Parasites and diet as main drivers of the Malagasy gut microbiome richness and function

Interactions between the prokaryotic microbiome and eukaryotic parasites in the vertebrate gut may affect overall host health and disease. While intertropical areas exhibit a high rate of parasites carriers, such interactions are understudied in these populations. Our objectives were to (1) describe the gut microbiome of individuals living in Madagascar, (2) identify potential associations between bacterial taxa and parasites colonizing the digestive tract and (3) highlight main determinants of the gut microbiota composition in this developing country. Metadata (socioeconomic, diet, clinical) and fecal samples were collected from 219 volunteers from North-West Madagascar (Mahajanga). Fecal microbiome was assessed through 16S rRNA gene sequencing and metabolomics, and related to dietary habits and parasites carriage. We highlight important Malagasy gut microbiome peculiarities. Out of three detected enterotypes, only one is similar to that observed in Westernized countries (Ruminococcus-driven). Functions associated with the two others (Clostridium sensu stricto-driven and Escherichia/Shigella-driven) are mostly directed toward amino acids biosynthesis and degradation, respectively. Diet and protozoan carriage were the main drivers of microbiota composition. High protozoan carriage was associated with higher diversity, richness and microbial functionalities. The gut microbiome of Malagasy strongly differs from that of Westernized countries. Asymptomatic protozoan carriage and dietary habits are the external factors with the deepest impact on gut microbiome. Further studies are needed to understand whether gut microbial richness constitute a predilection niche for protozoans colonization, due to their gazing features, or whether the parasites themselves induce a higher bacterial richness.

www.nature.com/scientificreports/ Parasitic eukaryotes members of the human microbiome are often underestimated and underexplored, even though their importance in the gut ecosystem, mainly as immunity drivers, is widely accepted. Protozoans' colonization ranges from 5 to 100% of the population, depending on geographical origin, with many carriers remaining asymptomatic. Some parasitic protozoa, such as Entamoeba histolytica, Giardia intestinalis, and Tritrichomonas suis, can modify intestinal mucus abundance and composition, enabling them to penetrate the mucus barrier during pathogenesis 5 . Many gut protozoa can also damage the epithelium during pathogenesis (disruption of tight junctions, cell invasion and destruction) 6 . Both mucosal and epithelial damages will directly alter the host's interaction with their microbiome during infection. However, asymptomatic parasitic colonization also impacts the host microbiome and one of the most studied organisms is Blastocystis sp. (abbreviated as Blastocystis in the present paper). While historically considered as parasite and assumed to have a detrimental effect on the host organism, the vision of a potential commensalism and beneficial role of Blastocystis starts to emerge 7 . A recurrent finding is the robust link between gut colonization by Blastocystis and higher microbiome diversity and/or richness [8][9][10][11][12] , together with a negative association with the Bacteroides genus abundance and Bacteroides enterotypes 9,12,13 .
Still, these results are seen through the prism of westernized countries (France, Sweden, Belgium and Mexico) and compiled data from shotgun metagenomic studies. Considering the higher rate of parasitic carriers in intertropical area, we focused our research on volunteers from Madagascar and previously described a wide panel of intestinal parasites 14 . Protozoa were the most prevalent with 72.8% of the population colonized whereas helminths and microsporidia were detected in less than 10% of the volunteers. Blastocystis was the most prevalent protozoa followed by various amoebas and flagellates.
The objectives of the present study were to (1) describe the gut microbiota composition of individuals living in Madagascar, (2) identify potential associations between gut bacterial taxa and parasites colonizing the digestive tract and (3) highlight main determinants of the gut microbiota composition in this developing country.
It is, to our knowledge the first paper intending to describe the general gut microbiome profiles of healthy adults from Madagascar, and, together with the Flemish gut project 12 , linking parasitic status to microbiome in a cohort of more than 200 individuals.
Intestinal colonization by parasites strongly affects the gut microbiome. The contribution of demographical (n = 5), medical (n = 8), dietary (n = 7) and parasitism status (n = 14) factors on inter-individual microbial composition (β-diversity) was assessed by PERMANOVA analysis. At a false discovery rate of 0.1, 19 factors were associated with dissimilarity in microbial composition (Fig. 1D). Altogether, these factors explained 25.79% of the observed variation between individuals. Neither bacterial diversity nor bacterial richness correlated with age, gender or BMI, while variation in the microbiota composition within Malagasy population was slightly but significantly affected by individual's age (1% of total microbiota variance).
Parasitic status accounted for most of the bacterial microbiome dissimilarity (8.39%) and especially Entamoeba coli (1%), Blastocystis (0.94%), Entamoeba histolytica/dispar (0.73%) and Chilomastix mesnili (0.7%). Within our cohort, 85% of individuals were colonized by one or more parasites. Blastocystis was the most prevalent (77.6% of the cohort) and frequently associated with Dientamoeba fragilis (22.8%). While most parasites were associated with higher bacterial richness, microbiome diversity was also significantly increased in individuals colonized by Blastocystis sp., Entamoeba coli, and Enteromonas (Additional file 2: Fig. S2A). On the other hand, individuals positive for Blastocystis, Ameoba and C. mesnili displayed lower bacterial load. Strikingly, cumulative parasite colonization of the Malagasy gut increased bacterial richness and diversity ( Fig. 2A). Relative abundances of 21 taxa shifted according to the presence of parasites with Blastocystis and E. coli having the greatest impact on gut microbiome. Colonized patients had higher relative abundance of commensal bacteria but lower proportions of Streptococcus and Lactobacillus (Fig. 2C). A higher parasitic cumulative diversity was associated with a significantly lower abundance of Lactobacillus (Fig. 2B), and with a low sanitary level (Additional file 2: Fig. S2B).  Fig. S2C). On the contrary, most of the individuals colonized by three or more different parasites clustered within the Ruminococcus-driven Ent1 (Additional file 2: Fig. S2D).
We further investigated the relative importance of parasites symptomatology and carriage on the gut bacterial composition. While for most parasites, asymptomatic and symptomatic carriers share a similar microbiome, individuals colonized with Blastocystis and presenting gut symptoms of blastocystosis exhibited lower relative abundance of both Acetanaerobacterium and Flavonifractor (Additional file 2: Fig. S2E) than asymptomatic carriers.
Dietary habits strongly impact the Malagasy gut microbiome. Dietary habits had a significant impact (6.74%) on the inter-individual microbiota composition (Fig. 1D). From all collected dietary information, daily liquid, rice and fruit consumptions had the most significant influence on gut microbiota variance (1.54%, 1.54% and 1.64% respectively). Low liquid consumption (< 1 L/day) was associated with significantly higher proportions of Senegalimassilia, Roseburia, Haemophilus and Actinomyces (Fig. 3A). Microbiota of individuals not eating fruits harboured higher percentages of bacteria from Actinomyces, Bifidobacterium and Acetanaerobacterium (Fig. 3A) whereas moderate fruit eaters (1-2) displayed a higher abundance of Prevotella, Haemophilus and Lactobacillus. Daily consumption of 3 or more fruits increased the abundance of Sporobacter, Oscillibacter, Flavonifactor, Coprococcus, Clostridium IV and Butyricicoccus genera.
Malagasy gut bacterial load was significantly lower in individuals eating either meat or fish meals (Fig. 3C). Individuals eating both meat and fish harboured higher proportions of Eubacterium and Olsenella than www.nature.com/scientificreports/ non-eaters. Microbiota of meat eaters was characterized by less Actinomyces, whereas fish eaters' microbiota exhibited more Acetanaerobacterium and Senegalimassilia. Low to moderate daily rice consumers (< 400 g/day) had higher relative abundance of bacterial species related to Lactobacillus, Clostridium XIVa and Sporobacter (Fig. 3A). Yogurt consumption had also a deep impact on the microbiota. Both homemade and industrialized yogurt consumption led to significantly higher abundance of Bifidobacterium and Lactobacillus. Thirteen other genera were significantly associated with homemade yogurt consumption, most of them being less represented. Dietary habits were overall similar between Ent1 and Ent2 individuals whereas Ent3 was characterized by higher volume drinkers, lesser fruit and rice eaters and more homemade yogurt consumers (Fig. 3B).

Metabolism of the gut microbiome in Malagasy individuals. A global metabolomic approach has
been applied on a subset of fecal samples (nEnt1 = 5, nEnt2 = 7, nEnt3 = 6) and 256 peaks were detected. Similarly to gut microbiota composition, metabolomic profiles allowed distinguishing three groups of individuals (Fig. 4A). A total of 59 peaks, belonging to six classes (Amino Acids, Peptides, and Analogues-Aromatic Compounds-Aliphatic Compounds-Carboxylic acids and derivatives-Lipids-Organic Acids and Derivatives) were differentially represented between enterotypes (Fig. 4B). Specific metabolites associated with each enterotype are described in Additional file 4 (Table S2). Metabolites relating to (1) phenylalanine, tyrosine and tryptophan biosynthesis, (2) tryptophan metabolism, (3) valine, leucine, isoleucine biosynthesis and (4) lysine biosynthesis were specific of Clostridium-driven Enterotype (Ent2). On the other hand, Pantothenate and CoA biosynthesis as well as beta-Alanine metabolism were distributed in both Ent2 and Escherichia-driven Ent3. Finally, metabolites associated with Ruminococcus-driven Ent1 related to metabolism more than biosynthesis (Glycine, serine, threonine metabolism-Cyanoaminoacid metabolism-Riboflavin metabolism).
A statistical inference of metabolic functions with 16S rRNA data was also performed with PICRUSt to cover the overall population. The most important shift in terms of microbiome gut functions was observed between Ent3 and the two other Enterotypes (Fig. 4C and Additional files 4 and 6: Tables S2 and S4). Xenobiotics biodegradation and metabolism such as aminobenzoate, fluorobenzoate, toluene degradation were enriched in Ent3. The metabolism of Terpenoids and Polyketides in Ent3 was marked by the biosynthesis of siderophore group non-ribosomal peptides whereas Ent1 and Ent2 harboured more functions classified in the biosynthesis of ansamycins, butirosin and neomycin. Ent3 was characterized by a lower abundance of functions related to pentose phosphate pathway, starch and sucrose metabolism. The amino acid metabolism was totally directed www.nature.com/scientificreports/ to the catabolism of gluthatione, lysine, tryptophan, valine, leucine and isoleucine. On the opposite, Ent1 and 2 were more prone to the biosynthesis of these amino acid units. The increased amino acid biosynthesis observed in Ent1 and 2 was interconnected with enriched protein export capabilities. Functions related to replication and repair, transcription and translation were less represented in Ent3. Similar results were observed for membrane transport and signal transduction functions. Metabolic functions deduced from PICRUSt were also screened in Malagasy gut microbiomes in accordance to increasing asymptomatic protozoans carriage ( Fig. 4D and Additional files 5 and 7: Tables S3 and S5). Microbial functions such as replication and repair, translation, amino acid and lipid metabolism, protein export and terpenoid backbone biosynthesis were enriched in individuals with higher protozoans carriage (n ≥ 3). On the opposite, xenobiotics biodegradation and metabolism such as aminobenzoate and chlorocyclohexane and chlorobenzene degradation together with unclassified amino acid metabolism related functions were increased within microbiome of low protozoans carriers (n < 3).

Discussion
The gut microbiome of people from the tropical region of Mahajanga, an urban community of Madagascar, strongly differs from that of westernized countries or isolated populations. While westernized countries microbiota can be clustered into three or four enterotypes 15 driven by Bacteroides, Prevotella and Ruminococcus, only the latter genus could be described as an enterotype driver in this population. However, the microbiome was still Red, blue and green boxes indicate whether the abundance is higher, equal or lower in individuals positive for the listed food type. ANOVA (group > 2) and Wilcoxon test (group = 2) were run all over genera count dataset according to diet. Only boxes tagged with a star are considered statistically informative. The heatmap was generated using R software (https:// cran.r-proje ct. org/) with in-house code that uses basic functions included in "stats" and "graphics" packages. It uses genera abundance (as listed in rows) and highlights, for daily consumption of liquid, rice and fruit items, the category that are the most significantly different from others. For Meat, Fish, Yogurt-HM and Yogurt-I items, the heatmap indicates genera whose abundances are increased/ decreased in individuals consuming these items. www.nature.com/scientificreports/ statistically stratified into three enterotypes, the other two being driven by Escherichia/Shigella and Clostridium sensu stricto. Yet while Prevotella still accounts for approximately 7% in average of total bacteria in the overall population (median = 3%), proportions of Bacteroides are lower than 5% (4.4% and median = 0.99%). Similarly, Pasolli et al. applied shotgun metagenomic gut microbiome profiling of two rural communities with non-Western lifestyles from north-eastern Madagascar and detected that the species-level genome bins profiles of the Madagascar population were profoundly different from that of Western-style populations 16 . A substantial proportion of genomic species related to Prevotella while Bacteroides was nearly absent as compared to westernized countries. Both our study and Pasolli et al. highlighted a depletion of Alistipes, Parabacteroides and Akkermansia in the Malagasy gut microbiomes.
Diet strongly explains this distribution. High proteins and animal fat intake has previously been associated with higher abundance of the Bacteroides dominant enterotype 17 . Its absence in Madagascar probably reflects a diet mostly based on rice. Fruits consumption had the highest impact on microbiome variance within the population. Fruits and vegetables are important sources of dietary fibers that will undergo microbial metabolism in the digestive tract. Fibers are known to increase microbiota diversity 18,19 . Consuming two or more fruits per day in the present population was associated with higher levels of Firmicutes and butyrate-producing bacteria such as Prevotella or Coprococcus. The gut microbiota of individuals ingesting important rice portions (≥ 500 g/ day) was enriched in Ruminococcus, Sporobacter and rice consumption was higher in individuals belonging to Ent1. Individuals within this enterotype also ingested less yogurt (both home-made and commercial). Yogurt was associated with significantly higher proportions of Lactobacillus and Bifidobacterium, and lower levels of Prevotella and Enterorhabdus. www.nature.com/scientificreports/ At a functional level, Ent1 and Ent2 were mostly enriched in functions dedicated to the biosynthesis of several amino acids as well as central bacterial cell mechanisms such as replication and repair, translation and transcription. On the opposite, numerous functions sustaining degradation of amino acid units and xenobiotics metabolism were enriched in Ent3 as compared to Ent1 and Ent2. This observation is coupled with higher metabolism of nitrogen compounds. The energy metabolism was more directed towards nitrogen metabolism in Ent3 while carbon fixation in photosynthetic organisms and methane metabolism were increased in Ent1 and 2. Previous reports have indicated that aminobenzoate metabolism influences the microbial composition and intermediaries of aminobenzoate degradation are able to enhance a stress response 20 and promote the growth and virulence of Enterobacteriaceae 21 . Similarly, the metabolism of aromatic amino acid such as tryptophan, over-represented in Ent3, has been documented for the noxious potential of derivative compounds issued from tryptophan degradation on the immune system and neuronal functions 22,23 .
Ent3 harboured less functions classified in the biosynthesis of ansamycins, butirosin and neomycin. Neomycin is a well-characterized antibiotic with excellent activity against Gram-negative bacteria. This could partly explain the Escherichia/Shigella specificity of Ent3.
A main finding of our study is the robust link between the cumulative number of colonizing parasites and the increase gut microbial diversity and richness. An association between Blastocystis colonization and microbiota diversity has previously been described in both westernized and developing countries 12,13 . In more than 600 healthy European (Flemish adults cohort), Tito et al. reported a higher bacterial diversity associated with the detection of Blastocystis. Moreover, Blastocystis prevalence was unevenly distributed over Westernized countrytype enterotypes, with the Bacteroides clusters comprising significantly less Blastocystis-carriers compared with both Ruminococcaceae and Prevotella enterotypes. This is in line with our observation since > 75% of our cohort was colonized by Blastocystis and we would not depict a Bacteroides cluster in this population, underlying the link between this enterotype and the absence of Blastocystis. We here highlight for the first time that gut colonization by a multi-species community of parasites is associated with higher microbiota richness, increasing with the number of different parasites (from 1 to more than 5 different parasites species). Interestingly, while microbial richness increased with the number of parasites colonizing the gut, this was inversely associated with proportions of Lactobacillus, which were significantly increased in yogurt consumers. Hence, one could speculate that yogurt consumption might, in this specific population, be disadvantageous for parasitic colonization and microbiome diversity and richness. An impact of Lactobacillus on fungi and parasites growth was previously described, mostly in the vagina, as a protective therapeutic option against candidiasis 24 . A key role of hydrogen peroxide and lactic acid production in this antifungal function is well documented. However, lactic acid was not detected in the metabolomic dataset. At the functional level, higher parasites carriage was associated with enrichment of metabolic activities related to replication and repair, translation and transcription. On the opposite, xenobiotics biodegradation and metabolism (amino acid metabolism, aminobenzoate and chlorocyclohexane/chlorobenzene degradation) were enriched in individuals harbouring low parasite diversity.
Finally, our results strengthen the observation made by Audebert et al. 8 on a potentially positive association between parasitic colonization and microbial diversity and richness. The features observed in Ent3 individuals, at both compositional and functional levels, emphasize potential health complication notably gut inflammation.
Eventhough one cannot presently decipher the causal link between these two types of microorganisms, presence of gut parasites may partly be responsible or associated with the specific microbiome described in this population. While Pasolli et al. suspected a role of the nutrition to explain the very low proportion of Bacteroides, toward increased proportions of Proteobacteria, mostly Enterobacteriaceae and Succinivibrionaceae (previously undescribed bacterial genus Succinatimonas) 16 , we demonstrate that parasitism is a very strong driver of the human gut microbiome composition, diversity and functionality, which overtake the impact of diet.

Conclusion
On the overall, our study highlights the existence of three specific enterotypes in the Malagasy population. Whereas diet, as previously described in other reports, is one of the most important drivers of the gut microbiota composition, individual protozoans carriage (understudied in westernized countries) accounts for most of the dissimilarity observed in the microbiome of people living in Madagascar. Further studies are needed to understand whether gut microbial richness constitute a predilection niche for protozoans colonization, probably due to their gazing features, or whether the parasites themselves induce a higher bacterial richness. Animal studies may help answering this question.

Materials and methods
Study population and design. The study was conducted prospectively from February to June 2015 in Mahajanga, North-West Madagascar, and its rural neighbourhood 14 . The region is characterized by tropical climate. Fecal samples from 219 Malagasy individuals, either healthy volunteers or patients hospitalized in the gastroenterology department of the university hospital PZaGa of Mahajanga, were collected. Two grams of feces were aliquoted and stored at − 20 °C before being routed to France in dry ice. Physicians concomitantly documented individuals' socioeconomic, demographic and medical information, as well as dietary habits from each individual (Table 1). Exclusion criteria included the use of antibiotic or probiotic 2 months prior to inclusion and/or an age under 18 year-old.
Data are summarized according to protozoan carriage. Individuals' age and BMI are expressed as the mean value computed for each groups. Sex is expressed as the male to female ratio. Protozoan detection, dietary habits, sanitary levels and geographic areas are expressed as the percentage of individuals encountered in each group of parasites load. Information related to the geographic areas can be found in Greigert et al. 14 27 . Reads were further corrected for known sequencing error using SPAdes 28 and then merged using PEAR 29 . OTUs were identified using a Vsearch pipeline 30 set up to dereplicate (-derep_prefix -minuquesize 2), cluster (-unoise3), chimera check (uchime3_denovo) the merged reads. OTU taxonomical classification was performed using both classifier and seqmatch from RDPTools suit 31 . The training sequence data set required for OTU classification at species level was downloaded from EzBioCloud 16S database for QIIME 32 . Metagenomic profile inference from 16S dataset was determined using PICRUSt 33 .
Parasitological examination. All fecal samples were processed in the 30 min following stool emission using both culture (Dobell and Laidlaw biphasic medium) and culture-independent approaches as previously described 14 . Briefly, culture-independent assays relied on microscopy examinations of both non-and concentrated faeces (MIF and Faust methods). Molecular diagnosis was carried out for Blastocystis, D. fragilis, Cryptosporidium, E. bieneusi and E. intestinalis. We discriminated between E. dispar and E. histolytica by using conventional PCR. Detailed information, including primer sequences, can be found in Greigert et al. 14 . Stool samples were considered positive if helminth eggs, larvae, cysts and/or trophozoites of protozoans were detected by at least one of the four conventional methods and/or a positive molecular diagnosis.
Microbial metabolomics. Eighteen fecal samples were selected for metabolomics analysis by Profilomic (http:// www. profi lomic. com/ fr). Aliquots (50 mg) were suspended in 50 µL of water and 400 µL of methanol, centrifuged at 4 °C and 10.000g for 15 min and precipitated for 1 h 30 at 4 °C. Following second centrifugation and cleaning, supernatant was evaporated under N 2 atmosphere. The pellet was suspended in 250 µL mix of water and AcN (40/60) containing internal standards. The solution was centrifuged at 20,000g and 4 °C for 15 min and injected in the mass spectrometer (Q-Exactive-ThermoFisher Scientific) coupled to liquid chromatography (aSequant ZIC-pHILIC-Merck) machine (LC-HRMS). The analysis was run at high resolution (70,000 FWHM) with alternating negative and positive ionisation mode. Signals were annotated using mass and retention time from each spectrum and a metabolite database issued from Profilomic.
Statistics. Statistical analyses were run using R programming language and software together with gplots, gdata, vegan (http:// cran.r-proje ct. org/ packa ge= vegan), ade4 and phangorn 34 packages. OTU counts were normalized via simple division to their sample size and then multiplication by the size of the smaller sample. α-Diversity and richness were estimated using diversity and estimateR. Distance matrix for β-diversity analysis was computed using vegdist and Bray-Curtis method. Principle Component Analysis and Principle Coordinate Analysis were computed either on compositional data or on distance matrix data using dudi.pca and dudi.pco respectively. Effects of demography, diet, presence of parasite on β-diversity were assessed using adonis function together with Bray-Curtis distance matrix. Wilcoxon test, Kruskal Willis test and anova were used as required to detect differences between groups of variables. P-values were corrected as necessary using False Discovery Rate correction. Enterotypes were defined as previously published 15 .
Ethics approval and consent to participate. This study was approved by the Malagasy Ministry of Public Health (reference number 33-MSANP/CE). Written informed consents were obtained from all participants. This study was conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki).

Data availability
Data are available from the corresponding author. 16S rRNA gene reads are publicly available from NCBI SRA under Bioproject accession number PRJNA600229.