According to the World Health Organization (WHO), there are nearly 1.7 billion cases of diarrheal disease every year, and 760,000 deaths, representing the second cause of death in children under 5 years of age (World Health Organization, 2014). Diarrheagenic Escherichia coli (DEC) are the most common bacterial cause in both developing and industrialized regions, primarily affecting children under 5 years of age (Gomes et al., 2016). DEC infection prevalence has been underestimated, largely due to the lack of surveillance studies and the limited availability of molecular diagnostic techniques. Implementation of molecular techniques allowing simultaneous detection of multiple enteric pathogens, including DEC pathotypes, have shown that the prevalence of these pathotypes in diarrheal infections is higher than previously thought (Buss et al., 2015; Farfán et al., 2016).

DEC strains comprise a group of E. coli strains that have acquired several genes conferring the ability to cause a broad spectrum of infections (Nataro and Kaper, 1998). According to the detection of the virulence factors, six classical categories or pathotypes have been described: enteropathogenic E. coli (EPEC), Shiga toxin-producing E. coli (STEC), enterotoxigenic E. coli (ETEC), enteroaggregative E. coli (EAEC), enteroinvasive E. coli (EIEC), and diffusely adherent E. coli (DAEC). In the last years, new E. coli pathotypes have emerged as a consequence of a unique combination of virulence traits (Rasko et al., 2011). In general, DEC infection involves three steps: (i) adherence and colonization of the intestinal surface, (ii) production and secretion of virulence factors, and (iii) diarrhea and inflammation (Nataro et al., 1996). These steps are tightly regulated, either by environmental conditions or bacterial regulators that induce or silence the expression of virulence factors (reviewed in Gomes et al., 2016). The expression of specific virulence genes and the ability to efficiently utilize nutrient in the intestinal milieu are key events for the successfully colonization of DEC to intestinal epithelia. However, this phenomenon is not completely understood for DEC, as most studies have focused on molecular mechanism occurring inside the bacteria, with only a few evaluating environmental factors, such as, gut microbiota (Pacheco and Sperandio, 2015; Singh et al., 2015). In the last years, several groups have focused on the role components of the gut microbiota and its association with diarrheal disease caused by enteric pathogens, including DEC strains (Nelson et al., 2012; Curtis et al., 2014; Iversen et al., 2015; Lindsay et al., 2015).

The gut microbiota is proposed to play an important role in human health by acting as a barrier against pathogens, stimulating metabolic function and promoting the development of the immune system (Hooper et al., 2012). Conversely, recent observations strongly suggest that diarrheal infections caused by DEC induces changes in the intestinal bacteria microbiota, and that specific components of gut microbiota can modulate the expression of virulence genes leading to enhancement of the infectious process (Curtis et al., 2014; Singh et al., 2015). Next Generation Sequencing (NGS) platforms have facilitated the quantitative phylogenetic identification of the microbiota at various taxonomic levels (Fraher et al., 2012). Analysis of 16S rRNA gene sequences based on percent identity, using Operational Taxonomic Units (OTUs) or Operational Phylogenetic Units (OPUs) for taxonomic assignment, has facilitated the identification of known and potentially new microorganisms (Yarza et al., 2014).

The epidemiological relevance of DEC infections unpaired by only a small number of studies evaluating DEC associated gut microbiota lead us to characterize the presence and relative abundance of indicative taxa associated with DEC infection in the intestinal microbiota of children under 5 years of age. The main goal of this study was to identify a distinctive gut microbiota in diarrheal samples positive for DEC pathotypes compared to diarrheal samples positive for enteric virus and stool samples from healthy children.

Dysbiosis of the gut microbiota has been identified in patients with acute diarrhea, but association with specific pathogens is lacking. In this study, we found that stool samples of children positive for DEC pathotypes harbored a defined microbiota composition different to that found in children suffering an enteric virus associated diarrhea episode or in healthy children.

Our methodological approach was based on comparison of OPU composition differing from most studies which have been based on an OTU approach. The accuracy of OTUs for proper taxa assignment has been questioned (Guazzaroni et al., 2013) and in previous reports, the use of OPUs for taxa inference resulted in more reliable taxon assignment reducing overall diversity, providing clearer taxonomic data (Vidal et al., 2015). In our study, rarefaction curves showed that OTUs overestimated diversity while OPU based taxa assignment reached a maximum that was not influenced by adding more sequences (Figure 1). A high threshold for taxa assignment allowed us to identify 852 indisputable OPUs and manual review for identity between sequences, including the best available references (Yilmaz et al., 2014) and phylogenetic environment allow high confidence in assignments. The OPU approach allowed building a phylogenic tree for gastrointestinal microbiota from Chilean children. The tree was based on alignments with well-classified sequences, followed by taxon assignment by phylogenetic inference.

In contrast to other studies, ß-diversity was similar between groups (Hill and Artis, 2010; Pop et al., 2014). In our case, the similarity between ß-diversity indexes does not represent a similar diversity of groups. As previously postulated (Jost, 2006), true diversity cannot be explained by indexes, while a deep analysis of taxa abundance and richness may allow better understanding of group compositions. In our study, RDA analysis on OPUs allowed to show clear differences between groups, and identified a unique gut microbiota associated with DEC infections compared to enteric virus infections and healthy children. Some taxa were highly abundant and others were rare in DEC samples, probably due to different DEC pathotypes present in the stool samples. We normalized the data by double-square root transformation allowing the use of multivariate analysis based on correlation of data rather than raw distances. RDA and indicative taxa analysis confirmed a differential composition and the presence of indicative taxa for each etiological group (Figure 4). Although all DEC pathotypes share a common E. coli ancestor, the virulence mechanisms employed by each pathotype are different and depend on the expression of particular virulence genes (Kaper et al., 2004). RDA analysis in DEC group showed a clustering by pathotypes (Figure 5A); such clustering was not observed in the enteric viral group (Figure 5B). Due to low number of samples positives for each pathogen, it was not possible to perform statistical analyses. Further studies including more samples from each of the various DEC pathotypes would allow to determine the microbiota associated to each DEC pathotypes and define if these infections should be analyzed individually or as a group.

Among healthy children the gut microbiota has proven to be highly diverse during the early months of life. As age increases, microbial communities become more similar converging toward a generic adult-like profile, characterized by a preponderance of Bacteroides and Firmicutes and a low abundance of Proteobacteria (Palmer et al., 2007; Tojo et al., 2014). In diarrheal episodes, a change in phylum composition has been observed, with an increase in the abundance of Proteobacteria and Bacteroidetes and a decrease in Firmicutes (Braun et al., 2017). Our results are in agreement with these previous reports, as we found high proportion of Bacteroidetes and Proteobacteria in the DEC group compared to healthy group (Figure 2A). The low abundance of Firmicutes in diarrheal group could have an important impact in this children. Metatranscriptomic studies have shown that Lachnospiraceae and Ruminococcaceae families belonging to Firmicutes phylum, play an important role in dietary fibers metabolism. Likewise, Ruminococcaceae has been associated to antibiotic biosynthesis, important on host defense against pathogens (Gosalbes et al., 2011).

At the family level, a loss of diversity in the intestinal microbiota is observed during intestinal infections caused by diverse pathogens, with an increased abundance of Enterobacteriaceae compared to healthy subjects (Stecher et al., 2007; Hill and Artis, 2010; Arumugam et al., 2011). Concordant with these observations, in our DEC group all indicative taxa belong to the Enterobacteriaceae family. The presence of the genera Citrobacter, Enterobacter, and Haemophilus are consistent with previous observations (Pop et al., 2014). The abundance of Escherichia/Shigella species in the DEC group compared to viral and healthy groups might be attributable to the presence of DEC pathotype in the diarrheal samples.

The majority of the indicative species in the DEC group identified in this study have been previously associated with diarrheal episodes. Escherichia albertii is an emerging human enteropathogen and avian pathogen (Huys et al., 2003; Ooka et al., 2015), that shares the ability of EPEC and STEC to promote attaching-effacing (A/E) lesions due to the presence of locus of enterocyte effacement (LEE) genes (Yamamoto et al., 2017). Yersinia enterocolitica is a well-known gastrointestinal pathogen that causes yersiniosis, an illness characterized by diarrhea, ileitis, and mesenteric lymphadenitis (Gupta et al., 2015). Some species of the genus Klebsiella have been associated with diarrheal disease, for example, Klebsiella oxytoca is one of the ethological microorganism of antibiotic-associated hemorrhagic colitis (Zollner-Schwetz et al., 2008). It is important to note, that our analysis could not distinguished between pathogenic or commensal bacteria belonging to the same species. Previous reports have shown that components of the gut microbiota impact the expression of virulence factors. Bacteroides tethaiotaomicron, a predominant member of the gut microbiota in healthy subjects, enhanced the expression of several genes of STEC reference strain in co-culture experiments, including virulence genes located in the LEE pathogenicity island. Moreover, an increase of A/E lesions and STEC adherence to epithelial cells was found (Curtis et al., 2014; Iversen et al., 2015). Considering the above, further studies should focus on the role of species described here as indicative of DEC group in the pathogenicity of DEC pathotypes.

One of the strengths of our study is the construction of a phylogenetic tree. To our knowledge, this is the first phylogenetic tree built from reference sequences to study the gut microbiota based on OPU identification in diarrheal and non-diarrheal stool samples of children. This tree was manually supervised and can serve as a tool to characterize the taxa composition not only in samples associated with diarrhea, but also other gastrointestinal diseases. We used FilmArray® GI testing for pathogen detection; this diagnostic tool that has been increasingly used for detection of enteric pathogens, as it provides rapid, sensitive results for the most relevant enteric pathogens including bacteria, viruses and parasites. It is important to note, that FilmArray® GI testing is FDA-approved exclusively for diarrheal stool samples, and its use in stool samples from healthy individuals has not been validated.

Our study has limitations. Inclusion criteria were children aged 1–5 years with an episode of diarrhea. It is well known that several environmental and host factors, such as, age, food consumption, economic status, antibiotic treatment, and others might influence the gut microbiota composition (Sartor, 2012; Bäumler and Sperandio, 2016). Our healthy children were from the city of Colina, a middle-low income town in the outskirts of Santiago while the diarrhea cases were children consulting in a private clinic within the high socioeconomic sector of Santiago. Lack of matching for the other variables may also be playing a role in microbiome differences between groups. Even though we did not find significant difference in age between groups, the mean age of children in DEC group was lower than the other groups. Considering that variability of gut microbiota in children <3 years is high, age-matched studies with stricter inclusion and exclusion criteria are necessary to validated our findings.

In conclusion, we found a distinctive gut microbiota that is associated with DEC infections compared to enteric virus infections and healthy children. The identification of indicative species in the DEC group are relevant in advancing our understanding as to the involvement of the gut microbiota in DEC pathogenicity.

PG and MI processed the samples, performed sequence analysis, interpreted the data, and participated in manuscript writing. MF participated in study design, acquisition of data, interpretation of data, manuscript writing, and final approval of the manuscript. NC participated in data analysis. RR supervised our OPU approach, reviewed taxonomic inference, and helped with data analysis. RV participated in study design and data analysis. MO participated in the study design and data analysis.