Gut 1994; 35: 1090-1097

Role of intestinal microflora in chronicinflammation and ulceration of the rat colon

S Videla, J Vilaseca, F Guarner, A Salas, F Treserra, E Crespo, M Antolin,J-R Malagelada

AbstractBacteria and their products stimulateinflammatory responses. The effectsof different antimicrobial regimens(amoxicillin/clavulanic acid, tobramycin,imipenem, vancomycin, metronidazole)were investigated on the course ofexperimental colitis induced bytrinitrobenzenesulphonic acid (TNB) inthe rat. On day 7 and 21 after the induc-tion of colitis, matched groups of controland antibiotic treated rats were subjectedto colonic dialysis to measure eicosanoidrelease, and killed for morphologicalassessment of the colonic lesions (macroand microscopic scores). Stool sampleswere cultured. Selective antibiotictreatment against Gram positive, Gramnegative or anaerobic bacteria had noeffect on colonic lesion scores. Bycontrast, certain broad spectrum anti-biotics (amoxicillinlclavulanic acid orthe association of imipenem plusvancomycin) significantly reduced macroand microscopic scores. Rats receivingthese antibiotics did not develop chroniccolitis as shown by the virtual absence ofcolonic strictures, adhesions, fibrosis, andgranulomas. On day 21 after TNB, theintracolonic release of prostaglandin E2,thromboxane B2, and leukotriene B4 wassignificantly higher in control than inantibiotic treated rats. Control stoolcultures showed abundant colony formingunits of both aerobic and anaerobicbacteria. Amoxicillin/clavulanic acid andimipenem plus vancomycin inducedappreciable reductions in luminalbacteria. In conclusion, certain broadspectrum antibiotics prevent chroniccolitis. The normal colonic flora seems toplay an important pathogenetic part in theprogression of inflammatory coloniclesions to chronicity.(Gut 1994; 35: 1090-1097)

The need for a reliable experimental model ofinflammatory bowel disease has led to con-siderable attention being given in recent yearsto the trinitrobenzenesulphonic acid (TNB)model of intestinal inflammation in the rat.1-10Inflammatory lesions in the distal colon thatpersist for up to five to six weeks can beinduced by a single intracolonic dose of TNBdissolved in ethanol.2 Chronic colitis by TNBis characterised by large mucosal ulcerationsand transmural inflammation with serosalinvolvement and adhesions to surrounding

tissues. The inflammatory infiltrate consists ofboth polymorphonuclear and mononuclearcells with occasional granulomas. At high TNBdoses, most animals develop colonic stricturesthat eventually result in voluminous dilatationof the proximal bowel with large chroniculcers. Spare areas of normal mucosa, how-ever, are seen in the vicinity of the lesions.Thus, the model shows some morphologicalfeatures that resemble human Crohn's disease.The mechanism by which TNB induces the

lesions is not clear. It was first suggested thatTNB acts as an hapten,2 as it can modify cellsurface proteins by forming covalent bondswith lysine groups.11 A macrophage mediatedresponse against such modified cells woulddamage the colonic mucosa triggering theinflammatory reaction.12 Subsequent studieshave also shown, however, that TNB exertsdirect toxicity on intestinal epithelial cellmonolayers in vitro without intervention ofmacrophages.10 The metabolism of TNB mayyield superoxide and hydrogen peroxideradicals with proinflammatory and cytotoxicproperties that could initiate the inflammatoryreaction.8 Either macrophage mediated cellinjury or formation of reactive oxygen species,or both, may thus induce acute inflammationof the rat colon, but these events do not explainthe most emblematic peculiarity of the TNBmodel of colitis - that is, its spontaneousprogression to chronicity.We hypothesised that bacterial antigens

present in the colonic lumen could play a partin the development of chronic colitis afterTNB. Bacterial cell wall fragments can elicit aresponse of chronic inflammation wheninjected into the bowel wall, as shown bySartor et al.13 The intestinal microflora couldinvade the colonic wall after disruption of theepithelium by the challenge with TNB, and thepresence of bacteria would perpetuate theantigenic insult and the inflammatory responsewithin the wall towards chronicity. Thus, theaim of our study was to test this hypothesis byexamining the effect of antimicrobial treatmenton faecal flora, release of inflammatorymediators, and morphological features ofTNBcolitis.

Methods

ANIMALSMale Sprague-Dawley rats weighing 225 to250 g were used in this study. The animalswere maintained in a restricted access roomwith controlled temperature (23°C) and light-dark cycle (12 h: 12 h), and were housed in

Digestive SystemResearch UnitS VidelaJ VilasecaF GuarnerM AntolinJ-R Malagelada

Department ofMicrobiology, HospitalGeneral Valld'Hebron,AutonomousUniversity ofBarcelona, SpainE Crespo

Department ofPathology, HospitalMutua de Terrasa,SpainA SalasF Treserra

Correspondence to:Dr F Guarner, DigestiveSystem Research Unit,Hospital General Valld'Hebron, Barcelona 08035,Spain.Accepted for publication23 November 1993

1090

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Role of intestinal microflora in chronic inflammation and ukeration of the rat colon

rack mounted cages with a maximum of fiverats. Standard rodent chow pellets (Biocenter,Barcelona, Spain) and tap water were providedad libitum.

EXPERIMENTAL COLITISChow pellets were withdrawn for 36 hours andrats were given drinking water containing 20%glucose and electrolytes. On the third day,chow pellets were again permitted. The ratswere lightly anaesthetised with ether and colitiswas induced by intracolonic instillation of 1 mlof a solution containing 60 mg of 2,4,6-trini-trobenzenesulphonic acid (Carlo Erba, Milan,Italy) in 10% ethanol (Merck, Darmstadt,Germany) using a rubber cannula (8 cm long,external diameter 2 mm) inserted through therectum. Subsequently, 0 5 ml of air wasinjected to completely clear the TNB ethanolsolution from the cannula.

EXPERIMENTAL DESIGN

ProtocolA: effect of antimicrobials on the acutestage of TNB colitisThree hundred and fifteen rats were includedin this study. All experiments consisted of atest group (n= 15) that received antibiotictreatment, and a matched control group(n= 15) that received the correspondingvehicle. The following antimicrobial drugswere tested: amoxicillin (160 mg/kg per day)plus clavulanic acid (40 mg/kg per day) giventwice daily in 5 ml saline enemas (controlsreceived normal saline enemas); tobramycin(2 mg/kg per day) plus vancomycin (50 mg/kgper day), by enema twice daily; tobramycin(2 mg/kg per day), by enema twice daily;metronidazole (20 mg/kg per day), by enematwice daily; imipenem (50 mg/kg per day) plusvancomycin (50 mg/kg per day), either byenema twice daily, by intraperitoneal injectiontwice daily, or orally, diluted in the drinkingwater as these drugs are not absorbed;vancomycin (50 mg/kg per day), in thedrinking water; and imipenem (50 mg/kg perday), in the drinking water. The volume ofwater consumed in every cage of rats wasmeasured daily. Addition of drugs did notchange the water volume consumed, whichaveraged 40 ml/day/rat. Drug treatmentsstarted three days before TNB and the ratsreceived the same drug up to day 7 after TNB.To test the effect of antibiotic treatment givenafter the induction of colitis, an additionalexperiment included a control group and twogroups receiving imipenem (50 mg/kg) plusvancomycin (50 mg/kg) given by mouth.Treatment started either three days beforeTNB (pre-treatment group) or 24 hoursafter TNB (post treatment group) and wascontinued up to day 7.

In all experiments, weight controls wereperformed on days 0, 3, and 7 after TNB. Onday 7, the rats were given intracolonicdialysis, and they were then killed by cervicaldislocation. The distal colon was removed,opened longitudinally, rinsed with physio-

logical saline, and pinned out on syntheticcork. The lesions were assessed macroscopi-cally and the whole piece was then fixed in asolution containing 2% formaldehyde and1% glutaraldehyde for later histologicalexamination.

Protocol B: effect of antimicrobials on the chronicstage of TNB colitisSixty rats were included in this part of thestudy. The experiments consisted of a testgroup (n = 15) that received antibiotic treat-ment, and a matched control group (n = 15)that received vehicle. The following drugs weretested: amoxicillin (160 mg/kg per day) plusclavulanic acid (40 mg/kg per day) given twicedaily by intracolonic enema in saline (controlsreceived saline enemas twice daily); andimipenem (50 mg/kg per day) plus vancomycin(50 mg/kg per day), diluted in the drinkingwater. Treatment started three days before theinduction of colitis and was continued up today 7 after TNB. Thereafter, no treatmentwas given up to day 21, which was the endpoint of the experiments. The rats were thengiven intracolonic dialysis, and killed bycervical dislocation. Colons were exposed,scored macroscopically, and processed forhistological assessment as above.

ASSESSMENT OF COLONIC LESIONSThe macroscopically visible damage wasassessed with a stereomicroscope by two

TABLE I Morphological criteria for assessment ofcolonicdamage

Macroscopicscore

AdhesionsNoneMinimalInvolving several bowel loops

StricturesNoneMildSevere, proximal dilatation

UlcersNoneIinear ulceration < 1 cm lengthTwo linear ulcers <1 cmMore sites of ulceration or one largeulcer > 1 cm

Wall thicknessLess than 1 mm1-3 mmMore than 3 mmMaximum possible score

UlcerationNo ulcer, epithelisationSmall ulcers <3 mmLarge ulcers >3 mm

InflammationNoneMildModerateSevere

Depth of the lesionNoneSubmucosaMuscularis propriaSerosa

FibrosisNoneMildSevereMaximum possible score

012

023

0123

01210Histologicalscore

012

0123

0123

01210

1091

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Videla, Vilaseca, Guarner, Salas, Treserra, Crespo, Antolin, Malagelada

10

a)0

0n

O _ I _ II-1 _ I

Macroscopic Histological Macroscopic Histological Macroscopic HistologicalFigure 1: Effect of antibiotic treatment on colonic lesion scores in the acute stage of TNBinduced colitis (day 7 after TNB). Scores were obtained according to the criteria shown inTable L Imipenem plus vancomycin treated rats (open bars) exhibited significantly lowermacroscopic and histological lesion scores than controls (solid bars) when treatment wasgiven either orally or by enema (*p<005 v controls). In contrast, intraperitoneal injectionof the antimicrobials was not effective. No mortality was seen in rats treated with theseantibiotics, and a single ratfrom each control group died.

observers unaware of the treatment (JV andFG). Each colon was assigned a score on ascale ranging from 0 to 10 based on thecriteria stated in Table I, which includecolonic adhesions to surrounding tissues,strictures, mucosal ulcerations, and wallthickening.For the histological studies, samples were

processed by routine techniques beforeembedding in paraffin wax. Sections wereobtained from areas showing macroscopicdamage, stained with haematoxylin and eosinand coded for blind examination by twopathologists (AS and FT). Both pathologistsexamined and scored all sections accordingto the presence of ulcerations, degree ofinflammation, depth of the lesions, and fibrosis(Table I). The highest score obtained byeach specimen was recorded for the statisticalanalysis.

INTRACOLONIC DIALYSISTo measure luminal eicosanoid release, ratswere given intracolonic dialysis for one hour,6under ketamine anaesthesia (100 mg/kg,intraperitoneal). Dialysis bags were preparedusing Visking seamless cellulose tubing (8/32,6-3 mm diameter, 7 cm long; Medicell,London, UK) attached by a 8 cm rubbercannula to an external syringe. After insertingthe entire cannula into the distal colon, thedialysis bag was filled with 1 ml of dialysissolution, consisting of 0 3% bovine serum

albumin in a solution of 120 mmol/l NaCl and30 mmol/l KHCO, (pH 7 9). One hour later,the fluid was withdrawn and stored at - 20°Cuntil assayed.

EICOSANOID RADIOIMMUNOASSAYSEicosanoid concentrations in the dialysateswere measured by specific radioimmunoassaysfor prostaglandin E2, thromboxane B2, andleukotriene B4. Tritiated standards were

purchased from Amersham International(Buckinghamshire, UK). Prostaglandin E2antiserum was kindly donated by Dr J A

Salmon (Wellcome Research Laboratories,Beckenham, UK). Antiserum for leukotrieneB4 radioimmunoassay was purchased fromAdvanced Magnetic (Cambridge, MA).Samples were diluted 1:10 to 1:25 in assaybuffer and measured without prior extractionand high performance liquid chromatographypurification.6

PLASMA CONCENTRATIONS OF ANTIMICROBIALDRUGSPlasma concentrations of vancomycin weremeasured by a standard technique(Vancomycin Reagent Pak, Abbot, AbbotPark, IL) using a TD analyser (Abbot) insamples from rats that had been on imipenem(50 mg/kg per day) plus vancomycin (50 mg/kgper day) for four days given either by oralgavage (n=4) or by intraperitoneal injection(n = 4). Blood samples were obtained one hourafter the last dose of vancomycin.

MICROBIOLOGICAL STUDIESStudies on faecal flora were performed sevendays after induction of colitis in two controls,two imipenem (50 mg/kg per day, orally),two imipenem plus vancomycin (50 + 50mg/kg per day, orally), and two amoxicillinplus clavulanic acid (160 + 40 mg/kg per day,enema) treated rats. In addition, two normalrats not given TNB were studied. Faecalspecimens were obtained directly from thecolonic lumen after laparotomy underanaesthesia (100 mg/kg ketamine) and kept inanaerobic containers. The samples werehomogenised, diluted in normal saline, andappropriate dilution volumes were incubatedunder aerobic or anaerobic conditions.Aerobic media consisted of blood agar, bloodagar with nalidixic acid, McConkey agar,CLED agar, Sabouraud agar, and mannitolagar. Anaerobic media consisted of lacqueredblood agar enriched with hemin and vitaminK,, lacquered blood agar with phenyl ethanol,and agar aztreonam. Plates inoculated forobligate anaerobes were incubated in ananaerobic chamber for 48-72 hours at 37°C,and plates for aerobes in air at 37°C. Afterincubation, colonies were enumerated andidentified. Final counts of colonies are referredto as gram of dried faeces.

STATISTICAL METHODSResults are presented as mean (SEM). Thestatistical difference between means wasdetermined using one way analysis of variancefor multiple comparisons and Student's t testfor single comparisons. Death rates werecompared by the Fisher's exact test.

Results

EFFECT OF ANTIMICROBIALS ON THE ACUTESTAGE OF TNB COLITISOn day 7 after TNB, control rats presenteda thickened colonic wall with tight serosal

1092

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Role of intestinal microflora in chronic inflammation and ulceration of the rat colon

TABLE II Effect of antimicrobials on the acute stage ofTNB colitis (day 7)

Lesion scores

Group Macroscopic Histological

Control (2)* 6-5 (0-8) 7 1 (0 6)Amoxicillin/clavulanic acid (4) 3.8 (0O8)t 5 3 (0 3)t

by enemaControl (1) 5-5 (0-7) 6 8 (0 3)Tobramycin (0) by enema 5 8 (0 7) 6 6 (0 4)Control (1) 7 2 (0 7) 8 9 (0 3)Impinenem (1 )t orally 7 2 (1-6) 8 5 (0 5)Control (0) 7 0 (0 8) 7 8 (0 5)Vancomycin (3) orally 5-1 (0-9) 7 1 (0 6)Control (1) 5 8 (0 7) 6 9 (0 3)Metronidazole (0) by enema 5 1 (0-7) 6 1 (0 3)Control (1) 6 8 (0 7) 7 5 (0 3)Tobramycin plus vancomycin (1) 5-6 (0-6) 7 0 (0 2)

by enema

*Each group consisted initially of 15 rats. The death rate onday 7 after TNB is shown in brackets; tp<0 05 v controlgroup. Data shown as mean (SEM).

adhesions to surrounding organs. The mucosa

seemed oedematous with transversal andlongitudinal linear ulcerations that conferredon it a characteristic 'cobblestone' aspect. Bylight microscopy, large areas of necroticmucosa, intense submucosal oedema, andtransmural inflammation was recognised inmost cases. Treatment with imipenem plusvancomycin or amoxicillin plus clavulanic acidsignificantly reduced macroscopic and histo-logical lesion scores (see Fig 1 and Table II),but particularly scores related to the depth ofthe histologically damaged area, the thickeningof the intestinal wall, and the presence ofadhesions.As Figure 1 shows, the way the antibiotic

treatment was given was critical for a beneficialeffect. The combination of imipenem withvancomycin acid was effective when given

3-

0

0 2 -

0

20

0

Adhesions Strictures Ulcers Thickness3

0

0

Cu1

Cn00)I

Ulceration Inflammation Depth Fibrosis

Figure 2: Effect of oral imipenem plus vancomycin oncolonic lesion scores in the chronic stage of TNB inducedcolitis (day 21). Treated rats (open bars) showedsignificantly lower macroscopic and histological lesion scoresthan controls (solid bars). Differences in adhesions,strictures, ulceration, depth of the lesion, andfibrosis wereparticularly noticeable (*p<O, 05 v controls).

orally or by enema, whereas intraperitonealinjection resulted in no significant effect(Fig 1). Both imipenem and vancomycin arepoorly absorbed when given orally. Plasmaconcentrations of vancomycin were undetect-able in four rats that received the drug by oralgavage. Therapeutic values of vancomycincould, however, be detected in plasma samplesone hour after intraperitoneal injection of thedrug (24 (7) p.g/ml), even though this methodyielded no beneficial effect on the colitis.Therefore, the beneficial effect of these anti-microbial drugs on colonic inflammation isunrelated to plasma bioavailability. Appro-priate antimicrobial concentration in thecolonic lumen seems to be critical for achievinga colonic therapeutic effect.When imipenem plus vancomycin treatment

was started 24 hours after induction of colitis,rats showed on day 7 lower macroscopic (7X1(08)) and histological (8-4 (0 5)) lesionscores than controls (9-8 (0 3)) and 9 7 (0 3),respectively, p<005). The effect was similarto that seen in rats pretreated with imipenemplus vancomycin (macroscopic score: 6-4(07); histological score: 8-1 (05). By day 7, 3of 15 rats had died in the control group, and 2of 15 rats in each imipenem plus vancomycingroup (pretreatment and post treatmentgroups).

Table II shows the macroscopic andhistological scores of colonic lesions in ratsthat received other antimicrobials againstdifferent bacterial populations. Treatmentwith tobramycin enemas was directed againstGram negative bacilli. Tobramycin did notchange the colonic lesion scores at day 7 afterTNB. Oral treatment with imipenem, a non-absorbable antibiotic active against most Grampositive cocci and Gram negative bacilli, andsome anaerobes, also failed to modify theseverity of the colonic lesion, but resulted in asignificantly higher death rate than in controlrats (p<001). Vancomycin is a bactericidalagent active against Gram positive organismsincluding anaerobic species. This drug givenorally reduced somewhat the macroscopiclesion scores but changes did not reachstatistical significance. Treatment withmetronidazole enemas, an agent active againstanaerobic species and several protozoa, wasnot effective in reducing colonic lesions.Finally, treatment with tobramycin plusvancomycin enemas resulted in a mildreduction in lesion scores without statisticalsignificance. This latter association is effectiveagainst Gram positive and Gram negativeaerobic species but is not active against mostanaerobic organisms.

EFFECT OF ANTIMICROBIALS ON THE CHRONICSTAGE OF TNB COLITISBroad spectrum antimicrobial drugs mitigatedthe progression of the colonic lesions tochronicity. As Figure 2 shows, treatment withimipenem plus vancomycin up to day 7 afterTNB significantly reduced macroscopic andhistological scores on day 21. Differencesbetween controls and imipenem plus vanco-

1093

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Videla, Vilaseca, Guarner, Salas, Tresetra, Crespo, Antolin, Malagelada

3

a)

0

202

0.

0

Adhesions Strictures Ulcers Thickness3 -

0

o

0

Ulceration Inflammation Depth Fibrosis

Fi'gure 3: Effect of amioxi'cillin/clavulani'c acid enlemas oticolonlic lesionZ scores i71 the chrotiizc stage of TNB indu4cedcoli'ti's (day 21). Treated rats (hatched bars) showedsignlificanltly lower niacroscopi'c atid hi'stological lesi'oti scoresthan controls (solid bars). Adhes crs, andfibrosiswere abseizt ziil rats treated wi'th amioxicillin1clavuxlanic aci'deaenias (*p<0 OS v controls).

mycin treated rats were particularly notable inulceration, depth of the lesions, fibrosis,strictures, and adhesions. The death rate wassimilar in both groups: 5 controls and 4imipenem plus vancomycin treated rats diedbefore day 2 l .

Figure 3 shows the effect of amoxi-cillin/clavulanic acid enemas given during thefirst week after TNB on the chronic coloniclesions on day 21 after TNB. Remarkably, noadhesions, fibrosis or strictures developed inthese rats. These are the pathological featuresthat are characteristic of the chronic stage ofTNB induced colitis. Indeed, in TNB colitisobvious inflammation and ulcers only persist

*

*

14 21

Days after TNB

Figure 4: Induction of colitis by TNB was associated with a significant body weight lossthat persisted up to day 7 in control rats (solid bars). In contrast, rats treated with oralimipenem plus vancomycin (open bars) showed only a mild body weight loss on day 3 afterTNB, and by day 7 changes were not significant. Subsequent body weight gain was higherin treated rats than in controls (*p<005).

beyond day 7 in dilated bowel segmentsproximal to strictures. It is therefore interestingto note that ulceration scores were higher incontrols than in antibiotic treated rats. Nomortality was seen in amoxicillin/clavulanicacid treated rats, and one rat from the controlgroup died.By light microscopy, microgranulomas were

found in 38% of the control rats whereas noneof the rats with amoxicillin/clavulanic acid andonly one rat in the imipenem plus vancomycingroup showed microgranulomas (p<0 05,Fisher's exact test).

Figure 4 shows the changes in body weightthat occurred after induction ofTNB colitis incontrol and in imipenem plus vancomycintreated rats. A net body weight loss wasdetected in every animal from the controlgroup three days after the TNB challenge andcontinued during the acute stage of colitispeaking by day 7. In contrast, antibiotic treatedanimals did not experience any significantweight loss. During the chronic stage ofcolitis, rats from the control group graduallyrecovered their lost body weight but, as shownby Figure 4, body weight gain remainedsignificantly higher in antibiotic treated ratsthan in controls.

COLONIC RELEASE OF INFIAMMA'I'ORYMEDIA'I'ORSThe release of inflammatory mediators by thecolonic mucosa was measured in vivo by intra-colonic dialysis. Figure 5 shows luminaleicosanoid release in controls and in ratstreated orally with imipenem-vancomycin.Before the induction of colitis, a baselineluminal release of prostaglandin E, throm-boxane B,, and leukotriene1 was detected inboth experimental groups. By day 7 afterTNB, luminal eicosanoid release had consider-ably increased in both groups of animalsreflecting colonic inflammation. No significantdifferences between groups were apparent. Byday 21 after TNB, however, luminal release ofall three eicosanoids in antibiotic treated ratshad decreased towards baseline values,whereas it remained significantly raised incontrol rats. These findings are consistent withour finding that during the chronic stage ofTNB colitis inflammatory activity onlypersisted in control rats but not in imipenem-vancomycin treated rats.

Similar results were obtained in rats treatedwith amoxicillin/clavulanic acid enemas.On day 21, luminal eicosanoid release inamoxicillin/clavulanic acid rats had returnedto baseline, while control rats showedconsiderable release of inflammatorymediators significantly above baseline values.

MICROBIOLOGICAL STUDIESTable III shows quantitative estimates ofcolony forming units after aerobic andanaerobic culture of stool samples. Analysis offaecal flora in normal rats not subjected tointracolonic TNB showed a considerablepredominance of Lactobacillus spp among the

120

80

40

Os

a)c

m

0)4-

-40 3 7

1 094

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Role of intestinal microflora in chronic inflammation and ukeration of the rat colon

50

cms0i

CD

a)

-J

25

0o0 7 14 21 0 7 14 21 0 7 14 21

Days after TNBFigure 5: Luminal release ofprostaglandin E2 (PGE,), thromboxane B2 (TXB2), andleukotriene B4 (LTBJ) as measured by intracolonic dialysis in rats treated with oralimipenem plus vancomycin (open bars) and controls (solid bars). No differences werefoundbetween control and antibiotic treated rats on day 7 after TNB. On day 21 after TNB,however, colonic release of all three eicosanoids returned to baseline in antibiotic treatedrats, while controls showed an enhanced release of the inflammatory mediators (*p<005 v

controls).

aerobic isolates. In TNB control rats, however,this species was not present on day 7 afterinduction of colitis. Quantitative analysisshowed an increase of other aerobic isolatessuch as Escherichia coli and Staphylococcus sppin control TNB colitis rats. Treatment withoral imipenem induced important changes infaecal flora, as aerobic isolates consistedmainly of Candida spp, and anaerobic countsincreased over values seen in TNB control ratsmostly because of noticeable predominance ofBacteroides vulgatus. Imipenem treated ratsshowed a high death rate (1 1 of 15) when com-

pared with TNB controls (1 of 15). Treatmentwith imipenem plus vancomycin also inducedan increase in Candida spp, but in contrastwith imipenem only, there was no growth ofanaerobes in stool samples from this group ofrats at the dilution used for the microbiologicalstudies. These results suggest that super-infection with Bacteroides vulgatus mayexacerbate the inflammatory lesions inducedby TNB.As Table III shows, enemas with amoxi-

cillin/clavulanic acid significantly reducedthe counts of both aerobic and anaerobicbacteria species in stools. When comparedwith cultures from TNB control rats, animalstreated with amoxicillin/clavulanic acid

showed a 50% reduction in colony formingunits of anaerobic species. In contrast with oralimipenem, amoxicillin/clavulanic acid enemas

did not induce qualitative changes in the faecalflora but only reduced colony counts.

DiscussionThe experiments described in this paper showthat treatment with certain broad spectrumantibiotics reduces the severity of the coloniclesions induced by intracolonic administrationof TNB. Acute colonic inflammation was notinfluenced by antibiotic treatment as shown bythe intracolonic release of eicosanoids at day 7after induction of colitis. The effect of broadspectrum antibiotics, however, was particularlymanifest during the chronic stage of TNBcolitis, as animals receiving antibiotic treat-ment rarely developed the characteristicpathological features seen in chronic colitissuch as colonic strictures with proximal boweldilatation, adhesions to surrounding tissues,fibrosis, and granulomas. Interestingly, duringthe chronic stage of colitis the intracolonicrelease of eicosanoids in antibiotic treated rateshad returned to baseline, showing that theinflammatory process was not active at thisstage. Taken together, these data suggestthat intestinal decontamination using widespectrum antibiotics inhibits the developmentof chronic inflammatory lesions. Instillation ofTNB seems to be responsible for the initialinflammatory stimulus and the luminal colonicflora influences the chronic response. Thenormal colonic flora seems to play a key part inthe progression of inflammatory coloniclesions to chronicity.Two different antibiotic associations could

prevent the development of chronic colitis,amoxicillin plus clavulanic acid and imipenemplus vancomycin. These drug regimens consistof chemically unrelated compounds that shareantimicrobial activity against a broad spectrumof bacteria species. It is therefore unlikely thatthe beneficial effect of the drugs resulted froma pharmacological action unrelated to theirantimicrobial activity. An interesting findingmade in this study was that the association ofimipenem plus vancomycin achieved a signifi-cant therapeutical effect in the prevention ofchronic colitis, whereas neither agent had anybeneficial effect when given independently.

TABLE III Faecalflora in TNB colitis

TNB

AmoxicillinlclavulanicNormal rats Control Imipenem Imipenem/vancomycin acid

Aerobic culture*7-78-7 79t 7-867-88 7 91-7 92 7 91-7-92 7 70-7-74Lactobacillus sppt E coli Candida sppt Candida sppt E coliStreptococcus spp Staphylococcus spp E coli E coli Streptococcus sppE coli Streptococcus spp Streptococcus spp Proteus mirabiis Proteus mirabilisAnaerobic culture8-10-8-30 7-907-92 8-22-8-50 <4§ 7-56-7-65Bacteroides spp Peptostreptococcus spp Bacteriodes sppt Bacteroides sppPeptostreptococcus spp Streptococcus intermedius Clostridium spp Clostridium spp

Clostridium spp Peptostreptococcus spp Peptostreptococcus sppBacteroides sppEubacterium spp

*Includes facultative anaerobic bacteria; tcolony forming units/g of dried faeces expressed as log,,; tpredominant organismamong the isolates; Sno anaerobes were isolated.

1095

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

1096 Videla, Vilaseca, Guarner, Salas, Treserra, Crespo, Antolin, Malagelada

Imipenem alone increased mortality. Thesefindings suggested that the drugs were notprimarily responsible for the anti-inflammatoryeffect, but that a broad spectrum bacterialdecontamination was required to improve thecolitis.The route of administration was critical to

achieve a therapeutic effect. Non-absorbableantibiotics such as imipenem and vancomycinwere effective when given orally or by enema,whereas intraperitoneal injections were noteffective. Only appropriate intracolonic con-centrations of the antimicrobial drugs wereassociated with the anti-inflammatory effect.Therapeutic plasma concentrations of theantimicrobials were irrelevant. Again, thesefindings preclude the possibility of an anti-inflammatory effect of the drugs unrelated totheir antibiotic activity.

Selective antimicrobial treatment againstGram negative bacilli (tobramycin), Grampositive enterococci (vancomycin) or anaerobicspecies (metronidazole) had no effect on thecolonic lesion scores. The association oftobramycin and vancomycin was also ineffec-tive. Imipenem, a drug active against a widerange of Gram negative and Gram positivespecies including some anaerobes, intensifiedthe colonic lesions and increased mortalityinduced by TNB. Our microbiological studiesshowed that treatment with imipenem onlyinduced overgrowth of Bacteroides spp,particularly Bacteroides vulgatus. These resultssuggest that superinfection with Bacteroidesvulgatus may exacerbate colonic lesionsinduced by TNB. Our findings are in agree-ment with previous studies by Onderdonket al,15 16 that extensively showed the implica-tion of Bacteroides vulgatus in the pathogenesisof carrageenin induced colitis in the guineapig. In contrast with our findings, however,these authors saw in the guinea pig modelthat pretreatment with metronidazole pre-vents the induction of colitis by carrageeninfeeding. 17The microbiological studies showed that

treatment with imipenem plus vancomycinor amoxicillin/clavulanic acid significantlyreduced the luminal content of bacteriaspecies. In stool specimens from rats treatedwith imipenem plus vancomycin, aerobic andanaerobic bacteria were almost absent and hadbeen replaced by overgrowth of Candida spp.Amoxicillin/clavulanic acid induced no impor-tant qualitative changes in the faecal flora, butreduced colony counts of anaerobes to 50%compared with controls. The fact that systemicantibiotics failed to prevent chronic inflam-matory colitis suggests that dissemination ofpathogens that could invade tissues is not acritical mechanism in the process of developingchronic lesions. Colitis induced by TNB doesnot evolve as a septic condition. Superinfectionof TNB induced injury by normal flora,however, may play an important part, becausewhen proliferation of the colonic flora isinhibited by appropriate intraluminal anti-biotics a remarkable therapeutic action oncolitis is achieved. Treatment with imipenemplus vancomycin was also effective when given

24 hours after the challenge with TNB,suggesting that the antibiotics enhanced repairmechanisms of established TNB inducedlesions. Thus, commensal bacteria seem to beresponsible for the chronicity of the inflam-matory response to TNB. This sort of bacteriacannot cross epithelial barriers and spreadbeyond the injured areas, but they canproliferate on the surface of the lesions andhamper the resolution of the healing process.The precise mechanism by which they do socannot be established from this study but ourdata clearly show that the presence of bacteriaand their products perpetuates the inflam-matory response. Perhaps, by sustaining theantigenic insult they stimulate the progressionof the colitis towards chronicity.The participation of luminal bacteria in the

pathogenesis of other inflammatory conditionsof the bowel has also been suggested in otheranimal models. Bacteria or bacterial productsseem to be critical to the induction ofmucosal ulcerations in the small bowel byindomethacin, as germ free rats developminimal lesions18 19 and antibiotics decreasethe severity of the ulcers.20 As mentioned,proliferation of anaerobes has been incrimin-ated in the pathogenesis of experimentalcolitis by chronic feeding of sulphated poly-saccharides such as carrageeninl4-17 or dextransulphate.2' This study provides significantevidence that the normal colonic flora plays akey part in the progression of inflammatorycolonic lesions to chronicity, and henceilluminates the potential relevance of this oftenoverlooked factor in human disease.

The study was supported by grant PB92-0733 from Direcci6nGeneral de Investigaci6n Cientifica y Tecnica, Madrid, Spain.

1 Boughton-Smith NK, Wallace JL, Morris GP, Whittle BJR.The effect of anti-inflammatory drugs in a chronic modelof inflammatory bowel disease in the rat.BrJ3Pharnacol 1988; 94: 65-72.

2 Morris GP, Beck PL, Herridge MS, Depew WT, SzewezukMR, Wallace JL. Hapten-induced model of chronicinflammation and ulceration in the rat colon.Gastroenterology 1989; 96: 795-803.

3 Wallace JL, MacNaughton WK, Morris GP, Beck PL.Inhibition of leukotriene synthesis markedly accelerateshealing in a rat model of inflammatory bowel disease.Gastroenterology 1989; 96: 29-36.

4 Allgayer H, Deschryver K, Stenson WF. Treatment with16,16'-dimethyl prostaglandin E2 before and after induc-tion of colitis with trinitrobenzene sulfonic acid in ratsdecreases inflammation. Gastroenterology 1989; 96:1290-300.

5 Rachmilewitz D, Simon PL, Schwartz LW, Griswold DE,Fondacaro JD, Wasserman MA. Inflammatory mediatorsof experimental colitis in rats. Gastroenterology 1989; 97:326-37.

6 Vilaseca J, Salas A, Guamer F, Rodriguez R,Malagelada J-R. Participation of thromboxane and othereicosanoid synthesis in the course of experimentalinflammatory colitis. Gastroenterology 1990; 98:269-77.

7 Vilaseca J, Salas A, Guamer F, Rodriguez R, Martinez M,Malagelada JR. Dietary fish oil reduces progression ofchronic inflammatory lesions in a rat model of colitis. Gut1990; 31: 539-44.

8 Grisham MB, Volkmer C, Tso P, Yamada T. Metabolismof trinitrobenzene sulfonic acid by the rat colon producesreactive oxygen species. Gastroenterology 1991; 101:540-7.

9 Pons L, Droy-Lefaix MT, Buerno L. Leukotreine D4participates in colonic transit disturbances induced byintracolonic administration of trinitrobenzene sulfonicacid in rats. Gaseroeneerology 1992; 102: 149-56.

10 Yamada T, Marshall 5, Specian RD, Grisham MB. A com-parative analysis of two models of colitis in rats.Gaseroenterologry 1992;- 102: 1524-34.

11 Little JR, Eisen HN. Preparation and characterization ofantibodies specific for the 2,4,6-t.rinitrophenyl group.Biochemistry 1966; 5: 3385-95.

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from

Role of intestinal microflora in chronic inflammation and ulceration of the rat colon 1097

12 Kunin S, Gallily R. Recognition and lysis of altered self cellsby macrophages, a modification of target cells by 2,4,6-trinitrobenzene sulfonic acid. Immunology 1983; 121:1711-8.

13 Sartor RB, Cromartie WI, Powell D, Schwab JH. Granulo-matous enterocolitis induced in rats by purified bacterialcell wall fragments. Gastroenterology 1985; 89: 587-95.

14 Onderdonk AB, Bartlett JG. Bacteriological studies ofexperimental ulcerative colitis. Am J Clin Nutr 1979; 32:258-65.

15 Onderdonk AB, Cisneros RL, Bronson RT. Enhancementof experimental ulcerative colitis by immunization withBacteroides vulgatus. Infect Immun 1983; 42: 783-8.

16 Onderdonk AB, Steeves RM, Cisneros RL, Bronson RT.Adoptive transfer of immune enhancement of experi-mental ulcerative colitis. Infect Immun 1984; 46: 64-7.

17 Onderdonk AB, Hermos JA, Dzink JL, Bartlett JG.

Protective effect of metronidazole in experimentalulcerative colitis. Gastroenterology 1978; 74: 521-6.

18 Robert A, Asano T. Resistance of germfree rats toindomethacin induced intestinal lesions. Prostaglandins1977; 14: 333-41.

19 Davis SW, Holt LC, Sartor RB. Luminal bacteria andbacterial polymers potentiate indomethacin-inducedintestinal injury in the rat. Gastroenterology 1990: 98:444A.

20 Kent TH, Cardelli RM, Stamier FW. Small intestinal ulcersand intestinal flora in rats given indomethacin. Am JfPathol 1969; 34: 237-45.

21 Okayasu I, Hatakeyama S,. Yamada M, Ohkusa T, InagakiY, Nakaya R. A novel method in the induction of reliableexperimental acute and chronic ulcerative colitis in mice.Gastroenterology 1990; 98: 694-702.

on March 25, 2021 by guest. P

rotected by copyright.http://gut.bm

j.com/

Gut: first published as 10.1136/gut.35.8.1090 on 1 A

ugust 1994. Dow

nloaded from