Enterococcus faecalis Metalloprotease Compromises Epithelial Barrierand Contributes to Intestinal InflammationNATALIE STECK,* MICHA HOFFMANN,* IRINA G. SAVA,* SANDRA C. KIM,‡,§ HANNES HAHNE,�
SUSAN L. TONKONOGY,§,¶ KATRIN MAIR,# DAGMAR KRUEGER,** MIHAELA PRUTEANU,‡‡ FERGUS SHANAHAN,‡‡
ROGER VOGELMANN,# MICHAEL SCHEMANN,** BERNHARD KUSTER,� R. BALFOR SARTOR,‡,§ and DIRK HALLER*
*Chair for Biofunctionality, ZIEL–Research Center for Nutrition and Food Science, CDD Center for Diet and Disease, �Chair for Proteomics and Bioanalytics, and*Chair for Humanbiology, Technische Universität München, Freising-Weihenstephan, Germany; ‡Department of Medicine and §Center for Gastrointestinal Biologynd Disease, University of North Carolina, Chapel Hill, North Carolina; ¶College of Veterinary Medicine, North Carolina State University, Raleigh, North Carolina;
#Klinikum rechts der Isar, II. Medizinische Klinik und Poliklinik, Technische Universität München, München, Germany; and ‡‡Alimentary Pharmabiotic Centre, University
College Cork, National University of Ireland, Cork, Ireland
iieocidcaccam
tlfamtr
wvst
d
BA
SIC
AN
DTR
AN
SLA
TIO
NA
LA
T
See editorial on page 795.
BACKGROUND & AIMS: Matrix metalloproteases (MMPs)mediate pathogenesis of chronic intestinal inflammation.We characterized the role of the gelatinase (GelE), a met-alloprotease from Enterococcus faecalis, in the developmentof colitis in mice. METHODS: Germ-free, interleukin-10 – deficient (IL-10�/�) mice were monoassociated withthe colitogenic E faecalis strain OG1RF and isogenic, GelE-mutant strains. Barrier function was determined by mea-suring E-cadherin expression, transepithelial electrical re-sistance (TER), and translocation of permeability markersin colonic epithelial cells and colon segments from IL-10�/� and TNF�ARE/Wt mice. GelE specificity was shown
ith the MMP inhibitor marimastat. RESULTS: Histo-ogic analysis (score 0 – 4) of E faecalis monoassociatedL-10�/� mice revealed a significant reduction in colonicissue inflammation in the absence of bacteria-derivedelE. We identified cleavage sites for GelE in the sequencef recombinant mouse E-cadherin, indicating that itight be degraded by GelE. Experiments with Ussing
hambers and purified GelE revealed the loss of barrierunction and extracellular E-cadherin in mice susceptibleo intestinal inflammation (IL-10�/� and TNF�ARE/Wt
mice) before inflammation developed. Colonic epithelialcells had reduced TER and increased translocation ofpermeability markers after stimulation with GelE fromOG1RF or strains of E faecalis isolated from patients withCrohn’s disease and ulcerative colitis. CONCLUSIONS:The metalloprotease GelE, produced by commensalstrains of E faecalis, contributes to development ofchronic intestinal inflammation in mice that are sus-ceptible to intestinal inflammation (IL-10�/� and
NF�ARE/Wt mice) by impairing epithelial barrier in-tegrity.
Inflammatory bowel diseases (IBD), comprising the 2main idiopathic pathologies of ulcerative colitis (UC)
nd Crohn’s disease (CD), are spontaneously relapsing
mmune-mediated inflammatory disorders of the gastro-ntestinal tract. An accepted hypothesis for the pathogen-sis of IBD is that the genetically susceptible host devel-ps an aggressive immune response toward the intestinalommensal microbiota. There are 2 general scenarios lead-ng to these pathologic circumstances: either the diseaseefect originates from the mucosal immune system asso-iated with the loss of tolerance against luminal microbialntigens, or the fundamental abnormality is based onompositional and/or functional changes in the gut mi-robiota.1 Several studies in experimental rodent modelss well as clinical trials suggest the crucial role of the guticrobiota in the development of IBD.2,3 With regard to
the intestinal epithelium, the importance of the barrierfunction in limiting access of luminal antigens to under-lying tissues is pivotal for the symbiosis between theintestinal microbiota and the host. In the context of IBD,intestinal barrier dysfunction has been associated withdisease pathogenesis,4,5 although additional triggers seemo be required for disease development.6 Intestinal epithe-ial cells (IECs) form a tight line of defense against harm-ul molecules and microorganisms. Tight junctions (TJs)nd adherence junctions provide barrier function andediate cell-cell contact between IECs. The pathogenicity
raits of a number of pathogens, including enterohemor-hagic Escherichia coli,7 Salmonella typhimurium,8 Shigella flex-
neri,9 or Campylobacter jejuni,10 target the paracellular path-ay to infect the host. It has been shown that specificirulence genes encoding toxins and other proteintructures disrupt epithelial barrier functions throughhe direct disassembly of TJ proteins11,12 or modulation
of intracellular pathways resulting in TJ protein redis-tribution.13 The involvement of virulence factors is pre-
ominantly reported for pathogens, whereas putative
Abbreviations used in this paper: GelE, gelatinase from Enterococcusfaecalis; IEC, intestinal epithelial cell; IFN, interferon; IL, interleukin;LC-MS/MS, liquid chromatography–mass spectrometry; MMP, matrixmetalloprotease; TER, transepithelial electrical resistance; TJ, tightjunction; TNF, tumor necrose factor; Wt, wild-type.
virulence genes from commensal bacteria are largelyignored in the pathogenesis of IBD.
Monoassociation of gnotobiotic interleukin (IL)-10 –deficient (IL-10�/�) mice with Enterococcus faecalis strainsevealed bacterial strain–specific effects on the develop-
ent of experimental colitis,14 possibly due to the expres-sion pattern of virulence genes. E faecalis is one of the most
bundant commensal bacterial species in the human guticrobiome15 but displays a dualistic character; in fact,
this species plays a role in a number of infectious pro-cesses such as endocarditis,16 bacteremia,17 and urinarytract infections.18
During disease progression in IBD, the degradation ofextracellular matrix and mucosal damage mediated byhost-derived matrix metalloproteinases (MMPs) is one ofthe most serious consequences.19 Recent studies associ-ated MMP activity with changes in epithelial barrier func-tion20 and mucin production.21 Within the large family ofzinc-dependent MMPs are 2 gelatinases, MMP-2 andMMP-9, which share their main matrix substrates (type IVcollagen, gelatin) and differ from other MMPs in terms oftheir structure.22 Whereas MMP-9 is absent from healthyissues, its expression and activity are increased in severalnimal models of colitis.23,24 In this study, we aimed at
characterizing the role of commensal bacteria-derivedproteases in the pathogenesis of IBD. We identified the31.5-kilodalton metalloprotease gelatinase E (GelE) fromE faecalis OG1RF as a putative candidate for our study. Itdisplays similar substrate specificity as the mammalianendopeptidase 24.11 and Streptococcus thermophilus thermo-lysin (EC3.4.24.4). GelE expression is regulated by the fsrABC(D) genes, which encode for a 2-component signaltransduction system that is involved in bacterial biofilmformation.25 To investigate the role of GelE in the devel-opment of chronic intestinal inflammation, we monoas-sociated germ-free wild-type (Wt) and IL-10�/� mice with
faecalis OG1RF and 2 isogenic mutant strains, TX5264nd TX5266, both lacking GelE expression. Ussing cham-er experiments with distal colon segments from nonin-amed IL-10�/� and TNF�ARE/Wt mice revealed the effectf GelE on epithelial barrier function before histologichanges have occurred in the tissue. In addition, GelE-roducing E faecalis strains isolated from patients withBD supported the role of bacterial proteases in the con-ext of epithelial barrier disruption.
Materials and MethodsBacteria Strains, Cultivation, and Preparationof Concentrated Conditioned MediaFor our study, we used the characterized E faecalis strain
OG1RF.26,27 The isogenic mutants, TX5264 (�GelE) andTX5266 (�fsrB), were kindly provided by M. Gilmore(Sheepens Eye Research Institute, Boston, MA). The strainsTX543928 (�gel reconstitution) and TX5266.0129 (�fsrB re-constitution) were provided by B. Murray (University of TexasMedical School, Houston, TX). The reconstituted strains wereaccomplished by introducing stable plasmids (pTEX5438 and
pTEX5249). These plasmids are not incorporated into the ge- p
nome of the bacteria and therefore require the presence oferythromycin (10 �g/mL). The noncharacterized E faecalis strainsCD11, CD28.1, UC7, UC18.1, and AH114 were isolated fromfecal samples of patients with IBD and healthy controls accord-ing to their ability to degrade gelatin.30 Cultures were spottedonto Tryptic soy agar supplemented with 0.5 g/L L-cysteine and1.6% Difco gelatin and displayed zonal changes around them.Bacteria were considered gelatinase positive and were identifiedby 16S to 23S ribosomal DNA intergenic spacer region sequenc-ing. Bacteria were cultivated in brain heart infusion medium(BD, Sparks, MD) at 37°C under aerobic conditions. For prep-aration of concentrated conditioned media, bacterial culturesupernatant from an overnight culture was concentrated (factor50) by using an Amicon Filter System (Millipore, Billerica, MA)with an exclusion size of 10 kilodaltons.
Mice and Bacterial MonoassociationThe experiments with IL-10�/� and Wt mice were per-
formed in collaboration with R. B. Sartor at the GnotobioticFacility of the University of North Carolina at Chapel Hill.Germ-free IL-10�/� and Wt mice on the 129S6/SvEv background
ere monoassociated at 15 to 18 weeks of age with E faecalisG1RF and the mutant strains �GelE and �fsrB by oral gavage
nd rectal swab. After 15-week colonization, mice were killednd bacterial colonization was evaluated by 10-fold dilutioneries of cecal contents plated on Luria Bertani Agar. Animal userotocols were approved by the Institutional Animal Care andse Committee of the University of North Carolina at Chapelill.Please see Supplementary Materials and Methods for infor-
ation on morphologic studies, isolation of primary mouseECs, quantitative reverse-transcription polymerase chain reac-ion and Western blot, immunohistochemistry, cell culturexperiments, protein purification, determination of proteasectivity with an azocasein activity assay and zymography,etermination of GelE antigenicity, Ussing chamber experi-ents, investigation of GelE expression in E faecalis strains,
degradation of recombinant E-cadherin, identification of pro-teins and cleavage sites with liquid chromatography–mass spec-trometry (LC-MS/MS) analysis, and statistical analysis.
ResultsE faecalis–Derived GelE Contributes to theDevelopment of Experimental Colitis in IL-10�/� MiceDeregulated expression of host-derived MMP
has been implicated in several diseases, including ar-thritis, atherosclerosis, colon cancer, and IBD.19,31,32 In
ur study, we investigated the effect of E faecalisG1RF-derived gelatinase, GelE, on the development of
hronic intestinal inflammation. Therefore, we mono-ssociated germ-free Wt and IL-10�/� mice for 15 weeksith E faecalis OG1RF and 2 isogenic mutants, �GelEnd �fsrB, both lacking GelE expression. Histopatho-ogic analysis (score 0 – 4) revealed a significant reduc-ion of proximal (Figure 1A) and distal (Figure 1C)olonic tissue inflammation (OG1RF: proximal, 2.0 �.3; distal, 3.2 � 0.2) in the absence of bacterial GelETX5264: proximal, 1.0 � 0.2; distal, 2.1 � 1.0; TX5266:
sections from proximal (Figure 1B) and distal (Figure1D) colon are shown from paraffin-embedded tissue.
t mice did not develop inflammation in any part ofhe colon, emphasizing the nonpathogenic nature ofhe E faecalis strains. Equal colonization was confirmedy 10-fold dilution series of cecal contents on LB agar,
Figure 1. Development of inflammation in Wt and IL-10�/� mice monoaof the (A) proximal and (C) distal colon are presented as mean values � Smonoassociated mice were assessed by one-way analysis of variance ashown from (B) proximal and (D) distal colon.
howing that the observed effects did not result from T
growth differences of the E faecalis strains (data notshown). Due to possible erythromycin-mediated immu-nosuppressive effects33 and technical implementationshat would compromise the interpretation of the ex-eriment, we have not determined the inflammatoryotential of the reconstituted E faecalis mutants
ciated with E faecalis OG1RF, �GelE, and �fsrB. Histologic scores (0–4)m 5 animals per group. Significant differences in comparison to OG1RF
marked as *P � .05. Representative tissue sections stained for H&E are
ssoD frond
X5439 and TX5266.01 in IL-10�/� mice.
sfGttn
coc
BA
SICA
ND
TRA
NSLA
TION
AL
AT
962 STECK ET AL GASTROENTEROLOGY Vol. 141, No. 3
E faecalis–Derived GelE Does Not TriggerAntigen-Specific CD4� T-Cell ResponseTo investigate the colitogenic potential of E faeca-
lis– derived GelE, we purified the protease from the culturesupernatant of E faecalis OG1RF and TX5439, which is theGelE reconstituted �GelE mutant strain, by using anionexchange chromatography. Furthermore, we purified theE faecalis lipoprotein EF1362 and used it as unrelatedprotein control in subsequent experiments. The puritywas confirmed by sodium dodecyl sulfate/polyacrylamidegel electrophoresis followed by Coomassie straining (Fig-ure 2A). The proteolytic activity of GelE was measuredusing an azocasein activity assay (Figure 2B). Bacterial
Figure 2. Purification and proteolytic assessment of GelE from E faecaculture supernatant of OG1RF and TX5439 (�GelE reconstituted straisodium dodecyl sulfate/polyacrylamide gel electrophoresis followed bconcentrated conditioned media of E faecalis strains were assessed by apreincubation with marimastat. To test antigenicity of GelE, splenic antigpurified GelE, or bacterial lysates from E faecalis OG1RF and �GelE (10T cells from E faecalis OG1RF monoassociated mice. Supernatants weassay. Values represent mean � SD of IFN-� concentration in triplicate
GelE activity can be inhibited by marimastat, known as a
broad-spectrum MMP inhibitor (Figure 2C). To determinewhether GelE evokes an antigen-specific T-cell response,we cultured MLN-derived CD4� T cells from E faecalisOG1RF monoassociated IL-10�/� mice with antigen-pre-enting cells that had been pulsed with bacterial lysatesrom E faecalis OG1RF and �GelE, as well as with purifiedelE and KLH as negative control. Interestingly, the an-
igenic capacity of GelE seems not to be associated withhe development of tissue pathology in OG1RF colo-ized IL-10�/� mice, because purified GelE did not
trigger interferon (IFN)-� responses in the coculturesystem (Figure 2D). In addition, lysates from Wt Efaecalis (Figure 2D, dark gray bar) and �GelE (Figure 2D,
OG1RF and its antigenic potential. (A) GelE was purified from bacterialsing anion exchange chromatography. The purity was determined byoomassie staining. (B) The protease activity of purified GelE and thezocasein activity assay. (C) GelE activity could be inhibited by 15-minutepresenting cells from Wt mice were pulsed with KLH (negative control),/mL) for 18 hours. Antigen-presenting cells were cocultured with CD4�
collected after 72 hours, and IFN-� was measured by immunosorbentulture supernatants from 4 animals (D).
lisn) uy Cn aen-�gre
light gray bar) triggered comparable IFN-� responses in
ad
sofoc
n3cad
lsv
BA
SIC
AN
DTR
AN
SLA
TIO
NA
LA
T
September 2011 BACTERIAL PROTEASES IN IBD 963
activated T cells, clearly supporting the hypothesis thatntigen-independent mechanisms are important for theisease-promoting effects of GelE.
GelE Triggers the Loss of ExtracellularE-Cadherin Protein Domain and TissueResistance in Disease-Susceptible MiceTo identify the disease-promoting mechanisms of
E faecalis– derived GelE, we next explored the role of GelEon the regulation of junctional proteins composing theepithelial barrier. Western blot analysis of isolated IECsfrom monoassociated Wt and IL-10�/� mice showedlightly modulated expression levels of the TJ proteinsccludin and JAM-A in the presence of GelE-producing E
aecalis. However, the presence of the extracellular domainf adherence junction protein E-cadherin was signifi-antly reduced in IL-10�/� mice monoassociated with E
faecalis OG1RF (Figure 3A). Most importantly, the loss ofextracellular domain of E-cadherin in E faecalis OG1RFmonoassociated IL-10�/� mice was confirmed by immu-
ofluorescence staining of distal colon sections (FigureB). Expression levels of the intracellular domain of E-adherin remained unaffected, and we could not detectny differences at the gene transcript level between theifferent monoassociated IL-10�/� mice, neither for E-
cadherin messenger RNA nor for any tested TJ protein(Figure 3C). Interestingly, there was a significant induc-tion of MMP-9 messenger RNA expression in the presenceof GelE. To address the question of a possible interactionbetween bacterial GelE and host-derived MMP-9, we in-cubated recombinant murine pro-MMP-9 with purifiedGelE. According to the results shown in Figure 3D, bac-terial GelE was not able to activate MMP-9 (marked asblue box in Figure 3D) comparable to the organic com-pound p-aminophenylmercuric acetate, suggesting thatthe MMP-9 induction in OG1RF monoassociated IL-10�/� mice was not a direct consequence of the presenceof GelE. Due to the fact that E-cadherin protein harborsvarious cleavage sites for GelE-mediated degradation, wenext explored the hypothesis that the loss of extracellularE-cadherin in the disease-conditioned tissue was mediatedthrough proteolytic degradation. Recombinant murineE-cadherin was specifically degraded in the presence ofproteolytically active GelE (Figure 3E). Furthermore, LC-MS/MS analysis identified various additional cleavagesites in the extracellular domain of recombinant E-cad-herin after GelE incubation, supporting the possibility ofa direct GelE-mediated degradation in vivo (Figure 3F andSupplementary Tables 1 and 2). These results showed thatthe loss of extracellular E-cadherin protein was associatedwith different levels of tissue inflammation and the pres-ence of GelE-producing E faecalis in the disease-susceptibleIL-10 – deficient host.
To further investigate the hypothesis that GelE acts onmucosal barrier integrity of disease-conditioned colonictissue, we performed Ussing chamber experiments withdisease-susceptible IL-10�/� and heterozygous TNF�ARE/Wt
mice, a mouse model for tumor necrosis factor (TNF)-
induced ileitis. We used distal colon segments from8-week-old mice that have not developed disease pathol-ogy at this age (Figure 4A and C). As shown in Figure 4Band D, the apical stimulation of the tissue with purifiedGelE from E faecalis OG1RF and TX5439 led to a signif-icant decrease of the transepithelial electrical resistance(TER) in both mouse models. Epithelial barrier disruptionwas not significant in Wt mice, suggesting that GelEfacilitates the loss of barrier function in disease-condi-tioned and genetically susceptible hosts. Western blotanalysis of GelE-stimulated tissue from Ussing chamberexperiments confirmed the loss of extracellular E-cadherinprotein domain in colonic tissue from disease-free IL-10�/� and TNF�ARE/Wt mice (Figure 4E).
E faecalis–Derived GelE Leads to theImpairment of Epithelial Barrier Function inPtk6 Transwell Cultures: Synergistic Effects ofGelE and Proinflammatory CytokinesWe next studied the effect of E faecalis GelE on
epithelial barrier function of fully differentiated Ptk6Transwell cultures by measuring TER and translocationof fluorescent-labeled markers. Ptk6 cell monolayers wereapically stimulated with concentrated conditioned me-dium of E faecalis OG1RF, the GelE-deficient strains�GelE and �fsrB, and the GelE reconstituted mutantsTX5439 and TX5266.01. As shown in Figure 5A, thepresence of GelE significantly decreased TER values incomparison to untreated control after 24 hours of incu-bation. The concentrated conditioned media of the GelE-lacking strains did not lead to a reduction of TER values,suggesting that GelE was the only component responsiblefor barrier disruption. Parallel to the GelE-mediated de-crease in TER, we detected an enhanced translocation ofsodium fluorescein from the apical to the basolateral sideof the Transwell system (Figure 5B). Proof-of-principleexperiments with purified GelE and inhibited proteolyticGelE activity, either through heat treatment or by usingmarimastat, a broad-spectrum MMP inhibitor, confirmedthe GelE-mediated loss of epithelial barrier function (Fig-ure 5C and D). In addition, the effect of GelE on barrierfunction of murine IECs could be confirmed by TERexperiments with the human colon cell line T84 (Supple-mentary Figure 1A). Furthermore, the use of differentpermeability markers revealed a size-dependent capabilityfor sodium fluorescein (376 daltons), fluorescein isothio-cyanate–Dextran 4000 (4000 daltons), and fluorescein iso-thiocyanate–Dextran 10 (10,000 daltons) to translocateacross the polarized Ptk6 Transwell culture (Supplemen-tary Figure 1B).
It has been shown that TNF and IFN-� disrupt epithe-ial barrier function34 through the modulation of cyto-keletal filaments35 and the activation of mitogen-acti-ated protein kinases.36 To mimic the inflammatory
environment, we prestimulated Ptk6 cells basolaterallywith TNF and IFN-� for 12 hours, followed by 24-hourstimulation with GelE from the apical side. GelE as well as
the combination of the proinflammatory cytokines signif-
e
dro
BA
SICA
ND
TRA
NSLA
TION
AL
AT
964 STECK ET AL GASTROENTEROLOGY Vol. 141, No. 3
icantly decreased TER values (Supplementary Figure 1C).Interestingly, the treatment of Ptk6 monolayers with TNFand IFN-�, in combination with GelE, potentiated thesodium fluorescein flux (Figure 5E), suggesting a syner-
Figure 3. (A) Impact of E faecalis GelE on junctional proteins in the intejunctional proteins was determined by densitometric analysis of Westerngroup, and representative blots are shown. (B) Extracellular (green, Alexwere stained in distal colonic tissue sections. Immunofluorescence was4’,6-diamidino-2-phenylindole (blue). (C) Transcript levels of junctionaxpressed as relative expression level to the germ-free situation. Data re
could be activated by p-aminophenylmercuric acetate (blue box), but noetermining the proteolytic activity (zymography). (E) In contrast to p-aecombinant murine E-cadherin after 5-second incubation. (F) Predictedf E-cadherin. Cleavage sites were determined by LC-MS/MS analysis.
gistic effect of host-derived cytokines and the bacterial
protease GelE in triggering the functional loss of barrierintegrity. Immunofluorescence (Figure 5F) and Westernblot analysis (Supplementary Figure 1D) of E-cadherinprotein expression revealed no changes after GelE treat-
al epithelium of monoassociated Wt and IL-10�/� mice. Expression ofts from isolated IECs. Data represent mean values � SD from 5 mice perluor 488) and intracellular (red, Alexa Fluor 546) domains of E-cadherinsessed by confocal laser microscopy. Nuclei were counterstained withoteins and host-derived MMP-2 and MMP-9 in IL-10�/� mice weresent mean values SD from 5 mice per group, *P � .05 (D) Pro-MMP-9rough E faecalis GelE demonstrated on protein level (silver stain) and byophenylmercuric acetate–activated MMP-9, E faecalis GelE degradeddetected cleavage sites for E faecalis GelE in the amino acid sequence
stinbloa Fas
l prpret thminand
ment using antibodies that specifically target the intracel-
sT
BA
SIC
AN
DTR
AN
SLA
TIO
NA
LA
T
September 2011 BACTERIAL PROTEASES IN IBD 965
Figure 4. E faecalis GelE disrupts colonic barrier integrity in noninflamed IL-10�/� and TNF�ARE/Wt mice. Distal colon segments from 8-week-oldIL-10�/� and TNF�ARE/Wt mice and their Wt counterparts were apically stimulated with purified GelE (10 �g/mL) for 5 hours in Ussing chamberystems. Histologic scoring and representative H&E staining show the absence of tissue pathology in the distal colon from (A) IL-10�/� and (C)NF�ARE/Wt mice at this age. Results from Ussing chamber experiments are expressed as percent change of TER values compared with the initial
value of the tissue (B and D). Data represent median with the 25th and 75th percentiles from 5 animals per group; statistical significance wascalculated by one-way analysis of variance (*P � .05). E-cadherin protein expression was assessed by Western blot and densitometric analysis of 3animals per group and treatment (E).
BA
SICA
ND
TRA
NSLA
TION
AL
AT
966 STECK ET AL GASTROENTEROLOGY Vol. 141, No. 3
pm
pCtOtishio(pngprutse
htmlit
h
tbmp
s
BA
SIC
AN
DTR
AN
SLA
TIO
NA
LA
T
September 2011 BACTERIAL PROTEASES IN IBD 967
lular or extracellular domain of E-cadherin. However, weobserved reduced levels of the TJ protein occludin.
To validate the role of GelE from the colitogenic Efaecalis strain OG1RF relative to E faecalis strains from
atients with IBD in the context of epithelial barrierodulation, we next isolated E faecalis strains from fecal
samples of patients with IBD (CD11, CD18.1 from pa-tients with CD; UC7, UC28.1 from patients with UC) andone strain (AH114) from a healthy control according totheir capability to produce GelE (Figure 6A). GelE expres-sion in the isolated E faecalis strains and GelE activity inthe concentrated conditioned media from these strainswere approximately half compared with GelE fromOG1RF (Figure 6B and C). Therefore, we adapted thequantity used for experiments with Ptk6 cells to GelEactivity of the concentrated conditioned media of OG1RF.Concentrated conditioned media from all strains reducedTER values comparable to OG1RF (Figure 6D). To furtherdescribe the effect of GelE produced by those strains, wepurified GelE from CD11 and UC7 according to the sameprotocol used for GelE from OG1RF. Proteolytic activitiesof GelE from human isolates were lower compared withGelE from OG1RF. Silver staining revealed additionalproteins, which were more abundant in the GelE samplesfrom IBD isolates (Figure 6E). Despite the fact that GelEwas the most abundant protein, LC-MS/MS analysis iden-tified a set of other proteins in those samples listed inSupplementary Table 3. Although equal proteolytic activ-ities were used, GelE from CD11 and UC7 did not reduceTER values to the same extent as GelE from OG1RF(Figure 6F). These differences might originate from theimpurities containing other E faecalis– derived proteins.
In conclusion, the bacteria-derived metalloproteaseGelE, in combination with other disease-conditioning fac-tors such as proinflammatory cytokines, may trigger theloss of barrier function mediated by E-cadherin degrada-tion and the subsequent induction of a colitogenic im-mune response as a consequence of enhanced transloca-tion of immunostimulatory bacterial components acrossthe epithelium (Figure 7).
DiscussionThis study provides mechanistic evidence for the
fact that a protease produced by commensal bacteriacould contribute to the development of experimentalcolitis through the impairment of epithelial barrierfunction. The presence of metalloprotease GelE was
4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™Figure 5. GelE from E faecalis OG1RF disrupts barrier function of Ptkmedia of E faecalis OG1RF, �GelE, TX5439 (�GelE reconstituted strain)ranslocation of sodium fluorescein were measured after 24 hours. Purifiearrier function. (C and D) Specificity of GelE-mediated disruption of barimastat and an E faecalis unrelated protein control (UPC). (E) To shermeability, cells were preincubated basolaterally with TNF (20 ng/mL)
24 hours. (F) Values represent triplicates from 1 of 3 independent experimof variance and marked as follows: *P � .05, **P � .001. After 24 hours, a
tained in green (Alexa Fluor 488) or red (Alexa Fluor 546).
shown to potentiate the development of experimentalcolitis in E faecalis monoassociated IL-10�/� mice, inde-
endent of antigen-specific activation of colitogenicD4� T cells. Previous studies demonstrated the coli-
ogenic character of the well-described E faecalis strainG1RF in IL-10�/� mice,37,38 but the structural charac-
eristics of this commensal strain responsible for diseasenitiation and progression remained unclear. In thistudy, isogenic mutant E faecalis strains lacking GelE ex-ibited significantly diminished potential to trigger exper-
mental colitis in monoassociated IL-10�/� mice. The col-nization with the reconstituted �GelE and �fsrB strainsTX5439 and TX5266.01) was not performed due to un-redictable erythromycin-mediated side effects and tech-ical limitations. Erythromycin is absorbed in the upperastrointestinal tract, and the dosage to induce the ex-ression of an erythromycin-sensitive promoter in bacte-ia colonizing the large intestine of mice is completelynknown. Moreover, erythromycin-related macrolide an-ibiotics have been shown to exert direct immunosuppres-ive effects on the host, independently from intestinalcological changes of the gut microbiota.33
Interestingly, GelE triggered degradation of the extra-cellular domain of adherence junction protein E-cadherinand the loss of epithelial barrier integrity in disease-sus-ceptible but not Wt mice, suggesting that the luminalmilieu in combination with host-derived factors such asproinflammatory cytokines or an intrinsic barrier defectare required for disease initiation. We could confirm thesefindings in colonic tissues from noninflamed IL-10�/�
and TNF�ARE/Wt mice. Of note, E-cadherin, as one regula-tor of epithelial barrier function and cell-cell adhesion, istargeted by a number of endogenous metalloproteases,including stromelysin-1, matrilysin,39 ADAM-10,40 andmeprin-�.41 GelE degraded murine recombinant E-cad-
erin, suggesting that the loss of E-cadherin in colonicissues may be a direct consequence of bacterial GelE-
ediated proteolytic degradation. In addition, the metal-oprotease toxin BFT from Bacteroides fragilis was shown tonduce the shedding of the E-cadherin ectodomainhrough an unknown IEC receptor-mediated induction of
�-secretase,42 suggesting the possibility that indirect pro-tease-mediated effects might be implicated in the loss ofE-cadherin. Together with the previous description of anintrinsic intestinal barrier dysfunction under IL-10 defi-ciency,43 we hypothesize that in the disease-susceptible
ost GelE might gain access to the epithelium, where it
™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™ells. Ptk6 cells were apically stimulated with concentrated conditionedrB, and TX5266.01 (�fsrB reconstituted strain) (50 �L). (A) TER and (B)elE (10 �g/mL) from E faecalis OG1RF and TX5439 significantly reducedr function could be shown by using heat-inactivated GelE or inhibitorsynergistic effects of GelE and proinflammatory cytokines on epithelialIFN-� (50 ng/mL) for 12 hours, followed by apical GelE stimulation for
ts. Statistically significant differences are calculated by one-way analysisl stimulation with active and inhibited GelE, and junctional proteins were
™™6 c, �fsd Garrieowanden
pica
BA
SICA
ND
TRA
NSLA
TION
AL
AT
968 STECK ET AL GASTROENTEROLOGY Vol. 141, No. 3
vpceIsduoowvd
ish
gb
BA
SIC
AN
DTR
AN
SLA
TIO
NA
LA
T
September 2011 BACTERIAL PROTEASES IN IBD 969
degrades E-cadherin or other unknown target proteins.Various studies have shown that the proinflammatorycytokines TNF and IFN-� alter barrier function of culti-ated epithelial cells,34,36 but fewer data are available forrimary tissue. Amasheh et al described changes of mu-osal morphology and alterations in TJ profile in rectalxplants from a rat model after exposure to TNF andFN-�,44 supporting the hypothesis for a combined andynergistic effect of luminal bacterial proteases and host-erived cytokines in the context of intestinal barrier mod-lation. We further showed a GelE-mediated disruptionf barrier integrity in vitro due to loss of the TJ proteinccludin. E-cadherin was not affected in cell culture; thus,e suggest additional or different events happening in
ivo that lead to the degradation of the extracellularomain of E-cadherin.
E faecalis strains are part of the intestinal microbiome15
but are also regarded as nosocomial pathogens and play arole in several infections (eg, endocarditis,16 bacteremia,17
and urinary tract infections18). A recent study showedncreased numbers of enterococci isolated from biopsypecimens from children with UC and CD compared withealthy controls.45 However, the microscopic analysis of
colonic biofilms showed that enterococci growing in mi-crocolonies occurred on mucosal surfaces of patients withUC, but not in healthy people, suggesting that spatial
4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™Figure 6. GelE-mediated effect of E faecalis isolates from patients witelatinolytic activity on gelatin-containing TSA agar were isolated fromacteria; (C) proteolytic activity in the concentrated conditioned media fr
Equal proteolytic activity was used for apical stimulation of Ptk6 cells, andexhibited less proteolytic activity and a higher abundance of other proteused to apically stimulate cells for 24 hours. TER values represent mea
Figure 7. Proposed mechanism of GelE-mediated loss of epithelial barmechanism regarding how E faecalis GelE impairs mucosal barrier funpreconditioned tissue of a susceptible host.
significant differences are calculated with one-way analysis of variance and
distribution and mucosal proximity of the microorgan-isms might be linked to the disease process rather thantheir absolute numbers.46 Furthermore, these investiga-tors revealed higher titers of E faecalis–specific antibodiesin the serum of patients with UC compared with healthycontrols, showing an immune-mediated recognition of Efaecalis under conditions of chronic inflammation. It hasbeen shown that GelE controls bacterial survival anddissemination through the regulation of coccal chainlength, suggesting that GelE contributes to the spreadingof bacteria in high-density environments such as the gutmucosa. The barrier disruption in cultivated IECs medi-ated by the concentrated conditioned media of our Efaecalis isolates from patients with IBD and a healthycontrol was comparable to the effect of OG1RF. Silverstain of purified GelE showed the presence of severaladditional proteins. LC-MS/MS analysis revealed that Efaecalis GelE was the most abundant protein in all purifiedGelE samples, but impurities were more potent in thesamples of clinical isolates. Membrane-associated proteinswere more frequently within the list of further identifiedproteins. We only can speculate that the additional pro-teins present in GelE samples from CD11 and UC7 mightdisturb the protease activity and therefore exhibit a lowerpotential to break the epithelial barrier. This might ex-plain that even equal proteolytic activities do not possess
™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™D on barrier function of Ptk6 cells. (A) E faecalis strains possessing al samples. (B) GelE production was assessed on transcript level in theE faecalis strains was determined using an azocasein activity assay. (D)R was determined after 24 hours. (E) Purified GelE from CD11 and UC7(F) Equal proteolytic activity of GelE from OG1RF, CD11, and UC7 was� SD from triplicates from 1 of 3 independent experiments. Statistical
function in the susceptible host. Schematic illustration of the proposedon. The commensal-derived protease target junctional proteins in the
™™h IBfecaom
TEins.ns
riercti
marked as follows: �P � .05, ��P � .001.
w
1
1
1
1
2
2
2
2
3
3
BA
SICA
ND
TRA
NSLA
TION
AL
AT
970 STECK ET AL GASTROENTEROLOGY Vol. 141, No. 3
the same capability to break barrier integrity. It should beemphasized that GelE production was not unique todisease-associated E faecalis isolates. Based on our results,
e hypothesize that commensal E faecalis strains from thehuman gut harbor the potential to produce GelE thatmediates disease only in the susceptible host.
In summary, our results show that a bacterial pro-tease from a commensal inhabitant of the gastrointes-tinal tract can contribute to the development of intes-tinal inflammation through mechanisms that impairepithelial barrier functions. The GelE-producing Efaecalis strain OG1RF induced a significantly higher co-lonic inflammation in IL-10�/� mice compared with GelE-deficient strains. Although GelE-expressing E faecalis strainswere present and functional in patients with IBD andcontrol patients, GelE-mediated loss of barrier integritywas only detectable in the histologically normal but pre-activated colonic tissues from IL-10�/� and TNF�ARE/Wt
mice, showing that the bacterial metalloprotease GelE canbe a disease-conditioning trigger in a susceptible host.
Supplementary Material
Note: To access the supplementary materialaccompanying this article visit the online version ofGastroenterology at www.gastrojournal.org and at doi:
0.1053/j.gastro.2011.05.035.
References
1. Strober W, Fuss I, Mannon P. The fundamental basis of inflamma-tory bowel disease. J Clin Invest 2007;117:514–521.
2. Sartor RB. Microbial influences in inflammatory bowel diseases.Gastroenterology 2008;134:577–594.
3. Nell S, Suerbaum S, Josenhans C. The impact of the microbiota onthe pathogenesis of IBD: lessons from mouse infection models.Nat Rev Microbiol;8:564–577.
4. Laukoetter MG, Nava P, Nusrat A. Role of the intestinal barrier ininflammatory bowel disease. World J Gastroenterol 2008;14:401–407.
5. McGuckin MA, Eri R, Simms LA, et al. Intestinal barrier dysfunctionin inflammatory bowel diseases. Inflamm Bowel Dis 2009;15:100–113.
6. Su L, Shen L, Clayburgh DR, et al. Targeted epithelial tight junctiondysfunction causes immune activation and contributes to devel-opment of experimental colitis. Gastroenterology 2009;136:551–563.
7. Li Q, Zhang Q, Wang C, et al. Invasion of enteropathogenic Esch-erichia coli into host cells through epithelial tight junctions. FEBSJ 2008;275:6022–6032.
9. Sakaguchi T, Kohler H, Gu X, et al. Shigella flexneri regulates tightjunction-associated proteins in human intestinal epithelial cells.Cell Microbiol 2002;4:367–381.
0. Chen ML, Ge Z, Fox JG, et al. Disruption of tight junctions andinduction of proinflammatory cytokine responses in colonic epi-thelial cells by Campylobacter jejuni. Infect Immun 2006;74:6581–6589.
1. Viswanathan VK, Koutsouris A, Lukic S, et al. Comparative anal-ysis of EspF from enteropathogenic and enterohemorrhagic Esch-erichia coli in alteration of epithelial barrier function. Infect Immun
2004;72:3218–3227.
2. Amieva MR, Vogelmann R, Covacci A, et al. Disruption of theepithelial apical-junctional complex by Helicobacter pylori CagA.Science 2003;300:1430–1434.
13. Babbin BA, Sasaki M, Gerner-Schmidt KW, et al. The bacterialvirulence factor lymphostatin compromises intestinal epithelialbarrier function by modulating rho GTPases. Am J Pathol 2009;174:1347–1357.
14. Hoffmann M, Kim SC, Sartor RB, et al. Enterococcus faecalisstrains differentially regulate Alix/AIP1 protein expression andERK 1/2 activation in intestinal epithelial cells in the context ofchronic experimental colitis. J Proteome Res 2009;8:1183–1192.
15. Qin J, Li R, Raes J, et al. A human gut microbial gene catalogueestablished by metagenomic sequencing. Nature. 2010;464:59–65.
16. Nallapareddy SR, Singh KV, Sillanpaa J, et al. Endocarditis andbiofilm-associated pili of Enterococcus faecalis. J Clin Invest2006;116:2799–2807.
17. Graninger W, Ragette R. Nosocomial bacteremia due to Entero-coccus faecalis without endocarditis. Clin Infect Dis 1992;15:49–57.
18. Fisher K, Phillips C. The ecology, epidemiology and virulence ofEnterococcus. Microbiology 2009;155:1749–1757.
19. Baugh MD, Perry MJ, Hollander AP, et al. Matrix metalloproteinaselevels are elevated in inflammatory bowel disease. Gastroenterol-ogy 1999;117:814–822.
20. Takehara M, Nishimura T, Mima S, et al. Effect of claudin expres-sion on paracellular permeability, migration and invasion of co-lonic cancer cells. Biol Pharm Bull 2009;32:825–831.
21. Garg P, Ravi A, Patel NR, et al. Matrix metalloproteinase-9 regu-lates MUC-2 expression through its effect on goblet cell differen-tiation. Gastroenterology 2007;132:1877–1889.
22. Overall CM. Molecular determinants of metalloproteinase sub-strate specificity: matrix metalloproteinase substrate binding do-mains, modules, and exosites. Mol Biotechnol 2002;22:51–86.
23. Salmela MT, MacDonald TT, Black D, et al. Upregulation of matrixmetalloproteinases in a model of T cell mediated tissue injury inthe gut: analysis by gene array and in situ hybridisation. Gut2002;51:540–547.
24. Medina C, Videla S, Radomski A, et al. Increased activity andexpression of matrix metalloproteinase-9 in a rat model of distalcolitis. Am J Physiol Gastrointest Liver Physiol 2003;284:G116–G122.
25. Qin X, Singh KV, Weinstock GM, et al. Characterization of fsr, aregulator controlling expression of gelatinase and serine proteasein Enterococcus faecalis OG1RF. J Bacteriol 2001;183:3372–3382.
6. Murray BE, Singh KV, Ross RP, et al. Generation of restriction mapof Enterococcus faecalis OG1 and investigation of growth require-ments and regions encoding biosynthetic function. J Bacteriol1993;175:5216–5223.
7. Bourgogne A, Garsin DA, Qin X, et al. Large scale variation inEnterococcus faecalis illustrated by the genome analysis of strainOG1RF. Genome Biol 2008;9:R110.
8. Zeng J, Teng F, Murray BE. Gelatinase is important for transloca-tion of Enterococcus faecalis across polarized human enterocyte-like T84 cells. Infect Immun 2005;73:1606–1612.
9. Sifri CD, Mylonakis E, Singh KV, et al. Virulence effect of Entero-coccus faecalis protease genes and the quorum-sensing locus fsrin Caenorhabditis elegans and mice. Infect Immun 2002;70:5647–5650.
0. Pruteanu M, Hyland NP, Clarke DJ, et al. Degradation of theextracellular matrix components by bacterial-derived metallopro-teases: implications for inflammatory bowel diseases. InflammBowel Dis 2011;17:1189–1200.
1. Itoh T, Matsuda H, Tanioka M, et al. The role of matrix metallo-proteinase-2 and matrix metalloproteinase-9 in antibody-induced
32. Dubois B, Opdenakker G, Carton H. Gelatinase B in multiplesclerosis and experimental autoimmune encephalomyelitis. ActaNeurol Belg 1999;99:53–56.
3. Tamagawa H, Hiroi T, Mizushima T, et al. Therapeutic effects ofroxithromycin in interleukin-10-deficient colitis. Inflamm Bowel Dis2007;13:547–556.
4. Fish SM, Proujansky R, Reenstra WW. Synergistic effects of inter-feron gamma and tumour necrosis factor alpha on T84 cell func-tion. Gut 1999;45:191–198.
5. Turner JR. Molecular basis of epithelial barrier regulation: frombasic mechanisms to clinical application. Am J Pathol 2006;169:1901–1909.
6. Patrick DM, Leone AK, Shellenberger JJ, et al. Proinflammatorycytokines tumor necrosis factor-alpha and interferon-gamma mod-ulate epithelial barrier function in Madin-Darby canine kidney cellsthrough mitogen activated protein kinase signaling. BMC Physiol2006;6:2.
7. Balish E, Warner T. Enterococcus faecalis induces inflammatorybowel disease in interleukin-10 knockout mice. Am J Pathol 2002;160:2253–2257.
8. Kim SC, Tonkonogy SL, Albright CA, et al. Variable phenotypes ofenterocolitis in interleukin 10-deficient mice monoassociated withtwo different commensal bacteria. Gastroenterology 2005;128:891–906.
9. Noe V, Fingleton B, Jacobs K, et al. Release of an invasionpromoter E-cadherin fragment by matrilysin and stromelysin-1.J Cell Sci 2001;114:111–118.
0. Maretzky T, Reiss K, Ludwig A, et al. ADAM10 mediates E-cad-herin shedding and regulates epithelial cell-cell adhesion, migra-tion, and beta-catenin translocation. Proc Natl Acad Sci U S A2005;102:9182–9187.
1. Huguenin M, Muller EJ, Trachsel-Rosmann S, et al. The metallo-protease meprinbeta processes E-cadherin and weakens intercel-lular adhesion. PLoS One 2008;3:e2153.
2. Wu S, Rhee KJ, Zhang M, et al. Bacteroides fragilis toxin stimu-lates intestinal epithelial cell shedding and gamma-secretase-dependent E-cadherin cleavage. J Cell Sci 2007;120:1944–1952.
disruption allows nondisease-causing bacteria to initiate and sus-
tain IBD in the IL-10 gene-deficient mouse. Inflamm Bowel Dis2007;13:947–954.
44. Amasheh M, Grotjohann I, Amasheh S, et al. Regulation of muco-sal structure and barrier function in rat colon exposed to tumornecrosis factor alpha and interferon gamma in vitro: a novel modelfor studying the pathomechanisms of inflammatory bowel diseasecytokines. Scand J Gastroenterol 2009;44:1226–1235.
45. Fyderek K, Strus M, Kowalska-Duplaga K, et al. Mucosal bacterialmicroflora and mucus layer thickness in adolescents with inflam-matory bowel disease. World J Gastroenterol 2009;15:5287–5294.
46. Macfarlane S, Furrie E, Kennedy A, et al. Mucosal bacteria inulcerative colitis. Br J Nutr 2005;93(Suppl 1):S67–S72.
Received August 10, 2010. Accepted May 12, 2011.
Reprint requestsAddress requests for reprints to: Dirk Haller, PhD, Chair for
Biofunctionality, Technische Universität München, Gregor-Mendel-Str.2, 85350 Freising-Weihenstephan, Germany e-mail:[email protected].
AcknowledgmentsThe authors thank Prof Barbara Murray from the University of
Texas Medical School for providing the reconstituted E faecalismutant strains TX5439 and TX5266.01 as well as Fany Reffuveilleand Alain Rince from Laboratoire de Microbiologie del’Environnement, University of Caen, for providing the E coli strainproducing the recombinant E faecalis lipoprotein EF1362.
Conflicts of interestThe authors disclose no conflicts.
FundingSupported by German Research Foundation (Deutsche
Forschungsgemeinschaft) grant GRK 1482, the European Union grantwith agreement no. 202020 (IPODD) the NIH grants R01 DK053347and P40 RR018603 as well as the Crohn’s and Colitis Foundation
