and microbiota-based products represent a fast-growing market today, with an estimated 275–400 $ million worldwide [4].

Current research on microbiota has primarily focused on bacteria, most of which belong to a small number of phyla [5]. Firmicutes and Bacteroidetes represent around 85–90% of the total microbiota, whilst Actinobacteria, Verrucomicrobia, and Proteobacteria appear in smaller proportions [5]. However, this composition is highly variable between individuals and in the same individual over time and can be affected by several factors, with diet being the most researched of all [6]. Both long-term and short-term dietary patterns can produce changes in the structure of microbiota by providing or removing nutrients, and modifying the conditions of the gastrointestinal ecosystem surrounding bacteria e.g., pH, bile acid content, and so forth. [6]. Similarly, antibiotics and psychotropic drugs such as SSRIs have been found to have an impact on the microbiota-gut-brain axis [6], and several studies have shown the composition may vary depending on age, stress, BMI, and physical activity [6], among other factors.

The surge in the interest in microbiota has been motivated by its physiological potential, with an estimated 100 bacterial genes for every human gene [1]. Moreover, gut microorganisms have been considered key players in the gut-brain axis [7], and it has been suggested they may favour communication with the brain via the endocrine, immune, or nervous system [7]. Typically, the prime candidates for establishing this connection are bacterial metabolites (especially SCFAs), neurotransmitters or immune intermediates [6, 7]. Hence, it is reasonable to suggest that an imbalance in gut microbiota composition or dysbiosis may be linked to the clinical features of several mental health disorders [7]. While the idea that gut microorganisms may affect behaviour is not novel, dating back to Bouchard’s autointoxication theory [8], it is only with the emergence and cheapening of sequencing techniques and bioinformatics that research in the field has developed [7].

Several methods have been applied to the analysis of the microbiota-gut-brain axis [7]. Microbiome studies have employed massive sequencing techniques such as 16S sequencing to compare profiles of microbial communities in individuals with and without pathology [9]. Alternative approaches include Faecal Microbiota Transplantation (FMT) [7] or humanised gnotobiotic rodents, i.e., rodents that have been transplanted with the faecal microbiota of individuals with or without pathology [10]. Thus, research has examined the causal role of microbiota on mental health [7] and, to date, associations for autism, psychotic disorders, anxiety, depression, ADHD, post-traumatic stress disorder and, more recently, eating disorders, including Anorexia Nervosa (AN), have been proposed [7].

AN affects 1% of the population [11] with 90% of cases being women [12]. Its onset is usually associated with puberty or early adulthood and, despite its low prevalence, it has the highest mortality rate among mental disorders [11]. According to the DSM-5, AN is characterized by energy restriction resulting in low body weight, fear of gaining weight, and distorted body image [12]. Two subtypes have been established i.e., restrictive and purgative.

Overall, AN is often associated with physical signs derived from malnutrition such as bradycardia, hypotension, lanugo, hypothermia, and hyperactivity [11]. The latter appears in approximately 70–80% of cases [13] though estimates vary given the lack of consensus in its definition. Furthermore, gastrointestinal symptoms, marked by abdominal pain and discomfort, appear in the literature since the first descriptions [14]. In fact, it has been estimated that 90% of AN cases present some type of gastrointestinal complaint [15].

At present, there is a lack of efficacious treatments for adults with AN [11, 16, 17]. Although international guidelines [18] recommend psychological intervention such as CBT (cognitive behavioural therapy), the MANTRA protocol (Maudsley Model of Anorexia Nervosa Treatment for Adults), and SSCM (Specialist Supportive Clinical Management) as the treatment of choice, a recent meta-analysis conclude that: “No significant differences between psychological treatment and control condition were found on weight gain, on eating disorder pathology, and on quality of life.” [16, p.2]. In fact, this could be the outcome of a poorly understood aetiology.

In an attempt to progress, Bulik et al. [19] suggested reformulating/reframing AN as a metabo-psychiatric disorder [19, 20]. This reformulation was grounded in the recent study published by the Eating Disorders Working Group within the framework of the Psychiatric Genomics Consortium (PGC), which concluded the existence of genetic associations between AN and metabolic/anthropometric traits [20].

Under the premise of a metabo-psychiatric disorder, microbiota would emerge as a potential explanatory factor for key features of AN [21]. Therefore, dietary reduction, psychosocial stress, and marked weight loss are considered to contribute to the establishment of a selective intestinal environment for bacterial survival favouring the development of dysbiosis [19]. So far, it has been suggested that this might explain low weight gain [19, 22], affective symptomatology [19, 22, 23], altered eating behaviour [23], and even hyperactivity [22]. Thus, modifications of microbiota composition using pro-, prebiotics [22], FMT, or precision nutrition [19] have been proposed as promising therapeutic interventions.

In an effort to determine whether altered composition is more than a mere consequence of malnutrition, several authors have examined the underlying causal mechanisms of microbiota in the aetiology of AN. Intestinal dysbiosis has been suggested to favour tryptophan deficiency by altering serotonergic pathways that increase physical activity [22]. In the same vein, affective symptomatology has been associated with altered gut-brain communication due to increased uremic toxins [22], and reduced SCFAs [21]. Fetissov & Hökfelt [23] define eating disorders as the result of altered communication between microbiota, the immune system, and the neuroendocrine system. Thus, these authors focus on the ClpB protein (caseinolytic peptidase B) generated by members of the Enterobacteriaceae family, presenting a mimetic sequence with α-MSH [24]. These authors claim that a dysbiosis involving an increase in these bacteria would cause the segregation of ClpB affecting the aetiology of AN either by (a) a direct effect on the peripheral satiety routes -increasing PYY [25, 26], or centrally through MC4R [27]; or (b) an indirect effect from the formation of anti-ClpB IgG antibodies cross-reactive with α-MSH that would form immune complexes, activating chronically the central MC4R receptor and inducing increased satiety and anxiety, or both. Recently, Frostad [28] has emphasized the role of CCK-4 as a complementary mechanism integrated into this model; malnutrition would cause disruption of the blood–brain barrier, increasing the sensitivity of CCK2 receptors to the CCK-4 peptide generated by enteroendocrine I cells during meals. Thus, anticipatory anxiety would occur that would favour psychological maintenance mechanisms such as the pursuit of thinness and weight overvaluation.

Bearing in mind the above findings on the treatment and aetiological understanding of AN, this work aimed to assess the current scientific findings on the role of intestinal microbiota in the development and maintenance of the disorder in human and animal studies; hence, both human and animal studies were included.

Method

Search strategy and study selection

A systematic search was conducted on November 5, 2021 (updated June 13, 2022) of all papers referring to the topic from 2009 to 2022 in the databases PubMed, Web of Science (Wos), PsycINFO, and ScienceDirect. The search consisted of the following combination of terms ((microbiota) OR (microbiome)) OR (dysbiosis)) AND (Anorexia nervosa). PICOS criteria for inclusion and exclusion of studies are shown in Table 1. This review was neither registered nor was a protocol published.

Table 1 PICOS study inclusion criteria
Full size table

Only original peer-reviewed published scientific papers were included. The exclusion criteria were as follows: Theses, conference papers, letters to the editor, manuals, literature reviews; studies with authors having conflicts of interest. Articles are written in a language other than English, Spanish, or French. Studies based on populations with anorexia due to other medical conditions. Articles that exclusively evaluated communities of microorganisms other than bacteria or archaea. The search yielded a total of 400 papers, of which 18 papers were finally selected following the guidelines of the PRISMA 2020 protocol [29], as shown in Fig. 1. Moreover, a manual search was performed from the bibliography of the selected paper without adding further references. Article screening was carried out using the Rayyan® web application [30]. Titles and abstracts were independently reviewed by two authors (NG and EG).

Fig. 1
figure 1

PRISMA flowchart (Page et al., 2021) showing literature search

Full size image

Data extraction and synthesis

Data were extracted to an Excel spreadsheet. The following publication information was sought: sample size and description, intervention or treatment (if applicable), sequencing and analysis methods, gut microbiota outcome data (taxa, diversity) or other outcomes (e.g., metabolite levels, associations between gut microbiota and AN symptoms, FMT outcomes). No statistical analysis was performed due to the heterogeneity of the included studies.

Risk of bias

The Newscastle-Ottawa scale adapted for cross-sectional studies [31], the SYRCLE’s RoB tool for animal studies [32], and the JBI critical appraisal tools (for case report and case series) [33] were used to assess the potential bias of included studies.

Results

Study selection and risk of bias

Of the initial 400 papers, 147 duplicate papers were excluded as shown in Fig. 1. After the primary screening, 187 studies were excluded for not being related to the subject of the study or not being specifically focused on AN. 65 of the remaining 66 were recovered. Finally, a full-text review of these was carried out and 29 reviews were excluded: 2 papers written in a language other than English, French or Spanish; 1 doctoral thesis; 5 conference abstracts; 1 letter to the editor; 1 article with population other than the target; 7 papers with objectives different to the aims of this review, and 1 research protocol.

The quality of the included studies is summarized in Supplementary Tables S1, S2, S3.

Studies in humans

The present review included five longitudinal studies [34,35,36,37,38], six cross-sectional studies [39,40,41,42,43,44], two single case reports [45, 46], and a case series design [47] the main features of which are summarized in Table 2.

Table 2 Main characteristics of human studies included in the review
Full size table

One study aimed to analyse the composition of intestinal microbiota in patients with AN [45]. The remaining studies, apart from single-case and case-series studies [45,46,47], assessed differences in intestinal microbiota between AN and a normal weight control group (NW) adjusted for age and sex. Other studies also compared microbiota composition in an obese group [39, 41, 44], an overweight group [41, 44], and a group of athletes [44]. Seven studies assessed changes in microbiota composition in patients with AN in response to FMT [46], or nutritional rehabilitation [34,35,36,37,38, 47]. Finally, two studies assessed differences in intestinal microbiota composition in AN subtypes in comparison to a control group [42, 43].

Most studies included female participants. In one study a male was included in the AN sample [41], and in another, the sex of participants was not specified [39]. Regarding the AN subtype, most of the studies included patients with both AN subtypes, except for three studies that only included patients with the restrictive subtype [40, 42, 45], one study including patients with the purgative subtype [46], and three studies with unspecified subtype [34, 39, 47]. As for diagnosis, five studies [37,38,39, 42, 47], used the DSM-5 diagnostic criteria [12], two studies [34, 43] employed the DSM-IV-TR [48], one study [39] used the DSM-IV [49], another [44], the ICD-10 criteria [50], and two studies failed to specify the diagnostic criteria employed [36, 45].

As for symptomatology, an array of instruments had been employed such as SCID-5-CV [36, 42], SCID-I [34, 47], BSI [36, 42], SCL-90 [40], EDI-2 [38, 40], EDE-Q [34, 36,37,38, 42, 46], EDE [38], SCAS [36, 38, 42], STAI-Y form [40], BDI-II [40, 46], BDI [34, 38, 44], HDRS or HAMD [44], and BAIFootnote 1 [34, 46].

As shown in Supplementary Table S4, the exclusion criteria also differed with two studies using a standard diet for both control and AN groups [36, 42], one monitoring food intake [37], whilst the remaining studies either collected dietary information from structured interviews 24 h diet [35, 44], 3 day self-reports [40], food frequency questionnaires [35], or no type of control was performed [34, 38, 39, 41, 43, 47].

The treatment applied in the studies involved nutritional rehabilitation, except in one case involving FMT [46]. Three studies combined treatment with psychological therapy, either unspecified [35, 38], or cognitive-behavioural therapy [36].

In all studies intestinal microbiota analysis was performed on a faecal sample using real-time qPCR [39, 41], 16S rRNA sequencing [34,35,36,37,38, 40, 42, 44], a combination of both [40], cultivation techniques [45], or reverse transcription qPCR [43]. A variety of primers were used for 16S rRNA sequencing, using both the V1-V2 hypervariable region [38, 44], V1-V3 [34], V3–V4 [37, 40], and V4 only [35, 36, 42, 46, 47]. Furthermore, several studies examined other parameters such as faecal pH [43], short-chain fatty acids (SCFAs) in faecal samples [35, 37, 40, 43], other faecal metabolites [36, 37, 42], intestinal permeability [46], anthropometric parameters [37, 40, 44], metabolic indicators [47], biochemical parameters from a blood sample [40, 43, 44], two studies evaluated fungal communities [41, 46], one study evaluated the severity of functional gastric symptomatology and its post-treatment variation [35], and six studies assessed relationships between microbiota and psychological symptomatology [34, 35, 37, 38, 40, 44].

Results of microbiota composition analysis

Two parameters were mainly reported: the relative abundances of different taxonomic groups, and significant differences of these between AN and control groups; as well as the alpha and beta diversity indices. One study reported significant differences at OTU levels [37]. Other studies gathered data on increased Methanobrevibacter smithii counts [40], or decreased counts of Clostridium coccoides, Clostridium leptum, Bacteroides fragilis, Lactobacillus plantarum in AN patients versus NW [43].

The main results of the comparative analysis on the relative abundance of microbiota in AN patients and NW controls are summarized in Table 3. In spite of certain contradictory results, this analysis showed a lower ratio of the genus Roseburia.Footnote 2 In patients with AN [35, 40], a higher relative abundance of Verrucomicrobia [35, 42], Actinobacteria [35, 36], Coriobacteriaceae [34, 44], and M. smithii [35, 39], the latter being negatively correlated to BMI [40, 41].

Table 3 Significant differences in the relative abundance of intestinal bacterial taxa in individuals with AN as compared to NW controls
Full size table

After treatment studies have shown a lower relative abundance of the Coriobacteriaceae family [34, 36], and a higher abundance of Ruminococcaceae and the Firmicutes phylum as compared to an NW control group [34, 38]. Finally, a lower relative abundance of the genus Bifidobacterium was observed, and a higher ratio of Haemophilus, and the Pasteurellaceae family in the restrictive subtype than in the purgative subtype [42].

Two studies examined the correlations between bacterial ratios and total scores in different psychometric tests [36, 40]. Thus, a negative correlation was observed between the relative abundance of Clostridium spp and total BDI scores in patients with AN [40], as well as between different bacterial genera and scores on the BSI, and EDE-Q [36]. However, one study found no correlation with anxiety, depression, or eating disorders symptomatology [38]. One study reported a negative correlation between BMI and a relative abundance of Bacteroides uniformis [44].

Ten of fourteen studies reported inconsistent results in microbiota alpha and beta diversity in AN patients in comparison to an NW control group [34,35,36,37,38, 40, 42, 44, 46, 47]. Three studies found lower levels of alpha diversity in AN patients [34, 36, 42], whereas five studies could not replicate these results [35, 37, 38, 40, 44]. As for beta diversity, three studies found significant differences [34, 35, 38], but four did not [36, 40, 42, 44].

Post-treatment comparison of these parameters with respect to NW controls also showed inconsistent results. Thus, one study showed an increase in alpha diversity [38], whilst another found no significant differences [34]. One study found significant differences in beta diversity [38], whereas two did not [34, 36]. Two studies reported negative correlations between alpha diversity and BDI scores [34, 44], but another study only found this correlation for the whole group included in the study [44]. Alpha diversity was negatively correlated to the weight and figure concern subscales of the EDE-Q in one study [34], but this was not replicated in another study [37].

Results of analysis of faecal metabolites

A comparative analysis of SCFAs and BFCAs faecal concentrations revealed inconsistent results, as shown in Table 4. One study found a positive correlation between butyrate concentrations and a relative abundance of Roseburia, and a greater abundance of valerate and BFCAs concentrations in AN0 that was maintained or even increased following treatment with respect to NW [35]. Butyrate concentrations were negatively correlated to anxiety and depressive symptomatology [40]. Three studies analysing other faecal metabolites [36, 37, 42] found lower levels of/ amino-acids and metabolite concentrations from intestinal microbiota in AN patients [36] while establishing differences in the faecal metabolomic profile between ANR and ANP [42].

Table 4 Comparison of SCFAS levels (propionate, acetate, butyrate, and valerate), and BFCAS between AN and NW
Full size table

Miscellaneous results

One study reported a new species in the gut of AN patients using a culture technique [45], and still another study of an FMT in a case of AN found significant changes in structure, intestinal permeability, diversity, and relative abundances of A. muciniphila and M. smithii that were unrelated to improved gastrointestinal symptomatology, mood, or eating behaviour [46].

One study evaluating daily changes in intestinal microbiota composition and diversity in three patients observed these changes were specific to each individual [47]. Another study assessing gastric symptomatology and its evolution in response to nutritional rehabilitation found no improvement or even a worsening of symptoms of the upper tract after treatment such as abdominal pain, intestinal noises, and a sensation of incomplete evacuation despite the changes in relative abundances [35]. Finally, another study observed a reduction in faecal GABA and dopamine concentrations in ANR0 relative to NW controls [37].

Animal studies

As shown in Table 5, four studies with different animal models examined the causal role of microbiota in the development of AN [64,65,66,67].

Table 5 Main characteristics of animal studies included in the review
Full size table

Two of these studies employed germ-free mouse models (Mus musculus), C57BL/6 [65], and BALB/c [66]. One study employed C57Bl/6JRj mice [64], while another used Wistar rats [67]. There were also differences regarding the sex of the animals i.e., two studies employed females [66, 67], one study used males [64], and another mixed groups [65].

Two studies performed human faecal microbiota transplants in animal models [65, 66]. One study colonized the same generation orally from frozen faeces [65], while another colonized the parent generation from fresh faeces, using offspring as part of the experiment [66]. The main objectives of these studies were to examine differences in weight gain [65, 66], and behaviour [66] in mice reconstituted with faecal microbiota from AN patients (gAN) as compared to mice with normal weight control microbiota (gNW); and to analyse the relationship between bacterial taxa and weight gain or behaviour [65].

Moreover, two studies evaluated differences in intestinal microbiota composition in animal models under the ABA protocol as compared to control groups in different nutrition and activity conditions [64, 67]. Thus, data are reported for different control groups: a control group without food or activity restrictions (C) [64, 67]; a control group with activity and food restriction [64]; a reduced body weight control group (reduction of food to maintain a 25% lowering in body weight without physical activity), and a control group without food restriction and wheel access (CRW) [67].

In all experiments, microbiota analysis was performed by 16S rRNA sequencing using the V3-V4 region [66, 67], V4 [65] as the primer, or V5–V6 [64]. Furthermore, 5-HT and 5-HIAA concentrations in brain tissue [66], reductions in brain volume and in the number of astrocytes [64], and differences in anxiety levels were assessed using behavioural tests (open field test or marble burying test), as well as the effects of probiotic treatment (Bacteroides vulgatus) on behaviour [66].

Faecal microbiota transplant experiments yielded inconsistent results i.e., one study found gAN animals exhibited significantly less weight gain than gNW animals, with longer times in the peripheral quadrants in the open field test, and a higher number of buried marbles, indicating higher levels of anxiety [66], but another study found no significant differences between the two groups [65].

Studies evaluating faecal microbiota in animal models under the ABA protocol showed that in the rat model there were significant differences in composition and diversity between the experimental groups, which were mainly linked to reduced intake rather than hyperactivity [67]. However, no significant differences in alpha diversity were observed in mice, but a higher relative abundance of Clostridium clocleatum was found in the ABA model, as well as a positive correlation between the relative abundance of Burkholderiales and body weight, food intake, and the percentage of lean mass [65].

Discussion

In recent years, numerous studies have claimed microbiota is a potential causal explanatory factor for AN. This research is based on the description of an intestinal dysbiosis associated to pathology, a deduction underlying most of the human studies included in this review (Supplementary Table S5), which contrasts with the inconsistency in the results obtained (Table 2). The definition of dysbiotic state in AN differs from one study to another and has been explained in different ways: due to the immaturity of the techniques, the methodological differences between the studies, and the ambiguity of the term itself.

As indicated by Hooks & O'Malley [68], “the broader the definition, the easily it is detected.” (p.6). The concept of dysbiosis in the current literature poses problems in both precision and consensus and is usually associated with microbiota “imbalance” in terms of differences in the relative abundances and diversity parameters. However, owing to the high inter-individual variability in microbiota composition in humans [5, 9], there is a lack of normative data on what is healthy or normal [68]. Furthermore, the comparison with healthy individuals creates a circular fallacy whereby a pathological process causes dysbiosis which, in turn, leads to a pathological process [4]. Hence, it would be reasonable to ask whether the comparison with a normal weight group does really provide relevant information regarding the aetiology of AN. For microbiota to be regarded as an aetiological factor in AN, it not only has to be different from that of normal weight people but also different from other states of malnutrition.

Furthermore, field studies are still in their infancy, with no clearly established protocols and with major obstacles to overcome in the analyses [69, 70]. Massive sequencing techniques provide data of a compositional nature, collected as relative abundance, which poses a challenge for statistical analysis [69,70,71]. Thus, an absolute abundance cannot be inferred since the total number of reads per sample provided by the sequencing platforms reflect a technical artifact, unrelated to an actual biological composition [69, 71]. Consequently, the data are not independent of each other, and the increase in the relative abundance of one taxon automatically causes changes in that of others [69, 71, 72]. This implies the need to be prudent in analysing the results. It can be misleading to say that a taxon increases because of a physiological disturbance, without considering a specific reference point [71, 72]. In addition, the probability of false positives, and spurious correlations will increase [69]. At present, even though different analytical solutions have been proposed, the control of these variables has not been completely resolved [70, 71], raising serious doubts regarding the authenticity of the correlations reported in the studies reviewed, as well as the differences obtained between groups.

Moreover, quantitative techniques (qPCR) have been used to obtain absolute data with respect to a specific number of taxa. Though several studies have found a high abundance of M. smithii, the biological implications of this finding remain controversial. Thus, it may be involved in constipation often observed in AN, or function as an adaptive factor to low nutrition increasing energy efficiency [73].

In the establishment of causal relationships, that is, in the search for explanatory mechanisms, two methodologies have been used i.e., humanized gnotobiotic animal models and FMT. As for the former, Walter et al. [10] recently pointed out the lack of rigor in these designs. A small number of donors are usually used, whose microbiota is transferred to a much larger number of animals, artificially inflating the sample size. This, on the one hand, does not allow representing the high inter-individual variability existing in the composition of the human microbiota and, on the other hand, favours pseudo replication, increasing the probability of false positives [10]. The humanized gnotobiotic animal studies included in this review [65, 66] showed this weakness in design, in addition to divergent results that cast reasonable doubt on the conclusions drawn.

In this review, the FMT treatment was administered to only one patient in one study [46], without positive results in food intake, weight, or affective and gastrointestinal symptoms regardless of the modification in microbiota composition. In the literature, another publication reporting positive results with the same methodology was excluded due to a declared conflict of interest [74].

Bearing in mind the above findings, claims of an intestinal dysbiosis associated to AN are at best speculative given that the current evidence on the proposed theoretical mechanisms is rather scarce [1, 7]. Thus, even though multiple studies have underscored the possible role of SCFAs, the results were highly inconsistent, and failed to support this assumption, as shown in Table 3. As for Fetissov & Hökfelt’s [23] proposal, increases in Enterobacteriaceae cannot be substantiated due to its widespread distribution in the human intestine, though it might be a source of antigens even in the absence of any alteration. Thus, this model has two main weaknesses. First, there is the assumption of a common origin for AN, bulimia nervosa, and binge eating disorder [23]. This assumption has been strongly challenged [75, 76], on the grounds of hindering the integration of biological evidence [76], and the homogenization of eating disorders [75] that dilute the importance of fundamental factors for understanding the aetiology of AN, such as malnutrition and its interaction with physical activity [77]. Thus, there is no consistent evidence supporting Fetissov & Hökfelt’s [23] assumption as the data was obtained from heterogeneous samples composed of several of these three categories. Second, correlations between autoantibodies or ClpB and AN have not been established with the most objective AN diagnostic signs, but with subjective symptoms [24, 78, 79], such as the pursuit of thinness. However, as pointed out by Gutiérrez & Carrera [14, 80], this feature is not representative of all AN cases.

Hence, these two shortcomings may explain why IgG autoantibodies against α-MSH have been found in AN patients and normal weight controls [24, 81,82,83], the high inter-individual variability that has been observed [79, 83], and the inconsistent data regarding the very existence of a greater concentration of this bacteria in AN [24, 79, 83]. Moreover, differences in molecular affinity properties are inconclusive, as sometimes the dissociation rate may be higher [84], or lower [83] in AN than in a normal weight control group.

Moreover, further research on the role of ClpB in peripheral and central satiety pathways is required since its involvement in PYY secretion comes from in vitro studies [26], or from studies with low sample size [25, 26]. Furthermore, one study reported a fragment of the ClpB molecule with an α-MSH homologous sequence with no activity on MC3R and MC4R receptors [85], and no significant differences in concentrations between AN patients and normal weight controls [78].

Finally, further research is required to ascertain if these molecules are transported through the blood–brain barrier, and to determine the mechanisms enabling them to cross the intestinal barrier, as the evidence regarding intestinal hyperpermeability in AN patients is inconsistent [83, 86, 87]. Likewise, the recent integration of the role of CCK-4 in this model [27] is also controversial. The anxiogenic character of this peptide was observed in the 70 s of the XX century and was attributed to an agonistic action on the CCK2 brain receptor [88]. However, tetragastrin is a synthetic molecule whose natural endogenous synthesis has not been demonstrated [88, 89]. In addition, the main forms of cholecystokinin (CCK) present in plasma (CCK-58, 33, 22 and 8) [90] have been found in low concentrations, in the range of pMol/L, far from peak in the nanomolar range generated by CCK-4, and it remains to be determined whether they can cross the blood–brain barrier significantly to affect CCK2 receptors and have panicogenic activity [88].

A further controversial issue is Sudo’s [22] explanation of hyperactivity in AN, which was inconsistent with the findings of the present review. The animal studies reviewed [64, 67] showed a bacterial composition mainly associated with dietary restriction quite different from the restrictive feeding schedule of the ABA model. Despite observing a lower level of serotonin in the brainstem of the gAN group, Hata et al. [66] reported no significant differences in motor activity. Alterations in the serotonergic pathways associated with AN have been described in the literature, but they have not been established as exclusive to the condition [91]. Serotonin (5-HT) could be involved in increased activity through participation in the aversive motivational system, opposite to the dopaminergic one [92]. However, 5-HT-related interventions tested in the ABA model were partially effective, whereas in humans, they were not [91,92,93].

Strength and limits.

The main limitation of this review rest on its qualitative nature, as well as difficulties in interpreting the results given the broad methodological diversity, and the lack of conceptual consensus. Moreover, the dearth of studies evaluating the functionality of microbiota, and the small samples mostly composed of European women, provide a partial view of the issue. Nevertheless, this review seeks to provide insight into the current shortcomings in the field, and to address the explanatory difficulties associated with the assumption of microbiota as a central aetiological agent of AN.

What we already know on this subject?

In recent years, the etiological role of microbiota in AN has undergone a revival, and studies have underscored the role of intestinal dysbiosis, which has prompted its reconceptualization as a metabo-psychiatric disorder.

What does this study add?

This review includes clinical and preclinical evidence published in June 2022 regarding the relationship between gut microbiota and AN. The inconsistency in the results failed to substantiate the view of microbiota as an explanatory factor of AN. Shortcomings such as poor consensus, methodological limitations, and ambiguous definitions are inherent to a field that is just starting to emerge/move.

Conclusions

Owing to the lassitude of the term dysbiosis, and the lack of studies evaluating differences between AN and other cases of malnutrition, the existence of a pathologically specific microbiota composition associated with AN has not been substantiated to date, which undermines its aetiological role. Further research and new protocols are required to advance are our understanding and generate new data to fully elucidate the role of microbiota. However, this accomplishment should not overlook the cautious words of Hanage [94], when he warned, “In pre-scientific times when something happened that people did not understand, they blamed it on spirits. We must resist the urge to transform our microbial passengers into modern-day phantoms” (p. 248).