Clinical evidence of the role of Methanobrevibacter smithii in severe acute malnutrition

Gut microbial dysbiosis has been shown to be an instrumental factor in severe acute malnutrition (SAM) and particularly, the absence of Methanobrevibacter smithii, a key player in energy harvest. Nevertheless, it remains unknown whether this absence reflects an immaturity or a loss of the microbiota. In order to assess that, we performed a case–control study in Mali using a propensity score weighting approach. The presence of M. smithii was tested using quantitative PCR on faeces collected from SAM children at inclusion and at discharge when possible or at day 15 for controls. M. smithii was highly significantly associated with the absence of SAM, detected in 40.9% controls but only in 4.2% cases (p < 0.0001). The predictive positive value for detection of M. smithii gradually increased with age in controls while decreasing in cases. Among children providing two samples with a negative first sample, no SAM children became positive, while this proportion was 2/4 in controls (p = 0.0015). This data suggests that gut dysbiosis in SAM is not an immaturity but rather features a loss of M. smithii. The addition of M. smithii as a probiotic may thus represent an important addition to therapeutic approaches to restore gut symbiosis.

www.nature.com/scientificreports/ adiposity and that the relationship between M. smithii and body mass index is an inverted U curve. M. smithii has also been shown to improve the production of acetate 9 as well as butyrate, a key molecule for human health 3 , through microbe-microbe interactions 22 . All these unique properties make M. smithii one of the best candidates as a marker of healthy digestion and nutritional status. Severe acute malnutrition is a major public health problem affecting nearly 20 million children under five and causing up to 1 million deaths annually in low-or middle-income countries in Africa and South Asia 22 . It is a severe disease primarily related to inadequate diet. Micronutrients as well as protein-energy deficiencies have been linked to severe acute malnutrition [23][24][25] . However, cases refractory to a therapeutic diet 26 and the fact that antibiotics improve mortality rates 26,27 suggest other instrumental factors, such as gut microbiota dysbiosis. A delay in maturation of the digestive microbiota has been reported 11,13 , suggesting a quantitative developmental abnormality or immaturity. However, this immaturity could be a consequence and not the main feature of gut dysbiosis. A preliminary study conducted using specific quantitative PCR has shown that no malnourished children were positive for M. smithii compared to 75% of healthy children 28 . This observation suggests an absence or a loss of M. smithii which could contradict the "immaturity hypothesis". This prompted us to launch a clinical investigation of the key role of M. smithii in severe acute malnutrition.
Accordingly, we performed a large case-control study to clarify the association between M. smithii and severe acute malnutrition. We particularly investigated whether the findings were consistent with immaturity and whether M. smithii remains undetectable after therapeutic renutrition. This hypothesis would suggest that oral intake as a probiotic and/or in organic dairy 29 of this cultivable archaeon, previously isolated from human milk and colostrum 21 , may be a useful addition in the treatment of severe acute malnutrition.

Results
Participant characteristics. A total of 180 malnourished children were screened. Of these, 16 could not be included because afflicted with moderate acute malnutrition. A total of 164 of them were eligible for study, but only 143 severely malnourished children were ultimately included, as 21 were excluded because of an unavailable or insufficient number of samples provided (Fig. 1a). Of 209 healthy children screened, 61 were excluded outright because they did not meet the adequate anthropometric criteria. Of the 148 eligible controls with adequate anthropometric data, 13 were excluded for the presence of clinical symptoms or diseases such as gastroenteritis, malaria, rhinitis/rhino bronchitis and chickenpox. We excluded 25 others for non-available stool samples. We finally included 110 control children (Fig. 1b).
Detection of M. smithii increased with age only in controls, while prevalence decreased with age among malnourished children (Figs. 3 and 4). The prevalence rose to 90% in controls but dropped to 0% in severely malnourished children at 24 months (Fig. 3). The predictive positive value of detection of the M. smithii curve gradually increased to reach its maximum (all three children aged more than 55 months were positive) before www.nature.com/scientificreports/ 60 months of age in the healthy controls, unlike that of the malnourished children, which decreased abruptly and was nil after 15 months of age (Fig. 4). Controlling for age and gender, we found that the detection of M. smithii was associated with the absence of severe acute malnutrition (OR = 0.06, 95% confidence interval [0.02-0.15], p = 1·6 *10 -9) . We therefore performed a linear regression (Fig. 5) that showed that the concentration of M. smithii DNA increased with age only in controls (slope was positive (0.088) and was significantly different from zero in controls (p < 0.001) but negative (-0.0085) in severe acute malnourished children and not different from zero (p = 0.33)), and the difference between the two slopes was significant (p < 0 0.001).

Discussion
There is a paradox that the systematic analysis of the presence of M. smithii reported here has not been carried out before since this archaeon is a candidate of choice to explain good or bad digestion. Here, we confirmed that M. smithii, the main human gut archaeon and a critical human commensal found in virtually all human adults 8,9,17 , is lost (rather than decreased) in severe acute malnutrition. More than 80% of healthy children were positive when older than 20 months, while only 6 of 143 severely malnourished children were positive, all younger than 15 months. These findings are robust thanks to the large sample size (n = 253) obtained in a different geographical area (Mali) than our previous work, which was located in Niger and Senegal 28 , and to the strict and rigorous selection of malnourished and healthy children, according to clinical data and WHO child growth standards 2006 30 . Selection bias was controlled using a propensity score weighting approach based on the different age groups. All children were sampled during the same period of inclusion to avoid time bias. Stool samples of cases were collected before treatment with antibiotics and renutrition. www.nature.com/scientificreports/ Using a DNA extraction protocol optimized for methanogen detection with quantitative real-time polymerase chain reaction, we showed that M. smithii was detectable before 24 months but not after 24 months, suggesting a loss of this microbe in children with severe acute malnutrition 31 . Rather than an immaturity of the microbiota, this suggests that the absence of M. smithii is a marker of dysbiosis, whether due to obesity or severe acute malnutrition 12,13 , and furthermore supports the association of this archaeon with weight gain in children 32,33 . Control samples were asymptomatic and without any known disease by definition as children selected as controls were excluded in case of gastroenteritis or a moderate or chronic form of malnutrition. In cases, the low prevalence of digestive symptoms such as diarrhoea, emesis as well as that of gastroenteritis could not explain  www.nature.com/scientificreports/ this loss of M. smithii as confirmed by the analysis excluding the children suffering from severe acute malnutrition exhibiting the aforementioned symptoms.
We have long believed that M. smithii plays an essential role in nutrition. Indeed, Gordon's work in the 2000s showed that the digestion of complex carbohydrates requires the combined action of multiple bacterial strains equipped with a large abundance and diversity of glycosidases 8,9 . Digestion is accompanied by the production of hydrogen, which ultimately inhibits the metabolic activity of these bacteria. This demonstration prompted us to develop a specific method for the extraction and detection of M. smithii in faeces 17 . Indeed, 16S rRNA amplifications may miss methanogenic archaea, which have thick walls containing lysozyme-resistant pseudopeptidoglycan and thus require specific DNA extraction procedures 17 . The result of this earlier work was that in a normal French population, 100% of people were methanogen carriers 17 , in contrast with an earlier report that only 30% of the population were carriers 34 .
Since then, we have developed tools for better detection of M. smithii. Using these tools, we have noted the absence of M. smithii in malnourished patients in Africa 28 , an observation that had escaped other teams working on malnutrition, paradoxically including those of Gordon, whose work inspired our research on M. smithii.
In practice, we confirmed the significant absence of M. smithii in the faeces of malnourished children. We have previously shown that the source of M. smithii in the digestive tract of newborns originated from colostrum and breast milk 21 . Here, we observed a loss of the methanogenic archaea M. smithii in the gut microbiota of children with severe acute malnutrition. Severe acute malnutrition is an acute disease with an acute risk of death. Whether this loss is reversible after discharge remains unknown. It is noteworthy that M. smithii colonization is associated with organic dairy consumption 29 . Accordingly, organic dairy consumption may help these children recolonize their gut with M. smithii after the acute disease. This could be tested in further longitudinal studies with longer observation period post-discharge. Moreover, the acute loss of M. smithii may be associated with a sudden collapse of digestion, fermentation and butyrate production which is associated with death in these children 35 . Supplementation with missing microbes 36 , including M. smithii by probiotics/organic dairy may be required to prevent death in such a situation.
We believe that it is justified to consider the reintroduction of M. smithii in malnourished subjects in the form of a probiotic additive. Breastfeeding or the addition of milk-borne probiotics could be adequate since colonization has been associated with organic dairy consumption 29 . Future milk-based microbiome-directed therapies should investigate the potential benefit of the addition of M. smithii to seed the children's gut and preserve this critical commensal as a key to restoring host-microbial mutualism. In addition, future studies should investigate the possible mechanisms leading to the loss of M. smithii and determine whether this loss is the cause of the consequence of other characteristics of the severe acute microbiota associated dysbiosis (loss of aero-intolerant bacterial species among others).

Methods
Participants/study design. This case-control study was reported according to the indications of the STROBE statement 37 (STROBE checklist provided in Table S2) from May 2015 to January 2017 in the periurban area of Kalaban Coro located southeast of Bamako in Mali. The cases were severely malnourished children under five years of age and were recruited from the unit of recovery and nutritional education (URENI) of the Kalaban Coro reference health centre. The controls were children of the same age group with no form of acute or chronic illness that can modify the gut microbiota, such as fever, diarrhoea, and/or antibiotic intake, within fifteen days before inclusion.
To test the reversibility of the absence of M. smithii in cases and controls, we attempted to obtain a second sample. In the group of malnourished children, this second sample was collected after the acute phase, i.e., after recovery from acute clinical complications (diarrhoea, dehydration, fever, gastroenteritis, respiratory infections), return of appetite with an intake of at least 75% of the daily RUTF ration required for the child and a weight gain of 15 to 20% compared to the entry weight 29 . For control children, a second sample was collected 15 days after the first.
Malnourished infants who did not give stool samples before renutrition were excluded as well as those with acute or chronic illness that may explain their nutritional status. Cases of refusal of consent were also excluded. Case and control children were classified by age range of 0-6 months, 7-12 months, 13-24 months and > 24 months. Data sources/measurement/definitions. Anthropometric parameters, including weight, height, midupper arm circumference (MUAC) and age, were measured for all participants to determine the nutritional status of the children. We also calculated the weight-for-age, weight-for-height and height-for-age z-scores using the WHO Anthro software (https ://www.who.int/child growt h/softw are/en/) according to the date of inclusion, gender, date of birth, height measurement recumbent or not, and the presence or absence of oedema. Based on the WHO 2009 severity criterion on acute malnutrition (30), including weight-for-height Z-score (WHZ), weight-for-age Z-score (WAZ), height-for-age Z-score (HAZ), mid-upper arm circumference (MUAC) and the presence of oedema, cases were defined by WHZ <-3 standard deviations (SD), by the presence of nutritional oedema and/or by the MUAC < 115 mm for children over 6 months. Clinical data including temperature (to www.nature.com/scientificreports/ detect fever), respiratory symptoms and digestive symptoms among which diarrhoea, emesis and gastroenteritis were collected. Moreover, the presence of HIV infection as well as malaria was recorded. Control children were enlisted in health centres during health monitoring or in the Kalaban Coro health district with anthropometric parameters meeting WHO standards including WAZ > -2 SD, HAZ > -2 SD, WHZ > -2 SD, and MUAC ˃ 125 mm in children older than 6 months and without known disease and who were asymptomatic and without oedema. All the children were screened during the same period and in the same geographical area to address potential sources of bias. A multivariate analysis including age and gender as confounding factors was also performed. In addition, we calculated the size of our sample by referring to the proportion of children positive for M. smithii in controls and in children with severe acute malnutrition in our previous study from Niger and Senegal (15/20 in controls vs 0/20 in cases) 28 . According to Fleiss with continuity correction 38 , we needed 8 malnourished compared to 8 controls to be powerful enough (80%) to confirm or infirm the results published in our previous work using a 95% two-sided confidence level 28 . Here, we included 143 cases and 110 controls, i.e. sample sizes well beyond what was required for this study.
Management of severe acute malnutrition. Management of severe acute malnutrition in Malian URENI (Unités de Récupération et d'Education Nutritionnelle Intensive) consists of three phases, which include, on one hand, nutritional cure with therapeutic milk and ready-to-use therapeutic food (RUTF) and, on the other hand, medical treatment with antibiotics, antiparasitic drugs including antimalarial, and vitamin A supplementation as described by the PECIMA, a programme on the integrated management of severe acute malnutrition (Fig. S1). To avoid therapeutic gut microbiota alteration, faecal samples were collected at admission before administration of any anti-infectious drugs.
Variables and parameters collected/techniques. Gut methanogenic archaea quantification. We performed M. smithii detection by targeting the 16S rRNA gene using real-time quantitative polymerase chain reaction using the optimized protocol of Dridi et al. 17 .
Real-time quantitative polymerase chain reaction. DNA was extracted manually from 30 mg of feces using the E.Z.N.A. Tissue DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer's instructions. The total DNA extracted was pure and diluted to the tenth and one-hundredth for real-time quantitative PCR, mainly targeting the 16S rRNA gene of M. smithii and F. prausnitzii, the sohB gene of E. coli and the nucA gene of S. aureus (Table S3). All PCRs were performed with a positive control series (plasmids) and negative controls (mix). Specific primers and probe systems were used for amplification (Table S2). Real-time PCR was performed in a total volume of 20 μL, including 10 μL of master mix (Roche Diagnostics GmbH, Mannheim, Germany), 3 μL of distilled water, 0.5 μL of primer Fwd, 0.5 μL of primer Rev, 0.5 μL of probe, 0.5 of uracil-DNA glycosylase (UDG) and 5 μL of DNA. The amplification reactions were performed using the Roche protocol, which consisted of in two minutes at 50 °C, five minutes at 95 °C followed by 40 cycles of five seconds at 95 °C and 30 s at 60 °C and analysed using the CFX96 real-time PCR detection system (Bio-Rad Life Science, Marnes-La-Coquette, France). The real-time PCR results were considered negative in the absence of an amplification curve.
Statistical methods. The data collected on a questionnaire (supplementary material) were entered in Microsoft Excel and analysed using SPSS software version 20.0 (IBM, Paris, France), SAS 9.4 statistical software (SAS Institute, Cary, NC) and GraphPad Prism 8.0 (GraphPad software, La Jolla, USA). Descriptive statistical analyses were performed for all parameters. The normality test of Shapiro-Wilk was first used for the distribution of quantitative data to apply parametric tests (t tests) or nonparametric tests (Mann-Whitney-Wilcoxon test). The chi-square test was used to test the differences in proportion between groups. All tests were two-tailed. The threshold of significance was set at a value of p ≤ 0.05.
Despite our attempt to recruit controls within the same age group and sex as the cases, controls were still not matched within age categories (p < 0.001, Table S1). In order to control for this confounding factor, we used a propensity score weighting approach on our entire study population. The propensity score was calculated using logistic regression on the age groups. The predicted probabilities from the propensity-score model were used to calculate the stabilized inverse-probability-weighting weights 39 . Associations between groups (cases/controls) and the different variables were then estimated using weighted regressions (normal or logistic depending on the outcome). The positive predictive value (PPV) of M. smithii according to age for healthy and malnourished children was calculated. PPV is an estimate of the specificity and sensitivity of a variable, calculated using the following formula: number of true positives/(number of true positives + number of false positives) 40 . We performed linear regression models on the DNA concentration to analyse the dynamics of M. smithii relating to age and to compare the speed of expected increase. For this purpose, the slope difference with the horizontal was determined to evaluate to what degree the DNA concentration of M. smithii was different from zero in each group and between groups. We performed these calculations on the slope of linear regression following the method described by Zar 41