Plant cell wall carbohydrates

Plant cell walls are complex insoluble structures that contain different types of carbohydrates (Table 1) plus non-carbohydrate constituents (mainly protein and lignin, approximately 10 % of dry weight)^{(Reference Englyst, Quigley and Hudson15,Reference Fry and Ulvskov16)} . Cellulose microfibrils are crosslinked by a range of other carbohydrates collectively designated as hemicellulose (excluding α-galacturonate-rich carbohydrates) or pectin (α-galacturonate-rich carbohydrates)^{(Reference Fry and Ulvskov16)}. Pectin also serves as an adhesion layer between adjacent cells, called the middle lamella. As a rough rule of thumb, each of the three major cell wall components accounts for approximately 30 % of dry weight in many dietary plants belonging to dicotyledons (e.g. apple, berries, carrot, legumes, nuts) and monocotyledons (e.g. asparagus, bananas, onions), with their primary cell walls being designated type I cell walls^{(Reference Fry and Ulvskov16,Reference Zavyalov, Rykov and Lunina17)} . Pectin consists of four different structural domains, homogalacturonan (approximately 15 % of total cell wall dry weight), rhamnogalacturonan I (approximately 10 %), rhamnogalacturonan II (approximately 1–4 %) and xylogalacturonan (usually very low amounts) (Table 1). The exact cell wall composition differs between plants and also depends on other factors, such as plant growth conditions, ripeness and plant storage^{(Reference Holland, Ryden and Edwards18)}. Monocotyledon plants belonging to the Poales (including the dietary grains barley, maize, oats, rice, rye and wheat) have type II primary cell walls^{(Reference Fry and Ulvskov16,Reference Zavyalov, Rykov and Lunina17)} . They have a much lower pectin and xyloglucan content (xyloglucan, a hemicellulosic carbohydrate, constitutes approximately 20–25 % of total dry weight in type I and 4 % in type II cell walls). Xylans (including arabinoxylans and glucuronoarabinoxylans), conversely, constitute approximately 30 % of total dry weight in type II cell walls compared to about 5–8 % in type I. Furthermore, type II cell walls contain approximately 30 % total dry weight of β-glucans, which are absent in type I cell walls^{(Reference Fry and Ulvskov16,Reference Zavyalov, Rykov and Lunina17)} (Table 1).

Table 1. Main characteristics of major plant dietary fibre carbohydrate constituents^{(Reference Englyst, Liu and Englyst1,Reference Ndeh and Gilbert5,Reference Fry and Ulvskov16,Reference Buckeridge, Pessoa dos Santos and Tiné26)}

PCW, plant cell wall; Ac, acetyl ester; Me, methyl ester; Kdo, (2-Keto) – 3-deoxy-β-d-manno-octulosonic acid; Dha, (2-Keto) – 3-deoxy-β-d-lyxo-heptulosaric acid.

* Plant exudates and mucilages (including galactans and glucuronomannans)^{(Reference Ndeh and Gilbert5,Reference Klassen, Xing and Tingley14,Reference Fry and Ulvskov16 )} are not listed separately here as they typically constitute a relatively small fraction of dietary fibre.

^† All monosaccharides in D configuration unless specified otherwise.

Storage carbohydrates

A major plant storage carbohydrate present in cereals, legumes, rhizomes, roots and tubers is starch^{(Reference Bertoft19)}, a polymer consisting of linear (amylose) and branched (amylopectin) α-linked glucose residues (Table 1). Starch is principally digestible in the human upper gut by pancreatic α-amylase, but some starch, termed resistant starch (RS), can escape host digestion due to its physicochemical properties. Starch digestibility depends on several factors, which form the basis for the classification of RS^{(Reference Lockyer and Nugent20,Reference Cerqueira, Photenhauer and Pollet21)} . RS1 is physically inaccessible within the food matrix, for example, within intact plant cells; RS2 is inaccessible due to the native starch conformation, for example, high amylose starches that have a more crystalline structure; RS3 is generated during food processing, such as cooking and cooling (retrogradation), which leads to a change in physicochemical properties, such as an increase in its crystallinity; RS4 is chemically modified, for example, by cross-linking or esterification, to reduce its digestibility; RS5 includes amylose-lipid complexes and this category has recently been proposed to be extended to include natural or manufactured self-assembled complexes of starch with other macromolecules^{(Reference Gutiérrez and Tovar22)}. Only a small fraction of the total starch within foods escapes upper gut digestion (typically within the range of 0–20 %), with large differences between plants, food processing and preparation techniques^{(Reference Capuano, Oliviero and Fogliano23)}.

Other plant storage carbohydrates also contribute to dietary fibre, including inulin-type fructans and raffinose-family oligosaccharides (Table 1). Both contain a terminal sucrose residue, as plants synthesise them starting with sucrose^{(Reference Van den Ende24)}, which is extended either with fructose residues in the case of fructans or with galactose residues in the case of raffinose-family oligosaccharides (also called α-galactosides). Raffinose-family oligosaccharides are present in legumes and are mostly comprised of raffinose, stachyose and verbascose, containing 1–3 galactose residues^{(Reference Englyst, Liu and Englyst1)}. Different types of fructans are present in plants^{(Reference Van den Ende24,Reference Young, Latousakis and Juge25)} , but in dietary fibre, inulin-type fructans are the predominant form, with the main food sources being onions, Jerusalem artichoke, chicory and wheat^{(Reference Englyst, Liu and Englyst1)}. They are often designated as non-digestible oligosaccharides, but this only includes molecules of a degree of polymerisation of up to nine units^{(Reference Englyst, Liu and Englyst1)}. As inulin-type fructans include molecules of up to degree of polymerisation of 60, small non-digestible carbohydrates are alternatively classified as resistant short-chain carbohydrates, whereas larger polysaccharides that do not contain α-(1→4)-linked glucose are referred to as non-starch polysaccharides (NSP)^{(Reference Englyst, Liu and Englyst1)}. Whilst not a major contributor to dietary fibre, it should be noted that some hemicellulosic carbohydrates also take on storage functions in seeds^{(Reference Buckeridge, Pessoa dos Santos and Tiné26)} (Table 1).

Biochemical and physicochemical complexity of dietary fibre

Considering the number of different monosaccharides, the presence of non-sugar constituents (such as methyl- and acetyl-groups, phenolic compounds) and the number of different glycosidic linkages present in dietary fibre (Table 1), a multitude of microbial enzymes are required for its degradation. In addition to the biochemical complexity, physicochemical factors also need to be considered when assessing microbial fibre fermentation. A large fraction of fibre arrives in the large intestine in the form of complex insoluble particles, such as intact plant cells, cell wall fragments or granular macromolecular aggregates, especially on diets containing mostly whole plant-based foods with little processed ingredients^{(Reference Stephen, Champ and Cloran13,Reference Capuano, Oliviero and Fogliano23)} , thus limiting access to the individual carbohydrate molecules for microbial degradation. The intrinsic solubility of the different constituents also differs and depends on their specific properties in different plants. For example, the solubility of pectins, which are negatively charged due to the presence of galacturonic acid residues, is affected by pH and by their degree of methylation, as the methyl groups render carboxylic acid residues neutral^{(Reference Fry and Ulvskov16)}. The solubility of xyloglucans differs depending on the plant source, as type I cell wall xyloglucans are typically highly branched and therefore more soluble than cereal type II xyloglucans^{(Reference Fry and Ulvskov16)}. Further structural differences between the two different cell wall types include a lower galactose-, arabinose- and fucose-content in type II cell wall xyloglucans and more extensive oligosaccharide side chains and esterification with acetyl, feruloyl and 4-coumaroyl groups in type II cell wall xylans^{(Reference Fry and Ulvskov16)}.

The importance of the type of glycosidic linkage in determining physicochemical properties of carbohydrates is exemplified by fibre constituents exclusively composed of glucose monosaccharides, namely cellulose, β-glucans and RS. The β-(1→4)-linkages in cellulose result in linear molecules that tightly align with each other via hydrogen bonds and form highly insoluble microfibrils, which makes cellulose an excellent scaffolding material to provide strength to the plant cell wall^{(Reference Fry and Ulvskov16)}. Cereal β-glucans also contain β-(1→4)-linkages, but those are interspersed with β-(1→3)-linkages (which is the basis for their alternative designation as mixed-linkage glucans), which results in more flexible molecules that do not form highly ordered microfibrils and are more soluble, but relatively viscous^{(Reference Fry and Ulvskov16)}. The α-(1→4)-glucose linkages in amylose-fractions of starch can adopt different conformations including helical structures, and the α-(1→6)-branchpoints in amylopectin result in very complex structures of the overall starch molecule. Starch granules contain both amorphous and crystalline regions, and the overall starch structure differs between dietary plants^{(Reference Bertoft19)}.

Genetics and physiology of fibre breakdown strategies in gut microbes

Much of what is currently known about fibre degradation by individual members of the gut microbiota has been learned from in vitro investigations with cultured isolates in the laboratory and in silico analyses of their genomes. Fibre breakdown genes and their regulation have been most extensively investigated in Bacteroides species belonging to the dominant phylum Bacteroidetes. Members of this phylum contain numerous (often over a hundred) genetic polysaccharide utilization loci, which are operons that encode CAZymes required for the breakdown of specific dietary fibre carbohydrates together with corresponding carbohydrate binding, transport and regulatory functions^{(Reference Ndeh and Gilbert5)}. This enables the bacteria to sense the presence of many different types of carbohydrates and induce the corresponding functions for their degradation and uptake. Thus, Bacteroides species are regarded as generalists that are able to access many different potential growth substrates, although the level of metabolic flexibility differs between species^{(Reference Cockburn and Koropatkin3,Reference Briggs, Grondin and Brumer6)} . It appears that Bacteroides species with overlapping substrate spectra limit competition with each other by prioritising different carbohydrates when grown together on a mix of substrates^{(Reference Tuncil, Xiao and Porter30,Reference Patnode, Beller and Han31)} . The initial polysaccharide degradation in Bacteroidetes takes place at the cell surface and oligosaccharides are imported across the outer membrane into the periplasmic space for further degradation and transport into the cytoplasm^{(Reference Briggs, Grondin and Brumer6)}.

Species within the other dominant phylum, the Firmicutes, encode fewer CAZymes on average than Bacteroidetes species^{(Reference el Kaoutari, Armougom and Gordon27)} and often have smaller genomes overall. However, there is also large variation between the many different species^{(Reference Cockburn and Koropatkin3,Reference Briggs, Grondin and Brumer6)} . For example, a study of genomes from eleven strains belonging to five Firmicutes species within the Roseburia spp./Eubacterium rectale group of the Lachnospiraceae family showed that most strains harboured between fifty-six and eighty-six GH genes, whereas the three Roseburia intestinalis strains contained between 102 and 146^{(Reference Sheridan, Martin and Lawley32)}. Many CAZymes present in this group of Firmicutes are also organised as operons including regulatory and transport functions, but there are differences to the polysaccharide utilization locus organisation found in Bacteroidetes, reflecting the Gram-positive cell surface architecture of the Firmicutes. Gram-positive cells lack an outer membrane and periplasmic space, leading to differences in the composition and organisation of the carbohydrate-degrading machinery^{(Reference Cockburn and Koropatkin3)}. CAZyme operons found in Firmicutes have therefore been designated Gram-positive polysaccharide utilization loci^{(Reference Sheridan, Martin and Lawley32)}.

Some bacteria within the Ruminococcaceae family of Firmicutes employ a number of different CAZymes encoded across several sites of the genome to build multienzyme complexes on the bacterial cell surface. This has been extensively studied in Ruminococcus champanellensis, the only bacterium from the human gut described so far able to degrade crystalline cellulose^{(Reference Ben David, Dassa and Borovok33,Reference Moraïs, Ben David and Bensoussan34)} . Multiple enzymes form a protein complex with structural scaffoldin proteins via protein–protein binding between dockerin and cohesin domains, and scaffoldin proteins also tether the complex to the cell surface. In addition, individual proteins often contain complex multi-modular domain structures, which may include several catalytic and carbohydrate-binding domains. The resulting cellulosome complex contains enzymes for the degradation of cellulose as well as hemicellulosic carbohydrates. The close proximity of the different enzymatic functions likely leads to synergism and enables the degradation of highly recalcitrant crystalline cellulose as well as complex particulate plant cell wall matter^{(Reference Ben David, Dassa and Borovok33)}. Some of the CAZymes present in the R. champanellensis cellulosome are strongly upregulated during growth on cellulose compared to cellobiose^{(Reference Moraïs, Ben David and Bensoussan34)}.

Another Ruminococcus species, Ruminococcus bromii, also makes use of scaffoldins, dockerin and cohesin domains to build multienzyme complexes on its cell surface, but those are amylosomes rather than cellulosomes, as their GH are amylases that target starch rather than cellulose^{(Reference Ze, ben David and Laverde-Gomez35)}. R. bromii is a highly specialised starch-degrading species, as analysis of several strains showed that they contain less than 30 GH in their genomes, the majority of which are involved in starch breakdown^{(Reference Mukhopadhya, Moraïs and Laverde-Gomez36)}. The genes are scattered around the genome and mostly not linked to other GH. Amylase activity was constitutively expressed in R. bromii L2-63^{(Reference Ze, ben David and Laverde-Gomez35)}, which further confirms it to be an extreme specialist adapted to starch breakdown. Indeed, R. bromii may play a keystone role in RS degradation, as was discovered during human dietary intervention studies involving a dietary period with very high intakes of RS^{(Reference Ze, le Mougen and Duncan37,Reference Abell, Cooke and Bennett38)} . In a trial with fully controlled diets comparing a high NSP to a high RS intake, the relative abundance of R. bromii increased in faecal samples of the volunteers within a few days on the high RS diet, and quickly decreased again after its discontinuation^{(Reference Walker, Ince and Duncan39,Reference Salonen, Lahti and Salojärvi40)} . Two volunteers who had low or undetectable levels of R. bromii excreted a large fraction of the ingested RS in their faeces, whereas faecal starch levels were very low for all other volunteers^{(Reference Walker, Ince and Duncan39)}. In vitro incubations of faecal microbiota from one of the two volunteers and addition of individual known starch degraders (Bacteroides thetaiotaomicron, Bifidobacterium adolescentis, E. rectale, R. bromii) revealed that only R. bromii was able to restore starch degradation to levels seen in healthy volunteers^{(Reference Ze, Duncan and Louis41)}. As the genome of R. bromii does not contain an exceptional number of starch-degrading enzymes compared to other starch-degrading bacteria from the human gut, it appears that it is their organisation into amylosomes that provides its enhanced ability to degrade recalcitrant RS^{(Reference Mukhopadhya, Moraïs and Laverde-Gomez36)}.

Dockerin-cohesin pairs and other protein domains likely to be involved in the formation of cell surface CAZyme complexes have also been identified in other bacteria, including in the host mucin-degrading opportunistic pathogen Clostridium perfringens ^{(Reference Low, Smith and Abbott42)}. The Ruminococcaceae pectin-degrading specialist Monoglobus pectinilyticus contains some putative dockerin domains in proteins of unknown function, whereas several of its CAZymes contain other domains that may facilitate the assembly of multi-enzyme complexes^{(Reference Kim, Healey and Kelly43)}, suggesting that further biochemical variations on the theme of multifunctional enzyme complexes exist in nature.

Within the other Gram-positive phylum that is commonly detected in the human gut, the Actinobacteria, most research has been carried out on Bifidobacterium species. There is diversity in which types of fibre are utilised by different species, but many species appear to be adapted to utilise mainly oligosaccharides or monosaccharides rather than complex insoluble fibre, and some species utilise host-derived carbohydrates^{(Reference Briggs, Grondin and Brumer6,Reference Turroni, Milani and Duranti44,Reference Kelly, Munoz-Munoz and van Sinderen45)} . Furthermore, RS-degrading species such as B. adolescentis have also been reported^{(Reference Cerqueira, Photenhauer and Pollet21,Reference Ze, Duncan and Louis41)} . Regulators have been found associated with the corresponding genes for substrate breakdown, suggesting that the bacteria can sense and respond to the available substrates and have preference hierarchies for different carbohydrates^{(Reference Kelly, Munoz-Munoz and van Sinderen45)}.

Prediction of microbial function from genomic sequence information

Genome sequence information is invaluable in providing hypotheses on the likely physiology and behaviour of different microbes, but function cannot always be deduced from sequence alone. Thus, it can be difficult to establish substrate specificity of CAZymes from their amino acid sequences, as several CAZyme families include enzymes targeting different substrates^{(Reference Lombard, Golaconda Ramulu and Drula28)}. The limitations of establishing the ecological niche of a bacterial species from its genome sequence are exemplified by a recent study of Coprococcus eutactus within the Lachnospiraceae family of the Firmicutes phylum. It was found to contain two GH9 genes, a GH family containing mainly cellulases^{(Reference Ravachol, Borne and Tardif46)}. They are relatively rare in human gut bacterial genomes and are mostly present in bacteria with confirmed cellulose-degrading ability, especially when more than one GH9 gene is present^{(Reference Alessi, Gray and Farquharson47)}. Four GH5 genes were also present in C. eutactus ART55/1, another GH family containing many cellulases^{(Reference Aymé, Hébert and Henrissat48)}, suggesting that this species may be able to degrade cellulose. However, when growth tests were performed on a range of soluble and insoluble substrates, no growth was detected on cellulose^{(Reference Alessi, Gray and Farquharson47)}. Instead, growth profiles and gene expression analyses suggest that β-glucans are the preferred growth substrate for this species, with lower growth on gluco/galactomannans, galactan and starch. Interestingly, a closely related species, Coprococcus sp. L2-50, was more specialised towards β-glucan, showing only limited growth on starch and no growth on mannan, glucomannan, galactomannan or galactan^{(Reference Alessi, Gray and Farquharson47)}. Thus, phylogenetically closely related bacteria can exhibit major functional differences. This is usually not well captured in studies that analyse microbiota changes based on 16S rRNA gene amplicon sequencing, as this often does not allow for phylogenetic resolution down to species level.

Another limitation of deducing microbial function from sequencing-based microbiota profiling is the fact that many bacteria share the same genus name despite not being phylogenetically closely related, as they were originally misclassified based solely on phenotypic characteristics before phylogenetic classification based on genome sequence information was available. For example, several species currently within the genus Coprococcus require taxonomic reclassification as they are not sufficiently closely related to C. eutactus, which is also reflected in functional differences, such as differences in their growth substrate profiles^{(Reference Alessi, Gray and Farquharson47)}. Thus, when sequence-based studies find associations between certain bacterial genera (including Firmicutes such as Clostridium, Coprococcus, Eubacterium, etc.) and health outcomes or nutritional factors, it can be difficult to deduce function if it is not clear which specific species, or even phylogenetically related taxa, this actually represents.

The functionality of a given species can also depend on its environmental context at the time, which has to be taken into consideration when assigning function based on presence in microbiota sequence-based profiles. For example, Coprococcus catus produces butyrate from fructose, a breakdown product of fructans provided by primary fructan degraders within the microbiota. It can alternatively also grow on the fermentation acid lactate, but produces mainly propionate instead of butyrate on this substrate^{(Reference Reichardt, Duncan and Young49)}. Thus, the balance between butyrate and propionate production of this species depends on its ecological context within the complex community, including the abundance of cross-feeders providing the different growth substrates, as well as competitors for those substrates.

Factors affecting microbial competition

Direct competition for dietary fibre substrates between different microbes depends on the substrate specificity of their CAZymes (including the chain length of oligosaccharides and substitution with non-carbohydrate ligands^{(Reference Leth, Ejby and Workman52)}) and also seems to be influenced by their biochemical organisation on the cell surface. Thus, close proximity of different enzymes likely leads to synergism between them to facilitate the breakdown of insoluble complex substrates^{(Reference Ben David, Dassa and Borovok33,Reference Mukhopadhya, Moraïs and Laverde-Gomez36)} . Differences in the efficiency of substrate binding and transport also need to be considered to understand competitive interactions between gut microbes. For example, it has been hypothesised that the four carbohydrate-binding domains of an R. intestinalis xylanase give this species superior ability to compete for insoluble xylans over Bacteroides species in co-culture competition assays^{(Reference Leth, Ejby and Workman52)}. Transporter specificities for xylan breakdown products also vary between the different species, likely enabling their co-existence on a pool of xylo-oligosaccharides of varying lengths^{(Reference Leth, Ejby and Workman52)}. Detailed investigation of a mannan utilisation locus in Bifidobacterium animalis subsp. lactis revealed high affinity transport of manno-oligosaccharides, which enables the bacterium to effectively compete with Bacteroides ovatus on carob galactomannan in co-culture. This was found despite the fact that its β-mannanase for extracellular mannan breakdown is secreted rather than cell-attached, which suggests that galactomannan breakdown is likely more physically distant from its cell surface transporters than that of Bacteroides species with their cell surface-associated CAZymes and transporters being in close proximity^{(Reference Ejby, Guskov and Pichler53)}.

Other aspects of bacterial physiology should also be considered when examining competitive relationships. The pH in the gut fluctuates with the level of microbial activity due to the formation of acidic fermentation products. It tends to be mildly acidic in the proximal gut, where dietary fibre substrate concentrations are high and acid production exceeds the uptake capacity of the gut wall. It shifts to a more neutral pH in the distal colon, as carbohydrate fermentation slows down due to exhaustion of easily fermentable fibre^{(Reference Walker, Duncan and Carol McWilliam Leitch54)}. Different bacteria vary in their tolerance of acidic pH, as was exemplified in continuous culture studies of human faecal microbiota on different carbohydrates, which showed higher levels of Bacteroidetes at pH 6⋅5 and of Firmicutes at pH 5⋅5^{(Reference Walker, Duncan and Carol McWilliam Leitch54,Reference Chung, Walker and Louis55)} . However, this broad categorisation is somewhat simplistic and there can be large differences in acid tolerance between closely related species. For example, E. rectale within the Lachnospiraceae family of the Firmicutes exhibited good growth in media with an initial medium pH of as low as 5⋅1, whereas growth of a relatively closely related species, Roseburia inulinivorans, was severely curtailed below pH 5⋅5 and absent at pH 5⋅1^{(Reference Duncan, Louis and Thomson56)}. This potentially poor competitiveness at lower pH values may partially explain why R. inulinivorans was not found to be stimulated within the microbiota by fructans in vivo ^{(Reference Ramirez-Farias, Slezak and Fuller57)} or in vitro ^{(Reference Reichardt, Vollmer and Holtrop58)}, despite showing good growth on fructans of different chain lengths when grown in pure culture^{(Reference Scott, Martin and Duncan51)}. The requirement for other growth factors (minerals, amino acids, vitamins, etc.) may also disadvantage certain microbes if they are not available in sufficient quantities in the gut environment. For example, a recent study found several vitamin auxotrophies in a range of butyrate-producing Firmicutes from the human gut^{(Reference Soto-Martin, Warnke and Farquharson59)}.

Microbial cooperation by metabolic cross-feeding

Microbial cross-feeding plays an important role in providing growth substrates to the wider microbial community, as only some species, termed primary degraders, are able to degrade the fibre as it arrives in the large intestine (Fig. 1). For example, the previously described keystone role of R. bromii in making RS available to other bacteria has been demonstrated in vivo and in vitro ^{(Reference Cerqueira, Photenhauer and Pollet21,Reference Ze, le Mougen and Duncan37–Reference Ze, Duncan and Louis41)} . The level to which primary degraders share their resource with other gut bacteria varies^{(Reference Briggs, Grondin and Brumer6)}. R. bromii releases extensive amounts of glucose and maltose from RS during in vitro growth, which can be utilised by other microbes. As R. bromii cannot utilise glucose itself and prefers longer oligosaccharides over maltose, it is a cooperative cross-feeder benefitting other microbes^{(Reference Ze, Duncan and Louis41)}. Nutritional cooperation has also been established for Bacteroides ovatus when grown on inulin^{(Reference Rakoff-Nahoum, Foster and Comstock60)}. Despite the fact that B. ovatus takes up intact inulin molecules without extracellular breakdown, it also expresses two extracellular enzymes that make shorter oligosaccharides available to other bacteria. Co-culture and in vivo studies suggest that B. ovatus receives benefits from the cross-feeding beneficiaries in return, in this case Bacteroides vulgatus ^{(Reference Rakoff-Nahoum, Foster and Comstock60)}. Other primary degraders seem to have a much more selfish approach to external degradation of fibre. For example, co-culture studies of B. thetaiotaomicron wild type with mutant strains that had a deletion in amylopectin- and levan-targeting extracellular CAZymes showed that there was only limited cross-feeding of carbohydrate degradation intermediates from the wild type to the mutant^{(Reference Rakoff-Nahoum, Foster and Comstock60)}.

Fig. 1. Main routes of metabolic cross-feeding of dietary fibre by the human gut microbiota and major factors affecting the activity of individual microbes. CAZyme, carbohydrate-active enzyme.

Cross-feeding also takes place at the level of fermentation products^{(Reference Louis and Flint61)} (Fig. 1). Hydrogen is produced by many fermentative gut bacteria and consumed by three different microbial groups, sulphate-reducing bacteria (which can also convert fermentation acids), acetogens and methanogenic Archaea^{(Reference Smith, Shorten and Altermann62)}. Formate cross-feeding was also established between R. bromii and the acetogenic bacterium Blautia hydrogenotrophica in continuous culture. Transcriptomic analysis revealed further metabolic interactions, including amino acid catabolism and vitamin acquisition, between the two species^{(Reference Laverde Gomez, Mukhopadhya and Duncan63)}. Cross-feeding can have considerable benefits for host health. For example, lactate is produced by many different gut microbes, but is known to have a range of potentially deleterious effects on the host, and can have de-stabilising effects on gut microbiota composition by lowering pH and inhibiting the growth of other gut bacteria^{(Reference Wang, Rubio and Duncan64)}. Fortunately, lactate can be utilised and converted to either butyrate or propionate by other gut bacteria, although this activity is limited to certain species^{(Reference Reichardt, Duncan and Young49,Reference Louis and Flint61,Reference Duncan, Louis and Flint65,Reference Belenguer, Duncan and Calder66)} . These lactate-utilising bacteria therefore play an important role in preventing the build-up of detrimental concentrations of lactate in the colon^{(Reference Wang, Rubio and Duncan64,Reference Belenguer, Holtrop and Duncan67)} . Microbes may also benefit from the production of other compounds such as vitamins by co-inhabitants, based on in vitro evidence^{(Reference Soto-Martin, Warnke and Farquharson59)}. Furthermore, metabolic interactions also likely take place during the breakdown of secondary compounds (xenobiotics, phytochemicals). Thus, an in vitro study of wheat bran degradation by human faecal microbiota suggested that the release and biotransformation of the abundant phenolic phytochemical, ferulic acid, was due to the action of several different microbial species. The primary wheat bran-degrading bacterial species responsible for breaking down the fibre and releasing ferulic acid only showed very limited further transformation of this compound^{(Reference Duncan, Russell and Quartieri68)}. Overall plant-derived metabolite pools in the human gut are therefore dependent on both primary degraders of plant material and the wider gut microbiota, which can further biotransform released metabolites.