INFEcrION AND IMMUNITY, Apr. 1992, p. 1295-13010019-9567/92/041295-07$02.00/0Copyright © 1992, American Society for Microbiology

Interleukin-6 Response of Epithelial Cell Lines toBacterial Stimulation In Vitro

SPENCER HEDGES,* MAJLIS SVENSSON, AND CATHARINA SVANBORG

Department of Clinical Immunology, University ofLund, S-223 46 Lund, Sweden

Received 2 July 1991/Accepted 7 January 1992

This study demonstrated that epithelial cell lines secrete interleukin-6 (IL-6) in response to stimulation withgram-negative bacteria. Human epithelial cell lines of urinary tract origin (A-498 and J82) and of intestinalorigin (HT-29 and Caco-2) were analyzed for the secretion of IL-6 by using the B9 bioassay. The supernatantsfrom cells maintained with culture medium were used to assess the constitutive production of IL-6. Thesupernatants from cells exposed to Escherichia coli strains, lipopolysaccharide, lipid A, and isolated fimbriaewere used to quantitate the IL-6 response to these stimulants. The urinary tract epithelial cell lines were foundto constitutively secrete IL-6. The IL-6 activity in the supernatants of the bladder cell line (J82) increased aboveconstitutive levels after 2 h of stimulation by most of the bacterial strains tested. The IL-6 activity in thesupernatants of the kidney line (A-498) accumulated at a constant rate over the 24-h assay period. The role ofbacterial adherence for the induction of IL-6 production was investigated by comparing the responses torecombinant E. coli strains expressing different fimbriae. In addition, isolated P and S fimbriae wth andwithout the receptor binding domain were also used as stimulants. The IL-6 activity in the supernatants of thebladder cell line increased after exposure to bacteria and bacterial products regardless of their adhesiveproperties. In contrast, the kidney cell line was stimulated to secrete significantly more IL-6 by adheringbacteria and by adhesin-positive P fimbriae than by nonadhering bacteria or adhesin-negative P fimbriae. TheS-fimbrial preparations had no specific effects on the IL-6 activity of the cell supernatants. These results are

consistent with our hypothesis that epithelial cells can be a major source of IL-6 when stimulated by bacteriaand that the adhesive properties of the bacteria can influence this response.

Interleukin-6 (IL-6) activates hepatocytes to secreteacute-phase reactants (7), stimulates mucosal B lympho-cytes (2), and acts as an endogenous pyrogen (14). Deliber-ate instillation of Eschenchia coli bacteria into the humanbladder activated an IL-6 response which could be measuredin urine a few hours later (10). The IL-6 levels also increasedin urine and serum samples during episodes of naturalurinary tract infection (13). A similar pattern of IL-6 secre-

tion was observed during experimental infection in mice,where urinary IL-6 levels increased within minutes afterintravesical E. coli challenge (6, 12). On the basis of thispattern of IL-6 secretion and its spectrum of biologicalactivities, we proposed IL-6 as a likely messenger from localsites of infection or injury (such as mucosal surfaces) tosystemic compartments which respond to these events.The inflammatory response in the urinary tract depends on

the properties of the infecting E. coli strain. The pyelo-nephritogenic E. coli clones, which are characterized by theexpression of P fimbriae (17), cause higher levels of cyto-kine-mediated reactions such as fever and the production ofC-reactive protein than other E. coli clones do (5). The Pfimbriae are not only markers of the inflammatogenicity ofthese clones but are directly involved in the induction of theresponse. Isolated P fimbriae elicited a urinary IL-6 responsein mice provided that the P fimbriae retained the receptor-binding adhesin (20).The rapid kinetics of the urinary IL-6 response, and the

ability of mucosally administered bacteria and bacterialproducts to induce such a response, suggested that themucosal lining was the first site of IL-6 prodi, tion. Prelim-inary studies from our group have suggested ,hat epithelial

* Corresponding author.

cells of urinary tract origin can be activated by bacteria toproduce IL-6 (11). The aim of the present study was todetermine whether epithelial cell lines in culture produceIL-6 in response to bacterial challenge and to analyze theinfluence of bacterial adhesins on IL-6 secretion by epithelialcells.

MATERIALS AND METHODS

Cell lines. The cell lines A-498 (ATCC HTB44), J82(ATCC HTB1), HT-29 (ATCC HTB38), and Caco-2 (ATCCHTB37) were grown in RPMI 1640 (Flow Laboratories,Irvine, United Kingdom), supplemented with 0.05 mg ofgentamicin per ml, 2 mM glutamine, and 5% fetal calf serum.For each experiment, the cell lines were taken from liquidnitrogen and cultured in 50-ml cell culture flasks (Nunc,Roskilde, Denmark) until confluent. The cells were trypsintreated (0.25%; Sigma Chemical Co., St. Louis, Mo.) for 10min at room temperature. The trypsin was removed bycentrifugation at 500 x g for 10 min. The supernatant was

discarded, and the cells were resuspended and diluted to 5 x

104/ml in media. The cell suspension was added to 96-wellmicrotiter plates (0.1 ml per well; Nunc) and allowed to growto confluency before use (1 to 2 days). Immediately beforecell stimulation, the medium in each well was aspirated andreplaced by 0.1 ml of fresh medium.

Induction of IL-6 secretion. IL-6 secretion was induced bythe addition at time zero of 10 ,ul of a stimulatory agent to thetest wells. Supernatants were removed from triplicate wellsfor each stimulus at the four experimental time points. Thetriplicates were combined into one sample and immediatelyfrozen at -20°C. A control sample obtained from unstimu-lated cells was treated similarly. Samples from a singleexperiment were analyzed in the same IL-6 assay.

1295

Vol. 60, No. 4

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

1296 HEDGES ET AL.

Bacteria. Bacteria differing in adherence characteristicswere used to stimulate the epithelial cells. We selected as acontrol stimulant E. coli Hu734 which had elicited an IL-6response in the mouse urinary tract (6, 12). This lac mutantof the wild-type strain E. coli GR12 (27) (serotype 075:K5:H-, Hly-, ColV+) expressed adhesins mediating the attach-ment to two types of receptors, the globoseries of glycolipidsand mannose-containing glycoproteins. The wild-type strainE. coli C1212 (serotype 06:K2:H1) expressed P fimbriae ofthe F71 and F72 antigen types, specific for the globoseries ofglycolipid receptors. E. coli 536 (serotype 06:K15:H31),isolated from a urinary tract infection (9), was the S-fimbrialprototype strain from which the sfa sequences were cloned(23). Staphylococcus saprophyticus 998 was previously iso-lated from a patient with a urinary tract infection.The transformants in E. coli 83972 had previously been

constructed for colonization of the human urinary tract (10).Intravesical instillation of these strains had been shown toactivate a urinary IL-6 response. The wild-type recipientstrain was phenotypically negative for both mannose- andGalao1-4Galp3-specific adhesins and did not attach to humanuroepithelial cells in vitro. It was genotypically negative forthe pap DNA sequences but positive for the pil DNAsequences. E. coli Hu1061 had received the plasmidpRHU845 with the pap DNA sequences encoding fimbriaeand adhesins specific for the globoseries of glycolipids. E.coli Hu1053 had received the plasmid pRHU960, with thepilDNA sequences encoding type 1 fimbriae with mannose-specific adhesins. E. coli Hu1193 had received the plasmidpREG without insertion sequences.The bacteria were maintained on tryptic soy agar plates

with the appropriate antibiotic. For epithelial cell stimula-tion, the bacteria were cultured overnight in Luria broth andfixed by formalin treatment. Bacterial cultures were treatedwith 0.5% formalin at room temperature for 4 h and leftstanding overnight at 4°C. Immediately before use thetreated bacteria were centrifuged at 4,000 x g for 15 min,resuspended in phosphate-buffered saline (PBS), recentri-fuged, and resuspended in PBS to a final concentration of10 /ml. The binding properties of the strains were examinedbefore epithelial cell stimulation.

Isolation of fimbriae. Adhesin-positive (adh+) and adhesin-negative (adh-) fimbriae were kindly provided by K. Jannand H. Hoschutzky (Max Planck Institute, Freiburg, Ger-many). The procedure used for the isolation of these fimbriaewas described previously (15). Briefly, fimbriae were elutedby heating to 65°C for 30 min and then precipitated withammonium sulfate. To remove contaminating lipids, thefimbriae were suspended in 50% ethanol and precipitatedwith LiCl (250 mM). After centrifugation, the pellet wasresuspended in 10 mM Tris (pH 7.8). Residual lipopolysac-charide (LPS) was removed by deoxycholate (0.5%) treat-ment at 60°C for 30 min. The fimbriae were reprecipitatedwith LiCl (250 mM), dissolved in a small amount of water,and stored at -70°C. To obtain fimbriae devoid of theadhesin, a fimbrial solution (10 mg/ml) in 10 mM Tris (pH 8)was treated with Zwittergent 3-16. The mixture was heatedto 80°C for 30 min, causing the dissociation of the fimbriaeinto rods and adhesins. The fimbriae devoid of adhesins werecollected by precipitation with LiCl (250 mM) and centrifu-gation.P fimbriae of serotype F71 were obtained from E. coli

21624, a recombinant in HB101 with pDAL 201B, carryingas an insert the chromosomal pap DNA region from thewild-type uropathogenic strain E. coli C1212. S fimbriaewere obtained from a recombinant in E. coli HB101 with

pANN 801-13, carrying as an insert the sfa DNA sequencesencoding S fimbriae and adhesins specific for NeuAcat2-3Galfrom the wild-type strain E. coli 536 (23). Fimbriae wereused as a stimulant at a final concentration of 10 p,g/ml.

Bacterial binding properties. The known adhesin expres-sion of the wild-type bacteria, transformant bacteria, andfimbrial preparations was confirmed by hemagglutinationand receptor-specific tests. The adhesins of E. coli 734 andC1212 caused P-blood-group-dependent, mannose-resistantagglutination of human erythrocytes (A1P1, A1P-) and ag-glutinated Galaol-4GalpI latex beads. Both strains also causedmannose-sensitive agglutination of guinea pig erythrocytes.E. coli 536 agglutinated human and bovine erythrocytes at4°C in a P-blood-group-independent manner. The S. sapro-phyticus strain did not hemagglutinate human or other eryth-rocytes under the conditions tested. adh+ P fimbriae agglu-tinated human and sheep erythrocytes in the presence ofmannose and agglutinated Galal-4Galp latex beads. adh+ Sfimbriae agglutinated human, bovine, and sheep erythro-cytes in the presence of mannose at 4°C but not the Galtl-4Galp latex beads. adh- fimbriae of either type did notagglutinate human erythrocytes or the Galal-4Galp latexbeads.

Bacterial adherence. For adherence testing, bacteria werecultured on tryptic soy agar with or without the antibioticselection required to maintain the plasmids or overnight instatic Luria broth without antibiotics to induce type 1fimbriae (as appropriate). Bacterial attachment was tested onthe cultured epithelial cell lines by using an earlier method(26) with slight modifications. Briefly, confluent cells (5 daysof culture) were scraped from the culture flask with a rubberpoliceman and resuspended in PBS. The cells were washedfree of culture medium by three cycles of centrifugation andresuspension in PBS. After the third washing, the cellconcentration was adjusted to 106/ml. A 0.1-ml volume of thecell suspension was mixed with 0.1 ml of the bacteria(107/ml), and the final volume was adjusted to 1 ml with PBS.The mixture was incubated for 1 h at 37°C with end-over-endrotation. Unattached bacteria were eliminated by repeatedwashing in PBS. The number of attached bacteria wascounted by interference contrast microscopy. Adherence isgiven as the mean number of bacteria attached to 40 epithe-lial cells.Other stimulants. LPS from E. coli 0111 was purchased

from Difco, Detroit, Mich. Lipid A from E. coli 0111 LPSwas kindly provided by I. Mattsby-Baltzer, Department ofBacteriology, Goteborg University, Goteborg, Sweden. LPSand lipid A were used at a final concentration of 10 ,ug/ml.The concentrations of endotoxin in the samples and in thetissue culture medium were determined by using the Limuluslysate assay (Hassle, Molndal, Sweden).

IL-6 assay. The cell line B13.29 which is dependent onIL-6 for its growth was described previously (18). For IL-6determinations, the more sensitive subclone B9 was used (1,14). The B9 cells were harvested from the tissue cultureflasks, and 5,000 cells per well were seeded into microtiterplates (Nunc) containing dilutions of sample or recombinanthuman IL-6 (as a standard) (4) and Iscoves modified Dulbec-co's medium (Flow Laboratories), supplemented with 50 ,iM2-mercaptoethanol, 5% fetal calf serum, and gentamicin (0.1mg/ml), in a total volume of 0.2 ml. [3H]thymidine was addedbetween 68 and 72 h of culture, and the incorporatedradioactivity was measured and compared with a standardcurve. The standard curve was prepared by using stockrecombinant human IL-6 (8,000 U/ml; 1 U = concentrationof IL-6 required for half-maximal thymidine incorporation).

INFEC-r. IMMUN.

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

CELL LINE IL-6 RESPONSE TO BACTERIAL STIMULATION 1297

800

E 700U)E 600

__ 500

m 400

0 3000.CD 200

CD 100-J n

6 1 2 1 8

Time (hours)FIG. 1. Constitutive IL-6 production by epithelial cells. Meansstandard errors of six to nine separate experiments are shown.

Symbols: El, A-498; 0, J82.

The activities of each sample are given in units per milliliteras compared with the standard curve. The assay specificityfor IL-6 was tested on random samples by inhibition by usinganti-IL-6 monoclonal antibody 6B4 (immunoglobulin Gl)(kindly provided by J. Van Snick) (28). Supernatants fromthe 6B4 cell line abolished the IL-6-dependent growth of theB9 cells in response to the recombinant standard and to thesamples.The IL-6 in the epithelial cell supernatants induced by the

stimulant was calculated as follows: (IL-6 in the sample fromthe stimulated cells) - (IL-6 activity from the unstimulatedcontrol cells from the same time point) - (B9 cell prolifera-tion induced by the stimulant).

Statistics. The Mann-Whitney U test and the Wilcoxonrank sum test were used where appropriate. Differenceswere considered significant for P values of less than 0.05.

RESULTS

Epithelial cell lines of urinary tract origin secrete IL-6. Theepithelial cell lines were grown to confluency in 96-wellplates. At time zero, the culture medium was removed andreplaced with new medium or medium and stimulant. Thecells were incubated for 24 h, with samples taken at 0, 2, 6,and 24 h. The IL-6 activity in the supernatants of cellsmaintained with culture medium was used as a measure ofthe constitutive production of IL-6 by these cells. Underthese conditions, IL-6 activity was detected in the superna-tants of the A-498 and J82 but not the Caco-2 and HT-29 celllines (Fig. 1). The IL-6 activity of the J82 cell line reached a

maximal level by 6 h, with a mean value of 644 U/ml (meanof nine experiments). The IL-6 activity of the A-498 cellsreached a maximal level after 24 h, with a mean of 254 U/ml(mean of six experiments). The IL-6 activity in the supema-tants of the J82 cell line was significantly higher than that ofthe A-498 cells at all time points.

Bacteria stimulate epithelial cell IL-6 secretion above theconstitutive levels. The ability of E. coli bacteria to stimulateepithelial cell IL-6 production was first tested with E. coliHu734. This strain was known to stimulate IL-6 secretion inmice (6, 12). E. coli Hu734 was formalin killed, and 10 RI ofthe bacterial suspension was added to the cells (final con-

centration, 108/ml). The stimulus was maintained for 24 hwith samples taken at 0, 2, 6, and 24 h. The IL-6 activity inthe supernatants of the cells exposed to bacteria was subse-quently compared with the constitutive IL-6 activity from

the same experiments and sample times. After exposure toE. coli Hu734, the IL-6 activity in the supernatants from theA-498 and J82 cells increased (Fig. 2). At 2 h, the IL-6 levelsof the J82 cells were significantly higher than the constitutivelevels in five of six separate experiments (P < 0 05,Wilcoxon rank sum test). The IL-6 activity in the superna-tants from the A-498 cells was significantly higher than theconstitutive levels in the 24-h sample (P < 0.05, Mann-Whitney U test). The HT-29 and Caco-2 cell lines were notstimulated by E. coli Hu734 to secrete IL-6.

Bacterial adhesion and IL-6 secretion. The influence ofbacterial adhesion on the stimulation of IL-6 secretion wasanalyzed. Wild-type bacteria and recombinant E. coli strains(which differed in adhesin expression) were used to stimulatethe epithelial cell lines under conditions analogous to thoseused for E. coli Hu734. The ability of these bacteria toadhere to the J82 and A-498 cell lines is shown in Table 1. E.coli C1212, Hu734, and Hu1053 attached equally well to boththe kidney and bladder epithelial cells. The IL-6 levels werecalculated as the difference between the IL-6 from thesample and the constitutive IL-6 from the same day and timeand from the effect of the stimulant on the B9 cells. Table 1shows the IL-6 activity at the time of maximum difference.The IL-6 response of the J82 bladder cell line to bacterialstimulation appeared to be independent of bacterial adhe-sion. The highest IL-6 levels (>200 U/ml, above the consti-tutive levels) were detected after stimulation with the S-fim-briated strain E. coli 536, the P- and type 1-fimbriated strainE. coli Hu734, the type 1-fimbriated strain E. coli Hu1053,and the gram-positive strain S. saprophyticus 998. E. coliHu1061, Hu1193, and C1212 induced low IL-6 activity (<100U/ml). In contrast, the highest IL-6 levels from the A-498cell line were detected after stimulation with the strainsexpressing either both P and type 1 fimbriae or solely type 1fimbriae (E. coli Hu734, C1212, and Hu1053). Lower levelsof IL-6 were detected after exposure to the S-fimbriated oradhesin-negative E. coli strains and the S. saprophyticusisolate. The results are compiled for the adhering versusnonadhering strains by using the IL-6 data after subtractionof the constitutive and control IL-6 levels (Fig. 3). Theadhering strains stimulated significantly higher IL-6 levels at24 h in the A-498 cells than in the nonadhering strains. Nosuch difference was observed for the J82 cells. The intestinalcell lines Caco-2 and HT-29 did not secrete IL-6 in responseto any of the bacteria tested. They were considered to benonresponders and were omitted from further experiments.

Stimulation of IL-6 secretion by bacterial components. LPSis known to activate IL-6 production by nonepithelial cells(6). We first analyzed whether LPS contamination of thetissue culture medium might be a stimulant for the constitu-tive IL-6 production by the epithelial cells. The LPS con-centration in the medium was <100 ng/ml. LPS and lipid Awere subsequently added to the cells at concentrations of0.1, 1.0, and 10.0 ,ug/ml, and the 24-h supernatants weretested for the presence of IL-6. The IL-6 levels of the A-498cells only increased above constitutive levels after exposureto LPS or lipid A at 10.0 ,ug/ml. In contrast, the IL-6 activityin the supernatants of the J82 cells increased above theconstitutive levels in a dose-dependent fashion, with 10.0,ug/ml giving the highest response. The responses to LPSstimulation are shown in Fig. 4. In all experiments, lipid Ainduced IL-6 levels and secretion kinetics similar to thoseinduced after LPS stimulation (data not presented). Thelevels of IL-6 produced by both cell lines were significantlyhigher in the 24-h sample than the constitutive levels were (Pc 0.05).

VOL. 60, 1992

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

1298 HEDGES ET AL.

E600-

'E 400-

CS1200-

CO 12000

CD 800

=600-T

200.r

0 6 12 18 24Time (hours)

FIG. 2. EPithelial Cell IL-6 PrOdUCtiOn after StimUlatiOn With E. coli HU734 (108/ml). MeanS ± Standard errOrS frOm three tO SiX SeParateeXPerimentS are ShOWn. SYmbOlS: C1, A-498 COnStitUtiVe; *, A-498 StimUlated; O, J82 COnStitUtiVe; *, J82 StimUlated.

Isolated P fimbriae have previously been shown to stimu-late IL-6 secretion in mice (20). Since the E. coli strains usedhere either expressed several fimbrial types or overex-pressed fimbriae of a single receptor specificity, the questionof whether P fimbriae could trigger a parallel IL-6 responsein the human cell lines could not be directly examined byusing whole bacteria as the stimulant. Therefore, the abilityof fimbriae to trigger epithelial IL-6 production was furtheranalyzed, by using isolated P and S fimbriae either with(adh+) or without (adh-) the receptor binding domain. TheIL-6 activity in the J82 cell supernatants was not significantlyincreased compared to constitutive levels by any fimbrialpreparation (Fig. 5). In contrast, stimulation of the A-498 cellline with adhesin-positive P fimbriae resulted in significantlyhigher IL-6 activity after 24 h of stimulation (P c 0.01) (Fig.5). Neither the adhesin-negative P fimbriae nor the S-fimbrialpreparations had such an effect.

DISCUSSION

In this study, epithelial cell lines were shown to produceIL-6 during in vitro culture. IL-6 was secreted constitutivelyby epithelial cell lines from the urinary tract (J82 and A-498).These cell lines also increased their IL-6 secretion aboveconstitutive levels in response either to whole, killed bacte-ria or to bacterial components. This confirmed our prelimi-nary observation that epithelial cells can participate in thecytokine network (11) and parallels a growing body ofevidence that the epithelial cell does not merely provide amechanical barrier to infection but is an essential part of thecellular immune response. These results are consistent withour hypothesis that urinary tract epithelial cells may be amajor source of the IL-6 produced in response to mucosalbacterial infection and that bacterial adherence may influ-ence this response, especially in the kidney.

TABLE 1. Bacterial efficiency of inducing epithelial cell IL-6 secretionA-498 cell line J82 cell line

Bacterial strain(adhesin expression) Peak IL-6a Time of peak" Bacterial Peak IL-6a Time of peak' Bacterial

(U/ml) (h) adherencec (U/ml) (h) adherencec

E. coliC1212 (P, type I) 458 24 35 74 6 21Hu734 (P, type I) 952 24 25 222 2 10536 (S) 95 6 0 409 2 0Hu1053 (type I) 371 24 100 211 6 100Hu1193 96 6 0 78 2 0Hu1061 (P) 46 6 1 63 2 1

S. saprophyticus 998 89 24 0 285 2 0a Peak IL-6 levels induced by each bacterial strain above constitutive secretion levels.b Time after exposure to bacteria.c Mean number of bacteria adhering to each cell line.

INFECF. IMMUN.

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

CELL LINE IL-6 RESPONSE TO BACTERIAL STIMULATION 1299

CD

c

C)._-

Q

0.0-C

cD-J

6005004003002200:100

0600 -

500.'400:300

200100

010 6 12

Time (hours)18

FIG. 3. Epithelial IL-6 production above constitutive levels afterstimulation by adhering or nonadhering strains (108/ml). Means --

standard errors from five to seven separate experiments are shown.Symbols: *, A-498 exposed to adhering strains; El, A-498 exposedto nonadhering strains; 0, J82 exposed to adhering strains; 0, J82exposed to nonadhering strains.

Previous observations of patients with urinary tract infec-tions, and of experimentally infected mice (6, 10, 12),suggested that mucosal surfaces are a major source of IL-6.The urinary IL-6 activity in both humans and mice increasedfrom undetectable levels before infection to levels in therange of 20 to 500 U/ml during infection. The elevation ofurinary IL-6 levels was not always accompanied by a serumIL-6 response. This dissociation between local and systemicIL-6 levels gave rise to the concept that IL-6 was locallysecreted but did not demonstrate which cell type was secret-ing the IL-6. We considered it likely that epithelial cells

1000

800

E 600

* 400

_ 200

CDcu1000

a 800600

o 400.J

2UU

produced the IL-6, since they dominate the surface of themucosa and since local IL-6 secretion preceded the influx ofgranulocytes and monocytes to the site of infection. Theresults presented here support this contention. The rele-vance of the transformed epithelial cell lines as a model formucosal IL-6 production may be questioned. The use of celllines for studies of human urinary tract infections wasinevitable since, for ethical reasons, biopsy specimens fromacutely infected urinary mucosa could not be made availablefor study. Because of the nature of transformed cells, theresults need to be confirmed in normal epithelial cells.The urinary tract epithelial cell lines constitutively pro-

duced IL-6. At this point, it is not clear whether thisproduction was due to the transformed nature of the cells orto stimulants present in the medium. The LPS content of themedium was below the threshold for stimulation of the A-498cells but in the range which marginally stimulated the J82cells. The LPS concentration was not sufficient to triggerIL-6 levels comparable to the constitutive production. Al-though IL-6 has been detected in transformed epithelial celllines (8), transformation per se was not solely responsible forthe response seen here. Under the assay conditions usedhere, neither the HT-29 nor the Caco-2 cell line producedIL-6. However, the Caco-2 cells have been reported tosecrete IL-6 under different experimental conditions (22a).The IL-6 production by the intestinal epithelial cells maydepend on the subclone used.The kinetics of the constitutive IL-6 production differed

between the cell lines. The IL-6 production increased morerapidly after a medium change in the J82 cells than it did inthe A-498 cells. This pattern was also observed after micro-bial stimulation. These results have been confirmed byimmunostaining of the epithelial cells with monoclonal anti-IL-6 antibodies. The fraction of J82 cells staining for IL-6increased at 2 h, while the peak staining for the A-498 cells

0 6 12 18 24

Time (hours)FIG. 4. Epithelial cell IL-6 production after stimulation with LPS (10 ,ug/ml). Means + standard errors from five to seven separate

experiments are shown. Symbols: El, A-498 constitutive; *, A-498 stimulated; 0, J82 constitutive; 0, J82 stimulated.

VOL. 60, 1992

I

.L

i

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

1300 HEDGES ET AL.

1000 1A750

Cu

*aE 50O0cn

25O0

.0-0 I

1 000

cnC: 75000._

c 2500O--

B

0 6 12 18 24Time (hours)

FIG. 5. Epithelial IL-6 secretion after stimulation by adhesin-positive or adhesin-negative P and S fimbriae. Results are the means ±

standard errors from three to four separate experiments. (A) A-498; (B) J82. Symbols: A, constitutive; *, F7 fimbriae, adhesin positive; 0,

F7 fimbriae, adhesin negative; 0, S fimbriae, adhesin positive, 0, S fimbriae, adhesin negative.

occurred later (la). Furthermore, IL-6 did not consistentlyaccumulate in the supernatants of the J82 cells as it did forthe A-498 cells. The decline in IL-6 levels, after the earlypeak responses, to near constitutive levels at 24 h suggestedthat the J82 cells may reabsorb IL-6, possibly in an autocrinemanner.The ability to stimulate IL-6 production differed between

the bacterial strains used. E. coli Hu734 and the otheradhering E. coli strains enhanced IL-6 production by thekidney cell line A-498 significantly more than the otherbacteria did. The levels of IL-6 correlated with the degree ofadherence rather than with the receptor specificity of thefimbrial type expressed by a strain. The three strains whichtriggered the highest IL-6 levels in the A-498 supernatants allexpressed type 1 fimbriae, and two of them also expressed Pfimbriae. The enhanced IL-6 response to attaching bacteriamay reflect the approximation of stimulatory bacterial com-ponents to the cell surface rather than an activation viaadherence per se. This question was addressed by exposingthe epithelial cells to fimbrial preparations with or withoutthe receptor binding domain. The P fimbriae were shown toactivate the cells. These results are consistent with previousobservations in the mouse, where P-fimbrial preparationselicited a urinary IL-6 response (20). The dependence on theadhesin in this reaction suggested that the binding to thegloboseries of glycolipids can activate the receptor-bearingcell to secrete IL-6. The inability of S fimbriae to stimulateIL-6 secretion in this and earlier studies does not reflect ageneral defect in their ability to trigger cellular responses. Sfimbriae cause the release of mediators like leukotriene B4from human polymorphonuclear leukocytes (22). We havespeculated that the difference between P and S fimbriae isdue to the molecular nature of the receptors. Type 1 and Sfimbriae bind to secreted glycoproteins (24, 25). In contrast,the globoseries of glycolipids are integral membrane compo-

nents. It is possible that such membrane association isrequired for the induction of IL-6 secretion.LPS has not previously been shown to induce epithelial

cell IL-6 production. LPS, however, is known to triggercytokine production, including tumor necrosis factor, IL-1,and granulocyte-macrophage colony-stimulating factor, inother cell types such as monocytes and endothelial cells (3,16, 21). The epithelial cell cytokine response to LPS wasdifferent from those of other cell types in that no tumornecrosis factor production was observed (25a). This mayindicate that epithelial cells have a more restricted cytokineresponse window than other cell types involved in cellularimmune responses. It should also be emphasized that theadh+ P fimbriae contained small but detectable amounts ofLPS (0.7%) (20). P fimbriae and lipid A have previously beenshown to interact synergistically to induce mucosal polymor-phonuclear leukocyte influx in mice (19). Thus, the directstimulatory effect of the P-fimbrial adhesin could be claimedto be secondary to, or synergistic with, the effect of LPS forIL-6 production. However, the amount of LPS present in theadh+ P-fimbrial preparation was not stimulatory for theA-498 cells. Furthermore, evidence for LPS-independentP-fimbrial IL-6 activation (via the adhesin) was obtained inLPS-nonresponding mice which are refractoxy to the effectsof LPS (20). However, in the in vivo situation, it is likely thatIL-6 production is a combined effect of the fimbriae andLPS.The finding that epithelial cells participate in the network

of cytokine responses provides a new aspect to mucosalimmunity (11). It has become established that mucosallylocated microbial antigens can signal homing mucosal Blymphocytes to proliferate and/or can mature to secreteimmunoglobulin A (2). Cytokines have the same effect on Blymphocytes in vitro, especially in LPS-driven cultures. Thepresent study provides an explanation for how a bacterial

I

I

INFECT. IMMUN.

0

I

I

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from

CELL LINE IL-6 RESPONSE TO BACTERIAL STIMULATION 1301

stimulus in the lumen can influence the human mucosalimmune system in the subepithelial compartment as well as

at distant sites. As a result of microbial activation of theepithelial cells from the lumenal side, cytokines may besecreted both apically and basolaterally. Locally releasedcytokines may act on lymphocytes and other cells in themucosal compartment and provide the local signal requiredto expand the homing immunoglobulin A-committed B-cellpopulation. Cytokines which reach the bloodstream mayactivate host responses at distant sites and explain thesymptoms of infection.

ACKNOWLEDGMENTS

We thank Lucien Aarden and Cees van Kooten for the generousgifts of human recombinant IL-6 and the B9 cell line and Richardand Sheila Hull for development of the transformant strains.These studies were supported by the medical faculty at the

University of Lund, The Swedish Medical Research Council (grant7934), and the Osterlund, Tesdorph, and Crawford foundations.

REFERENCES1. Aarden, L. A., E. R. de Groot, 0. L. Schaap, and P. M.

Lansdorp. 1987. Production of hybridoma growth factor byhuman monocytes. Eur. J. Immunol. 17:1411-1416.

la.Agace, W., et al. Unpublished data.2. Beagley, K. W., J. H. Eldridge, F. Lee, H. Kiyono, M. P.

Everson, W. J. Koopman, T. Hirano, T. Kishimoto, and J. R.McGhee. 1989. Interleukins and IgA synthesis. Human andmurine interleukin 6 induce high rate IgA secretion in IgA-committed B cells. J. Exp. Med. 169:2133-2148.

3. Beutler, B., J. Mahoney, N. LeTrang, P. Pekala, and A. Cerami.1985. Purification of cachectin, a lipoprotein lipase-supressinghormone secreted by endotoxin induced RAW 264.7 cells. J.Exp. Med. 161:984-995.

4. Brakenhoff, J. P. J., E. R. de Groot, R. F. Evers, H. Pannekoek,and L. Aarden. 1987. Molecular cloning and expression ofhybridoma growth factor in Escherichia coli. J. Immunol.139:4116.

5. de Man, P., U. Jodal, and C. Svanborg-Eden. 1992. Dependenceamong host response parameters used to diagnose urinary tractinfection. J. Infect. Dis. 163:331-335.

6. de Man, P., C. van Kooten, L. Aarden, I. Engberg, and C.Svanborg Eden. 1989. Interleukin-6 induced at mucosal surfacesby gram-negative bacterial infection. Infect. Immun. 57:3383-3388.

7. Gauldie, J., C. Richards, D. Harnish, P. Lansdorp, and H.Baumann. 1987. Interferon 121B-cell stimulatory factor type 2shares identity with monocyte-derived hepatocyte-stimulatoryfactor and regulates the major acute phase response in livercells. Proc. Natl. Acad. Sci. USA 84:7251.

8. Grace, H., W. Wong, and D. V. Goeddel. 1986. Tumour necrosisfactors a and P inhibit virus replication and synergize withinterferons. Nature (London) 323:819.

9. Hacker, J., G. Schmidt, C. Hughes, S. Knapp, M. Marget, andW. Goebel. 1985. Cloning and characterization of genes in-volved in the production of mannose-resistant, neuraminidase-susceptible (X) fimbriae from a uropathogenic 06:K15:H31Escherichia coli strain. Infect. Immun. 47:434 440.

10. Hedges, S., P. Anderson, G. Lidin-Janson, and C. Svanborg.1991. Interleukin-6 response to deliberate gram-negative colo-nization of the human urinary tract. Infect. Immun. 59:421-427.

11. Hedges, S., P. de Man, H. Linder, C. van Kooten, and C.Svanborg-Ed6n. 1990. Interleukin-6 is secreted by epithelialcells in response to Gram-negative bacterial challenge, p. 144-

148. In T. MacDonald (ed.), Advances in mucosal immunology.International Conference of Mucosal Immunity. Kluwer, Lon-don.

12. Hedges, S., H. LUnder, P. De Man, and C. Svanborg-Eden. 1990.Cyclosporin A dependent, nu independent, mucosal interleu-kin-6 response to Gram-negative bacteria. Scand. J. Immunol.31:335-343.

13. Hedges, S., K. Stenqvist, I. Brolin, J. Martinell, T. Sandberg,and C. Svanborg. Submitted for publication.

14. Helle, M., J. P. J. Brakenhoff, E. R. de Groot, and L. A. Aarden.1988. Interleukin 6 is involved in interleukin-1 induced activi-ties. Eur. J. Immunol. 18:957-959.

15. Hoschutzky, H., F. Lottspeich, and K. Jann. 1989. Isolation andcharacterization of the a-galactosyl-1,4-p-galactosyl-specific ad-hesin (P adhesin) from fimbriated Escherichia coli. Infect.Immun. 57:76-81.

16. Jirik, F. R., T. J. Podor, T. Hirano, T. Kishimoto, D. J.Loskutoff, D. A. Carson, and M. Lotz. 1989. Bacterial lipopoly-saccharide and inflammatory mediators augment IL-6 secretionby human endothelial cells. J. Immunol. 142:144.

17. Klemm, P., I. 0rskov, and F. 0rskov. 1982. F7 and type 1-likefimbriae from three Escherichia coli strains isolated from uri-nary tract infections: protein, chemical, and immunologicalaspects. Infect. Immun. 36:462-468.

18. Lansdorp, P. M., L. A. Aarden, J. Calafat, and W. P. Zeijle-maker. 1986. A growth factor dependent B-cell hybridoma.Curr. Top. Microbiol. Immunol. 132:105-112

19. Linder, H., I. Engberg, I. M. Baltzer, K. Jann, and C. Svanborg-Eden. 1988. Induction of inflammation by Escherichia coli onthe mucosal level: requirement for adherence and endotoxin.Infect. Immun. 56:1309-1313.

20. Linder, H., I. Engberg, H. Hoschuitzky, I. Mattsby-Baltzer, andC. Svanborg. 1991. Adhesion-dependent activation of mucosalinterleukin-6 production. Infect. Immun. 59:4357-4362.

21. Loppnow, H., P. Libby, M. Freudenberg, J. Krauss, J. Weckes-ser, and H. Mayer. 1990. Cytokine induction by lipopolysaccha-ride (LPS) corresponds to lethal toxicity and is inhibited bynontoxic Rhodobacter capsulatus LPS. Infect. Immun. 58:3743-3750.

22. Marre, R., J. Hacker, W. Henkel, and W. Goebel. 1986. Contri-bution of cloned virulence factors from uropathogenic Esche-richia coli strains to nephropathogenicity in an experimental ratpyelonephritis model. Infect. Immun. 54:761-767.

22a.Mestecky, J. Personal communication.23. Moch, T., H. Hoschfitzky, J. Hacker, K. D. Kroncke, and K.

Jann. 1987. Isolation and characterization of the alpha-sialyl-beta-2,3-galactosyl-specific adhesin from fimbriated Eschenichiacoli. Proc. Natl. Acad. Sci. USA 84:3462.

24. 0rskov, F., I. 0rskov, B. Jann, and K. Jann. 1980. Tamm-Horsfall protein or uromucoid is the normal urinary slime thattraps type 1 fimbriated Escherichia coli. Lancet i:8173.

25. Parkkinen, J., R. Virkola, and T. K. Korhonen. 1983. Esche-richia coli strains binding neuraminyl a2-3 galactosides. Bio-chem. Biophys. Res. Commun. 11:456-461.

25a.Rydberg, J., et al. Unpublished data.26. Svanborg-Edin, C., B. Eriksson, and L. A. Hanson. 1977.

Adhesion of Escherichia coli to human uroepithelial cells invitro. Infect. Immun. 21:767-774.

27. Svanborg-Eden, C., R. Freter, L. Hagberg, R. Hull, S. Hull, H.Leffler, and G. Schoolnik. 1982. Inhibition of experimentalascending urinary tract infection by receptor analogue. Nature(London) 298:560.

28. Vink, A., P. G. Coulie, P. Wauters, R. P. Nordan, and J. VanSnick. 1988. B cell growth and differentiation activity of inter-leukin-HP1 and related murine plasmacytoma growth factors.Synergy with interleukin 1. Eur. J. Immunol. 18:607-612.

VOL. 60, 1992

on February 24, 2021 by guest

http://iai.asm.org/

Dow

nloaded from