Shifting from a gene-centric to metabolite-centric strategy to determine the core gut microbiome

Abstract

A key challenge in the area of determining how the microbiome communicates with the host's karyome is deciding which microbial functions should be studied. Ideally we would wish to look at functions which are not only important to the microbial host, but which also play roles in host physiology. Selecting the key microbial functions is essential to developing robust strategies to either promote or demote them, with the aim to enhancing host health. This commentary argues that the bottom-up approach is not providing the necessary gene-set from which we can start to develop a robust core microbiome and in fact we should adopt a top-down strategy in order to indentify the functions that are important and need further study.

Introduction

The current interest in the microbiota of the mammalian gut has been driven by the realisation that, what we thought were the major bacterial members of the community are actually those that are easily cultured1 and the gut microbiota may act as an extension of the host's karyome and potentially plays a role in maintaining a healthy host2,3 and possibly initiating diseases not only of the gut, but external to this organ.4 The first point has led to the implementation of culture-independent methods, which were mainly developed to explore the roles that microbes play in non-animal ecosystem function, while the second has resulted in a “gold-rush” to try and find microbes responsible for many of the developed countries' diseases, such as cardio-vascular disease,5 colorectal cancer,6 obesity,7 atopic diseases8 (asthma and eczema) and some behavioral diseases9 (stress and depression). The overall aim is to determine which functions are assets and which are liabilities and to maximise and minimise them respectively. In respect of this strategy we have quite a comprehensive knowledge base for the members of the gut microbiota responsible for synthesising butyrate and other short chain fatty acids,10–12 but beyond this role we have a limited understanding of the other functions which may be important to the host.

The Current Approaches Used to Map the Bacterial Landscape of the Gut

Many of the initial forays which aimed to measure gut bacterial diversity were based on community “profiling” methods such as denaturing/temperature gradient gel electrophoresis13 which provided an insight, albeit at low resolution, into the community structure of the gut microbiota. Surprisingly the gut microbiota were relatively stable over time and host specific, which made life difficult when comparing samples, as an individual generally became their own control, since inter-individual comparisons were almost meaningless when using species to discriminate between samples. The obvious next step was to develop inventories of the species present, by using the 16S rRNA gene as a proxy for a cultured isolate. While the technology was available (dideoxy based sequencing chemistry or 1^st generation sequencing technology), the costs, both time and financial, were prohibitive for large scale “deep” sequencing. However, small cohort projects were undertaken and one of the most cited (>1,000 citations in Web of Science) was able to capture the diversity in the colon. Eckburg and co-workers14 created inventories of 16S rRNA genes from 6 mucosal sites in the colon (cecum, ascending colon, transverse colon, descending colon, sigmoid colon and rectum) and a fecal sample from each of the 3 volunteers. In total they sequenced 13,355 partial genes products (average numbers of reads was 563, range 355–1,060) which provided the most in depth analysis of the distal gut and showed that this system is predominated by members of the Bacteria, predominantly from two phyla, Firmicutes and Bacteroidetes (Fig. 1) and principally one species of Archaea, namely Methanobrevibacter smithii. This distribution of the main phyletic groups in the distal colon has been accepted as a consensus for the general adult large intestine and regardless of the starting conditions (birth mode, feeding regimes, etc.,) the majority of infants will, after the first 18 months, show signs of reaching this climax community.15,16 Once we had established the range of phyla in the distal gut, which happened to be quite limited, it was hypothesised that there might be a set of bacterial species or core microbiota (which should be called the bacteriota, as no viral or micro-eukaryotic components were included in the census) which is shared by all humans. Several studies have attempted to determine what the members of this core would be and have reported mixed results17,18 while reports using 2^nd generation sequencing platforms (analysis of 16S rRNA genes primarily on the Roche 454 platform) have also addressed this topic,7,19–22 but with mixed conclusions. The consensus of opinion is that the core microbiota exists at a higher taxonomic levels, for example at the phylum or class, but at the lower levels the core is absent. So some researchers have started to move away from determining species and moved instead to looking at functions in the gut. The aim would be to determine whether there is a core microbiome rather than a core microbiota/bacteriota.

Functional Analysis of the Gut Microbiome

Determination of the functions in the gut can be assessed in one of two ways, assuming a culture independent approach is used. Either by sequencing random fragments of genomic DNA (gDNA) or by functionally screening a clone library of semi-random fragments of gDNA (mainly through heterologous expression of the gDNA from a surrogate host). Each method has its advantages and disadvantages which are outlined in Table 1 and will not be discussed any further in this commentary. Approaches based on sequencing have recently been used to undertake a comprehensive analysis of the distal gut microbiome23 and have shown that there exists a huge repertoire of unique genes in the distal gut, approximately 3.3 million and around 536,000 per individual gut. The authors concluded that the core microbiome consisted of over 6,300 functional orthologous groups which were found in all subjects analysed, however, we need to be clear on what was measured, as this is really the bacteriome and not the microbiome, no viral components or micro-eukaryotes were targeted. Interestingly they were able to determine that the shared species between all 124 subjects was very small, 18 species, and were all members of the Firmicutes or Bacteroidetes, thus supporting the notion that each person's gut is host-specific and a core bacteriota does not exist at the species level. However, the authors do stress that the depth at which the gut is sequenced sampled can strongly influence this set, so if groups are willing to spend the time and monies, we may answer this question in the near future. In the meantime the more interesting aspect of this study was the aim to establish the core functions in the distal gut, both at the individual bacterial genome and metagenomic levels. To this end the authors were able to show that a significant proportion of the functions they identified were involved in housekeeping roles and gut specific functions, such as adhesins and sugar harvesting, were noted, but this list was limited. In addition there were a significant number of genes which were unknown or hypothetical in the minimal gut genome and minimal gut metagenome, 73% and 80% respectively. This data, while of great value, does raise the central issue of this commentary, are these approaches actually addressing the issue of what microbial functions are important to both hosts? The short list of gut specific functions shows the difficulty in determining what is important to both hosts in this ecosystem. In respect of this it is suggested that we need to re-address the issue of what constitutes a core function. If we now view a core function as one that must benefit the bacterial host as well as interacting with the human host and is present in the majority of healthy distal guts (>50%), many simple housekeeping functions are not included. The need to undertake DNA replication, protein synthesis, amino acid metabolism or nucleotide metabolism are bacterial core functions, irrespective of where you find the bacterium. However, the need to process host molecules such as primary bile salts24 and metabolise exogenous and endogenous β-glucuronides25 are definitely core gut functions, since they benefit the bacterial host and interact with the host's physiology/biochemistry. While these two functions are ubiquitous in the gut bacteriome (Fig. 2) it can be difficult to determine the full repertoire of what should be considered in the list of core functions of the gut bacteriome. These two functions were established by analysis of cultured isolates and the diversity expanded using functional metagenomics. Similar studies have been undertaken to investigate the genes involved in butyrate synthesis26 and show a large degree of diversity and dispersion. However, it becomes difficult if no genes are identified, but there is a clear case for a microbial function which is gut specific and influences the host, for example being involved in cardio vascular disease or behaviour.5,27 In the example shown in Figure 2 genes have been chosen which have been clearly implicated in bacterial host function and at the same time eukaryotic host physiology, so some genes are considered to be involved in pathogenesis, e.g., genes from the Afa operon of the diffusely adhering E. coli,28 while the butyrate synthesising genes are mutualistic and are essential for the production of butyrate which is used by colonocytes in preference to any other energy source29 and the disposal of electrons by the bacterial host30 and thus fulfils all the criteria for being a core function. From this simple analysis it can be seen that we can start to define foundations of a core bacteriome, one that includes short chain fatty acid synthesis (proxies used were β subunit of CO dehydrogenase/acetyl-CoA synthase complex, butyryl-CoA:acetate CoA-transferase and formate-tetrahydrofolate ligase), bile processitivity (bile salt hydrolases) and β-glucuronide processing (H11G11 like β-D-glucuronidases). Moreover, genes involved in pathogenesis, e.g., from the AFA operon, are infrequent and not widely disseminated, which is to be expected. The key to establishing this set of genes is what should we include and what should we omit. Clearly a bottom-up approach does not help as we are swamped with house-keeping genes and unknown/hypotheticals. Hence a top-down approach needs to be implemented which guides us to the core bacteriome/microbiome.

Shifting to a Top-Down Approach to Help Define the Core Bacteriome

Current approaches work from a bottom-up, a gene to product information flow, which results in a bias in a metagenomic strategy towards genes which are common, but not necessarily true core functions. Working from the top of the information pile may deliver more relevant candidates which we can test as being members of the core microbiome. In this instance the information we are dealing with are the metabolites being produced by the microbes in the gut and which can be shown to be bioactive, which can either be small molecules such as the SCFA, proteins (BSHs, H11G11 like β-glucuronidases) or carbohydrates/polysaccharide.31 In order to determine these we need metabonomic and metabolomic approaches to determine the core metabolites in the gut and those that are due to diet, those that are variable in the population and those produced by the microbiota. Good examples of these lead metabolites are those derived from the metabolism of tryptophan metabolism,27,32 phosphatidylcholine,5 primary bile salts,33 amino acids5,34 and plant polyphenols.35 A list of candidate metabolites has been already been developed (Table 2) which would act as a starting point.36 A more comprehensive list of metabolites needs to be constructed, which are found in all guts studied and especially which are not found in humans, but are present in experimental animal models. For some members of the list we will have candidate genes and for others we will need to determine the functional genes responsible using a variety of approaches that are available and these will include culture-dependent as well culture-independent methods. A notable strategy to start and find organisms responsible for synthesising these metabolites would be to construct heat maps showing species or OTUs which correlate with specific metabolites37 and use this information to guide the culturing or metagenomic analysis.

Conclusions

The area of gut microbiology has come a long way since the early forays into this system using culturing, but the application of 2^nd and now 3^rd generation sequencing platforms has started to produce huge arrays of data and trying to determine what's important in the context of gut microbiology is daunting, we are struggling to see the wood for the trees. The metabolic phenotype of an individual, while very complex, contains valuable information that can be used to determine the core microbiome. In this commentary it has been suggested that the concept of a gut core function be restricted and that we should start to be more selective in what we call a core function. In addition we should use the metabolic data to inform us as to which functions we should be investigating and those to which we should be paying attention.

Figures and Tables

Figure 1 (A) Phylogenetic tree based on the combined human intestinal 16S rRNA gene sequence data set, the label for each clade includes, in order, the total number of recovered sequences, phylotypes and novel phylotypes (in parentheses; adapted from ref. 14). (B) Relative abundance of sequences from stool and pooled mucosal samples per subject: ■-other groups represents the Fusobacteria, Actinobacteria and unclassified near Cyanobacteria phyla, ■-Verrumicrobia and □-Proteobacteria; “M” denotes pooled mucosal sequences per subject and “S” refers to stool sample (adapted from ref. 16).

Figure 2 Tblastn analysis of the MetaHIT healthy dataset (n = 85) with functional genes from the distal gut (e ≤ 1 × 10⁻⁵). (A) Mb DNA screened for one “hit” to the target gene and (B) Percentage of samples in which the gene was found. Gene codes were: McrA-α subunit of methyl coenzyme M reductase from Methanobrevibacter smithii ATCC 35061 (YP_001273475); AFaB and AFaC; two genes from the Afa operon involved in adhesion of pathogenic diffusely adhering E. coli;28 DsrA-α subunit of dissimilatory sulfite reductase from Desulfovibrio piger ATCC 29098 (AAL57441); Acs-β subunit of CO dehydrogenase/acetyl-CoA synthase complex from Blautia hansenii DSM 20583 (EEX21742); Lipo-putative lipoprotein from fosmid clone 52B7 (ADR74178);38 But-Coa-butyryl-CoA:acetate CoA-transferase from Clostridium sp. SS2/1 (ZP_02439482);26 Gluc-β-D-glucuronidase from fosmid clone H11G11 (CBJ55485);25 Fhs-formate-tetrahydrofolate ligase from Bacteroides thetaiotaomicron VPI-5482 (AAO75844) and Bsh-bile salt hydrolase from Alistipes sp. HGB5 (ZP_08513987).

Table 1 Comparison of the pros and cons of the two metagenomics methods used to study the functions in an ecosystem

	Sequence based screen	Function based screen
Screen large amounts of DNA	Yes—with the use of 2^nd & 3^rd generation sequencing platforms	Yes—with the aid of colony picking and arraying robots
Provide novelty	No	Yes
Genomic context	Limited and relies on assembly of reads.	Yes
Toxic genes	Yes	No
Expression issues	No	Yes
Storage issues	No	Yes-physical storage issues.
Computational issues	Yes—BLAST searches and data analysis are becoming bottlenecks in the analysis	No

Table 2 List of common metabolites found in human, mouse and rat feacal water

The metabolites shaded in grey are at higher concentration the human samples than in the rodent samples, those in black are unique to the human gut and those in red are either bioactive or precursors to known bioactive agents such as trace amines, for example phenylalanine is decarboxylated to β-phenylethylamine (adapted from ref. 36).