If You Eat It or Secrete It, They Will Grow: the Expanding List of Nutrients Utilized by Human Gut Bacteria

In order to persist, successful bacterial inhabitants of the human gut need to adapt to changing nutrient conditions, which are influenced by host diet and a variety of other factors. For members of the Bacteroidetes and several other phyla, this has resulted in the diversification of a variety of enzyme-based systems that equip them to sense and utilize carbohydrate-based nutrients from host, diet, and bacterial origins.

ABSTRACT

In order to persist, successful bacterial inhabitants of the human gut need to adapt to changing nutrient conditions, which are influenced by host diet and a variety of other factors. For members of the Bacteroidetes and several other phyla, this has resulted in the diversification of a variety of enzyme-based systems that equip them to sense and utilize carbohydrate-based nutrients from host, diet, and bacterial origins. In this review, we focus first on human gut Bacteroides and describe recent findings regarding polysaccharide utilization loci (PULs) and the mechanisms of the multiprotein systems that they encode, including their regulation and the expanding diversity of substrates that they target. Next, we highlight previously understudied substrates such as monosaccharides, nucleosides, and Maillard reaction products that can also affect the gut microbiota by feeding symbionts that possess specific systems for their metabolism. Since some pathogens preferentially utilize these nutrients, they may represent nutrient niches competed for by commensals and pathogens. Finally, we address recent work to describe nutrient acquisition mechanisms in other important gut species such as those belonging to the Gram-positive anaerobic phyla Actinobacteria and Firmicutes as well as the Proteobacteria. Because gut bacteria contribute to many aspects of health and disease, we showcase advances in the field of synthetic biology, which seeks to engineer novel, diet-controlled nutrient utilization pathways within gut symbionts to create rationally designed live therapeutics.

INTRODUCTION

The importance of symbiotic gut bacteria to human health has been appreciated for decades. Advances in sequencing, culture, and other technologies have invigorated research into the relationships between the human gut microbiota (HGM) and various diseases, with new connections emerging rapidly over the last decade. As the mechanistic relationships between various altered HGM states and diseases become better understood (1–4), there is a desire to design rational interventions to restore or suppress certain microbes or even implant engineered bacteria that can uniquely exploit rare nutrients (5, 6). A key feature underlying the persistence of many gut bacteria is their ability to utilize fiber and some other carbohydrate-based nutrients consumed by the host, which exert a heavy impact on the lower gut ecosystem because they are not degraded or absorbed in the small intestine. Underscoring the importance of dietary fiber polysaccharides to gut microbial ecology and health, when these nutrients are absent or reduced, the microbiota shifts to foraging on mucosal glycans (7), which increases susceptibility to pathogens (1), changes the spatial organization of gut microbes (8), and leads to exacerbated disease in models of inflammatory bowel disease (9). Significant research has focused on the numerically abundant members of the Bacteroidetes phylum and their abilities to access carbon from the diverse array of polysaccharides presented through diet, host secretions, and microbial exopolysaccharides (EPSs) and capsules (10–17). This ability in Bacteroidetes is accomplished via the expression of polysaccharide utilization loci (PULs), genomic clusters of coexpressed genes that encode products that sense, cleave, transport, and metabolize polysaccharides and other nutrients. As research has extended into the often more abundant but also more fastidious members of the Firmicutes phylum, similar clusters of nutrient utilization genes have been observed (18–20), although their enzymes and other products are adapted to the Gram-positive cell surface.

The ability of the Bacteroidetes and members of other phyla to utilize complex carbohydrates has been reviewed in the past few years (19, 21, 22). However, a number of recent studies have expanded the known repertoire of polysaccharides and other nutrients that are targeted by gut bacteria, along with increased knowledge of the enzymes and local and global regulatory mechanisms involved. Some nutrients, such as Maillard reaction products (MRPs), formed between reducing sugars and amino acids during cooking (23), or commonly ingested food additives like trehalose are unexpectedly metabolized by some members of the gut microbiome using specifically adapted systems (24). Here, we review recent findings regarding the enzymatic and regulatory strategies through which gut bacteria degrade carbohydrates, with a focus on newly identified substrates and the regulation of these attributes within bacterial cells that are often equipped to utilize many different substrates. We approach this review with the idea that complex carbohydrates are of ubiquitous importance in shaping the ecology and physiology of the HGM and therefore present a convenient lever to intentionally manipulate these communities.

MECHANISMS OF POLYSACCHARIDE UTILIZATION BY HUMAN GUT BACTEROIDETES

Polysaccharides are arguably the most diverse category of biological molecules because of the number of different sugars that they may incorporate, the two different glycosidic linkage types (α or β) that connect them, the ability of some sugars to form 5 (furanose) or 6 (pyranose) atom rings, and the ability of polysaccharides to contain linkages to multiple carbons on constituent sugars. Bacteria that degrade these molecules as nutrients must possess enzymes that are specific for cleaving each of the unique linkages that they contain, although partial degradation may be possible by some bacteria. The degradative carbohydrate-active enzymes (CAZymes) that conduct these reactions are grouped into numbered families that reflect common structure and homology (25) and may also indicate activity. As an introduction to polysaccharide structure and bacterial degradation, Fig. 1 shows an example structure of a plant cell fiber, the hemicellulose galactomannan composed of mannose and galactose, a relatively simple bacterial glycoside hydrolase (GH) pathway for its degradation that has been determined in the gut symbiont Bacteroides ovatus, and a schematic of the genes involved (26, 27). A key aspect of this degradative pathway is the sequential action of multiple enzymes, some of which act in an endo fashion (i.e., cleaving within polymers) and others of which work in an exo mode (i.e., cleaving at the ends of polymers). A more detailed catalog of the sugars and linkages that are present in plant, host, and microbial polysaccharides and that impact gut bacteria, as well as a review of the bacterial degradative systems involved, is provided by Porter and Martens (21).

FIG 1.

Structural schematic of the plant polysaccharide galactomannan and the corresponding requirement for sugar- and linkage-specific enzymes for its degradation. (A) Schematic of the two sugars galactose (yellow) and mannose (green) that compose galactomannan as well as the α and β linkages that connect them at positional carbons (carbons that are involved in forming these glycosidic linkages are numbered in red). (B) Symbol schematic of a longer region of galactomannan with the region in panel A delineated in the dashed box and the sequence of enzymatic steps for its degradation. Like proteins and nucleic acids, polysaccharides have a directional orientation that includes a reducing end (right) and one or more nonreducing ends (the arrow at the left, which indicates further elongation of the polymer for which the final residue in the chain is the nonreducing end as well as the galactose branches). Polysaccharide-degrading enzymes, such as glycoside hydrolases (GHs), are sugar and linkage specific, and therefore, gut bacteria that degrade diverse dietary fiber structures often possess hundreds of linkage-specific enzymes in their genomes. GHs and other carbohydrate-active enzymes (CAZymes) are classified into numbered families based on structural homology (25), which may provide some indication of their function, although individual families often contain multiple activities. These enzymes may act on their substrate in an “endo” fashion (i.e., cleaving within a chain to release products longer than two sugars) or may be exo-acting (i.e., mostly cleaving at the nonreducing end to release individual sugars or disaccharides). Endo-acting enzymes generally work early in degradation to release smaller fragments. The example shown is from recent studies on the gut symbiont Bacteroides ovatus (26, 27). (C) A series of four enzymes that are coexpressed within a single polysaccharide utilization locus (PUL) are shown as a gene schematic. The sequence and sites at which the encoded enzymes cleave galactomannan are shown in panel B. The initial degrading enzyme is a cell surface GH26 endo-β-mannosidase that cleaves within the β1,4-linked mannose backbone (green circles) to produce a fragment with a new, nonreducing end. Cleaved fragments are then debranched by a GH36 family α-galactosidase that prefers removing these branches when they are present within chains (26). Finally, the remaining β1,4 mannooligosaccharide chains are broken down by specific exo-acting activities such as GH26 and GH130 family enzymes. As polysaccharide substrates become more complex and incorporate more or different sugars, the number and activities of the required enzymes and their sequence of action also change and may increase substantially. (D) Schematic of a recently described gpPUL from Roseburia intestinalis that is also involved in the degradation of galactomannan and other β-mannans. Note that several GH enzymes are common between the two PULs shown in panels C and D, but the Roseburia system relies on an ABC transporter instead of the TonB-dependent transport mechanism that is common in Bacteroidetes. (See reference 20 for a complete description of the Roseburia degradative pathway.)

The major Gram-negative and Gram-positive bacterial phyla in the human gut have evolved a variety of different systems to access nutrients, including carbohydrates, and these systems often incorporate various endo- and exo-acting CAZymes at the cell surface or in the periplasm and cytoplasm (Fig. 2). Members of the Gram-negative Bacteroidetes usually comprise a substantial proportion of the bacteria present in the HGM, with the genus Bacteroides typically being the most abundant in industrialized populations (28). Since several of these species are amenable to genetic manipulation, significant effort has focused on understanding the polysaccharide-degrading mechanisms of certain Bacteroides species, which sometimes devote up to 20% of their genomes to PULs (29–32). The first PUL to be discovered encodes the Bacteroides thetaiotaomicron starch utilization system (Sus), which has served as an archetype for understanding other PUL-encoded systems and is among the mechanistically best understood (33–35). This system includes a maltooligosaccharide-specific outer membrane TonB-dependent transporter (SusC) that partners with a starch binding protein (SusD) (36) (Fig. 2A). This transporter/binding protein pair works with a cell surface endo-acting amylase (SusG), two additional cell surface starch binding proteins (SusE and -F), and two periplasmic α-glucosidases (SusA and -B) to bind, transport, and saccharify starch into its component sugar glucose. Upon determining the B. thetaiotaomicron genome sequence (30), many Sus variants (that each target different polysaccharides) were identified in this species as well as many other gut Bacteroidetes and their environmental relatives. The original definition of a PUL arbitrarily included at least one set of susC-susD homologs (37), although they typically encode several CAZymes (38), carbohydrate binding proteins, regulators, and other enzymes (kinases, proteases, and sulfatases, etc.). Each PUL-specific functional suite tailors the system to target a particular substrate (39, 40). However, variant or degenerate forms of these gene clusters exist, and not all meet the original definition. Examples of variant PULs for bacteria or substrates discussed below include the presence of genes that are required for a single complex substrate spread across multiple distinct PULs, as is the case for B. thetaiotaomicron rhamnogalacturonan II degradation (14), and uncoupling of essential functions, including susC-susD-like genes, from the gene cluster that encodes the enzymes that they work with, as observed for Zobellia galactanivorans carrageenan utilization (41).

FIG 2.

Mechanisms of nutrient utilization for Gram-negative (Gram−) and Gram-positive (Gram+) bacteria. (A) Generic polysaccharide utilization locus (PUL)-encoded multiprotein system with major proteins displayed, including a glycoside hydrolase (GH) and SusC-, SusD-, and SusE/F-like outer membrane proteins. Different mechanisms of either sequestering sugars during degradation (“selfish”) or releasing some sugars (“sharing”) are illustrated. Outer membrane vesicles (OMVs) that contain acidic hydrolytic enzymes with signal peptidase II lipidation motifs (72) may help facilitate the sharing of polysaccharides or their degradation farther away from the producing cell. (B) Prominent carbohydrate utilization systems for some members of the Gram-positive Firmicutes and Bifidobacterium phyla. The ABC and other sugar/oligosaccharide transport systems encoded by Gram-positive PULs (gpPULs) are often coexpressed with degradative enzymes (GHs and carbohydrate esterases [CE]) and carbohydrate binding modules (CBMs), the latter sometimes encoded within the same polypeptide as the catalytic domain of glycoside hydrolases or other enzymes. (C) Some Gram-positive Ruminococcus species, and other related bacteria, produce complex multiprotein assemblies termed cellulosomes or amylosomes to degrade cellulose or resistant starch, respectively. In addition to the extracellular-facing enzymes and binding proteins, these strains often rely on ABC transporters to import the cleaved products. (D) Many proteobacteria rely on mono- and disaccharides for growth and therefore may rely on the sharing of these resources by primary degraders by the use of cytoplasmic membrane importers such as ABC transporters, major facilitator superfamily (MFS) proteins, and PTSs (blue). Once in the cytoplasm, kinases and isomerases further contribute to sugar metabolism. At least one strain of E. coli has been observed to steal or pirate cytoplasmic nutrients from host cells using a type III secretion system/injectisome. The bacterium delivers effector proteins or toxins into host cells, forcing nutrients into an adjacent tube that shuttles carbohydrates or amino acids back to the bacterium.

PUL-encoded systems from HGM Bacteroidetes or their environmental relatives have been demonstrated to target a remarkably diverse array of complex carbohydrates, spanning essentially all of the common terrestrial plant and host glycans. While we do not provide a comprehensive catalog of PULs and their corresponding nutrient targets, we provide an overview and highlight several PUL-substrate pairs that have been discovered in recent years. In addition to starch, the plant-derived substrates that gut Bacteroides PULs have been shown to target include individual substrates within three more diverse families of storage and cell wall polysaccharides: fructans, pectins, and hemicelluloses (27, 31, 42–46). Different Bacteroides PULs equip some species to target the host-derived O-linked glycans attached to mucus as well as N-linked glycans and glycosaminoglycans in other host glycoconjugates (7, 12, 29, 47, 48). The expression of some of the PULs that target sulfated host glycans, probably those in colonic mucus, has been associated with the ability of B. thetaiotaomicron to cause colitis in a sulfatase-dependent manner (49). Additional Bacteroides PULs have been shown to target human milk oligosaccharides (50), some bacterial exopolysaccharides (15, 51), the α-mannan (10, 52) and β-glucan (11) in fungal cell walls, and the seaweed-derived polysaccharides agarose (17), porphyran (5, 16), laminarin (53), and alginate (54).

PULs contained in non-HGM Bacteroidetes, such as those from the soil, rumen, marine, and insect gut ecosystems, extend the list of target substrates even further, having been linked to the degradation of chitin (55), cellulose (56), and additional seaweed polysaccharides (41, 57, 58). Underscoring the adaptability of this system to diverse substrates, this list has continued to expand in recent years, with a landmark study describing the enzymatic mechanism through which B. thetaiotaomicron degrades rhamnogalacturonan II—thought to be among the most complex polysaccharides known—through a process that requires 3 separate PULs and 26 different enzymes (14). Related studies on the also complex rhamnogalacturonan I and its attached side chains, which vary between botanical sources, revealed the requirement for at least 4 different B. thetaiotaomicron PULs encoding 30 enzymes (13). Extending the range of host glycans targeted via B. thetaiotaomicron PULs, another recent study investigated this species’ ability to access high-mannose N-glycans (59, 60), including those attached to mucosal immunoglobulin A through the action of several separate PULs (59).

Owing to the structure of their Gram-negative cell envelope, Bacteroidetes are able to transport and sequester oligosaccharides into the periplasm prior to full depolymerization (61–63). Efficient utilization of polysaccharides may be enhanced through the action of outer membrane-bound carbohydrate binding proteins such as SusE and SusF and related carbohydrate binding module (CBM)-containing proteins that could prevent diffusion from the cell surface (63, 64). Bacteroides species frequently digest polysaccharides using this “selfish” mechanism that is characterized by little release of free oligosaccharide or sugar into the extracellular milieu (Fig. 2A), which has been shown for α-mannan and some xylans (7, 52). However, other studies have implied that when certain substrates are degraded by different Bacteroides species, significant amounts of oligosaccharides can be released to be used by other community members (both related Bacteroides as well as those from non-Bacteroidetes phyla) and thought of as “common goods” (65–67). While released degradation products may become available to competitors, the broad existence of interbacterial antagonism strategies (e.g., type VI secretions systems plus others) may provide protection for the primary degrader, especially from closely related species/strains with similar abilities to access the same resources (68–70). Of particular significance are observations that invading pathogens such as Salmonella and Citrobacter rodentium are more readily able to colonize the host when they gain access to nutrients liberated from dietary fiber or host mucosal glycans by resident Bacteroides.

In some cases, nutrient “sharing” between bacteria may promote species-specific cooperation instead of competition. Highlighting how substrate and strain specificities play deciding roles, B. thetaiotaomicron was shown to degrade amylopectin and the fructan levan in a mostly selfish fashion, although some amount of product appeared to be released to cross-feed a non-enzyme-producing mutant of the same strain (66). In contrast, fragments of a different fructan, inulin, were produced in greater abundance by B. ovatus during degradation through the secretion of an extracellular enzyme, which was not required by B. ovatus itself, and the corresponding products were accessible to a strain of Bacteroides vulgatus but not Bacteroides fragilis (66). Interestingly, several Bacteroides species appear to preferentially pack hydrolytic enzymes into outer membrane vesicles (OMVs) (71, 72), which as these hydrolytic particles diffuse away from the producing cell have the potential to degrade carbohydrates and generate common-goods resources for other bacteria with the ability to use them (65) (Fig. 2A). The utilization of the oligosaccharide products generated by these enzymes could still be specific to strains that possess enzymes required to finish their degradation but lack the catalytic ability to degrade full-length polysaccharides. Such a phenomenon was recently observed in strains of B. vulgatus that can utilize xylooligosaccharides (XOSs) up to 4 sugars long but are unable to degrade larger xylans (73).

While it has been perhaps the best-characterized system, the prototypic B. thetaiotaomicron Sus continues to provide new mechanistic insights. All four of the outer membrane lipoproteins in this system have had crystal structures determined with a ligand, revealing 8 binding sites spread across 4 proteins, plus 1 additional active site in SusG, all for a substrate with low sugar/linkage complexity (36, 61, 74). Previous work demonstrated that a binding site mutation in SusD results in the loss of starch binding as well as the loss of the ability to activate growth on starch unless a small amount of maltose is provided (62). However, this B. thetaiotaomicron SusD mutant was still able to grow on maltooligosaccharides, suggesting compensatory roles for one or both of the related outer membrane proteins SusE and SusF. A recent study revealed that despite sharing significant sequence and structural homology, SusE exclusively confers this ability to grow on maltooligosaccharides (63). Additional work using live-cell imaging has revealed substantial variability in the surface mobility of Sus proteins. In live anaerobic cells, SusE remains stationary (75), while the hydrolytic enzyme SusG moves dynamically around the cell surface, spending more time in the locality of bound starch molecules (76). These studies provide important insight into the assembly and dynamic function of PUL-encoded machinery and serve as a reference for additional detailed studies that have begun to emerge for other Sus-like systems, including xyloglucan and cereal-, yeast-, and seaweed-derived β-glucans in other Bacteroides species (44, 53, 64, 77).

MARINE BACTEROIDETES USE PULs DURING UTILIZATION OF ALGAL POLYSACCHARIDES AND HAVE TRANSFERRED SOME OF THESE ABILITIES TO HUMAN GUT BACTEROIDES

Interestingly, several studies have shown that some human gut Bacteroides have the ability to degrade seaweed-derived polysaccharides such as porphyran, agarose, alginate, laminarin, and carrageenan (16, 17, 41, 54, 57, 58, 78, 79), which have been introduced into the human diet through cuisine or as food additives. In at least some cases, the PULs enabling these abilities appear to have been transferred into gut Bacteroides via mobile DNA elements, and the genes often have close relatives in marine Bacteroidetes, suggesting transfer between these environments (16). Additional PULs that target other algal polysaccharides have been identified in marine Bacteroidetes themselves, and these PULs often contain more genes, encode a more expanded enzyme repertoire than those present in gut Bacteroidetes, and include genes for adhesion proteins that may promote anchoring to the nutrient source (80). These expanded enzyme repertoires may reflect the complexity of the polysaccharides present in the cell walls of red, green, and brown algae as well as those from cyanobacteria and the variable EPS structures that they produce (81–84). An important role for marine Bacteroidetes that possess these PULs is to release fixed carbon during and after algal blooms, and the transcripts and corresponding proteins from these systems have become some of the most highly observed products in marine metatranscriptomic and corresponding metaproteomic samples (85, 86). Growth analysis of cultured organisms has been used to measure the activities of the bacterial members present in such blooms, leading to insight into the diverse polysaccharides being utilized as well as which species establish niches around these complex substrates (80, 87–91).

Similar to studies in gut Bacteroidetes, full or partial degradation mechanisms for the algal polysaccharides laminarin (91–95), alginate (54), agarose (96, 97), ulvan (98), and carrageenan (41) have been determined in some of their marine relatives. Some marine Bacteroidetes also appear to utilize α- and β-mannans (99), and the Bacteroidetes in this environment tend to utilize substrates more selfishly, perhaps because of the rapid loss due to dilution that might occur with extracellular degradation in the ocean (100, 101). The genes responsible for the utilization of ulvan, a sulfated heteropolymer of rhamnose and uronic acids, have been characterized in Formosa agariphila, with an impressive armament of 39 PUL genes and 20 additional non-PUL genes involved in processing this polysaccharide (102). While substrate-specific PULs exist in marine Bacteroidetes to target ulvan, furcellaran, and others (103, 104), they have not yet been detected in human gut Bacteroidetes, possibly because many algae have not become a routine part of the human diet and provided the corresponding selective pressure for environment-to-gut transfer of the genes involved. Although PULs specific to these additional algal polysaccharides may not have yet been transferred to the human gut, horizontal gene transfer (HGT) of PULs and other genomic islands is widespread among Bacteroides and gut bacteria in general, and the acquisition of these systems plays a role in niche partitioning (105). As discussed below, PULs involved in degrading seaweed polysaccharides could be utilized to engineer new “orthogonal niches” into recombinant human gut bacteria to create biotherapeutics that are coupled to the ingestion of polysaccharides that are rare in the human diet (5).

GRAM-POSITIVE BACTERIA AND SOME ENTEROBACTERIACEAE ACCESS POLYSACCHARIDES VIA ALTERNATIVE MULTIPROTEIN SYSTEMS

In contrast to the TonB-dependent transporter systems that are dominant in Bacteroidetes, polysaccharide degradation in the Gram-positive Firmicutes and Actinobacteria is accomplished through different cellular mechanisms that are often built around enzymes with catalytic domains similar to those of Bacteroidetes (106). Gram-positive gut symbionts often encode polysaccharide-degrading functions within genetic loci termed Gram-positive PULs (gpPULs) (18), which may encode ATP binding cassette (ABC), major facilitator superfamily (MFS), or phosphoenolpyruvate-phosphotransferase system (PTS) transporters that are coexpressed with associated degradative enzymes (22, 107–109) (Fig. 1D and Fig. 2B). Several examples have been investigated at a mechanistic level, including in Eubacterium rectale, which is able to grow on starch and its component oligosaccharides through functions encoded in five separate genomic loci that include three membrane-bound amylases, one of which has five appended carbohydrate binding modules (CBMs) (110, 111). Recently, two separate multienzyme loci were characterized in the butyrate producer Roseburia intestinalis that mediate the degradation of β-mannans (20). Similar to E. rectale, R. intestinalis employs membrane-bound enzymes that harbor carbohydrate binding modules (Fig. 2B) in the same polypeptide, which may fill roles analogous to those of the SusE/F binding domains in B. thetaiotaomicron Sus, which are encoded separately. After initial degradation at the membrane or extracellularly, R. intestinalis imports partially broken-down oligosaccharides into the cytoplasm using ABC transporters, and the oligosaccharides are further broken down by the action of additional esterases and glycoside hydrolases. R. intestinalis also uses dietary xylan by cleaving it into xylooligosaccharides (XOSs), mainly xylotetraose, at the cell surface and then imports these products through an ABC transporter into the cell for further intracellular processing and catabolism, similar to the strategy that it uses for β-mannan degradation, with the addition of an intracellular epimerase that helps shunt foraged sugar products into central metabolism (Fig. 2B) (112). Other species within the genera Bifidobacterium and Lactobacillus have the ability to utilize xylan, XOSs, or arabinoxylooligosaccharides (AXOSs) (113, 114) through more elaborate mechanisms, including the secretion of non-cell-attached, extracellular enzymes to break down insoluble or complex structures prior to binding and import into the cytoplasm (115). Like some Bacteroides noted above that can use only oligosaccharides derived from larger polymers (73, 116), some bifidobacteria can engage in interphylum cooperation with Bacteroides, with the former profiting from the breakdown of xylans from certain botanical sources into accessible xylooligosaccharides by species like B. ovatus (117).

Some gut Gram-positive bacteria have adapted another multienzyme system to target resistant starch (RS), cellulose, and possibly other substrates. The “cellulosome” is a mechanistically different strategy to organize polysaccharide-degrading enzymes and associated CBMs (118). Cellulosomes have been well characterized in environmental bacteria (118–120) and have more recently been identified in some human gut bacteria. As their name implies, these systems were originally discovered for their role in degrading cellulose (121) and consist of large, membrane-bound, multienzyme complexes (Fig. 2C). The most widely studied of these complexes was first described in the 1980s in the thermophilic soil bacterium Clostridium thermocellum (119, 120). Analogous cellulosomes exist in the gut symbiont Ruminococcus champanellensis that target cellulose (122). Interestingly, the same protein organization that is a hallmark of cellulosomes has been adapted by the gut species Ruminococcus bromii to equip it to degrade RS using amylases that presumably have gained added catalytic capacity toward this insoluble substrate (123). These so-called “amylosomes” contain component proteins similar to cellulosomes termed dockerins, cohesins, and scaffoldins adapted to display these RS-specific enzymes (123, 124). Although the amylosome of R. bromii is the best studied, metagenomics analysis suggests that more human Ruminococcus strains use this strategy for the breakdown of RS, cellulose, and perhaps other substrates (125). Given the taxonomic diversity of gut Firmicutes, additional polysaccharide degradation strategies employed by these Gram-positive bacteria will likely be uncovered as further investigations are performed on these often fastidious and therefore understudied species, a research horizon that will undoubtedly be propelled by resurgent culturing efforts (126–130).

In contrast to the polysaccharide-dominant nutrient utilization strategies of the more abundant Bacteroidetes and Firmicutes, most of the common gut Enterobacteriaceae such as Escherichia coli are predominantly confined to the utilization of more simple mono- and disaccharides. However, a few strains of Enterobacteriaceae have evolved strategies for degrading complex carbohydrates. Notably, the pathogen Yersinia enterocolitica is able to bind and degrade the plant cell wall pectin polygalacturonic acid via a series of periplasmic carbohydrate binding modules and outer membrane-bound and extracellular polysaccharide lyases (PLs), carbohydrate esterases (CEs), and GHs (131).

MONOSACCHARIDES AND OTHER OVERLOOKED NUTRIENTS

The importance of monosaccharides has been given less attention in studies of HGM species, perhaps because they are often broadly utilized by these bacteria, and many are assumed to be absorbed by the host. However, it is likely that relevant amounts of simple sugars or oligosaccharides may be transiently present, perhaps through the action of nonselfish bacterial catabolism, which might enable some bacteria to rely on consuming these simpler products. For E. coli, competition for nutrient niches has been primarily defined by the importance of various monosaccharides to support colonization by different pathogenic and commensal strains. While many species broadly utilize simple sugars, they sometimes vary in their preference for metabolizing these sugars in complex mixtures, leading to alterations in the importance of certain sugars to different strains (132). Studies that examined the ability of invading pathogenic bacteria to access nutrients during infection demonstrated that both enterohemorrhagic E. coli (EHEC) and Salmonella upregulate genes for the utilization of ribose and other simple sugars during infection of cows, chickens, and mice (133–135). EHEC prefers ribose to some other sugars such as fucose, arabinose, and maltose (132), while the related compounds deoxyribose and DNA can also be utilized during invasion by pathogenic E. coli (136, 137). Several pathogenic bacteria such as Clostridium difficile, Salmonella enterica serovar Typhimurium, Citrobacter rodentium, and E. coli have an additional strategy that involves scavenging simple sugars such as fucose, sialic acid, and galacturonic acid, respectively, that have been liberated from larger polysaccharides by the action of commensal Bacteroides (138–141).

Often, the loci involved in monosaccharide catabolism by Proteobacteria include a sugar kinase, a transporter (e.g., ABC transporter or MFS transporter), a regulator, and additional genes specific for the metabolism of that sugar (e.g., aldolase or isomerase). These systems are generally less complex than PUL-encoded mechanisms in Bacteroides (Fig. 2D). However, B. thetaiotaomicron has related gene clusters for the catabolism of the monosaccharides l-fucose, l-rhamnose, d/l-arabinose, d-ribose, and d-fructose. For fucose and rhamnose, the corresponding B. thetaiotaomicron loci look very similar to those found in Proteobacteria, without any additional PUL machinery such as Sus homologs or CAZymes (142–144). Curiously, the machinery for arabinose metabolism is similar to that found in Proteobacteria but is embedded within a larger PUL for the degradation of arabinan, a polysaccharide that also contains arabinose. Despite this genetic arrangement, the genes involved in depolymerizing arabinan are activated in response to α1,5-linked arabinooctaose oligosaccharides (31), while the genes for arabinose utilization are turned on by the free monosaccharide, which may also be liberated from other polymers (145). This observation suggests that some Bacteroides PULs may have evolved from simpler systems that enable sugar utilization but with the addition of enzymatic machinery to target larger polysaccharides that are sources of the cognate sugar. Indeed, two other B. thetaiotaomicron PULs support this idea. A PUL involved in degrading the β2,6-linked fructan levan contains substrate-specific CAZymes assembled around a core system encoding fructokinase, fructose permease, and a fructose-activated regulator (43). Interestingly, fructose-activated variants of this PUL exist in other Bacteroides species, including strains of B. thetaiotaomicron (146), that contain different enzymes for releasing fructose from the β2,1-linked polymer inulin (43). A PUL involved in the catabolism of ribose, nucleosides (NSs), and RNA (147) also contains two ribose kinases, a permease, and a predicted ribose-responsive regulator. Interestingly, the ability to degrade RNA and NS is dependent on the actions of the ribokinases and a non-PUL-encoded nucleoside phosphorylase (BT4554), despite other CAZymes and SusC/D homologs being present. Partially homologous variants of this ribose PUL are found throughout the phylum in human gut, marine, and soil isolates, with many different CAZyme and genetic architectures. While all of these retain the core regulator, kinase, and permease genes, the accessory genes within these PULs are often highly variable, suggesting that individual species have evolved these variant PULs to access ribose from diverse sources.

Some plants synthesize and store a number of additional carbohydrate-derived molecules such as glucosinolates, flavonoids, and sulfonoacid sugars, and several studies have highlighted the abilities of gut bacteria to metabolize these nutrients, in some cases activating bioactive compounds that then exert effects on the gut. One of these studies demonstrated that B. thetaiotaomicron and other Bacteroides species possess genes that enable them to cleave glucosinolates, a group of thioglucosides that are abundant in cruciferous plants like broccoli (148). This releases glucose that becomes available to bacteria but also releases an active isothiocyanate molecule, which is variable based on the particular glucosinolate cleaved and with some being associated with protective effects against intestinal cancer and other diseases (149). Another family of plant compounds that have been associated with protective health effects are flavonoids, which occur in free flavonol forms (e.g., quercetin) as well as flavonols with one or more attached mono- or disaccharide sugars (e.g., glucose or rutinose). Some bacteria such as Eubacterium ramulus and Flavonifractor plautii have developed the ability to ferment these molecules as a growth substrate, although the genes involved have not been identified (150, 151). Finally, strains of Eubacterium rectale and Clostridium are capable of growing on the plant-derived sugar sulfoquinovose, a sulfonoacid derivative of glucose that is abundant in plants and is a precursor to sulfolipid biosynthesis (152). While these abundant plant compounds can serve as direct nutrients for gut bacteria, the fact that several of them are associated with positive health effects points to the presence of the proper activating bacteria as a critical factor in unlocking their effects.

Finally, pathogens have also evolved strategies to access nutrients beyond carbohydrates during infection. When the intestine is inflamed, l-serine that is commonly present in the diet provides Enterobacteriaceae a competitive edge over resident bacteria through the actions of dedicated l-serine transporters and dehydratases that are upregulated and provide an inflammation-specific advantage (153). Enteropathogenic E. coli (EPEC) was recently shown to employ its type III secretion system (T3SS) injectisome, an apparatus long associated with the injection of effector proteins into the host cell, to steal or “pirate” amino acids from host cells and gain an advantage over commensal bacteria (154) (Fig. 2D). Taken together, it is clear from these studies that both gut pathogens and commensals have forged new nutrient niches by adapting strategies to scavenge or access both monosaccharides and amino acids.

TO COOK OR NOT TO COOK, LET’S ASK THE MICROBIOTA: COOKING AND FOOD PRESERVATION ALTER NUTRIENTS FOR THE GUT MICROBIOTA

Advanced glycation end products (AGEs) or Maillard reaction products (MRPs) are generated from reactions between sugars and proteins during cooking and can also influence the gut microbiota (23, 155, 156). The MRP fructose-asparagine (FA) is metabolized almost exclusively by Salmonella enterica during infections because S. enterica encodes a high-affinity transporter for FA. Due to S. enterica causing inflammation, other microbiota members such as Clostridia, which also possess high-affinity FA systems, are excluded from this nutrient niche (157). Another recent study revealed that commensal Collinsella intestinalis and Collinsella aerofaciens utilize the MRP fructose-lysine (FL) to various degrees based on the presence of other, repressive nutrients such as glucose. In the presence of glucose, C. intestinalis does not repress FL utilization genes, while C. aerofaciens does. These strains accomplish FL metabolism through the actions of a phosphotransferase system and a deglycase that converts FL into the central metabolite glucose-6-phosphate (158). Finally, the AGE N-ε-carboxymethyllysine is degraded by as-yet-uncultured members of the Oscillibacter genus, as inferred based on enrichment culturing experiments, and by a cultured Cloacibacillus evryensis strain through an unknown mechanism (159). These recent studies complement earlier reports of the effects that cooking food has on the microbiota (160, 161) through the formation of MRPs and AGEs and indicate that this area requires deeper investigation.

MRPs and AGEs are not the only substrates created or modified by cooking. Starch is perhaps the most important nutritional carbohydrate in the human diet, and its digestibility is dramatically impacted by cooking. While the starch in a raw potato may be 50 to 70% resistant to digestion by human amylases (162) (so-called “resistant starch” that reaches the colon and its resident HGM), cooking a potato increases starch digestibility by both human and bacterial enzymes through a process called gelatinization. Resistant starch (RS) is categorized into 5 types that vary based on whether it is derived from raw foods, foods that have been cooked and cooled to allow starch to retrograde into crystalline RS, or food containing starches that are chemically modified or have physical coatings that prevent enzymatic access (163). Some bifidobacteria and R. bromii strains degrade naturally occurring type 2 RS found in foods such as raw potatoes and other uncooked foods (124, 164). Chemical modifications through the process of esterification, cross-linking, transglycosylation, or acid hydrolysis of type 4 RS are thought to make the starch less accessible to bacterial degradation. However, studies have observed that dietary supplementation with this type of RS often changes the composition of the HGM, skewing toward more Bacteroidetes and Actinobacteria and fewer Firmicutes and specifically increasing Parabacteroides distasonis and Bifidobacterium adolescentis (165, 166). These observations suggest that some species may be able to access this nutrient through an as-yet-unknown mechanism. Other types of RS, such as type 1 (physically inaccessible granules due to being surrounded by a protein matrix or cell wall material) and type 3 (retrograded starch due to heating and subsequent refrigeration), have also been shown to alter the composition of in vitro synthetic communities derived from porcine microbiota. However, studies into direct mechanisms are still needed to determine which species are responsible for these effects (167).

The introduction of other food additives, such as the stabilizer trehalose, has implications for the HGM. Specific C. difficile ribotypes are able to sense very low concentrations of trehalose and have acquired a PTS for the high-affinity transport of this disaccharide sugar, which can also lead to greater upregulation of toxin B genes (24). One possible reason for increased C. difficile outbreaks, specifically caused by the ribotypes able to use trehalose, is the expanded use of trehalose in the diet since the early 2000s when it became inexpensive to produce, providing a selective pressure for these ribotypes to dominate globally (24). Collectively, these and other studies highlight that additional food modifiers such as emulsifiers (168) and artificial sweeteners (169) should also be considered for their potential to modulate the gut microbiota either because they are directly degraded or because of other indirect mechanisms.

“WATCHING” WHAT YOU EAT: HOW BACTEROIDES SPECIES REGULATE THEIR CARBOHYDRATE INTAKE

Given their abundance of carbohydrate-degrading systems, Bacteroides species have correspondingly developed a series of finely tuned local and global regulatory networks that enable surveillance for available nutrients without the costly expression of enzymes and other functions in the absence of their substrate. Most PULs are regulated by locally encoded regulators that respond to a polysaccharide’s presence and are usually transcription activators since eliminating them blocks PUL activation (12, 170, 171). For these systems to be capable of responding to their specific substrate, there must be a sensory state in which small amounts of critical PUL enzymes and other functions are expressed, an idea that is supported by immunofluorescence imaging of a small number of B. thetaiotaomicron Sus proteins on the cell surface during growth under noninducing conditions such as minimal medium supplemented with glucose (62). Upon exposure to starch, the primed Sus system initiates the degradation of starch, yielding the inducing molecule maltose, which initiates transcription activation through the dedicated inner membrane-spanning regulator SusR. The response to starch exposure occurs incredibly quickly, with transcript and surface protein levels reaching their respective maxima in 5 to 10 min and 30 min after exposure (172). This “priming effect” is likely to occur for most PULs, although the details of the promoters and potential signals that mediate or modify basal expression have not been investigated deeply and, as discussed below, could be a critical point at which global regulation intersects with individual PULs to impose catabolite repression-like effects.

SusR is predicted to be an inner membrane-spanning transcription activator and is one of several different types of PUL regulators that share the theme of sensing saccharide cues in the periplasm or outer membrane, a feature that may enable them to both sense these cues quickly and incorporate additional sugar and linkage information before oligosaccharides are degraded into “less informative” single sugars. Only two PULs with SusR-type regulators have been characterized and both target glucose-based polysaccharides, starch or dextran, and this is a minor regulator type (171). A more abundant mode of regulation that Bacteroides species commonly employ is a variant of the more typical two-component regulator system in which all of the domains of the sensor histidine kinase and DNA binding response regulator are combined into a single protein termed a hybrid two-component system (HTCS). HTCS regulators are found in PULs involved in degrading a variety of polysaccharides, including most PULs associated with dietary fiber degradation and others that are specific for fungal cell wall mannans, chondroitin and heparin sulfates, seaweed-derived polysaccharides, and mucin O-glycans (10, 16, 43, 173, 174) (Fig. 3A). A third type of regulation is governed by extracytoplasmic function (ECF)-σ/anti-σ regulator pairs (Fig. 3B), which so far have been almost exclusively associated with systems responsible for utilizing host-derived mucin O- and N-linked glycans (10, 12, 59, 170). These regulators function similarly to the E. coli FecI/FecR sigma/anti-sigma factor pair (175, 176) by physically coupling a periplasmic domain of the anti-sigma factor to an N-terminal domain of a corresponding TonB-dependent receptor and activating transcription when the substrate is transported across the outer membrane (170, 177, 178). Additional regulators found in PULs include OmpR/streptomyces antibiotic regulatory protein (SARP) family regulators that are also predicted to span the inner membrane (12, 170) and AraC, GntR, NrtR, and LacI-type regulators, which may function by engaging their regulatory cues in the cytoplasm, although few have been investigated in depth (39, 142–144). Given the similarities in PUL architectures between gut and environmental Bacteroidetes, emerging studies are revealing the mechanisms of some of these additional regulator types, including a GntR family member in marine Zobellia galactanivorans that acts as a repressor in the absence of the substrate alginate for its associated PUL (179).

FIG 3.

Local and global regulatory mechanisms involved in Bacteroides carbohydrate utilization. (A) Hybrid two-component system (HTCS) regulators are frequently associated with PULs involved in the utilization of dietary fiber polysaccharides, although they are also associated with other substrates. Unlike classical two-component systems composed of separate response regulator and histidine kinase domains, these functions are fused into one multidomain-containing protein, with an oligosaccharide signal being bound and sensed directly by the HTCS, triggering the upregulation of the adjacent PUL genes (31, 201). Highlighted is one additional regulatory strategy in which the polysaccharide being degraded, chondroitin sulfate, yields unsaturated disaccharides that are bound by the HTCS regulator but also further degraded by a periplasmic GH. The GH that cleaves the inducing disaccharide but not the HTCS is upregulated, serving as a brake on the positive-feedback loop initiated by chondroitin sulfate degradation. (B) ECF-σ/anti-σ factor regulators are commonly found in PULs responsible for degrading host polysaccharides. The anti-σ factor spans the inner membrane and directly interacts in the periplasm with a specialized N-terminal domain of SusC-like transporters (170) to transduce the signal generated by substrate transport, ultimately resulting in the activation of the bound ECF-σ factor. In a subset of these systems found in multiple species, antisense small RNAs are expressed from promoters that overlap the start codon of the susC-like gene and act to repress transcription by an unknown mechanism (180). RNA Pol, RNA polymerase. (C) Global mechanisms likely act to exert carbon catabolite repression-like effects in Bacteroides, although the complete mechanism has not yet been determined. These effects may explain the nutrient utilization hierarchies of some Bacteroides strains, which are illustrated in a simple schematic for the utilization of two different polysaccharides in the top right inset. While both polysaccharides initially trigger PUL transcriptional responses, we hypothesize that after sugars are released from the high-priority polysaccharide (black curve), they enter the cell and repress the expression of the low-priority PUL transcript (blue), possibly through the mechanisms described in the text. Additional global mechanisms that have not been fully resolved involve the CRP homolog (BT4338), which directly controls the transcription of some non-PUL sugar utilization genes and also appears to exert indirect effects on several PULs. For simplicity, several additional outer membrane and periplasmic proteins, such as additional CAZymes and surface glycan binding proteins, are not shown in each panel but are required for polysaccharide degradation. UTR, untranslated region.

While PUL-associated regulators are at the front lines of activating specific responses to polysaccharides, additional layers of local regulation exist. Antisense small RNAs (sRNAs) were recently described in some but not all B. fragilis PULs that are regulated by ECF-σ/anti-σ pairs, occur between the cotranscribed anti-sigma and susC-like genes, and have been found in other species (180) (Fig. 3B). These sRNAs repress PUL transcription since eliminating them by mutating the antisense promoter results in increased PUL transcription both at a basal state and after induction by the cognate polysaccharide. Although the mechanisms through which they work and whether or not these sRNAs themselves integrate additional regulatory cues are unknown, the elimination of one B. fragilis sRNA in a PUL that is repressed by glucose partially alleviated this effect (180). Interestingly, an intergenic region (IGR) that lies between an anti-sigma–susC-like gene pair (the same region in which sRNAs are found) is important for B. thetaiotaomicron to exert fructose-based catabolite repression on the PUL in which the IGR is contained. This effect appeared to be localized to just the PUL containing the deleted IGR, since similar loss of repression was not observed at other PULs (47). This observation further suggests that these sRNA-containing regions integrate global sugar availability information but exert their effect locally at their own PUL.

Additional regulatory mechanisms have emerged from the biochemical function of PUL-associated sensors and enzymes. A B. thetaiotaomicron HTCS required to activate a PUL involved in chondroitin sulfate utilization (12) was found to compete for its activating signal (a 4-5 unsaturated disaccharide generated by the action of chondroitin lyase) with a periplasmic glycoside hydrolase that cleaves this signal into its component sugars. Since PUL activation via the HTCS increases the expression of the gene encoding the cleaving glycoside hydrolase but not the HTCS itself, this is thought to exert a brake on the positive-feedback loop that is initiated by substrate recognition and balance PUL activation with the rate of substrate processing (Fig. 3A) (181, 182).

Finally, posttranscriptional regulatory effects have recently been discovered in PUL regulation and are likely another point at which global nutrient status is integrated locally at the PUL level. In one B. thetaiotaomicron PUL involved in host glycan degradation, a single adjacent HTCS (BT3172) is present to presumably activate expression in response to this substrate (12). Both glucose and fructose reduce the amount of BT3172 protein but not the corresponding transcript, an effect that is mediated by the presence of a 5′ leader on the BT3172 message (183). Deletion of this leader alleviates BT3172 repression, and its effect can also be transferred to other HTCS-coding genes to cause them to be repressed by the same sugars.

Several of the observations described above suggest that connections exist between local PUL expression and the global nutrient status of the Bacteroides cell. Within several Bacteroides species, there are clear nutrient utilization hierarchies, within which PULs that target different polysaccharides are expressed in a defined order, even when all of their respective substrates are present (172, 184, 185). Studies looking for carbon catabolite repression mechanisms similar to those in E. coli and other bacteria have failed to detect the presence of cyclic AMP (cAMP) in Bacteroides (186, 187), and adding exogenous cAMP to B. thetaiotaomicron cultures grown in a mixture of high- and low-priority glycans failed to relieve the repression of PULs involved in degrading lower-priority substrates (47). However, recent studies have determined that a regulator in the same family as the E. coli cAMP receptor protein (CRP), originally designated MalR because it plays a role in maltose metabolism (34), indeed plays a role in the utilization of several polysaccharides by B. thetaiotaomicron. Deletion of malR (BT4338) from B. thetaiotaomicron resulted in substantially diminished growth on several mono- and polysaccharides. Moreover, the utilization of some substrates, like polymeric arabinan and its component arabinose, was almost completely eliminated (145). A follow-up study showed that BT4338/MalR acts as a transcription factor by binding to promoter sequences upstream of several loci responsible for the catabolism of monosaccharides (fucose, arabinose, and xylose), for which a BT4338 mutant also exhibits growth defects (188). Reductions in mRNA abundance were observed for several PULs when BT4338 mutant cells were shifted to carbon starvation conditions and compared to wild-type cells subjected to the same treatment. However, direct binding sites for BT4338/MalR near these PULs were not detected, therefore suggesting an indirect mechanism. One of the strongest responses to carbon starvation was the activation of fusA2, a gene encoding a translation elongation factor whose transcription is dependent on BT4338. Comparison of growth defects on multiple carbohydrates shows that a fusA2 mutant does not exhibit growth defects for the same carbohydrates that the BT4338 mutant does. This decoupled response suggests that PUL expression upon carbon starvation is not directly mediated by FusA2 but is likely mediated by another signaling pathway involving BT4338, while FusA2 is likely important because it provides a distinct protein synthesis pathway during changes in nutrient pools.

While the complete mechanism(s) for orchestrating carbohydrate utilization by Bacteroides remains to be determined, it appears that global signals such as the amount and identity of simple sugars generated by polysaccharide catabolism influence local PUL expression through factors such as BT4338 and perhaps other unknown factors that connect global sugar metabolism to local PUL regulation through transcriptional and posttranscriptional mechanisms. Previous work has revealed that PULs involved in utilizing lower-priority polysaccharides in B. thetaiotaomicron are activated early in a mixture containing multiple substrates (172). However, this response quickly fades, and the transcription of lower-priority PULs is fully or partially repressed until later in growth, presumably after repressing substrates have been used up. This suggests a model in which each PUL is activated by its associated sensor/regulator, an event that occurs quickly upon exposure and often in the periplasm. Then, as different sugars are liberated from the polysaccharides in a mixture, certain dominant sugars enter the cell, where they repress the transcription of low-priority PULs (Fig. 3C, inset). These effects could be exerted at a number of points in the PUL expression process described above but are known to be dynamically initiated by introducing a repressing substrate(s) into a culture that is processing lower-priority polysaccharides, resulting in the rapid loss of transcripts for low-priority PULs (47, 172).

Interestingly, a recent study that directly monitored both B. thetaiotaomicron PUL transcription and carbohydrate depletion over time in a polysaccharide mixture revealed that PULs involved in degrading high-priority substrates remain activated even after their substrate is largely depleted (172). This suggests that PULs targeting high-priority substrates may lack signals to be silenced during the utilization of other substrates such that they remain activated in the presence of a very small amount of their substrate. Retained expression of these high-priority PULs may equip B. thetaiotaomicron to rapidly reengage the degradation of high-priority substrates because the enzymes and other protein apparatus are already present at higher levels on the cell surface. Finally, since there is no general theme connecting the source or component sugar composition of high- and low-priority polysaccharides, it remains to be determined what the biological consequences of this global regulation are for these bacteria in the HGM. Since at least two studies have revealed variations in priority between different species (47, 185), and repressive signals can be recombinantly moved between PULs (183), an intriguing possibility is that polysaccharides are not prioritized due to any intrinsic value differences of their sugars. Instead, these responses could be programmed within each strain or species based on a given set of PULs and their corresponding hierarchical organization. If individual species and strains are differentially programmed with nonoverlapping priorities despite being able to target some of the same substrates, such a phenomenon might decrease competition among gut bacteria and promote the persistence of more diverse HGM communities.

SYNTHETIC BIOLOGY TO ENGINEER GUT BACTERIA

Given the metabolic importance of the HGM and the vast degradative capabilities encoded within its members, there has been increasing interest in using synthetic biology to engineer gut bacteria to perform particular functions. Ethical, regulatory, and consumer considerations aside, a major barrier to approaching this goal is the realization that the majority of species within the HGM lack tools for genetic manipulation. In several foundational experiments, genetically tractable B. ovatus was engineered to couple PUL-mediated transcriptional responses to the ingestion of dietary xylan (189) and induce the production of human cytokines and proteins to treat inflammation, including transforming growth factor β (TGF-β), human intestinal trefoil factor 3, and keratinocyte growth factor 2 (190–192). While it involved only in vitro experiments with engineered bacteria, an elegant example of a recent synthetic biology advance was the optimization of 12 independent, small-molecule-inducible sensors that respond to substrates like arabinose, choline, cuminic acid, and naringenin, among others, into a single strain of E. coli (193). This system serves as an example of how common HGM strains could be engineered to respond to different nutrients present in the gut and couple these signals to the production of different therapeutics or the expression of compound pathways.

Even members of the genus Bacteroides, for which genetic systems have been fairly well developed, have benefited from new tools in recent years that are critical for intentional engineering. These include multichannel fluorescence imaging to distinguish species and strains in vivo (194), tunable expression of genes (195), and the ability to respond to endogenous signals via CRISPR-Cas9-mediated gene expression and record exposures to nutrients via a chromosomally encoded integrase (196). One impediment to the manipulation of a broad set of Bacteroides strains and species is the availability of broadly applicable selectable and counterselectable markers, which has recently been addressed by the introduction of a minimal PUL for inulin utilization into non-inulin-utilizing strains as a means of positive selection and the corresponding use of a type VI secretion system toxin/antitoxin pair for counterselection (197). Perhaps one of the most interesting approaches is the transfer and manipulation of large Bacteroidetes PULs, particularly those that permit the utilization of rare or unique nutrients, into other human gut Bacteroides strains. This strategy can be used to design (via diet or supplements) and fill (via engineered bacteria that use those supplements) “orthogonal niches” around specific nutrients like the seaweed polysaccharides discussed above. Strains that are able to utilize the introduced polysaccharide, sometimes almost exclusively, will gain a substantial colonization advantage, permitting them to be engrafted and supported via diet as new strains in an existing HGM ecosystem (6). This idea is not new and has roots in Rolf Freter’s nutrient niche hypothesis (198) but with modern goals of designing and filling a new niche rather than allowing natural competition for nutrients and adaptation to take place.

As a proof of concept for this idea, a strain of Bacteroides plebeius that was derived from a Japanese adult (199) and contains a PUL for degrading the seaweed polysaccharide porphyran was transferred into mice fed a diet containing the seaweed Porphyra yezoensis, from which it is derived, leading to long-term engraftment of the new strain (6). A similar PUL for porphyran degradation in a B. ovatus strain conferred similar engraftment into conventional specific-pathogen-free (SPF) mice or ex-germfree mice colonized with human gut microbiota (5). This PUL was also transferred into B. thetaiotaomicron and B. stercoris, followed by the successful engraftment of these recombinant strains into conventional SPF mice or ex-germfree mice colonized with complex human microbiomes (5). These effects were tunable based on the amount of purified porphyran added to the diet, and the introduction of this orthogonal niche allowed B. thetaiotaomicron to be supported long enough that it colonized colonic crypts and replaced another previously present and otherwise isogenic strain of B. thetaiotaomicron, which typically blocks the engraftment of incoming strains (194). Taken together, these emerging technologies that allow for the creation of niches and the prolonged introduction of beneficial strains have the promise to help treat a variety of inflammatory and metabolic disorders both within and outside the gut. A key aspect of the design, construction, and deployment of these engineered microbes is the inclusion of appropriate and often redundant biocontainment systems, such as engineered auxotrophies, that ensure that the modified organisms cannot survive and compete when they are naturally released into the environment through feces (200).

PROSPECTUS

Given the central importance of complex carbohydrates and related nutrients to human gut bacteria, in-depth mechanistic studies are continuing to uncover the intricate metabolism of these nutrients by our gut symbionts. Within the HGM, there are likely to be additional carbohydrate nutrients that can serve as nutrients for some of its members. One of the most chemically diverse groups of polysaccharides in the gut is likely to be those that are produced by endogenous bacteria themselves as capsular and exopolysaccharides, and these might be important nutrients for some HGM members. While few of these polysaccharides have been studied in chemical detail, they are likely to contain different linkages relative to those in common plant and host polysaccharides and even novel sugars. Support for the idea that gut bacteria could degrade some of these microbial polysaccharides could lie in the presence of numerous PULs in otherwise well-studied Bacteroides species that have not been connected to known plant-, fungus-, or host-derived polysaccharides. It is probable that the direct production of at least the most abundant and prevalent microbial polysaccharides in the distal gut has provided selective pressure for other gut bacteria to assemble the enzymes and other functions required to utilize them. If the polysaccharides produced endogenously by gut bacteria are capable of feeding others, there could be an unseen and perhaps personalized food web within each of us that sustains and provides resilience to our own HGM.