Bacteria used in commercial probiotic preparations are most commonly gram-positive lactic acid-producing species, although there are also some probiotic products which utilise gram-negative coliform bacteria. Characterising how the innate immune system responds to these bacteria in vitro may give an indication as to the likely immunomodulatory events that can be triggered following probiotic administration in vivo. Here, an established gram-positive probiotic (Lactobacillus casei Shirota) was compared against a novel gram-negative probiotic strain (Escherichia coli Nissle 1917) for its ability to induce cytokine production in a cell type representative of the innate immune system; in addition, responses were contrasted against those induced by an enteropathogenic coliform, E. coli 2282. We investigated the ability of these three bacterial strains to modulate production of interleukins-10, -12 and -18; tumour necrosis factor-α; interferon-α; and transforming growth factor-β, via a series of in vitro culture experiments involving the murine monocyte/macrophage cell line J774A.1. All bacteria induced marked secretion of IL-12 and TNFα by cells, while only coliforms induced production of IL-10; there was minimal or no induction of IL-18 or TGFβ. Activation of cells with recombinant γ-interferon promoted increased production of IL-12, but decreased production of IL-10, in response to the co-culture of coliform bacteria, indicating differential cytokine induction depending on the activation status of the target cell. In general, live bacteria stimulated higher levels of IL-10, IL-12 and TNFα secretion than heat-killed preparations, while only live coliforms induced IFNα. These findings are discussed in relation to the likely immunomodulatory effects of gram-positive and gram-negative bacteria on the innate immune system in vivo, with particular emphasis on the marked similarity in cytokine response patterns observed between probiotic versus pathogenic coliform bacteria.
Probiotics are live microbial preparations which, when consumed as a dietary supplement, are capable of surviving in the human gastrointestinal (GI) tract and modulating host physiological systems, including the immune system. Several preparations containing defined probiotic strains are available as commercial products, many of which purport pro-immunity or anti-inflammatory health claims which denote a modulating effect on the immune system , and in many cases these claims are supported by in vitro, animal model and/or human trial data [2,3]. However, it is evident that the manner in which probiotics impact on the immune system varies depending not only on the bacterial species involved, but also on the individual strain of bacterium that is contained in the product , its state of viability , and its level of consumption .
The vast majority of immunomodulatory probiotic bacteria are gram-positive lactic acid bacteria, lactobacilli and bifidobacteria. Strains such as Lactobacillus casei Shirota have a long history of inclusion in human food products, and a well-documented impact on the mammalian immune system, including an ability to skew a burgeoning Th2 antigen-specific immune phenotype toward a Th1 profile , to activate cell-mediated immune (CMI) defences , to combat respiratory tract viral challenge , and to enhance tumouricidal activity . However, interestingly, there are also some commercial strains of probiotics which utilise gram-negative bacteria [11,12]. One such probiotic is Escherichia coli Nissle 1917, which has been reported to activate murine macrophage tumouricidal activity in vitro  and to provide enhanced protection to mice against systemic Listeria or Candida challenge in vivo . Further, in in vitro cell culture studies involving a human intestinal epithelial cell line (HT29), direct co-culture of E. coli Nissle 1917 with cells has been shown to induce production of the leucocyte chemotactic cytokine IL-8 . Of more particular note, though, E. coli Nissle 1917 has been reported to offer clinical benefit as an alleviative agent in human GI tract inflammatory disorders, including ulcerative colitis  and Crohn's disease , suggesting that the probiotic may also exert immune down-regulatory effects. However, the underlying mechanisms by which E. coli Nissle 1917 may boost CMI responses on the one hand, and yet down-regulate inflammatory conditions on the other, are yet to be elucidated, particularly in terms of the underlying cytokine patterns which drive these effector responses.
The initial immune recognition of de novo gut-colonising probiotic bacteria takes place via the ligation of bacterial cytoskeletal components  or unmethylated DNA motifs  to cell surface pattern recognition receptors, including Toll-like receptors, CD14 and mannose receptor . Such binding promotes the rapid activation of NFκB and STAT transduction pathways, and the subsequent production of immunomodulatory cytokines by mononuclear leucocytes of the monocyte/macrophage lineage . Various strains of probiotic lactobacilli (including L. casei Shirota) have been well-characterised in terms of their ability to induce cytokine production following contact with mononuclear phagocytic or accessory cells [22,23], and this contributes to our understanding of the in vivo impact of these probiotics on the immune system. In contrast, however, little is known of the patterns of cytokine induction by coliform probiotic bacteria following contact with the immune system (although several studies have reported cytokine response patterns induced by enteropathogenic coliform bacteria or their cell products; reviewed in ).
In this study, we have utilised the murine monocyte/macrophage cell line J774A.1 as a model to investigate the possible interactions of probiotic bacteria with cells of the innate immune defence system. The main focus of the study was to compare the cytokine response patterns observed following contact of J774A.1 cells with L. casei Shirota (a well-characterised, gram-positive probiotic) with those observed following cell contact with the novel, gram-negative probiotic, E. coli Nissle 1917. An additional aim of the study was to contrast the cytokine response patterns following cell contact with a probiotic coliform against those observed following cell contact with an enteropathogenic coliform. We have focused on characterising the overall patterns of cytokines induced following contact of bacteria with the cells, paying particular attention to the relative levels of induction of known pro-inflammatory cytokines versus regulatory/anti-inflammatory cytokines.
2Materials and methods
Escherichia coli strain Nissle 1917 was obtained as a commercial probiotic product (MutaflorTM, Ardeypharm GmbH company, Germany), while the enteropathogenic E. coli K12 strain NZRM 2282 – CDC 711 was obtained from the Environmental Science Research laboratories culture collection (Porirua, New Zealand) following isolation from a patient with urinary tract infection (strain 2282 is a heat-stable, enterotoxin positive isolate). Both coliform bacteria were cultured to log-phase growth by 18 h culture in brain heart infusion broth (BHI, Difco Laboratories, Sparks, MD, USA), prior to harvesting and washing in PBS. Serial dilutions of freshly-prepared cultures were plated onto tryptic soy agar and cultured overnight for enumeration. Lactobacillus casei strain Shirota (Yakult Honsha Co., Tokyo, Japan) was cultured under anaerobic conditions to log-phase growth for 30 h in l-cysteine HCL-supplemented de Man Rogosa Sharpe (MRS) broth (Fort Richards Laboratories, Auckland, New Zealand) prior to harvesting and washing in PBS. Serial dilutions of freshly-prepared cultures were plated onto MRS agar/cysteine-HCL and cultured for 36 h under anaerobic conditions, prior to enumeration. For all experiments involving killed bacterial preparations, samples of freshly-prepared cultures were enumerated using the appropriate agar, while additional samples of the same cultures were heat-killed (70 °C/30 min); subsequently, the concentration of the heat-killed preparation was adjusted in lieu of the live plate counts and used for further experimentation. Successful heat-killing was confirmed by the absence of bacterial growth on appropriate agar plates.
For all bacterial strains, standard growth curves were produced by plotting OD610 vs agar plate counts of freshly-prepared, serially-diluted cultures. These curves were fitted with logarithmic expressions (in order to calculate viable bacterial counts in freshly-prepared cultures) which each yielded r2 values of >98.5% (data not shown).
The murine monocyte/macrophage cell line J774A.1 (ATCC TIB 67) was maintained in Dulbecco's Modified Eagle Medium (DMEM) complete medium, supplemented with 10% foetal calf serum (Gibco BRL, Auckland, New Zealand), 2 mM l-glutamine and 5 μM 2-mercaptoethanol. For experiments involving heat-killed bacteria, penicillin-G and streptomycin sulphate (Gibco) were incorporated into the medium at 100 U/ml and 100 μg/ml, respectively. For experiments involving live bacteria, cells were cultured in either antibiotic-free medium or in medium containing 50 μg/ml gentamicin (Sigma, Perth, Australia) as described. In all cases, J774A.1 cells were grown to confluence in 75 cc tissue culture flasks, prior to harvesting by scraping using a rubber spatula. Viable cells were enumerated using trypan blue exclusion, and 2 ml aliquots containing 106 viable cells were plated onto multiple wells of flat-bottomed 24-well tissue culture plates (Nunc Brand Products Ltd, Denmark) and allowed to adhere overnight prior to experimentation; non-adherent cells were removed the following day and the spent culture medium was replaced with 0.5 ml fresh DMEM complete medium immediately prior to experimentation.
2.3Co-culture of cells with heat-killed bacteria
For extended co-culture of cells with killed bacterial preparations, suspensions of heat-killed E. coli Nissle, E. coli 2282 or L. casei Shirota were prepared in DMEM complete medium to a concentration of 5×107 bacteria/ml. Aliquots of these suspensions (0.5 ml) were then added to wells of 24-well plates containing J774A.1 cells, thus the final concentrations were: 106 J774A.1 cells in 1 ml DMEM complete medium co-cultured with 2.5×107 heat-killed bacteria, representing a bacteria:cell ratio of 25:1 (preliminary studies had indicated that this concentration was optimal for the induction of monokine secretion; the aim was to expose cells to bacteria at a level which offered the best opportunity of measuring cytokine synthesis , should any particular bacterial preparation be capable of inducing such a response at all). Cells were co-cultured in the presence of heat-killed bacteria for 8 or 16 h at 37 °C in a tissue culture incubator (5% CO2), prior to harvesting the culture supernatants from the wells. Supernatants were cleared of particulate material by centrifugation, and the supernatants were stored frozen (at −80 °C, for approximately 3–4 weeks) prior to subsequent cytokine analysis (IL-10, IL-12, IL-18, IFNα, TNFα and TGFβ, as described below).
In a further co-culture experiment, the response of J774A.1 cells to γ-interferon activation was assessed, in terms of how this might affect bacteria-induced secretion of the major immunoregulatory cytokines IL-10 and IL-12. In this case, J774A.1 cells were harvested from 75 cc tissue culture flasks, enumerated and added to 24-well plates at 106 viable cells per well in DMEM complete medium, either in the presence or absence of 200 U/ml recombinant murine IFNγ (Peprotech Ltd, Massachusetts, USA). After overnight culture, the spent culture medium was replaced with fresh DMEM complete medium (with/without IFNγ) and suspensions of heat-killed E. coli Nissle, E. coli 2282 or L. casei Shirota were added to the wells at a bacteria:cell ratio of 25:1 as described above. Some wells alternatively received 10 μg/ml of the standard cytokine-inducing phagocytic stimulant SAC (Staphylococcus aureus Cowan strain [Pansorbin] Calbiochem Inc, San Diego, USA). In all cases, cells were co-cultured with heat-killed bacteria (in the continuing presence/absence of recombinant IFNγ) for a further period of 16 h, prior to the harvesting and storing of cell culture supernatants for later analysis of IL-10 and IL-12.
2.4Limited exposure culture of cells with live bacteria
In order to compare the cytokine response of J774A.1 cells to live versus killed bacterial preparations, cells were first grown in 75 cc flasks as described above, and then seeded into wells of 24-well tissue culture plates at 106 cells/well and allowed to adhere overnight. Spent culture medium was removed and the cell surfaces were washed with warm PBS, and then fresh DMEM complete medium without antibiotics was added to the wells. To one set of wells, heat-killed bacteria were added at a final bacteria:cell ratio of 25:1; to another set of cells, freshly-prepared viable bacterial suspensions were added at a final bacteria:cell ratio of 25:1 (the number of viable cells was assessed from the standard growth curves for each bacterial strain). J774A.1 cells and heat-killed or live bacteria were co-cultured for a limited time (2.5 h, 37 °C) before non-phagocytosed bacteria were removed from the cell cultures by twice washing with PBS. Fresh DMEM complete medium (containing gentamicin) was then added to the wells (gentamicin was used to prevent extracellular bacterial overgrowth by any residual bacteria that may have remained in the tissue culture wells, while avoiding the issue of exogenous anti-microbial agents per se inhibiting bacteria in the intracellular milieu). The cells were cultured at 37 °C/5% CO2 for a further 16 h period, prior to the collection of culture supernatants for analysis of the cytokines IL-10, IL-12, IFNα and TNFα, as described below.
2.5Measurement of cytokine levels
Levels of cytokines in particulate-free cell culture supernatants were determined by capture ELISAs, and measured in each case against a standard curve generated by employing known amounts of recombinant cytokine. IL-10, IL-12 and TNFα were measured using DuoSet ELISA detection kits (R&D Systems Inc, MN, USA), while IFNα was measured using commercial pre-prepared ELISA detection plates (PBL Biomedical Laboratories, Piscataway, NJ, USA). IL-18 and TGFβ were assessed using paired anti-murine IL-18 or TGFβ capture and biotinylated-detection monoclonal antibodies (MBL Ltd, MA, USA for IL-18; BD PharMingen Inc, San Diego, USA for TGFβ). In the latter case (TGFβ) test cell culture samples were first activated by exposure to 0.1 N HCL for 60 min, followed by re-neutralisation with 0.1 N NaOH, before introduction to the ELISA plate. In all cases, detection antibody binding was visualised using streptavidin-horse radish peroxidase conjugate and a TMB (tri-methyl benzidine) substrate system at an OD of 450 nm.
For each experiment, the levels of cytokine induced by exposure of J774A.1 cells to bacteria was determined by comparing against the cytokine levels observed in J774A.1 cells that had been cultured for the same period in DMEM complete medium alone (controls). Data were calculated as the mean cytokine response (in pg/ml of cell culture supernatant) of each treatment from triplicate wells, plus or minus the standard error of the mean.
Time-dependent increases in cytokine levels were observed for all cytokines except IL-18 (Fig. 1), with heightened levels recorded after 16 h co-culture of bacteria with cells as compared to 8 h co-culture. The most noticeable effects of bacterial co-culture were on secretion of IL-12, TNFα and IL-10 (Fig. 1). Co-culture of cells with L. casei Shirota induced secretion of IL-12 and TNFα, but not IL-10. In contrast, co-culture of cells with E. coli Nissle or E. coli 2282 induced secretion of IL-12, TNFα and IL-10. Levels of IL-12 and TNFα induced by coliform bacteria were on average 3–5 fold higher than levels induced by L. casei Shirota. IL-18 and TGFβ were not readily induced by co-culture of J774A.1 cells with heat-killed bacteria, and so were not studied in subsequent experimentation.
Activation of J774A.1 cells via exogenous recombinant IFNγ invoked increased production of IL-12 but decreased production of IL-10, in response to the standard phagocytic cytokine-inducing stimulant SAC (Fig. 2). Co-culture of coliform bacteria (E. coli Nissle or E. coli 2282) with activated or normal J774A.1 cells invoked IL-12 and IL-10 production to levels exceeding those observed with SAC, with increased production of IL-12 in activated cells but decreased production of IL-10. Co-culture of L. casei Shirota with normal J774A.1 cells invoked low (although still detectable) levels of IL-12 but not IL-10; IL-12 production in response to L. casei Shirota was not increased in activated J774A.1 cells.
Limited exposure culture of bacteria with J774A.1 cells also induced cytokine secretion (Fig. 3) although, as expected, overall levels of cytokines tended to be lower than those observed following extended co-culture. Live or killed L. casei Shirota induced low, but still readily detectable, levels of IL-12 as well as TNFα in J774A.1 cells; neither live nor killed L. casei Shirota induced production of IL-10 in J774A.1 cells (Fig. 3). Both live and killed E. coli Nissle and E. coli 2282 induced production of IL-12, TNFα and IL-10, with responses noticeably higher for cells exposed to live bacteria. Live, but not killed, coliform bacteria induced production of IFNα.
Several previous studies [22,23,25–27], including our own , have reported the ex vivo / in vitro cytokine response patterns of co-culturing cells of the innate immune defence system with probiotic lactobacilli. The Lactobacillus strain studied here (L. casei Shirota) has been reported previously to induce the production of IL-12 and TNFα in murine primary splenocyte cultures or cell lines [7,23,27]; in contrast, little is known of the patterns of cytokines induced by probiotic coliform bacteria, although given that these latter probiotics are sold with health-claims that purport to their use in combating GI tract inflammatory diseases, one might expect that their mode of action would be down-regulatory, i.e. to selectively invoke the production of IL-10 and possibly TGFβ. Perhaps surprisingly, the present study has indicated that, in contact with murine cells at least, the probiotic E. coli Nissle 1917 did not produce such a down-regulatory pattern. There was no discernible induction of TGFβ by E. coli Nissle 1917, and while IL-10 was induced to a high level, this was accompanied by similarly strong induction of the pro-inflammatory cytokines IL-12 and TNFα. The strong production of IL-10 sets E. coli Nissle apart from L. casei Shirota, since the latter was shown here to induce very little IL-10 production, which agrees with the results of previous reports of poor IL-10 induction following co-culture of lactobacilli with murine phagocytic cell lines , human PBMC or PBMC-derived monocytes [29,30].
In contrast to IL-10, L. casei Shirota readily induced production of both IL-12 and TNFα by J774A.1 cells, confirming previous reports of pro-interferon and inflammatory monokine induction by this probiotic strain in murine splenocytes  and monocyte/macrophage cell lines . However, levels of IL-12 and TNFα were far lower than those observed following co-culture of J774A.1 cells with E. coli Nissle 1917, again demonstrating that the probiotic coliform has the potential to induce high-level monokine secretion in mononuclear phagocytes. Taken together, these results indicate IL-12 and TNFα induction by L. casei Shirota, in the absence of co-induction of IL-10, thus confirming this probiotic's purported in vivo role as an oral biotherapeutic capable of stimulating NK cell activity , down-regulating Th2 immune phenotype expression in animal models of allergy , and promoting CMI responses against tumours and intracellular pathogens [9,10,32]. In contrast, the results identify E. coli Nissle 1917 as a strain which induces production of IL-10, concomitant with high-level induction of IL-12 and TNFα. It therefore appears incongruous that this strain is capable of down-regulating inflammatory responses in vivo [11,16,17], unless its mode of action is quite different when in contact with other immune cell types, such as lymphocytes undergoing MHC-restricted antigen stimulation (for example, von der Weid et al.  have reported induction of anti-inflammatory TGFβ by inclusion of the probiotic Lactobacillus paracasei strain NCC2461 during an in vitro mixed lymphocyte reaction). Here, E. coli Nissle 1917 was not shown to induce TGFβ following contact with monocultures of J774A.1 monocyte/macrophage cells, and our data so far indicate that E. coli Nissle 1917 induces a cytokine response pattern, following direct bacterial:leucocyte contact, virtually indistinguishable from that observed with an enteropathogenic coliform, namely E. coli 2282. However, despite these similarities in vitro, the outcome of oral exposure to pathogenic as opposed to probiotic coliform bacteria is clearly quite different in vivo, and thus the potentially beneficial effects of E. coli Nissle 1917 in down-regulating immunoinflammatory cytokines in vivo – as a means of exerting anti-inflammatory effects in the consumer – remains to be investigated further.
As expected , pre-culture of J774A.1 cells with exogenous, recombinant IFNγ was capable of modulating cytokine production, with an increase in IL-12 output and a concomitant decrease in IL-10 output in activated cells following stimulation with a standard phagocytic stimulant, namely SAC. Interestingly, IFNγ-activation did not increase the capacity of cells to secrete IL-12 following co-culture with L. casei Shirota, indicating that while this probiotic strain remains capable of inducing IL-12 secretion (and – as a consequence – modulating cell-mediated effector immune responses that are dependent upon IL-12 induction ), IL-12 output is not markedly increased by the activation status of the cells. In contrast, IFNγ-activated J774A.1 cells responded to the presence of E. coli Nissle 1917 by markedly down-regulating IL-10 output but increasing IL-12 secretion, thus producing a typical pro-inflammatory cytokine pattern. Once again, this pattern was virtually identical to that induced by exposure of normal or activated J774A.1 cells to the enteropathogenic E. coli strain 2282, suggesting that activated macrophages become hyper-responsive to coliform bacteria in general upon activation. Since some GI tract inflammatory conditions (including enterocolitis and Crohn's disease) are known to be influenced by the presence of endogenous pro-inflammatory factors, such as IFNγ, further research is needed to elucidate the mechanisms of cytokine signaling at the gut mucosal surface by the probiotic coliform E. coli Nissle 1917, paying particular attention to its ability to modulate pre-exiting inflammatory conditions under controlled conditions.
It has been reported that live probiotic lactobacilli are more potent inducers of cytokine production in mammalian leucocytes than killed forms [29,35]. This observation has been confirmed here, since in limited contact assays, live L. casei Shirota was capable of inducing J774A.1 cells to secrete marginally higher levels of IL-12 and TNFα than those observed using a comparable number of heat-killed bacilli; however, more noteworthy was the fact that even live L. casei Shirota did not induce IL-10 production in J774A.1 cells, indicating that – overall – the innate pattern recognition of this probiotic strain does not influence IL-10 production. The heat-treatment method employed here to kill bacteria is commensurate with the approach used in several previous in vitro [7,23,25,27,36] and in vivo studies  which have demonstrated immunomodulation by lactobacilli, although it would be interesting to compare different methods of killing lactobacilli to determine if this can consequently influence the cytokine response patterns in leucocytes. De Waard et al.  have demonstrated that Listeria-specific CMI responses in rats, modulated by L. casei Shirota, require the oral delivery of live probiotic bacteria, thus confirming that viability of L. casei Shirota cells contributes to protective immune responses in vivo, as fitting the paradigm of probiotic immunomodulation.
In contrast to live lactobacilli, live coliform bacteria (E. coli Nissle 1917 and E. coli 2282) induced secretion of noticeably higher levels of IL-12 and TNFα when compared to heat-killed bacteria, as well as higher levels of IL-10. Moreover, live coliforms were the only test bacterial agents shown capable of inducing secretion of IFNα above background levels. Induction of IFNα (in conjunction with TNFα) is one of the hallmarks of a pro-inflammatory monokine response to pathogenic bacteria, following activation via the NFκB-signaling pathway . Thus, it appears that E. coli Nissle 1917 bacteria can induce – in murine monocyte/macrophages at least – a potent pro-inflammatory cytokine response, and one which is elevated when cells are in contact with live bacteria. Given that probiotics are – by definition – live microbial food supplements that are able to survive gastric transit and at least transiently colonise the small and large intestinal mucosal surfaces, further characterisation of the effect of viable cells of this strain on immunocompetent cells seems warranted. This is especially important since recent research has highlighted the fact that gram-positive bacterial cell wall components can modulate the immunocompetence of intestinal epithelial cells (IECs) with respect to gram-negative bacteria , while IECs can – in turn – interact with adjacent leucocytes to either diminish or augment an ensuing immunoinflammatory reaction . Thus, while results here have clearly shown that a probiotic coliform is capable of inducing a pro-inflammatory cytokine response upon direct contact with mononuclear phagocytic leucocytes, the true in vivo picture needs to be more fully determined taking into account the contributing roles of other commensal gut bacteria and of immunocompetent non-leucocytic cells in the gut environment.
We thank Dr. Pramod Gopal of the Fonterra Marketing & Innovation Centre, Palmerston North, New Zealand, for provision of the original stock culture of Lactobacillus casei Shirota that was used in this study.