Impact of the gut microbiome on mucosal inflammation · 2016-02-08 · DC, dendritic cell; LAP, latency associated-protein. Microbiome IL-12 IFN-γ IL-23 IL-17 T cell DC Thickened,

Feature Review

Impact of the gut microbiome onmucosal inflammationWarren Strober

National Institute of Allergy and Infectious Diseases (NIAID), NIH, Bethesda, MD, USA

Review

In the past 10 years it has become increasingly apparentthat the gut microbiome has profound effects on theimmune system to which it is juxtaposed, the mucosalimmune system. Here, I explore recent studies in whichthe effects of the microbiota expand or facilitate anti-inflammatory or regulatory immunological machineryor which favor development of proinflammatory immu-nological machinery in this system. I then focus on howthese opposing processes play out in inflammatorybowel disease (IBD); a disease in which normal immunehomeostasis is disturbed and inflammation takes hold.

Gut microbiota drives IBDIn recent years, studies probing the composition and func-tion of the endogenous microbiota in the normal gastroin-testinal (GI) tract have greatly expanded our appreciationfor and understanding of how the microbiota shape muco-sal immune responses, as well as more global GI tractactivities. To some extent, these studies have been drivenby the desire to understand better IBD.

Crohn’s disease (CD) and ulcerative colitis (UC) arethought to result from a breakdown in mucosal unrespon-siveness to gut commensal organisms [1–3] (Figure 1). Thisconcept is based first on the fact that in the many existingmouse models of colonic inflammation, either those in-duced by various external agents or those occurring spon-taneously in genetically altered mice, one does not seeinflammation in the absence of colonic microbiota [1]. Inaddition, there is now solid evidence that the most promi-nent genetic polymorphisms associated with IBD causedisease (or prevent disease) by affecting responsiveness ofthe mucosal immune system. For example, NOD2 deletionin mice or CD-associated NOD2 polymorphisms in humanslead to increased Toll-like receptor (TLR) responses be-cause such responses are regulated by prolonged or repeat-ed stimulation of NOD2 [4]. Similarly, deletion of theATG16L1 gene in mice, another gene with CD-associatedpolymorphisms, results in hyperactivity of the Nod-likereceptor family, pyrin domain-containing 3 (NLRP3)inflammasome and a polymorphism in the IL-23R genein humans that is associated with decreased risk for devel-oping IBD leads to decreased T cell interleukin (IL)-17

1471-4906/$ – see front matter .

Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.it.2013.07.001

Corresponding author: Strober, W. ([email protected]).Keywords: Gut microbiome; Mucosal regulatory T cells; Anti-inflammatoryorganisms; Proinflammatory (colitogenic organisms); Inflammatory Bowel Disease.

responses [5,6]. Finally, as discussed below, although coli-togenic microbiota can be demonstrated in certain mousemodels of colonic inflammation, there is as yet little evi-dence that such microbiota can cause persistent disease inthe normal host.

In the review below I first summarize current informa-tion on how the microbiota of the GI tract either controls orprevents gut inflammation by the induction of regulatory Tcells (Tregs), and then discuss data suggesting thatchanges in the microbiota can also result in the opposite,namely the induction or aggravation colitis. I then reviewthe now extensive information on microbiota changes inIBD patients and its possible relation to the causation ofthis disease.

Mucosal homeostasis and Tregs induced by gutmicrobiotaIn recent years, considerable evidence has accumulatedsupporting the notion that the gut microbiota inducesmucosal Tregs that then play a vital role in maintaininggut homeostasis under normal conditions or in controllinginflammatory responses that would lead to disease. Evi-dence of this type was first obtained in studies in which thegut epithelial barrier was transiently perturbed by intrar-ectal administration of ethanol or Vibrio cholerae zonulaoccludens toxic hexapeptide; agents that cause increasedepithelial permeability and increased exposure of laminapropria cells to luminal commensal microbiota [7]. Suchtreatment was shown to result in barely perceptible andtransient inflammation accompanied by IL-10-dependentinduction of forkhead box (Fox)p3-negative, cell-surfacetransforming growth factor (TGF)-b-positive CD4+ Tregsthat could be shown to protect mice from induction oftrinitrobenzene sulfonic acid (TNBS) colitis. Importantly,the development of these Tregs requires the presence of thegut microbiota and the presence of TLR2; it was thusestablished that innate TLR2 responses initiated by themicrobiota are necessary for Treg development.

Further work confirming and expanding on theseresults utilized germ-free mice recolonized with an ‘alteredSchaedler’s flora’; a nonpathogenic mixture of commensalorganisms [8]. Here, one observes induction of CD4+ Tregsthat in this case are Foxp3-positive and IL-10-independent.The development of these Tregs is dependent on both innateand adaptive immune responses, because recolonizedmyeloid differentiation primary response gene 88/TIR-containing adaptor molecule-1 [MyD88/Ticam-1 (TRIF)]

Trends in Immunology, September 2013, Vol. 34, No. 9 423

Microbiome

DC

Foxp3+ Treg cells

Foxp3– LAP+ Treg cells

Innate s�muli

TRENDS in Immunology

Figure 2. Homeostatic regulation of mucosal responses. Although the epithelial

cell layer of the intestine is a barrier to the entry of luminal materials, it is porous

enough to allow entry of microbiome-derived components that mediate the

development of regulatory T cells (Tregs). Innate stimuli [Toll-like receptor (TLR) or

NOD-like receptor (NLR) ligands] can interact with dendritic cells in the lamina

propria and these, in turn, activate regulatory cells that maintain a quiescent state

in the unperturbed intestine despite the presence of adjacent microbiota. In

addition, innate stimuli may interact directly with the regulatory cells as well. Two

types of regulatory cells have been described: a T regulatory cell 1 (Tr1)-like

[interleukin (IL)-10-dependent] cell that bears surface transforming growth factor

TGF-b in the form of latency-associated protein (LAP). Under conditions in which

the epithelial barrier is made more porous, these cells increase in number, but in

this case their function is overwhelmed by a strong proinflammatory stimulus.

Abbreviations: DC, dendritic cell; LAP, latency associated-protein.

Microbiome

IL-12 IFN-γ

IL-23IL-17

T cellDC

Thickened,inflamed gut wall

Innates�muli

Innate s�muli

TRENDS in Immunology

Figure 1. The basis of Crohn’s disease (CD), a major form of inflammatory bowel

disease. The consensus view of CD pathogenesis is that the latter consists of, at

least in part, an excessive immunological response to some component of the

microbiota existing in the bowel lumen, most likely a microbial-associated

molecular pattern (MAMP) interacting with either a Toll-like receptor (TLR) or an

NOD-like receptor (NLR). Thus, MAMP stimulation gives rise to production of

cytokines such as interleukin (IL)-12 or IL-23 that induce T cell differentiation into T

helper (Th)1 or Th17 cells, the ultimate effectors of the inflammation. The

excessive response could be due to a direct disturbance in the induction of effector

cells or to an indirect disturbance in the regulatory cells that control such

induction. The mechanism underlying ulcerative colitis is thought to be similar

although the cellular processes resulting from the excessive response is different.

Abbreviations: DC, dendritic cell; IFN, interferon.

Review Trends in Immunology September 2013, Vol. 34, No. 9

double-deficient mice whose T cells bear a T cell receptor(TCR) transgene specific for lymphocytic choriomeningitisvirus (SMARTA mice) that do not respond to commensalorganisms, exhibit greatly impaired Treg responses. In addi-tion, in the absence of Treg development or IL-10, the micemanifest robust IL-17 and interferon (IFN)-g responses in thecolonic lamina propria, albeit in the absence of tangibleinflammatory changes. Finally, in studies parallel to thoseinvolving mice treated with ethanol, recolonized mice pre-exposed to a low dose of dextran sulfate to cause injury to theintestinal barrier, developed heightened Treg responses andminimal cytokine responses, whereas recolonized SMARTAor MyD88/Ticam1-deficient mice failed to develop heightenedTreg responses and exhibited vigorous cytokine responses;the latter associated with high mortality.

These studies, taken together, lead to the view thatintestinal homeostasis, that is, the noninflamed state ofthe normal intestinal, is dependent on Tregs induced bycommensal microbiota that gain entry into the laminapropria. Thus, although unfettered entry of commensalsinto the lamina propria due to gross epithelial damage[9] may cause severe inflammation, low level and tran-sient entry has the opposite effect of girding the laminapropria from inflammatory influences. These conceptscondition our understanding of IBDs because they makeit likely that the inflammation of the GI tract thatdefines these diseases must initially overcome two an-ti-inflammatory barriers: the barrier imposed by Tregsinduced by commensal microbiota and that created byTregs that are generated by the inflammation itself(Figure 2).

424

Induction of Tregs and prevention of colitis by specificcommensal organismsSo far, I have focused on the ability of the intestinalcommensal microbiota as a whole to induce regulatoryeffects rather than on the ability of individual membersof the commensal microbial community with a specialpropensity to induce such effects. However, there are, infact, studies that show that certain bacteria are moreeffective than others in inducing Treg activity in the GItract. Perhaps the most complete of these studies relate tothe regulatory function of nonenterotoxigenic Bacteroidesfragilis, a commensal organism existing within the colonicmicrobiota of both mice and humans. Early studies showedthat mice mono-colonized with B. fragilis give rise to IL-10-producing Tregs that can protect recombination activatinggene 2 (RAG2)- deficient mice from Helicobacter-hepaticus-induced colitis; furthermore, these cells are induced by apolysaccharide produced by B. fragilis known as polysac-charide A (PSA) [10]. Later studies expanded on thesefindings by showing that colonization of germ-free micewith PSA-producing B. fragilis but not PSA non-producingB. fragilis elicits induction of Foxp3+ T cells that produceIL-10 and exhibit a PSA-specific profile as they produceTGF-b2 and not TGF-b1, and do not manifest increasedexpression of cytotoxic T-lymphocyte antigen-4 (CTLA) orglucocorticoid-induced TNFR-related protein (GITR) whenstimulated by PSA [11]. The regulatory function of thePSA-induced Foxp3+ T cells has been revealed in mice withTNBS colitis treated with PSA. PSA elicits Foxp3 Tregsthat suppress effector cell responses and ameliorates coli-tis when administered before and after TNBS administra-tion. On this basis, PSA has been proposed as a possibletreatment of human IBD, but this possibility will have to

Review Trends in Immunology September 2013, Vol. 34, No. 9

await studies showing that it can be given in a way thatdoes not also induce T helper (Th)1 effector cell responses,as shown in some studies [12]. In addition, PSA may beantigenic and lose effectiveness when repeatedly adminis-tered.

An interesting aspect of B. fragilis PSA regulatoryactivity is that it does not increase the number of Foxp3+

regulatory cells in TLR2-deficient mice and is thus TLR2dependent [11]. This initially posed a problem in explain-ing its mechanism of action inasmuch as other bacteriaalso expressing TLR2 ligands did not have a similar ca-pacity to induce regulatory cell activity. This problem wasresolved in additional studies that showed that PSA, butnot conventional TLR2 ligands, interacts directly withTLR2 on Foxp3+ T cells rather than on dendritic cells toinduce production of IL-10, and to establish the profile ofPSA-induced Foxp3+ cells mentioned above [12]. The pos-sible biological significance of PSA stimulation of regula-tory T cells via TLR2 was revealed in studies that showedthat colonization of the gut with B. fragilis reduces IL-17production via PSA induction of Tregs, and that B. fragilisexpressing PSA (but not those not expressing PSA) arecapable of colonizing colonic crypts in close proximity to themucosal immune system, presumably because of theircapacity to reduce local IL-17 production [12]. This leadsto the concept that PSA is representative of a new class ofTLR ligands that induce regulatory responses rather thaninflammatory responses.

Although the ability of PSA to induce Treg activity isnow well established by the studies discussed above, itsoverall function in mucosal homeostasis remains unclear.It may be that under normal homeostatic conditions it hasa limited role in enabling B. fragilis to occupy a particularniche in the colonic crypts rather than a more global role ininducing suppressor T cells. This conclusion is suggestedby compelling recent evidence that Clostridium ratherthan Bacteroides are the primary drivers of Foxp+ T celldevelopment in the colon.

Turning now to this Clostridium-related regulatoryactivity, it was shown initially that germ-free mice provid-ed a specific-pathogen-free (SPF) flora that exhibited anincrease in the numbers of Foxp3+ T cells in the colon, andthat this change depends on the Gram-positive, spore-forming fraction of the SPF (i.e., a fraction that excludesB. fragilis) [13]. In subsequent studies designed to identifywhich bacterial species induce Tregs, the approach takenwas to reconstitute germ-free mice with various cocktails ofspecific organisms and determine the effect of each cocktailon Foxp3+ cell levels in the colon. Although segmentedfilamentous bacteria (SFBs), and large collections of Bac-teroides or Lactobacillus, had no capacity to increase thenumber of Foxp3+ cells, clostridial groups, particularlythose belonging to clusters IV and XIVa, had a strikingcapacity to increase the number of Foxp3+ T cells. Inaddition, colonization with Clostridium enhanced colonicTGF-b concentration as well as other Treg-inducing mole-cules. These effects could also be observed in MyD88-,receptor interacting protein kinase-2 (RIP2)- andCARD9-deficient mice and were thus independent ofTLR, NOD or Dectin receptor signaling. Finally, itwas shown that Clostridium but not other bacteria

(including Bacteroides) induced IL-10-producing cells inthe colon. One caveat to these various findings is that theyapply to colonic rather than small intestinal Foxp3+ T cellsas reconstitution with clostridial organisms affected nei-ther Foxp+ T cells nor IL-10-producing T cells in the latterlocation.

Finally, in studies of the clinical significance of theabove findings concerning clostridial organisms, it wasshown that mice colonized with Clostridium at an earlyage (Clostridium-abundant mice) developed less severedextran sodium sulfate (DSS) colitis than control mice,and exhibited a reduced tendency to mount Th2 cytokineand IgE antibody responses [13]. These findings mesh withthe fact that Clostridium clusters IV and XIVa are propor-tionally reduced in patients with IBD, as discussed ingreater detail below. Thus, overall, clostridial organismsemerge as a major inducer of Tregs, albeit by mechanismsthat are as yet undefined (Figure 2).

Colitogenic bacterial microbiotaThe mirror image (or rather, negative image) of bacterialmicrobiota that induce regulatory cells and thus protectorganisms from inflammation are the so-called colitogenicorganisms mentioned at the outset of this review, whichcause de novo intestinal inflammation in normal mice. Thefirst convincing evidence that such colitogenic organismsexist came from studies of recombination activating gene 2(RAG2)- deficient, T-bet-deficient mice that develop spon-taneous colitis and have been termed TRUC mice, reflect-ing both the immunodeficiency (TR) and the resemblanceof the colitis to human IBD (UC) [14]. The origin of thecolitis in these mice can be traced to the fact that T-bet is atranscriptional repressor of tumor necrosis factor (TNF)-ain dendritic cells and its absence leads to excessive TNF-aproduction by these cells, which then synergizes with IL-23to drive IL-17 production by innate intestinal cells; inaddition, the mice lack Tregs due to the RAG2 deficiency[14,15]. The TRUC model of colitis would not have beenespecially remarkable were it not for the fact that studies ofthis model showed for the first time that mice with experi-mental colitis could develop a colitogenic flora that trans-mitted colitis vertically to wild type (WT) pups nursed byTRUC mice and horizontally to co-housed WT mice thatthen exhibited some level of colitis for considerable lengthsof time, even when separated from TRUC mice (Figure 3).

Subsequently, molecular characterization of fecal flora(based on 16s rRNA analyses) was conducted to define theorganisms that were colitogenic in TRUC mice [16]. Themost important findings to emerge from this analysis wereevidence that two bacterial species, Klebsiella pneumoniaeand Proteus mirabilis, were likely to be components of thecolitogenic flora. This evidence consisted of the fact thatthese organisms are present in TRUC fecal flora as well asin the flora of WT mice fostered by TRUC mice, andantibiotic treatment of the TRUC mice that amelioratesthe colitis reduces the numbers of these bacteria in thefeces to levels below the limit of detection. However, thesebacteria appear to require interaction with other bacteriain the normal flora to cause colitis because they do notcause colitis in germ-free TRUC mice free of colitis, but docause colitis in wild type or RAG2�/� mice with a normal

425

Spontaneous coli�s Increased severity of DSS-coli�s

Colitogenic microbiotain intes�nal crypts

Colitogenic microbiota

TRUC mice (T cell/T-bet deficient) NLRP3-deficient mice

TNF TNF

IL-23 IL-23

TRENDS in Immunology

Figure 3. Two types of colitogenic microbiota. Studies of murine models of gastrointestinal inflammation have identified two types of colitogenic microbiota, that is,

microbiota that cause colitis in a normal host. Truly colitogenic microbiota, so far only identified in TRUC mice (see text) are organisms that cause spontaneous colitis in co-

housed normal mice. Partially, colitogenic microbiota are represented by Prevotella organisms that proliferate in the intestinal crypts of mice lacking epithelial cells that

produce NLRP3 and therefore are deficient in interleukin (IL)-18 production. These organisms are partially colitogenic because they cause intensification of a pre-existing

colitis and not de novo colitis. Abbreviations: DSS, dextran sodium sulfate; NLRP3, NOD-like receptor family, pyrin domain-containing 3; TRUC, T-bet/RAG2-deficient,

ulcerative colitis-like; TNF, tumor necrosis factor.

Review Trends in Immunology September 2013, Vol. 34, No. 9

(pathogen-free) flora. Thus, whether these organisms aretrue colitogenic bacteria that require help from otherbacteria or are instead necessary helpers of true colitogenicbacteria awaits additional study. This question is relevantto findings in a more recent study of bacteria in TRUC micein which it was found that a substrain of TRUC that doesnot develop colitis differs from mice that develop colitis bythe fact that their microbiota lacked Helicobacter typhlo-nius and develop colitis if they are administered the latterorganism [15]. Thus, it is possible that H. typhlonius is themost proximal colitogenic organism in TRUC mice.

In a final set of studies TRUC mice were administeredeither anti-TNF-a neutralizing antibody or Tregs to deter-mine if restoration of immune function influences colitis inmice [15]. Both therapies ameliorate colitis and, althoughadministration of anti-TNF-a diminishes K. pneumoniaeor P. mirabilis levels, administration of Tregs does not. Thelatter fact suggests that these organisms and/or otherorganisms in the colitogenic microbiota require the pres-ence of an appropriately abnormal mucosal immunologicalmicroenvironment to cause disease and are not intrinsi-cally colitogenic, despite their ability to cause disease inWT recipients for some period of time. It is possible, forinstance, that the colitogenic potential of these organismsis enhanced by interaction with proinflammatorycytokines.

Another set of studies revealing the presence of organ-isms that are associated with colitis involves studies ofmice with deficiencies of NLRP6. The latter is a member ofan NLR family of intracellular microbial recognition mole-cules that activate inflammasomes, that is, molecularcomplexes that result in the activation of caspase 1 andthus the proteolytic cleavage and then secretion of activeIL-1b and IL-18. NLRP6 is a unique member of the NLRinflammasome family because its expression is localized toepithelial cells and thus it is particularly likely to beactivated by microbial flora inhabiting the microbial–epi-thelial interface.

Initial studies of NLRP6 showed that mice deficient inthis molecule manifest more severe induced colitis (DSScolitis) or even mild spontaneous colitis [17,18]. Laterstudies have provided evidence that NLRP3 deficiencyleads to a change in the bacterial microbiota that mediates

426

these susceptibilities [19]. This consists of the fact that WTmice co-housed for prolonged periods develop DSS colitis ofequal severity to their NLRP6-deficient cohorts. Similarly,WT mice co-housed with mice deficient in apoptosis-asso-ciated speck-like protein containing a CARD (ASC), a keyinflammasome component necessary for caspase I cleavageor co-housed with mice deficient in IL-18 also developcolitis, establishing that the NLRP6 deficiency is in factan inflammasome-related defect. It is important to pointout, however, that the increased susceptibility to DSScolitis in WT mice is a transient abnormality and WT miceseparated from NLRP6-deficient mice eventually lose theincreased susceptibility.

Subsequent studies of the microbiota in NLRP6–ASC-and IL-18-deficient mice revealed that these mice or indeedWT mice co-housed with these mice exhibited a microbiotaenriched for the anaerobic taxa, Prevotellaceae and candi-date bacterial phylum TM7 and that treatment of thesemice with antibiotics abolished the transferability of DSScolitis susceptibility to co-housed WT mice [19]. Of interest,the Prevotella organisms in deficient mice were locatedadjacent to epithelial cells in the intestinal crypts, indicat-ing that lack of epithelial inflammasome activity led to adefect in the ability of the deficient mice to control theproliferation of certain potentially pathogenic intestinalorganisms normally occupying this micro-niche. The mech-anism of this defect is, however, not yet known.

The ‘dysbiosis’ occurring in NLRP6-deficient mice mayapply to mice with other inflammasome defects affectinghematopoietic cells rather than epithelial cells, and indeedto mice with other defects in innate immune responses; thispossibility is in fact suggested by the finding that NLRP3-deficient and NOD2-deficient mice also exhibit increasedseverity of induced colitis and develop an altered micro-biota [20,21]. In addition, there is evidence in the case ofNOD2 deficiency that this altered microbiota causes in-creased susceptibility to induced colitis in WT mice. In anycase, it is important to point out that the organismsoccurring in mice with NLRP6 inflammasome defects(and possibly other defects) differ from those associatedwith TRUC mice discussed above in that they predisposenormal (or deficient) mice to induced colitis rather thancausing spontaneous colitis; thus, these organisms are

Review Trends in Immunology September 2013, Vol. 34, No. 9

more properly called colitis-predisposing organisms ratherthan colitogenic organisms. The significance of this lies inthe fact that these organisms, even more than the TRUCorganisms, probably do not cause intestinal inflammationin themselves and thus cannot initiate colitis in humans inthe absence of an intrinsic immune defect (Figure 3). Afinal point relating to the above findings in NLRP6-defi-cient mice concerns a recent comprehensive study of theintestinal microbiota of mice deficient in various TLRs orMyD88 that disclosed that, although the microbiota in thevarious mice differed markedly from each other, theygenerally did not differ greatly from WT littermates[22]. This suggests that the differences originated froman ‘extended husbandry in isolation’ rather from differ-ences in innate immune responsivity. It can be argued thatthese findings call into question the idea that NLRP6deficiency actually causes changes in the microbiota asso-ciated with the epithelium as implied above; however, thisis not likely because the appearance of colitis–predisposingorganisms in the intestine occurs in mice with ASC and IL-18 deficiency as well as NLRP6, indicating that a particu-lar inflammasome-mediated mechanism in a variety ofmouse colonies leads to the proliferation of a colitis-pre-disposing organism even in mice of differing origins.

Gastrointestinal microbiome in IBDIn the light of the above studies detailing either the anti- orproinflammatory effects of gut microbiota, it becomes ofgreat interest to define the microbiota of patients with IBDand thus to determine if the organisms contained withinthis microbiota contributes to the occurrence of suchdisease.

A considerable number of surveys of the microbiota inpatients with IBD and control individuals using metage-nomic analyses of 16S rRNA in extracted gut specimens orfecal material have now been conducted. Although thesemolecular techniques are generally superior to culture-based techniques, because of inherent difficulties in cul-turing many members of the microbial community, theystill have some limitations. Chief among these is that theyrely on PCR-based techniques that may not detect bacteri-al species present in low abundance and they quantify copynumber of 16S rRNA species rather than true bacterialnumbers. Other difficulties in such surveys that adhere toculture-based surveys as well is that patient populationsare both environmentally and genetically heterogeneous sothat results obtained with studies (especially with smallnumbers of patients) may not reflect general abnormalitiesor may pertain to only a subclass of patients.

One landmark study conducted by Frank et al. in 2007in large groups of patients provides a general frameworkfor microbiota abnormalities observed in IBD patients,and serves as a basis for evaluation of changes found instudies of smaller patient groups [23]. These investiga-tors analyzed surgically obtained gut-wall biopsy speci-mens and found that in colons of patients with IBD (bothCD and UC) Bacteroidetes and Firmacutes (Lachnospir-aceae family) were depleted whereas Proteobacteria(which contain Escherichia coli species) and the Bacillusgroup of Firmacutes were increased. In small intestinesof patients, the Bacillus group of Firmacutes were

decreased and Proteobacteria were increased relativeto controls, whereas Bacteroidetes were unchanged. Up-on principle component analysis, however, these overalldifferences were due to microbiota changes in only one-third of CD patients and one-quarter of UC patients, andthe remainder exhibiting a normal microbiota pattern.In these subgroups, the decrease in Firmacutes (Lach-nospiraceae family) and Bacteroidetes were particularlyapparent in both colon and small intestine.

A notable feature of this analysis is that it did notdisclose a particular bacterium present at levels expectedof a pathological agent; in particular, the analysis detectedfew if any copies of Mycobacterium avium ssp. paratuber-culosis rRNA; an agent that has been linked to CD in somestudies [24]. In addition, the subset of IBD patients withabnormal intestinal microbiota was younger and morelikely to have disease associated with abscess formation.These associations suggest that the abnormal microbiotaare a feature of more severe disease and could thus be afactor that aggravates disease. This view fits with the factthat an abnormal microbiota was not found in the majorityof patients and is thus unlikely to be a primary etiologicalfactor. Finally, this analysis revealed that IBD small in-testinal microbiota was characterized by reduced diversityin the Bacteroidetes and Firmacutes phyla, meaning thatthe number of distinctly different bacterial clones in thesephyla was decreased.

Other groups of investigators, also using molecular meth-odology to assess bacterial populations associated with gutspecimens, obtained results that were similar in some waysand different in others (reviewed in [25]). In summary, thesestudies of smaller and mostly CD patient populations pro-vide data that add to those obtained in the study by Franket al., in that they emphasize that decreases in the popula-tion of Firmacutes include Faecalibacterium prausnitiziidecreases and increases in the population of Proteobacteriainclude E. coli increases; in addition, they suggest that theseFirmacutes/Proteobacteria abnormalities may be limited toCD patients with small bowel disease [26–31]. The picturewith respect to Bacteroidetes is somewhat less clear in thatseveral studies found decreased number of these bacteriawhereas in other studies, most notably those examining theilia, increased concentrations of Bacteroidetes (B. fragilis)were found [29,32,33].

The molecular analysis of the gut microbiome inpatients with IBD summarized above, as well as previousculture-based studies of the microbiota not discussed here,offers several insights into the role of commensal bacteriain the etiology of this disease. They provide strong evidencethat a single pathogenic bacterial organism is not the rootcause of IBD-related inflammation. This evidence beginswith the fact that the molecular analyses did not reveal thepresence of any known pathogen in sufficient numbers tocause inflammation [23], but also includes the fact that inseveral studies distortions in the microbiota were presentin uninvolved tissue and/or were absent from involvedtissue (such as colonic tissue), and that distortions tendedto disappear when patients were administered agents thatameliorated immunological abnormalities, suggesting thatthey were secondary effects of underlying immunologicaldefects [34].

427

Review Trends in Immunology September 2013, Vol. 34, No. 9

Distortions in the gut microbiota are not found in allpatients and tend to occur in only certain parts of theintestine and in patients with more severe disease[21,26]. In addition, such changes in microbiota are moreevident in CD than in UC [31]. These facts argue againstthe idea that the distortions are a universal driver of theIBD disease process and for the idea that they are second-ary and situational. However, these views need to betempered by the realization that the methods so far usedto analyze the gut microbiota may be grossly deficient inidentifying the presence of more subtle bacterial abnor-malities.

A consistent finding in CD is the selective loss ofFirmacutes and Bacteriodetes organisms that conceivablycould be members of the microbiota important in theinduction of regulatory cells, as suggested by the murinestudies of B. fragilis and Clostridium, as discussed above.The most striking of these findings relates to Firmacutesprausnitzii, which, as discussed above, is a clostridialorganism that is consistently decreased in patients withCD; particularly those with small bowel inflammation [29].In direct studies of the immunoregulatory function of thisorganism, Sokol et al. showed that human peripheral bloodmononuclear cells (PBMCs) cultured in the presence of F.prausnitzii exhibited a higher ratio of IL-10 to IL-12 pro-duction than cells cultured in the presence of several othercommensal organisms; moreover, supernatants from F.prausnitzii cultures suppressed IL-1b-induced IL-8 secre-tion and nuclear factor (NF)-kB reporter gene activity inCaCo-2 cells [35]. Perhaps more importantly, intragastricadministration of F. prausnitzii or a supernatant obtainedfrom its culture reduces the severity of TNBS colitis andlowers colonic IL-12 production; in addition, such treat-ment tends to correct the dysbiosis observed in mice withTNBS colitis. These studies support the idea that changesin F. prausnitzii in the microbiota are a significant factor inthe severity of CD, but this requires further investigationbecause the cytokine changes observed in vivo, althoughstatistically significant, are nevertheless rather marginal.Finally, with respect to B. fragilis, one can hardly claimthat lack of this bacterium is playing an immunoregulatoryrole in IBD, because, as noted above, these organisms aregreatly increased in the biofilm of IBD patients.

A final insight of the analyses of bacterial populations inIBD concerns the finding that E. coli is frequently in-creased in patients with CD [27]. This finding coincideswith a series of studies showing that a subtype of E. coliwith semipathogenic properties, called adherent–invasiveE. coli (AIEC), is frequently found in patients with CD andplay some role in the pathogenesis of this inflammation[36,37]. AIEC occurs in the inflamed ileum of �22% of CDpatients with chronic inflammation and a somewhathigher percentage in the terminal small bowel in postsur-gical patients, but in only 6% of the ilia of control patients[36]. However, it is usually not found in affected colons ofpatients and is found in 22% of the iliae of patients withoutileal inflammation; thus, AIEC is by no means a universalaccompaniment of Crohn’s inflammation. As indicated byhow they have been named, AIEC adheres to and colonizessmall intestinal epithelial cells [37]. This important prop-erty of the organism is explained by the fact that AIEC

428

expresses a unique type I pilus that has the capacity tobind to CEACAM6 on the surface of epithelial cells andthus facilitates AIEC attachment to such cells [38]. Notunexpectedly, such binding is dependent on the level ofcarcinoembryonic antigen-related cell adhesion molecule 6(CEACAM6) expression. This is shown by the fact thatmice bearing a CEACAM (CEABAC) transgene in epithe-lial cells and expressing large amount of epithelial CEA-CAM6 exhibit massive AIEC colonization [39]. Thisrelation between CEACAM6 expression and colonizationmight explain the occurrence of AIEC colonization in CDinasmuch as the secretion of inflammatory cytokines suchas TNF-a upregulates CEACAM6 expression.

Two important questions relating to the role of AIEC inCD concerns first its capacity to penetrate the epithelialbarrier and thus invade the mucosa, and second, the inflam-matory properties once such penetration is achieved.

Regarding their invasive properties, AIEC organisms arefound in epithelial cells and in the lamina propria of in-flamed CD patients, but it is not clear if such penetration isoccurring through an already damaged (and possibly ulcer-ated) epithelium or through an intact epithelium. The factthat there are no reports showing that AIEC occurs withincells or tissues of uninflamed mucosa suggests that somedamage to the epithelium is antecedent to such invasion.The latter conclusion is concordant with the fact that al-though the invasive (and proinflammatory) properties ofAIEC are impressive, these organisms do not manifest theinvasive properties of true pathogens such as Shigella andSalmonella because, if this were the case, they would causeinflammation throughout the bowel and not just in areas ofCrohn’s inflammation. These caveats concerning the inva-sive capacity of AIEC are counterbalanced by the fact thatmice expressing a transgene expressing human CEACAM6mentioned above do exhibit AIEC translocation and passageinto the lamina propria [39]. In addition, these mice expressincreased amounts of claudin in the plasma membrane (apore-forming molecule) and manifest increased epithelialpermeability [40]. Thus, when CEACAM6 is highlyexpressed, penetration of pre-existing normal epitheliummay occur and this may be the case in inflamed IBD tissueexhibiting upregulated CEACAM6.

Related to the fact that AIEC can gain access to themucosa via epithelial cells is recent evidence that AIEC, incommon with several gut bacterial pathogens, express longpolar fimbriae that allow them to bind to an M cell glyco-protein called GP2 and thus enter M cells [41]. Inasmuchas M cells are specialized epithelial cells that take up andtranslocate bacteria and soluble molecules into the Peyer’spatches, this property of AIEC is a means by which thesebacteria can interact with the organized mucosal immunesystem.

AIEC taken up by both epithelial cells and macrophagestends to replicate and survive within these cells and inducethe macrophages to produce inflammatory cytokines suchas TNF-a and IL-6 [42,43]. Recently, it has been shownthat such survival in macrophages is controlled, at least inpart, by autophagic machinery, and cells with variousautophagic defects due to impaired expression of genesassociated with CD that adversely affect autophagy, suchas ATG16L1 and NOD2, support increased intracellular

Review Trends in Immunology September 2013, Vol. 34, No. 9

AIEC survival and cytokine induction [44]. This raises thepossibility that CD patients with these genetic abnormali-ties are more prone to AIEC colitogenic effects.

The adherent, invasive, and proinflammatory proper-ties of AIEC discussed above have led to the suggestionthat these organisms are primary causative factors in CD,at least in some patients. However, several observationsmitigate against this possibility. First, as alluded to above,AIEC colonization is dependent on CEACAM6 expressionand it is known that the latter can occur as a result ofinflammatory cytokine secretion; thus, it is possible or evenlikely that AIEC colonization is secondary to an underlyinginflammation rather than its cause. To counter this argu-ment one would have to have evidence that AIEC caninduce CEACAM6 expression in cells that cannot them-selves produce cytokines that would have this effect. Sec-ond, as already mentioned, penetration of AIEC organismsin the Crohn’s mucosa lacks the uniformity that one mightexpect of a colitogenic organism, and the inflammation ofthe lamina propria in CEACAM6 transgenic mice, in whichpenetration is more, uniform is a neutrophil-dominantinflammation that does not resemble the granulomatousinflammation of CD [39]. Third and finally, if AIEC doesplay a major role in CD, there should be evidence that thosepatients with AIEC colonization exhibit responsiveness totreatment with antibiotics to which AIEC are sensitive. Todate, antibiotic treatment of CD patients with a variety ofagents has had only marginal success and, perhaps moreimportantly, no patient subgroups have emerged (presum-ably those with AIEC colonization) who are particularlyresponsive to such treatment [45].

On the basis of the above, AIEC is most likely a secondaryand aggravating factor in CD whose colitogenic potential isdependent on the presence of pre-existent and geneticallydetermined proinflammatory immune abnormalities [46](Figure 4). This view obtains additional support from arecent study of mice with TLR5 deficiency, that is, micethat do not bind flagellated organisms on the basolateral

Macrophage

TNF-α

Foxp3+ Treg cells

Immunoregulatorymicrobiota

(Clostridium organism?)

Infec�ousmicrobiota

(AIEC)

TRENDS in Immunology

Figure 4. The gut microbiota may cause or aggravate Crohn’s disease (CD) by

defective induction of regulatory T cells (Tregs) or by infection of the mucosa and

the induction of inflammatory cytokines. Studies of mice and humans with

gastrointestinal inflammation have led to the identification of two kinds of

microbiota that may cause or aggravate the inflammation of CD. In the first

category are organisms in the Firmacute phylum represented by Faecalibacterium

prausnitzii, a clostridial organism that may be related to those shown in murine

systems to induce Tregs in the gut. In the second category is an Escherichia coli

organism known as adherent–invasive E. coli (AIEC) that lacks pathogenic

properties in normal intestine but that colonizes the small intestine of some

patients with CD. This organism invades macrophages and induces the latter cells

to produce proinflammatory cytokines. Abbreviations: Foxp3, forkhead box p3;

TNF, tumor necrosis factor.

surface of epithelial cells due to absence of the TLR5 flagellinreceptor [47]. It was found that such mice exhibit spontane-ous colitis and that the latter is associated with the devel-opment of a microbiota that includes organisms that bind tothe intestinal surface and contains increased numbers ofproteobacteria and enterobacteria such as E. coli. That suchorganisms initiated the colitis was suggested by the fact thatgerm-free TLR5-deficient mice, but not WT mice, weresubject to the development of chronic colitis when exposedto a reference strain of AIEC, and that such colitis persistedafter this organism was no longer present. One explanationof these findings is that bacteria that occasionally penetratethe mucosal barrier interact with TLR5 and thus stimulateepithelial cell production of factors (such as chemokinescausing neutrophil infiltration) that ordinarily limit prolif-eration of such penetrant bacteria. Thus, in the absence ofsuch TLR5 function, local proliferation of bacteria in thelamina propria is more likely to occur and this causesgeneral changes in epithelial function that cause colitis.This may include loss of barrier function that leads on theone hand to the excessive TLR responses that define thecolitic state, and on the other hand, to changes in themicrobiota such as the appearance of AIEC-like organismsthat aggravate the proinflammatory process.

Overall then, these studies of TLR5-deficient mice againsuggest that organisms with the properties of AIEC arecolitogenic but only in the sense that they trigger inflam-mation in the presence of an underlying genetic defect (inthis case involving epithelial cell function). Thus, we comeback to the conclusion that whereas colitogenic organismscapable of initiating de novo inflammation in a normal hosthave been identified in experimental (murine) models of GIinflammation, such as TRUC mice, they have not yet beenidentified in human IBD.

Concluding remarksThe study of the relation of the gut microbiota to thedevelopment and maintenance of the mucosal immunesystem is a dynamic and rapidly expanding area of research.We know that the organisms comprising the gut microbiomeact severally and singly to shape both the anti-inflammatory(i.e., regulatory) as well as the proinflammatory aspects ofmucosal function. In addition, we know that the gut micro-biome can be altered by disease and that such alterationsmay be primary or secondary factors in disease pathogene-sis. Finally, we know that genetic factors underlying thedevelopment of disease in many or all instances affect thelatter because of defective responses to organisms in the gutmicrobiome. Thus, one of the major research challenges inthe area of gut inflammation is to define the precise mecha-nisms that underlie these defective responses.

References1 Strober, W. et al. (2001) The immunology of mucosal models of

inflammation. Annu. Rev. Immunol. 20, 495–5492 Kaser, A. et al. (2010) Inflammatory bowel disease. Annu. Rev.

Immunol. 28, 573–6213 Strober, W. et al. (2007) The fundamental basis of inflammatory bowel

disease. J. Clin. Invest. 117, 514–5214 Strober, W. and Watanabe, T. (2011) NOD2, an intracellular innate

immune sensor involved in host defense and Crohn’s disease. MucosalImmunol. 4, 484–495

429

Review Trends in Immunology September 2013, Vol. 34, No. 9

5 Saitoh, T. et al. (2008) Loss of the autophagy protein Atg16L1 enhancesendotoxin-induced IL-1beta production. Nature 456, 264–268

6 Sarin, R. et al. (2011) Inflammatory disease protective R381Q IL23receptor polymorphism results in decreased primary CD4+ and CD8+human T-cell functional responses. Proc. Natl. Acad. Sci. U.S.A. 108,9560–9565

7 Boirivant, M. et al. (2008) A transient breach in the epithelial barrierleads to regulatory T-cell generation and resistance to experimentalcolitis. Gastroenterology 135, 1612–1623

8 Geuking, M.B. et al. (2011) Intestinal bacterial colonization inducesmutualistic regulatory T cell responses. Immunity 34, 794–806

9 Hermiston, M.L. and Gordon, J.I. (1995) Inflammatory bowel diseaseand adenomas in mice expressing a dominant negative N-cadherin.Science 270, 1203–1207

10 Mazmanian, S.K. et al. (2008) A microbial symbiosis factor preventsintestinal inflammatory disease. Nature 453, 620–625

11 Round, J.L. and Mazmanian, S.K. (2010) Inducible Foxp3+ regulatoryT-cell development by a commensal bacterium of the intestinalmicrobiota. Proc. Natl. Acad. Sci. U.S.A. 107, 12204–12209

12 Round, J.L. et al. (2011) The Toll-like receptor 2 pathway establishescolonization by a commensal of the human microbiota. Science 332,974–977

13 Atarashi, K. et al. (2011) Induction of colonic regulatory T cells byindigenous Clostridium species. Science 331, 337–341

14 Garrett, W.S. et al. (2007) Communicable ulcerative colitis induced byT-bet deficiency in the innate immune system. Cell 131, 33–45

15 Powell, N. et al. (2012) The transcription factor T-bet regulatesintestinal inflammation mediated by interleukin-7 receptor+ innatelymphoid cells. Immunity 37, 674–684

16 Garrett, W.S. et al. (2010) Enterobacteriaceae act in concert with thegut microbiota to induce spontaneous and maternally transmittedcolitis. Cell Host Microbe 8, 292–300

17 Chen, G.Y. et al. (2011) A functional role for Nlrp6 in intestinalinflammation and tumorigenesis. J. Immunol. 186, 7187–7194

18 Normand, S. et al. (2011) Nod-like receptor pyrin domain-containingprotein 6 (NLRP6) controls epithelial self-renewal and colorectalcarcinogenesis upon injury. Proc. Natl. Acad. Sci. U.S.A. 108, 9601–9606

19 Elinav, E. et al. (2011) NLRP6 inflammasome regulates colonicmicrobial ecology and risk for colitis. Cell 145, 745–757

20 Hirota, S.A. et al. (2011) NLRP3 inflammasome plays a key role in theregulation of intestinal homeostasis. Inflamm. Bowel Dis. 17, 1359–1372

21 Couturier-Mallard, A. et al. (2013) NOD2-mediated dysbiosis pre-disposes mice to transmissible colitis and colorectal cancer. J. Clin.Invest. 123, 700–711

22 Ubeda, C. et al. (2012) Familial transmission rather than defectiveinnate immunity shapes the distinct intestinal microbiota of TLR-deficient mice. J. Exp. Med. 209, 1445–1456

23 Frank, D.N. et al. (2011) Disease phenotype and genotype areassociated with shifts in intestinal-associated microbiota ininflammatory bowel diseases. Proc. Natl. Acad. Sci. U.S.A. 17, 179–184

24 Chiodini, R.J. et al. (2012) Crohn’s disease and the mycobacterioses: aquarter century later. Causation or simple association? Crit. Rev.Microbiol. 36, 52–93

25 Vanderploeg, R. et al. (2010) Influence of intestinal bacteria in humaninflammatory bowel disease. Infect. Dis. Clin. N. Am. 24, 977–993

26 Gophna, U. et al. (2006) Differences between tissue-associatedintestinal microfloras of patients with Crohn’s disease andulcerative colitis. J. Clin. Microbiol. 44, 4136–4141

27 Baumgart, M. et al. (2007) Culture independent analysis of ilealmucosa reveals a selective increase in invasive Escherichia coli of

430

novel phylogeny relative to depletion of Clostridiales in Crohn’sdisease involving the ileum. ISME J. 1, 403–418

28 Willing, B. et al. (2009) Twin studies reveal specific imbalances in themucosa-associated microbiota of patients with ileal Crohn’s disease.Inflamm. Bowel Dis. 15, 653–660

29 Sokol, H. et al. (2009) Low counts of Faecalibacterium prausnitzii incolitis microbiota. Inflamm. Bowel Dis. 15, 1183–1189

30 Swidsinski, A. et al. (2008) Active Crohn’s disease and ulcerative colitiscan be specifically diagnosed and monitored based on the biostructureof the fecal flora. Inflamm. Bowel Dis. 14, 147–161

31 Schwiertz, A. et al. (2010) Microbiota in pediatric inflammatory boweldisease. J. Pediatr. 157, 240–244

32 Kang, S. et al. (2010) Dysbiosis of fecal microbiota in Crohn’s diseasepatients as revealed by a custom phylogenetic microarray. Inflamm.Bowel Dis. 16, 2034–2042

33 Swidsinski, A. et al. (2005) Spatial organization and composition of themucosal flora in patients with inflammatory bowel disease. J. Clin.Microbiol. 43, 3380–3389

34 Swidsinski, A. et al. (2002) Mucosal flora in inflammatory boweldisease. Gastroenterology 122, 44–54

35 Sokol, H. et al. (2008) Faecalibacterium prausnitzii is an anti-inflammatory commensal bacterium identified by gut microbiotaanalysis of Crohn disease patients. Proc. Natl. Acad. Sci. U.S.A.105, 16731–16736

36 Darfeuille-Michaud, A. et al. (2004) High prevalence of adherent-invasive Escherichia coli associated with ileal mucosa in Crohn’sdisease. Gastroenterology 127, 412–421

37 Chambrun, G. et al. (2008) Pathogenic agents in inflammatory boweldiseases Curr. Opin. Gastroenterol. 24, 440–447

38 Barnich, N. et al. (2007) CEACAM6 acts as a receptor for adherent-invasive E. coli, supporting ileal mucosa colonization in Crohn disease.J. Clin. Invest. 117, 1566–1574

39 Carvalho, F.A. et al. (2009) Crohn’s disease adherent-invasiveEscherichia coli colonize and induce strong gut inflammation intransgenic mice expressing human CEACAM. J. Exp. Med. 206,2179–2189

40 Denizot, J. et al. (2012) Adherent-invasive Escherichia coli induceclaudin-2 expression and barrier defect in CEABAC10 mice andCrohn’s disease patients. Inflamm. Bowel Dis. 18, 294–304

41 Chaissaing, B. et al. (2011) Crohn disease-associated adherent-invasive E. coli bacteria target mouse and human Peyer’s patchesvia long polar fimbriae. J. Clin. Invest. 121, 966–975

42 Glasser, A.L. et al. (2001) Adherent invasive Escherichia coli strainsfrom patients with Crohn’s disease survive and replicate withinmacrophages without inducing host cell death. Infect. Immun. 69,5529–5537

43 Bringer, M.A. et al. (2012) Replication of Crohn’s disease-associatedAIEC within macrophages is dependent on TNF-a secretion. Lab.Invest. 92, 411–419

44 Lapaquette, P. et al. (2012) Defects in autophagy favour adherent-invasive Escherichia coli persistence within macrophages leading toincreased pro-inflammatory response. Cell. Microbiol. 14, 791–807

45 Khan, K.J. et al. (2011) Antibiotic therapy in inflammatory boweldisease: a systematic review and meta-analysis. Am. J.Gastroenterol. 106, 661–673

46 Strober, W. (2011) Adherent-invasive E. coli in Crohn disease: bacterial‘‘agent provocateur.’’ J. Clin. Invest. 121, 841–844

47 Carvalho, F.A. et al. (2012) Transient inability to manageProteobacteria promotes chronic gut inflammation in TLR5-deficientmice. Cell Host Microbe 12, 139–152