Probiotics promote gut health through stimulation ofepithelial innate immunityCristiano Pagninia,1, Rubina Saeeda, Giorgos Bamiasa,b, Kristen O. Arseneaub, Theresa T. Pizarrob,c,and Fabio Cominellib,2

aDigestive Health Research Center, University of Virginia Health System, Charlottesville, VA 22908; bDigestive Health Center, Case Western Reserve University,Cleveland, OH 44106; and cDepartment of Pathology, Case Western Reserve University School of Medicine, Cleveland, OH 44106

Edited by Charles A. Dinarello, University of Colorado Denver, Aurora, CO, and approved November 18, 2009 (received for review September 11, 2009)

Probiotic formulations are widely available and have a variety ofproposed beneficial effects, including promotion of gut health. Themechanisms of action of probiotic bacteria in the intestine are stillunclear but are generally attributed to an antiinflammatory effect.Here,wedemonstrate that themultipleprobiotic formulationVSL#3prevents the onset of intestinal inflammation by local stimulation ofepithelial innate immune responses (i.e., increased production ofepithelial-derived TNF-α and restoration of epithelial barrier func-tion in vivo). We also demonstrate that probiotic bacteria stimulateepithelial production of TNF-α and activate NF-κB in vitro. Our re-sults support the hypothesis that probiotics promote gut healththrough stimulation, rather than suppression, of the innate immunesystem. Furthermore, our findings provide the perspective thatdefects in innate immunity may play a critical role in the pathogen-esis and progression of intestinal disorders, such as inflammatorybowel disease.

Crohn’s disease | cytokines

Crohn’s disease (CD) is a chronic inflammatory bowel disease(IBD) of unknown etiology (1). The current hypothesis forCD

pathogenesis suggests that it develops in response to an overlyaggressive adaptive immune response to the commensal flora (2).However, it has been recently suggested that CD may result fromdeficits in innate immunity (3–5). In this context, the intestinalepithelium seems to play a major role, allowing sensing of the lu-minal environment and differential regulation of the gut immunesystem (6). Although previous studies have investigated the effectsof probiotics on intestinal cell lines in vitro, little is known re-garding their actions on the intestinal epithelium in vivo.Probiotics are live microorganisms that seem to promote gut

health and regulate intestinal homeostasis (7). Different possiblemechanisms of probiotic action have been proposed, includingboth suppression and stimulation of host immune responses.VSL#3 is a probiotic mixture of eight different bacterial strainsthat has demonstrated beneficial effects in several intestinal con-ditions. Controlled studies have shown VSL#3 to be efficaciousfor patients with ulcerative colitis (UC) and pouchitis (8, 9), aswell as in animal models of colitis (10, 11). However, the precisemechanisms of action are not fully understood.In this study, we tested the hypothesis that administration of

VSL#3 to SAMP1/YitFc (SAMP) mice with chronic CD-likeileitis can prevent the onset of gut inflammation through stim-ulation of innate immunity in the gut. The SAMP mouse modelhas been extensively characterized by our group and others andhas unique characteristics closely resembling CD (12–14). Ourresults show that pretreatment of SAMP mice with VSL#3 for6 weeks before the onset of ileitis prevents the occurrence ofintestinal inflammation through a mechanism involving stim-ulation of TNF-α in the intestinal epithelium and restitution ofnormal barrier function. Therefore, we propose the concept thatprobiotic bacteria promote gut health and prevent chronic in-testinal inflammation through stimulation of epithelial innateimmunity.

ResultsHigh-Dose VSL#3 Prevents Ileitis in SAMP Mice but Does Not AffectEstablished Disease. We first investigated the effects of low-doseVSL#3 in preventing the onset of SAMP ileitis (preventionprotocol). Although some amelioration of villous distortion andthickening of the muscularis mucosa was observed, a severe in-flammatory infiltrate in the lamina propria was still present, withno significant differences compared with controls (Fig. 1 Bi andBii). By comparison, SAMP mice pretreated with high-doseVSL#3 showed a marked decrease in the severity of ileitis com-pared with age-matched, untreated controls (total score: 3.2 ± 3.1vs. 11.6 ± 6.1, P < 0.005); the three histologic components of thetotal inflammatory score (i.e., villous distortion, active inflam-mation, and chronic inflammation) were all significantly decreased(Fig. 1A). Interestingly, ileitis was completely prevented in 5 of11 mice (45%; total score: 0–1), whereas no evidence of chronicinfiltration (chronic score: 0) was observed in 10 of 11 of these mice(91%). Histologically, mice administered high-doseVSL#3 showedalmost complete protection of mucosal integrity and virtually noinflammatory cell infiltrates in the lamina propria (Fig. 1Biii).The efficacy of high-dose VSL#3 was then evaluated in 30-

week-old SAMPmice with established ileitis (treatment protocol).No beneficial effects were observed between high-dose VSL#3treated and untreated mice (Fig. S1).

VSL#3 Administration Changes the Composition of Probiotic DNAFound in Fecal Material and the Terminal Ileum. We next per-formed RT-PCR on total fecal DNA with primers specific forthree of the probiotic bacteria contained in VSL#3 (Streptococcusthermophilus, Bifidobacterium infantis, and Lactobacillus acid-ophilus). We found that administration of high-dose VSL#3markedly increased the concentration of each of these bacteria,starting from the first week after probiotic administration. Theamount of bacterial DNA for the three strains was elevatedthroughout the 6 weeks of treatment (Fig. 2 A–C). As expected, S.thermophilus, which is present at the highest concentration inVSL#3, had the highest relative increase.We next compared the bacterial composition in the feces and

terminal ileum of mice treated with low- and high-dose VSL#3after 6 weeks of treatment. In SAMP mice treated with low-doseVSL#3, only S. thermophiluswas detected in the stool, whereas all

Author contributions: T.T.P. and F.C. designed research; C.P., R.S., and G.B. performedresearch; C.P., R.S., K.O.A., T.T.P., and F.C. analyzed data; and C.P., K.O.A., T.T.P., and F.C.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1Present address: Department of Digestive and Liver Diseases II, School of Medicine, Uni-versity “La Sapienza,” Sant’Andrea Hospital, 00189 Rome, Italy.

2To whom correspondence should be addressed at: Division of GI & Liver Disease, Uni-versity Hospitals/Case Western Reserve University, 10900 Euclid Avenue, BRB 425, Cleve-land, OH 44106-4952. E-mail: fabio.cominelli@case.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0910307107/DCSupplemental.

454–459 | PNAS | January 5, 2010 | vol. 107 | no. 1 www.pnas.org/cgi/doi/10.1073/pnas.0910307107

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0

three strains were significantly elevated in the high-dose treatmentgroup (Fig. 2 D–F). In the ileum, increased trends were found forB. infantis and S. thermophilus in mice administered low-doseVSL#3 (3.7-fold for B. infantis, 10.8-fold for S. thermophilus)compared with untreated control mice. Conversely, mice ad-ministered high-dose VSL#3 showed a consistent and abundantincrease for the two bacteria (265-fold for B. infantis; 639-fold forS. thermophilus) compared with untreated mice (P = 0.005 forboth; Fig. 3A–C).Nodetectable levels ofL. acidophilusDNAwerefound in the ileum of untreated or VSL#3-treated mice.

VSL#3 Acts on the Intestinal Epithelium by Both Restoring EpithelialBarrier Function and Stimulating Production of TNF-α by EpithelialCells. To test the hypothesis that the intestinal epithelium is the

primary site of action of VSL#3, we studied the effects of high-dose VSL#3 (prevention protocol) on intestinal permeability andproduction of epithelial-derived cytokines compared with controlmice. After 6 weeks, VSL#3-treated mice showed a marked de-crease in small intestinal permeability compared with untreatedcontrols (0.35 ± 0.03 vs. 0.45 ± 0.03, P < 0.01; Fig. 4) to a levelsimilar to that measured in healthy mice (15).We next studied the expression of TNF-α in freshly isolated

epithelial cells from different experimental groups. Admin-istration of high-dose VSL#3 significantly increased TNF-αmRNA levels in isolated epithelial cells compared with untreatedcontrol mice (5-fold increase, P < 0.05; Fig. 5A). In addition,mRNA levels of the NF-κB inhibitory subunit IκBα were sig-nificantly increased in VSL#3-treated mice compared with un-treated controls (2-fold, P < 0.05; Fig. 5B), suggesting activationof the NF-κB pathway. To confirm that TNF-α production wasalso increased at the protein level, immunohistochemical analysiswas performed. A consistent and marked increase in TNF-α, im-munolocalized to the epithelium, was observed in mice ad-ministered high-dose VSL#3 compared with low-dose anduntreated control mice (Fig. 5 A–C).

VSL#3 Conditioned Media Exerts Specific Stimulatory Effects onEpithelial Cells In Vitro. To further explore the hypothesis thatprobiotics enhance the innate immune defense of the intestinalepithelium, we investigated the effects of VSL#3 conditionedmedia (CM) on epithelial cells in vitro. Terminal ileum organcultures were incubated for 24 h in the presence of CM alone or

A

B

0123456789

1011121314

0.00.51.01.52.02.53.03.54.04.5

141210

86420

Total Score0.0

1.0

2.0

3.0

4.0

Chronic Inflammatory Score

0.0

1.0

2.0

3.0

4.0

0.01.0

2.03.0

4.05.0

Ac�ve Inflammatory Score Villous Distor�on Score

Control Low-dose VSL#3 High-dose VSL#3

i10x 20x

ii

iii

**

**

**

*

Fig. 1. VSL#3 administration prevents the onset of ileitis in SAMP mice.Three-week-oldmicewere administered high- or low-dose VSL#3 for 6weeks.(A) The total inflammatory score was markedly decreased in SAMP mice ad-ministered high-dose VSL#3. (B) Representative photomicrographs of H&E-stained sections, 10× and 20× original magnification: (i) control untreatedmice (n = 12) display severe distortion of the villi, with intense leukocyte in-filtration of the lamina propria and thickening of the muscularis mucosa; (ii)SAMPmice administered low-dose VSL#3 (n = 12) show less distortion of tissuearchitecture but a high level of cellular infiltration and thickening of themuscularis mucosa; and (iii) SAMPmice administered high-dose VSL#3 (n = 11)show almost complete prevention of mucosal damage, with preservation ofthe normal villi morphology and minimal inflammatory infiltrate. Data areexpressed as mean ± SEM. *P < 0.01; **P < 0.005.

100

900

0 1 2 4 6

0 1 2 4 6

EB

ND

0 1 2 4 6

0

Low-dose

B.infantis

S.thermophilus

*

0

10

20400

650

L.acidophilus F

ND

50

150

L.acidophilus

200

High-dose

0

100

200

300

Rela

�ve

Ra�o

Rela

�ve

Ra�o

Week

B.infantisDA

500

1000

1500

0

Rela

�ve

Ra�o

0 1 2 4 6

Week0 1 2 4 6

Rela

�ve

Ra�o

Low-dose High-dose

S.thermophilus

C

Week0 1 2 4 6

Rela

�ve

Ra�o

0

50

100

150

200

0

50

100

150

200

Low-dose High-dose

Rela

�ve

Ra�o

Fig. 2. VSL#3 alters the fecal probiotic DNA composition of SAMP mice. RT-PCR with primers specific for VSL#3 probiotic bacteria was performed weeklyon fecal extracts to assess the extent of bacterial colonization. Mice ad-ministered high-dose VSL#3 (n = 6) showed a relative increase in probioticDNA for all three strains at each posttreatment time point compared withbaseline (A–C). At week 6, mice given high-dose VSL#3 had significantlyincreased relative increases in probiotic DNA for each of the three strains,compared with mice given low-dose VSL#3 (E–G). Data are expressed asmean ± SEM. *P < 0.01 vs. low-dose VSL#3.

Pagnini et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 455

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0

concomitantly with fecal extract (FE) from SAMP mice. A sig-nificant increase was measured in TNF-α secretion for CM andCM+FE organ cultures compared with media and FE alone (2.9-and 2.5-fold, respectively, P < 0.001 for both; Fig. 6A).Finally, to test whether VSL#3 CM directly activates NF-κB,

the NF-κB subunit p65 (RelA) and the inhibitory subunit IκBαwere measured in nuclear and cytoplasm extracts, respectively,isolated from SAMP terminal ileum organ cultures stimulatedwith VSL#3 CM alone or in combination with FE. The levels ofp65 were increased in nuclear extracts and IκBα levels reduced incytoplasmic extracts from epithelial cells isolated from SAMPterminal ileum organ cultures stimulated with VSL#3 CM aloneand CM+FE, indicating activation of the NF-κB pathway (Fig.6B). These data confirm a stimulatory effect of VSL#3 probioticson the intestinal epithelium, as demonstrated by the increasedproduction of TNF-α and activation of the NF-κB pathway.

Anti–TNF-α Treatment Abrogates the Beneficial Effects of VSL#3 inSAMP Ileitis. To further assess that the preventative effects ofVSL#3 in SAMP ileitis are mediated by TNF-α, we studied theeffects of TNF-α blockade administered together with VSL#3 in3-week-old SAMP mice using the prevention protocol. SAMPmice treatedwith high-doseVSL#3 concomitant with anti–TNF-αmonoclonal antibody administration did not show significant im-provement in the severity of ileitis compared with untreated con-trolmice (total score: 9.0± 1.9 vs. 9.5± 1.0, not significant; Fig. S2).Interestingly, 3-week old SAMP mice treated with anti–TNF-αmonoclonal antibody alone also did not show any significant im-provements in ileitis severity compared with untreated mice (totalscore: 7.9 ± 1.7 vs. 9.5 ± 1.0, not significant; Fig. S2). These data

further demonstrate that the preventative effects of VSL#3 aremediated by stimulation of TNF-α.

DiscussionAlthough the defensive response of the intestinal epitheliumagainst pathogenic bacteria has been extensively explored, lessinformation is available regarding interactions with the commen-sal bacteria. A commonly held belief is that nonpathogenic andprobiotic bacteria exert opposing effects to those induced bypathogenic bacteria, including suppression rather than stimulationof proinflammatory cytokines in epithelial cells and macrophages(16). However, several studies underscore the importance of“physiologic inflammation” induced by the commensal flora, bothfor the development of the immune system and for the response topathogenic bacteria in the gut. Accordingly, the possibility thatprobiotic bacteria elicit the same kind of proinflammatory re-sponse is currently being explored.In this study we report that the probiotic mixture VSL#3

completely prevents the onset of intestinal inflammation in theSAMP mouse model of CD-ileitis. These effects were associatedwith a local effect on the intestinal epithelium (i.e., stimulation ofTNF-α production and NF-κB activation in epithelial cells, cou-pled with restitution of normal intestinal epithelial barrier func-tion). In addition, blockade of TNF-α by monoclonal antibodytreatment given concomitantly with VSL#3 administration com-pletely abrogated the beneficial effects of VSL#3 in this model.Analysis of bacterial DNA in the stool and ileal mucosa of treatedmice demonstrated a significant increase of probiotic strains,suggesting that effective bacterial colonization is required for thebeneficial effects of probiotic treatment. Taken together, our studysupports the hypothesis that probiotics promote gut healththrough a mechanism involving stimulation of epithelial innateresponses (i.e., stimulation of TNF-α expression), rather thansuppression of inflammation. In addition, our results provide im-portant insights into the pathogenesis of IBD by supporting theconcept that CD may be caused by a deficit in innate immunity.Several studies support the concept that probiotics exert stim-

ulatory rather than suppressive effects on the innate immunesystem. For example, DNA from the bacteria found in VSL#3exerts immunostimulatory effects onmacrophages, as indicated byTLR9-mediated activation of NF-κB and JNK. Despite theameliorative effects of VSL#3 administration observed in twoanimal models of colitis, bone marrow–derived macrophagesisolated from BALB/c mice produce the proinflammatory cyto-kines IL-6 and IL-12 when cultured with DNA from VSL#3 bac-teria (11). In addition, several studies showed that activation of

contro

l

low dose VSL#3

high dose VSL#3

rela

tive

ratio

A

B.infantis

S.thermophilus

C

B

*

S.thermophilus

B.infantis

*

Rela

�ve

Ra�o

Rela

�ve

Ra�o

0

10

20250350450550650

025

50

1000

3500

Fig. 3. VSL#3differentially alters the ilealprobioticDNAcompositionofSAMPmice. At the end of the study period (6 weeks), terminal ilea were collectedfrom SAMP mice and DNA isolated. (A) RT-PCR with primers specific for pro-biotic bacteria showed a significant increase of B. infantis and S. thermophiluscomparedwith untreatedmice. (B) Visualization of PCR products resolved on a2% agarose gel. Data are expressed as mean ± SEM. *P < 0.005 vs. control.

0 20.25

0.30

0.35

0.40

0.45

0.50

6

*

Lac/

Man

Ra�

o

Control

VSL#3

Week of Treatment

Fig. 4. VSL#3 pretreatment restores small intestinal epithelial barrier func-tion in SAMP mice. In vivo paracellullar permeability was determined bymeasuring the fractional urinary excretion of orally administered, region-specific (small intestinal) sugar probes (lactulose/mannitol ratio). Three-week-old SAMP mice were treated with VSL#3 for 2 weeks (n = 4) or 6 weeks (n = 6)and then compared with age-matched SAMP mice fed a normal diet. VSL#3administration significantly decreased small intestinal epithelial permeabilityby the end of the treatment period. Data are expressed as mean ± SEM.*P < 0.05 vs. control.

456 | www.pnas.org/cgi/doi/10.1073/pnas.0910307107 Pagnini et al.

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0

NF-κB plays a beneficial role in epithelial cells. In fact, when micethat lack epithelial expression of theNF-κB activating enzyme IκBkinase β (villin-Cre/Ikkβ F/Δ mice) were administered dextransodium sulfate (DSS), inflammation was of equal or greater se-verity as that observed in control mice (17). Activation of NF-κB issimilarly protective for the development of systemic inflammation,but not for the preventionof local injury to themucosa in an animalmodel of intestinal ischemia–reperfusion (18). Finally, the ob-servations that TNF-α–deficient mice or mice deficient in TNFRIsignaling in the intestinal epithelium are more susceptible to DSSacute colitis strongly support the concept that TNF-α may exertprotective functions in normal gut homeostasis and intestinal epi-thelial integrity (19, 20). Altogether, these observations suggestthat the beneficial effects of probiotics are associated withimmunostimulatory effects rather than immunosuppression. Ourresults provide direct evidence that probiotics prevent the onset ofexperimental ileitis by mechanisms involving stimulation of TNF-αand activation of NF-κB in the intestinal epithelium. However,stimulation of other innate immunity pathways, including the in-testinal production of IL-10 and TGF-β, may also be involved.TNF-α is a proinflammatory cytokine that plays an important

role in CD pathogenesis, as evidenced by the therapeutic benefitsthat are observed after administration of anti–TNF-α antibodies(e.g., infliximab) (21). During the inflammatory process, TNF-αexerts its function at the apex of the inflammatory cascade. It is

responsible for the recruitment and activation of lymphocytes andgranulocytes, the local expression of adhesion molecules on en-dothelial cells at the site of inflammation, the secretion of proin-flammatory mediators (IFN-γ and IL-6) and radical oxygenspecies, and the formation of edema and granulomata (22). Not-withstanding, TNF-α is protective for various cell types (23), inparticular for the intestinal epithelium (24). Our results supportthe importance of TNF-α production by epithelial cells in theprevention of disease, although no beneficial effect of probioticadministration was found in established disease.Our findings support the emerging concept that CD occurs as a

progression through distinct phases (disease initiation, establish-ment, and maintenance), each one characterized by specificimmunologic features and uniquemolecularmediators. This leadsto a model whereby a single mediator, such as TNF-α, is funda-mental in establishing the inflammatory disease yet can have anopposing, protective effect during disease initiation. Because ep-ithelial TNF-α is an integral part of the innate immune system, onehypothesis is that TNF-α production is partly directed against lu-minal antigens that would otherwise threatenmucosal integrity. Inthis scenario, aberrant penetration of luminal molecules (i.e.,bacterial or food antigens) into the lamina propria compartmentcould initiate a persistent inflammatory process. In a predisposedorganism, this could self-amplify exponentially and lead to thechronic inflammatory state characteristic of CD.

control CM FE CM+FE0

10

20

30

ratio

(wha

t/wha

t?)

control CM FE CM+FE0

1020304050607080

TNF

(pg/

ml)

**

Nuclearp65 (rel A)

Histone H1

control CM FE CM+FE0.0

0.5

1.0

1.5

ratio

(p65

/H1)

Control CM FE CM+FE

010203040506070

0.0

0.5

1.5

2.0

Control CM FE CM+FE

Control CM FE CM+FE

0

10

20

30

Ra�o

p65

/H1

CytoplasmicIkBa

B-actin

TNF

(pg/

ml)

A

B

Fig. 6. VSL#3 CM stimulates TNF-α production and activates NF-κB in smallintestinal organ cultures. Ilea from SAMP mice were harvested and culturedfor 24 h with either VSL#3 CM, FE, or combination CM+FE and comparedwith controls (media alone). (A) Secreted TNF-α production was measuredfrom resulting supernatants by ELISA and showed significantly increasedlevels in CM and CM+FE compared with controls. (B) Cytoplasmic and nuclearextracts analyzed by Western blot showed increased nuclear p65 (RelA) anddecreased cytoplasmic IκBα levels in CM and CM+FE compared with controls,indicating activation of the NF-κB pathway by VSL#3. Data are representa-tive of three independent experiments with similar results. *P < 0.001.

C

0123456789

controlVSL#3

rela

tive

ratio

*

0123456789

Rela

�ve

Ra�o

ControlVSL#3

A

0

1

2

3controlVSL#3

rela

tive

ratio

**

0

1

2

3 ControlVSL#3

Rela

�ve

Ra�o

B

Control

VSL#3

Fig. 5. VSL#3 administration stimulates epithelial cells in vivo. (A and B) In-creased expression of TNF-α and IκBα mRNA in epithelial cells after VSL#3administration.mRNAwas extracted from freshly isolated epithelial cells fromSAMPmice at the end of the study period. RT-PCRwas performedwith specificprimers and data normalized to β-actin. All values are expressed as mean ±SEM. *P < 0.05. (C) Representative photomicrographs of immunohisto-chemical staining demonstrated that TNF-α immunolocalized to the epi-thelium and was markedly increased in high-dose (Bottom Right) comparedwith low-dose (Bottom Left) VSL#3-treated mice, whereas untreated controls(Top Right) showed no TNF-α staining. Absence of primary detecting antibodyconfirmed specificity of TNF-α staining (Top Left). ×40 original magnification.

Pagnini et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 457

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0

More support for a unique role of TNF-α in maintaining theintegrity of the intestinal epithelium in the early phase of diseasedevelopment comes from our demonstration that there was re-duced small-intestinal permeability in mice treated with pro-biotics. Mice treated with VSL#3 in the prevention protocoldemonstrated a complete restitution of intestinal barrier functionto a level comparable to that of healthy mice. Several studies haveoutlined the important role played by impaired epithelial barrierfunction inCDanddemonstrated that increased permeability is anearly, and perhaps initiating, feature of IBD (15, 25). Our resultswere unexpected given that several studies have described a directeffect of TNF-α in producing leakage of the intestinal epithelium(26). We observed an increase in TNF-α production that corre-sponded to decreased epithelial permeability in the VSL#3-treated group. Because decreased intestinal permeability by pro-biotic bacteria has previously been described (10), it is likely thatthe improvement we observed is directly caused by the probioticbacteria, although an indirect effect of reduced mucosal in-flammation in theVSL#3-treated group should not be completelyexcluded.Interestingly, in the present study we found that chronic ileitis

was effectively prevented by innate immune stimulation in theearly phase of the disease. This supports the emerging theory thatattributes CD susceptibility to an initial defect of innate immunity(3–5). This may, in turn, cause recruitment and dysregulated ac-tivation of the acquired immune system, leading to the establish-ment of the typical inflammatory features characteristic of CD.Our findings raise the attractive hypothesis that probiotic bacteriacan restore the breach of the innate immunity system and therebyprevent the onset of the inflammatory disease.One of the open issues regarding probiotics is whether the

bacteria exert their effects through modification of the residentflora or whether they act as antiinflammatory agents. Our find-ings indicate that probiotic bacteria do indeed exert a pharma-cologic effect upon epithelial cells, albeit stimulatory rather thansuppressive, that requires the presence of temporary probioticcolonization of the intestinal lumen. Indeed, mice administeredlow-dose VSL#3 did not show significant increases in bacterialconcentration in the feces or ileum, and disease was not pre-vented or attenuated in these mice despite the fact that this dosewas comparable to that previously used to treat experimentalcolitis. The dose used for the prevention of ileitis is therefore animportant consideration.To assess the efficacy of bacterial colonization, we developed a

culture-independent technique that used specific primers for threerepresentative bacterial strains (L. infantis, L. acidophilus, andS. thermophilus) in theVSL#3 preparation. It is worth noting that,even though the different species displayed a similar time-coursepattern for colonization, the fecal and/or ileal probiotic DNAcontent differed for each species, as well as with respect to the doseof VSL#3 administered. Detection of only S. thermophilus in thefeces and intestinal mucosa of SAMP mice treated with low-doseVSL#3 likely reflects insufficient colonization of the other bac-terial species. It would be interesting to evaluate the beneficialcontribution of each separate VSL#3 bacterial species and howthis relates to the concentration of DNA in the fecal matter andthe ileum.In conclusion, we have demonstrated that administration of

high-doseVSL#3effectively prevents the development ofCD-likeileitis in SAMP mice. This preventive effect was mediated by in-creased epithelial innate immunity. As a result, cells of the ac-quired immune system are not recruited, and the onset of theinflammatory cascade is prevented. Our results support the hy-pothesis that probiotics promote gut health through a mechanisminvolving stimulation rather than suppression of the epithelialinnate immune system.

Experimental ProceduresVSL#3 Administration Protocols. Mice were maintained in an institutionalanimal facility under specific pathogen-free conditions. All protocols wereapproved by the UVA Institutional Animal Care and Use Committee. VSL#3administrationwas initiated in either 3-week-old SAMPmice before the onsetof ileitis for the prevention protocol, or in 30-week-oldmice for the treatmentprotocol. For the former, micewere administered either low-dose VSL#3 (17 ×108 cfu/day) by gavage (10, 11), or high-dose VSL#3 (50 × 109 cfu/day) mixed intheir food. Age-matched control mice were fed unsupplemented chow. Forthe treatment protocol, 30-week-old mice were fed high-dose VSL#3 daily intheir food. For both protocols, VSL#3 was administered for 6 weeks, afterwhich mice were killed and ilea and MLNs collected for further analysis. Insome prevention protocol experiments, SAMP mice were administered amonoclonal antibody against murine TNF-α (0.5 mg weekly) together withhigh-dose VSL#3 (27). Control mice were either administered anti–TNF-αmonoclonal antibody alone or vehicle.

Histologic Assessment. Histologic evaluation was performed on H&E-stainedileal sections by a single blinded pathologist using a validated scoring system(28). Briefly, inflammation was measured using a total inflammatory indexcomposed of three components: active and chronic inflammation, as well asvillous distortion.

Isolation of Fecal Extract and Intestinal Epithelial Cells. Fecal extract was pre-pared using a method previously described (29), and total protein concen-tration was evaluated by the Bradford assay. Samples were stored at −80°Cuntil further analysis. In the organ culture stimulation assay, fecal extractswere diluted to 400 μL/mL, of which 14 μg/mL per reaction well were used.Freshly isolated epithelial cells were harvested from ileal segments of ex-perimental mice as previously described (23) and processed for real-time RT-PCR as described below.

DNA Isolation and RT-PCR. Total DNAwas isolated and purified from feces andterminal ileum specimens using the QIAamp DNA Stool Mini Kit and QIAampDNAMini Kit (Qiagen). Fresh feces were collected weekly and stored at−80°Cuntil processed. Terminal ilea were washed in PBS containing 1% penicillin/streptomycin, and specimens were stored at −20°C until assay. Samples werediluted to a concentration of 20 pg/μL (feces) or 60 pg/μL (ileal tissue) andamplifiedusing SYBRGreen PCRmastermix (Applied Biosystems)with primersspecific for VSL#3 probiotic bacteria (B. infantis, L. acidophilus, and S. ther-mophilus) to determine levels of probiotic bacterial DNA (30, 31).

To measure expression of epithelial-derived mRNAs, RNA was extractedfrom epithelial cells using the RNAeasy Miniprep Kit (Qiagen). Reversetranscription was performed using the GeneAmp RNA PCR kit (AppliedBiosystems). IκBα (forward: 5′-AGACTCGTTCCTGCACTTG-3′; reverse: 5′-CCTGGCTGGTTGGTGATC-3′) and TNF-α (forward: 5′-GCGGTGCCTATGTCT-CAG-3′; reverse: 5′-GCCATTTGGGAACTTCTCATC-3′) were designed usingBeacon Designer 2 software (Premier Biosoft International).

PCR reactionswereperformed ina total volumeof20μL (16 μL of SYBRGreenPCR master mix and 4 μL of DNA) in an iCycler iQ detection system (Bio-RadLaboratories). Reactionswere rununder the following conditions: 2min at95°C,followedby 40 cycles of 15 s at 95°C, 15 s at 60°C, and 15 s at 72°C. Sampleswerethen run on a 2% agarose gel with ethidium bromide for visualization. TargetmRNA was normalized to β-actin for each sample, and all samples were run induplicate.Resultswereexpressedasa ratio relative to the lowest control sample.

Measurement of Intestinal Permeability. In vivo epithelial paracellular per-meability was measured as previously described (15), before and after VSL#3therapy. Permeability was evaluated after administration of high-doseVSL#3 for 2 (n = 4) and 6 weeks (n = 6) in comparison with age-matched,untreated mice (n = 4). Mice were fasted for 2 h then fed a sugar-probemixture (60 mg/mL lactulose, 40 mg/mL mannitol, and 30 mg/mL sucralose)via orogastric gavage, and urine collected after 22 h. Urine volume wasmeasured, and the concentration of each of the sugar probes was de-termined by HPLC. Fractional excretion of each probe was determined as aratio of the amount of probe in the urine to the amount administered.

Immunohistochemical Analysis. Tissue samples were prepared in Bouin’s fix-ative (LabChem), processed as previously described (15), cut into 3-μm-thicksections, and mounted on superfrost/PLUS glass slides (Fisher Scientific).Sections were deparaffinized in xylene (twice for 10 min) and dehydrated in100% ethanol (twice for 5 min) and 95% ethanol (twice for 5 min). Sectionswere then blocked for endogenous peroxidase activity using 1% hydrogenperoxide in methanol for 30 min, washed twice with PBS, then blocked with

458 | www.pnas.org/cgi/doi/10.1073/pnas.0910307107 Pagnini et al.

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0

3% BSA/PBS (Sigma Chemical) for 20 min. Tissue sections were incubatedovernight at 4°C with mouse anti-TNF (R&D Systems) primary antibody, fol-lowed by a biotinylated horse anti-goat antibody, and then incubatedwith anavidin/biotin complex (Vectastain Elite ABC Kit; Vector Laboratories) for30min. Immunoreactive cellswere visualizedbyadditionofdiaminobenzidinesubstrate and counterstained with hematoxylin. All incubations were con-ducted at room temperature unless otherwise noted. Negative controls wereprepared using the same conditions as described above, but replacing theprimary antibody with an equivalent volume of PBS.

Organ Culture Stimulation Assay. VSL#3 CM was prepared as previously de-scribed (32), and the CM diluted 1:10 in unconditioned media for organculture stimulation assays. The production of cytokines from epithelial cellswas measured in an in vitro assay as described by Rakoff-Nahoum et al. (24),and the concentration of TNF-α was evaluated by ELISA. TNF-α levels werecorrected for total protein concentration measured by the Bradford assay.

Western Blot. Nuclear and cytoplasmic extracts were prepared using the NE-PER kit and protein concentrationsmeasured using BCA protein assay reagent(Pierce). Equal amounts of proteins were electrophoresed using precast

NuPAGE Triacetate gels and separated proteins transferred using 0.2-μmPVDF membrane (Invitrogen). Proteins were then probed with primary an-tibody, followed by incubation with HRP-conjugated secondary antibody(Santa Cruz Biotechnology). Proteins were visualized using the ECL detectionsystem (Amersham Pharmacia). Densitometry was performed using Image-quant software, and band intensity ratios were measured.

Statistical Analysis. Statistical analysis was performed using a Mann-WhitneyU test. Data were expressed as mean ± SEM. An α level of 0.05 was consideredsignificant (P < 0.05).

ACKNOWLEDGMENTS. This work was supported by US Public Health Service/National Institutes of Health Grants DK-42191, DK-44540, DK-55812 (to F.C.),and DK-56762 (to T.T.P.) and by the University of Virginia Digestive HealthResearch Center (UVA DHRC) (DK-67629). We thank Sharon B. Hoang andJames R. Mize of the Morphology/Imaging Core of the UVA DHRC forhistologic evaluation and scoring of samples;WilliamG. Ross for FACS analysis;Dr. Claudio De Simone for providing the VSL#3 preparation; Dr. John Med-dings for the permeability assay; andDr. Robert Schreiber for kindly providingthe monoclonal antibody against TNF-α.

1. Podolsky DK (2002) Inflammatory bowel disease. N Engl J Med 347:417–429.2. Fiocchi C (1998) Inflammatory bowel disease: Etiology and pathogenesis.Gastroenterology

115:182–205.3. Bamias G, Nyce MR, De La Rue SA, Cominelli F; American College of Physicians;

American Physiological Society (2005) New concepts in the pathophysiology ofinflammatory bowel disease. Ann Intern Med 143:895–904.

4. Cobrin GM, Abreu MT (2005) Defects in mucosal immunity leading to Crohn’s disease.Immunol Rev 206:277–295.

5. Marks DJ, et al. (2006) Defective acute inflammation in Crohn’s disease: A clinicalinvestigation. Lancet 367:668–678.

6. Sansonetti PJ (2004) War and peace at mucosal surfaces. Nat Rev Immunol 4:953–964.7. Fuller R (1989) Probiotics in man and animals. J Appl Bacteriol 66:365–378.8. Bibiloni R, et al. (2005) VSL#3 probiotic-mixture induces remission in patients with

active ulcerative colitis. Am J Gastroenterol 100:1539–1546.9. Gionchetti P, et al. (2000) Oral bacteriotherapy as maintenance treatment in patients

with chronic pouchitis: A double-blind, placebo-controlled trial. Gastroenterology119:305–309.

10. Madsen K, et al. (2001) Probiotic bacteria enhance murine and human intestinalepithelial barrier function. Gastroenterology 121:580–591.

11. Rachmilewitz D, et al. (2004) Toll-like receptor 9 signaling mediates the anti-inflammatory effects of probiotics in murine experimental colitis. Gastroenterology126:520–528.

12. Kosiewicz MM, et al. (2001) Th1-type responses mediate spontaneous ileitis in a novelmurine model of Crohn’s disease. J Clin Invest 107:695–702.

13. Matsumoto S, et al. (1998) Inflammatory bowel disease-like enteritis and caecitis in asenescence accelerated mouse P1/Yit strain. Gut 43:71–78.

14. Rivera-Nieves J, et al. (2003) Emergence of perianal fistulizing disease in the SAMP1/YitFc mouse, a spontaneous model of chronic ileitis. Gastroenterology 124:972–982.

15. Olson TS, et al. (2006) The primary defect in experimental ileitis originates from anonhematopoietic source. J Exp Med 203:541–552.

16. Jijon H, et al. (2004) DNA from probiotic bacteria modulates murine and humanepithelial and immune function. Gastroenterology 126:1358–1373.

17. Greten FR, et al. (2004) IKKbeta links inflammation and tumorigenesis in a mousemodel of colitis-associated cancer. Cell 118:285–296.

18. Chen LW, et al. (2003) The two faces of IKK and NF-kappaB inhibition: Prevention ofsystemic inflammation but increased local injury following intestinal ischemia-reperfusion. Nat Med 9:575–581.

19. Naito Y, et al. (2003) Enhanced intestinal inflammation induced by dextran sulfatesodium in tumor necrosis factor-alpha deficient mice. J Gastroenterol Hepatol 18:560–569.

20. Mizoguchi E, et al. (2008) TNF receptor type I-dependent activation of innate responsesto reduce intestinal damage-associated mortality. Gastroenterology 134:470–480.

21. Hanauer SB, et al.; ACCENT I Study Group (2002) Maintenance infliximab for Crohn’sdisease: The ACCENT I randomised trial. Lancet 359:1541–1549.

22. Papadakis KA, Targan SR (2000) Tumor necrosis factor: biology and therapeuticinhibitors. Gastroenterology 119:1148–1157.

23. Bohan A, ChenWS, Denson LA, Held MA, Boyer JL (2003) Tumor necrosis factor alpha-dependent up-regulation of Lrh-1 and Mrp3(Abcc3) reduces liver injury in obstructivecholestasis. J Biol Chem 278:36688–36698.

24. Rakoff-Nahoum S, Paglino J, Eslami-Varzaneh F, Edberg S, Medzhitov R (2004)Recognition of commensal microflora by toll-like receptors is required for intestinalhomeostasis. Cell 118:229–241.

25. Hollander D, et al. (1986) Increased intestinal permeability in patients with Crohn’sdisease and their relatives. A possible etiologic factor. Ann Intern Med 105:883–885.

26. Söderholm JD, et al. (2004) Increased epithelial uptake of protein antigens in theileum of Crohn's disease mediated by tumour necrosis factor alpha. Gut 53:1817–1824.

27. Sheehan KC, Ruddle NH, Schreiber RD (1989) Generation and characterization ofhamster monoclonal antibodies that neutralize murine tumor necrosis factors.J Immunol 142:3884–3893.

28. Burns RC, et al. (2001) Antibody blockade of ICAM-1 and VCAM-1 amelioratesinflammation in the SAMP-1/Yit adoptive transfer model of Crohn’s disease in mice.Gastroenterology 121:1428–1436.

29. Cong Y, Weaver CT, Lazenby A, Elson CO (2002) Bacterial-reactive T regulatory cellsinhibit pathogenic immune responses to the enteric flora. J Immunol 169:6112–6119.

30. Furet JP, Quénée P, Tailliez P (2004) Molecular quantification of lactic acid bacteria infermented milk products using real-time quantitative PCR. Int J Food Microbiol 97:197–207.

31. Vitali B, Candela M, Matteuzzi D, Brigidi P (2003) Quantitative detection of probioticBifidobacterium strains in bacterial mixtures by using real-time PCR. Syst ApplMicrobiol 26:269–276.

32. Petrof EO, et al. (2004) Probiotics inhibit nuclear factor-kappaB and induce heat shockproteins in colonic epithelial cells through proteasome inhibition. Gastroenterology127:1474–1487.

Pagnini et al. PNAS | January 5, 2010 | vol. 107 | no. 1 | 459

PHYS

IOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 27

, 202

0