Anti-Inflammatory Effect of Lactobacillus casei on Shigella-Infected Human Intestinal Epithelial Cells¹

More than 400 species of bacteria are present in the human gut, ranging from 10⁵ to 10⁷ CFU/g of intestinal contents in the jejunum to 10¹¹ to 10¹² CFU/g in the colon (1). To maintain the homeostatic mechanisms responsible for tolerance of the intestinal mucosa vis-a-vis its commensal flora, and at the same time to permit the development of an innate response to pathogenic bacteria, intestinal epithelial cells (IECs),⁴ which are in close contact with luminal bacteria, are expected to play a key regulatory role (2, 3). Intestinal homeostasis has been shown to be under the control of TLRs. Indeed, mice deficient in TLR2, TLR4, or the adaptor molecule MyD88 are more prone to develop intestinal inflammation after dextran sodium sulfate-induced colitis (4). Tolerance to the commensal flora is similarly dysregulated in the case of inflammatory bowel disease (IBD) (5).

Lactobacilli and Bifidobacteria are natural components of the colonic microbiota, and as probiotic agents, they have been tested in the prevention and treatment of IBD (6, 7). A strain of Lactobacillus casei, DN-114 001, has been shown to decrease the secretion of TNF-α from the inflammed ileum of Crohn’s disease patients (8, 9). The beneficial effect of probiotics is proposed to be due, at least in part, to interference with the innate immune system and possibly the orientation of adaptive immunity. Indeed, the probiotic preparation VSL#3 (a mixture of eight different Gram-positive bacteria) as well as different lactobacilli strains were shown to increase production of the anti-inflammatory cytokine IL-10 by dendritic cells (10, 11). This anti-inflammatory effect was also observed using DNA extracted from these probiotics; VSL#3 DNA was able to reduce IL-8 production by epithelial cells exposed to pro-inflammatory stimuli via a mechanism involving I-κB stabilization (12). Using a nonvirulent Salmonella strain, Neish and colleagues showed that regulation of epithelial responses occurs by inhibition of I-κBα ubiquitination (13), which is likely mediated by the Salmonella-secreted protein AvrA (14). Finally, regulation of multiple intestinal functions, such as nutrient absorption, mucosal barrier enhancement, and angiogenesis, is also modulated by the commensal Bacteroides thetaiotaomicron (15, 16).

Shigella flexneri, an entero-invasive, Gram-negative bacterium, causes the inflammatory destruction of the intestinal epithelium (17), leading to bacillary dysentery, an acute recto-colitis causing lethal complications, mostly in infants and toddlers in the most impoverished areas of the planet. The annual mortality attributed to bacillary dysentery is about one million. The mechanisms leading to acute mucosal inflammation need to be understood to develop efficient methods of prevention and cure. A number of studies have demonstrated that S. flexneri infection leads to the induction of several markers of acute inflammation, such as the chemokines IL-8 and CCL20. These pro-inflammatory genes are under the control of the NF-κB pathway (18), and recently, the mechanism of NF-κB activation in Shigella-infected cells has been uncovered. Indeed, our group has demonstrated that after infection, epithelial cells sense the presence of invading S. flexneri through Nod1, an intracellular pattern recognition molecule that specifically detect peptidoglycan from Gram-negative bacteria (19, 20).

Because probiotics such as L. casei represent a subclass of commensals that modulate mucosal innate responses and possibly exhibit an anti-inflammatory function (8), we aimed to analyze the effect of L. casei on epithelial inflammatory signaling induced by S. flexneri. Our results demonstrate that L. casei potently modulates pro-inflammatory pathways induced by S. flexneri, and by using global microarray analyses, we identify the ubiquitin/proteasome complex as the main target of L. casei action. Together, this study allows for a better understanding of how the commensal microflora contributes to the homeostasis of the host intestinal tract.

Caco-2 and HEK293T epithelial cells lines were cultured with DMEM supplemented with 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 1% nonessential amino acids (Invitrogen Life Technologies).

Invasive colonies of the S. flexneri serotype 5a strain M90T were isolated on Congo red agar plates (Congo red-positive colonies). One colony was then grown overnight in 7 ml of Trypticase soy broth (Difco) medium at 37°C under rotation. After dilution in fresh Trypticase soy broth, a subculture was then performed for 2 h to bring the culture to the exponential phase (OD, 600 nm; O.4). The strain SC301, a derivative of M90T after its transformation with plasmid pIL22 (encoding the afimbrial adhesin AfaE) from uropathogenic Escherichia coli, was used in this study (21).

The probiotic strain L. casei DN-114 001 was provided by Danone Vitapole. A colony of L. casei isolated on a Mann-Rogosa-Sharpe (MRS) agar plate was picked and cultured overnight at 37°C in 10 ml of MRS broth without agitation in absence of oxygen.

To obtain a polarized monolayer of IECs, Caco-2 cells were grown to confluency and then for 2 additional weeks to reach full differentiation and polarization. The culture medium was changed every 2 d. For the infection step, Caco-2 cells were incubated overnight without serum, with L. casei at a multiplicity of infection (MOI) of 100, before the addition of strain M90T-AfaE at the same MOI for 4 h. For the time-course experiment, nonpolarized Caco-2 cells were incubated at a MOI of 100 with L. casei for 2, 6, or 24 h. Biological duplicates were done for this time-course experiment. After washing Caco-2 cells with cold PBS, RNAs were extracted with the RNEasy mini kit (Qiagen). The quantity and quality of RNA preparations were determined, respectively, by absorbance lecture and electrophoresis using the Agilent nanochip technology.

To detect and quantify cell invasion, overnight culture of HEK cells seeded on coverslips, in the presence or absence of L. casei, were infected with M90T-AfaE as described above. After 1 h of infection, cells were washed in PBS and fixed with 4% paraformaldehyde. To identify intracellular vs extracellular Shigella, after saturation with BSA (1% in PBS), a first staining was made in the absence of permeabilization with a rabbit polyclonal Ab against S. flexneri LPS, followed by the addition, after washing, of a rhodamine-labeled goat Ab against rabbit Ig. The cells were then permeabilized with 0.1% Triton X-100. The same incubation with the Ab was performed, with a FITC-labeled goat Ab. Once labeled intracellular bacteria appeared green-labeled, whereas twice-labeled extracellular bacteria appeared yellow-labeled. Fluorescence detection was done using an Olympus BX50 microscope (at a ×400 magnification) and a Leica DC350F camera. Pictures were analyzed using the Leica IM50 software (version 1.2; (Leica Microsystems). Determination of the green and red pixels were performed using Photoshop 7.0 software.

The PCR products of 1050 human genes were spotted in duplicate on positively charged nylon membranes as described previously (22). The macroarray design can be found on the ArrayExpress web site (www.ebi.ac.uk/arrayexpress) with the accession number A-MEXP-141. cDNA labeling and hybridization scanning were also as described previously (22).

After recording of the signals for each gene with ArrayVision and quality control of hybridizations using the luciferase signal intensity, data corresponding to all of the membranes were transformed in a log₂ scale and normalized by a method derived from the variance analysis (ANOVA) to give to all membranes an equal mean signal. This statistical method estimates the weight and significance of variability sources on experimental data. For each condition, eight biological replicates were performed and hybridized onto macroarrays in which PCR products were spotted in duplicate; 16 signals for one gene were used for one experimental condition. Comparative analyses between baseline (noninfected cells) and experiment (infected cells) were done with the dChip software (23), using an unpaired Welch t test with a p-value threshold of 0.05. This software was also used for hierarchical clustering using Euclidian distance and average as a linkage method. Before clustering, the expression values for one gene across all samples are standardized to have a mean of zero. Increased or decreased values are then ranged compared with this mean. In the clustering picture, each row represents a gene and each column represents a sample. For the color scale, blue and red represent, respectively, low and high signal expression values.

Starting from 5 μg of total RNA, the first strand of cDNA was synthetized using T7-(dT24) oligonucleotides and Superscript II reverse transcriptase (Invitrogen Life Technologies) for 1 h at 37°C. The double-strand cDNA was then synthetized at 16°C for 2 h by the addition of DNA polymerase, DNA ligase, and RNaseH. A terminal step by T7 DNA polymerase was performed for 5 min. The resulting DNA was used to synthetize biotin-labeled cRNA via an in vitro transcription labeling kit (Enzo Diagnostics). Ten micrograms of fragmented cRNA were hybridized with the GeneChip U133A for 16 h at 60 rpm in a 45°C hybridization oven. The chips were washed and stained with streptavidin PE (Molecular Probes) in the Affymetrix fluidics station. To amplify staining, an anti-streptavidin biotinylated Ab was introduced between two streptavidin PE staining steps. After hybridization, staining, and scanning (Agilent), the chip images were analyzed for quality control before signal analysis. The chips were then normalized using the scaling factor method (with a target intensity of 100). The analysis was performed using the statistical algorithms from the MAS-5 software (Affymetrix) to obtain absolute signals, detection calls, and the associated p values. Transcripts that were absent both in control and experimental conditions were removed from further analysis. Comparative analysis, between control and experimental conditions, which give a signal log ratio, a change call, and an associate p value, was performed with MAS-5 software using the t test and Mann-Whitney U test. Because the experiments were done in duplicate, to choose the modulated genes, we selected the probe set, the change call (increase or decrease) of which appeared to be four times (each experimental point vs each control point). Hierarchical clustering was performed using dChip software with the same parameters than those described above.

cDNAs were synthetized from a template of 5 μg of total RNA using oligo(dT) primers (Promega) and Superscript II (Invitrogen Life Technologies) for 1 h at 42°C. PCR was performed using 5 μl of a 1/20 dilution of the cDNA as a template in a final volume of 50 μl. The standard program used was as follows: denaturation for 5 min at 94°C, 35 cycles of 45 s treatment at 94°C, 45 s at 55°C, and 90 s at 72°C, followed by a final elongation for 7 min at 72°C. One unit of Eurobiotaq (Eurobio) was used for one PCR. After electrophoresis in a 2% agarose gel containing ethidium bromide, PCR products were quantified using the ImageQuant software (Amersham Biosciences). The primers used are described in Table I.

Experiments were conducted as reported previously (19). Briefly, 5 × 10⁵ HEK293 cells were transiently transfected using 75 ng of the NF-κB luciferase reporter system, and NF-κB-dependent luciferase activity was measured 20 h after transfection. NF-κB-dependent luciferase assays were performed in duplicate, and data represent at least three independent experiments. Data show mean ± SE.

For experiments in which L. casei was added before stimulation with either S. flexneri or TNF-α, the following procedure was set up: HEK293 cells were seeded at 5 × 10⁵ cells/ml and either left unstimulated (for Western blotting experiments) or transfected with 75 ng of the NF-κB luciferase reporter gene (for NF-κB-dependent luciferase assays) as described above. Twenty hours later, the incubation medium was changed and replaced by fresh medium containing either MRS medium alone or L. casei grown in MRS buffer (5 × 10⁷ bacteria/ml). After an overnight incubation, cells were rinsed three times with PBS, and fresh medium was added. HEK293 cells were then left unstimulated, stimulated with TNF-α (100 ng/ml), or infected with S. flexneri M90T-AfaE (5 × 10⁷ bacteria/ml) for variable periods (4 h in the case of NF-κB luciferase reporter assays), before lysis of the cells.

After overnight incubation of Caco-2 cells in the presence or absence of L. casei, the cells were stimulated either with M90T-AfaE or with TNF before lysis by the addition of 200 μl of Laemmli solution. After heating 5 min at 90°C, 10 μl of lysate were loaded in a 10% acrylamide SDS-PAGE. After migration, proteins were transferred onto nitrocellulose by semidry transfer. After blocking by PBS/5% milk, the membrane was incubated overnight with anti-I-κBα, antiphosphorylated I-κBα, anti-Rbx1, or anti-poly-ubiquitin Abs (all purchased from Santa Cruz Biotechnology) or with anti-β-tubulin (Sigma-Aldrich) at 1/500 in PBS/milk. After three washes in PBS/Tween, the membrane was incubated with a peroxydase-labeled secondary Ab (1/1000) for 1 h. After washing, the membrane was incubated for 5 min with ECL⁺ chemiluminescence reagent (Amersham Biosciences). Kodak film was then exposed for different time periods to the membrane.

Until recently, GeneChip studies have been hampered by some technical concerns, including the absence of key genes and redundancy. In an effort to characterize with precision the global response of cells infected by a bacterial pathogen, we have developed a macroarray-based procedure in which PCR products of 1050 genes of interest, including cytokines, chemokines, transcription factors, etc. (22), have been spotted. Using this tool, we aimed to identify the repertoire of genes modulated after infection of polarized Caco-2 cells with the invasive S. flexneri strain M90T-AfaE. Radioactive cDNA was hybridized onto macroarray nylon membranes, and after normalization plus comparative analysis using dChip, 36 genes modulated during infection with M90T with statistical significance (p < 0.05) were identified (Table II). Interestingly, genes encoding chemoattractant chemokines with pro-inflammatory functions such as CXCL1–3 and CCL20 were found up-regulated after Shigella infection. Importantly, results obtained by macroarray analyses confirm and further expand results obtained during Shigella infection of nonconfluent Caco-2 cells using Affimetrix technology (24).

To observe the effect of L. casei DN-114 001 on the inflammatory process occurring during Shigella infection of IECs, polarized Caco-2 cells were preincubated overnight with L. casei before infection with M90T-AfaE. After RNA extraction, hybridizations to the macroarray membranes were performed. Interestingly, hierarchical clustering identified L. casei-mediated modulation of Shigella-induced responses of epithelial cells (Fig. 1). Among these, some key pro-inflammatory genes were found to be down-regulated when Caco-2 cells were preincubated with the probiotic bacteria before infection with Shigella. This was the case for the expression of CXCL1 and CXCL2, two chemokines involved in chemoattraction of polymorphonuclear cells, and for the expression of the dendritic cell chemoattractant CCL20.

When macroarray-based technologies are used, it is crucial to further confirm the results obtained by individual RT-PCR. Using this approach, we could validate that expression of CXCL1 and CXCL2 was down-regulated (both by ∼50%) in conditions of coinfection with L. casei and M90T-AfaE compared with M90T-AfaE alone (Fig. 2). Moreover, in a previous study (24), we demonstrated that the expressions of ICAM-1 and amphiregulin were up-regulated during Shigella infection. In this study, we report that preincubation of Caco-2 cells with L. casei prevented increased transcription of these genes after infection with Shigella (Fig. 2).

The above results pointed to a role of L. casei in the specific modulation of a selected repertoire of Shigella-induced genes. However, to ascertain that these observations were not an indirect consequence of the experimental set-up used, we aimed to investigate whether coculture of Caco-2 cells with L. casei interferes with Shigella invasion. Using double-fluorescence labeling, followed by counting of intracellular vs extracellular Shigella, we observed that regardless of the absence or presence of L. casei in the culture medium, the ratio of intracellular vs extracellular Shigella remained unchanged (1.91 ± 0.74 or 1.70 ± 0.66, respectively).

Together, these results demonstrate that L. casei selectively down-regulates the expression of some key factors induced after infection of IECs with Shigella.

Because a number of the Shigella-induced genes identified above are under the control of the NF-κB transcription complex, we aimed to investigate whether L. casei acted directly on this signaling pathway. Strikingly, using a NF-κB-dependent luciferase reporter gene assay, we observed that overnight preincubation of HEK293 cells with L. casei repressed NF-κB activation induced by Shigella (Fig. 3). We then investigated the nature of this blockage at the molecular level. In epithelial cells, the signaling pathway leading to the activation of NF-κB after S. flexneri infection was recently identified. It involves the intracellular peptidoglycan-recognition molecule Nod1 and a signaling cascade involving Rip2/RICK/Cardiak, the IKK complex, and the inhibitory protein I-κBα (25). We analyzed whether preincubation with L. casei could modulate the expression of selected members of this signaling cascade. Although the expressions of Rip2 and IKKα were not modified by pretreatment with L. casei (Fig. 4,A), we noticed a substantial increase of I-κBα expression after overnight coculture of HEK293 cells with L. casei (Fig. 4 A).

Infection of epithelial cells with S. flexneri induces rapid and dramatic degradation of I-κBα, resulting in the release of a free NF-κB complex that translocates to the nucleus to induce transcription of NF-κB-dependent pro-inflammatory genes. Indeed, we could confirm these observations in a kinetic study of I-κBα degradation ranging from 10 to 40 min after infection (Fig. 4,B, top panel). In similar conditions, overnight preincubation of HEK293 cells with L. casei resulted in a strong inhibition of S. flexneri-induced I-κBα degradation. Instead, degradation of I-κBα was observed only at later time points and remained partial (Fig. 4,B, bottom panel). Interestingly, a slower migrating band likely corresponding to a phosphorylated form of I-κBα was transiently detected, but only in cells preincubated with L. casei. I-κBα degradation is mediated by the ubiquitin-proteasome system and the phosphorylation of I-κBα is a signal that targets the protein to degradation. In unstimulated cells, phospho-I-κ Bα is hardly detectable because the protein is immediately targeted for degradation. Because in epithelial cells incubated with L casei we could detect phospho-I-κBα (Fig. 5), we anticipated that the effect of L. casei on I-κBα may be related to altered degradation of the protein by the ubiquitin-proteasome system.

These results suggested that the anti-inflammatory effect of L. casei is largely mediated through specific modulation of the I-κBα protein level. However, I-κBα lays relatively downstream in the NF-κB signaling cascade, and several pro-inflammatory stimuli lead to its degradation. We therefore investigated whether L. casei could down-regulate NF-κB activation induced by TNF-α, another well characterized pro-inflammatory stimulus. Indeed, we could observe that overnight preincubation of HEK293 cells with L. casei potently repressed NF-κB activation induced by TNF-α (Fig. 6,A). Moreover, a kinetic study of I-κBα degradation ranging from 15 to 45 min after stimulation with TNF-α showed that I-κBα degradation was strongly decreased if HEK293 cells had been preincubated overnight with L. casei (Fig. 6 B). Together, these results strongly suggest that the anti-inflammatory properties of L. casei largely depend on the modulation of I-κBα expression and/or stability.

The above results demonstrate a role for L. casei in the specific modulation of NF-κB-driven genes induced after Shigella infection. However, it remains possible that other key signaling pathways are targets of L. casei-induced modulatory properties. To gain more insight into the global effects mediated by L. casei on host cell signaling, we performed a large-scale gene expression study (14,500 well characterized human genes; Affymetrix Genechips U133A) on Caco-2 cells cocultured with the probiotic strain for 2, 6, or 24 h. The entire set of results is available online (supplemental Table S1) ⁵, and Table III represents the number of probe sets modulated during Caco-2 cell culture in the presence of L. casei. Interestingly, most of the modulation was observed after 24 h of incubation. Because all of the results are available online, and to avoid to present an extensive list of modulated genes, we have focused in this study on four functional clusters (defined using NettAffyx, the database from Affymetrix) that appeared to be modulated most extensively by L. casei (Table IV and see below).

L. casei induced both positive and negative modulation of expression of genes involved in the regulation of the cell cycle. Indeed, tumor suppressor genes, such as tumor rejection Ag (gp96) and sel-1 suppressor of lin-12, were up-regulated during incubation with L. casei, whereas the expression of the gene encoding prohibitin, a negative regulator of cell proliferation, was down-regulated.

Expression of two genes encoding proteins involved in the inhibition of apoptosis (Fas apoptotic inhibitory molecule and cytokine-induced apoptosis inhibitor) were down-regulated, and expression of BCL2/adenovirus E1B was increased. It should be noted that the expression of Nalp2, encoding a part of the inflammasome (26), decreased during the incubation of Caco-2 cells with L. casei, suggesting its potential involvement in the anti-inflammatory properties of the probiotic bacteria. In addition, the expression of programmed cell death 4, a tumor suppressor (27), was positively modulated.

Expression of two genes under the control of the hypoxia-inducible factor-1 (HIF-1) transcription factor (hypoxia up-regulated 1 and HIF-1 response RTP801) was induced during the incubation of Caco-2 cells with L. casei.

Twenty-five genes involved in the protein ubiquitination and/or degradation process were found to be positively or negatively modulated during incubation of Caco-2 cells with L. casei. The transcription of genes encoding three ubiquitin-conjugating enzymes, E2D, E2L, and E2N, and four subunits of the 26S proteasome were down-regulated. Another gene, the expression of which decreased during this incubation, was rbx-1/roc-1/hrt-1, which encodes a protein belonging to the E3 ligase complex. We confirmed by individual PCR on a set of these selected genes their down-regulated expression after treatment with L. casei (Fig. 7 A). It is striking to note that, among all of the genes modulated by L. casei, the group of genes implicated in ubiquitination/degradation (group 4) is, by far, the most represented. This observation suggests that L. casei is capable of establishing a program leading to the down-regulation of the ubiquitin-mediated degradation of proteins. Moreover, because the ubiquitin system is responsible for the degradation of I-κBα, this observation strongly supports our conclusions that modulating the stability of I-κBα represents a major target of L. casei.

Among the 25 genes of the ubiquitination/degradation pathway that were identified, we noticed that Rbx-1 expression was down-regulated to a greater extent (∼4-fold decrease) by L. casei than the others (Table IV). This prompted us to investigate whether the effects of L. casei on the NF-κB pathway could be accounted for by the specific down-regulation of Rbx-1. First, the expression of Rbx-1 at the protein level was monitored in L. casei-treated cells and was found to be reduced by ∼50% (Fig. 7,B). Second, overexpression of exogenous epitope-tagged Rbx-1 in L. casei-treated cells was performed, and TNF-mediated (as well as Shigella-mediated) degradation of I-κBα was monitored (data not shown). However, restoring Rbx-1 expression was found to be insufficient to recapitulate the effects of L. casei on the ubiquitin/degradation pathway. As a consequence, it is likely that L. casei down-regulates this cellular process via the targeting of multiple effectors. Accordingly, we observed that the global pattern of ubiquitinated proteins after L. casei treatment closely resembles the one of cells treated with MG132, a broad inhibitor of multiple ubiquitin/proteasome-dependent pathways (Fig. 7 C). Together, this large-scale analysis of L. casei-modulated genes in Caco-2 cells strongly supports the conclusion that this probiotic bacteria displays anti-inflammatory properties through the targeting of the ubiquitin/proteasome system.

In this study, we have investigated the effect of coculture of IECs with the probiotic bacteria L. casei. We were interested in two questions of fundamental importance: 1) Does this probiotic bacteria display protective properties on IECs infected with a pathogenic entero-invasive bacteria? and 2) What is the pattern of transcriptional responses of the epithelial cells cocultured for various periods with L. casei? To gain insight into these questions, we embarked on a project to identify the global repertoire of host responses to L. casei using, for the first time on this topic, the technologies of macroarray analysis (1,050 genes selected) associated with DNA chip hybridization (Affymetrix; 14,500 genes) and biochemical analysis. Our results clearly identify a role for L. casei in the specific modulation of a repertoire of genes associated with the ubiquitin/proteasome pathway. Importantly, the down-regulation of this set of ubiquitin system-associated genes appears to play a crucial role in the modulation of pro-inflammatory pathways in IECs. Indeed, when pro-inflammatory stimuli, such as Shigella infection or TNF treatment, were used to stimulate these cells, we observed a key role of L. casei preculture in dampening these responses. Moreover, we could pinpoint the molecular basis of this modulation by the observation that L. casei specifically targets the stability of I-κBα, the specific NF-κB inhibitor, thereby shutting down this major pro-inflammatory pathway.

Our study confirms and also extends the hypothesis that certain commensal microorganisms have the potential to actively influence the homeostatic control of intestinal inflammation (2). Indeed, coculture of human tissues with L. casei significantly reduced TNF-α release from Crohn’s disease patients’ inflamed mucosa (8) even in presence of a nonpathogenic E. coli that alone enhanced TNF-α release from inflamed tissues (9). In addition, in a mouse model of skin hypersensitivity, ingestion of L. casei was demonstrated to reduce skin inflammation induced by dinitrofluorobenzene (28). Exposure of IECs to nonpathogenic Salmonella spp. could mediate anti-inflammatory signals by preventing ubiquitination of I-κBα (13). Recently, another anti-inflammatory mechanism induced by a commensal bacteria strain has been described. B. thetaiotaomicron attenuates pro-inflammatory cytokine expression by inducing the nuclear export of complexes formed by NF-κB and peroxisome proliferator-activated receptor-γ (29). This set of results indicates that preincubation with certain probiotic strains such as L. casei decreases NF-κB-driven activation of pro-inflammatory genes. In contrast, it has been shown that other commensals, such as E. coli MG 1655, induce NF-κB activation in IECs through the TLR5 signaling pathway (30). It must be noted that, in the present study, the two bacterial strains used, L. casei and S. flexneri, are not flagellated; thus, signal transduction via flagellin and TLR5, a major pro-inflammatory pathway described at the intestinal epithelial level (30), does not occur.

Among the genes that have been found to be modulated by L. casei through the Affymetrix study, we have identified a set of genes encoding proteins that control the cell cycle. Hence, the expressions of sel-1 and gp96, two tumor suppressor genes, were found to be up-regulated during incubation with L. casei. A high level of Sel-1 has been suggested to correlate with the decrease of breast tumor growth (31), and gp96 not only elicits immune responses to tumors by inducing a CD8⁺ T cell response (32) but also plays an adjuvant role in antitumor immune responses (33).

Another important pathway modulated by L. casei was shown to involve hypoxia-related genes. The level of transcription of vegf was found increased during incubation of Caco-2 cells with L. casei. This observation could prove of importance because VEGF is a growth factor that plays a key role in angiogenesis (34). Transcription of vegf is under the control of the transcription factor HIF-1 (35). Interestingly, two other hypoxia-responsive genes, both involved in apoptosis, were found up-regulated when Caco-2 cells were incubated with L. casei: the hypoxia up-regulated 1 gene (ORP150), which displays a protective function in hypoxia-conditioned cells (36), and RTP801 gene, which inhibits hypoxia-mediated apoptosis (37). The transcription of other genes known to be under the control of HIF-1, such as erythropoietin, was not modulated during such incubation. This is likely due to the restricted transcriptional program of Caco-2 cells. Significance as well as the mechanisms involved in the establishment of hypoxia sensing by Caco-2 cells during incubation with L. casei remain to be further analyzed. In particular, it will be of interest to determine whether these microorganisms actively induce this process or whether cellular hypoxia is an indirect consequence of extracellular consumption of oxygen or Fe³⁺ by the bacteria.

Because the expression of genes involved in different pathways (i.e., cell cycle, response to hypoxia, NF-κB-dependent transcription) was found to be modified during incubation of Caco-2 cells with L. casei, we investigated whether a common signaling pathway could account for these diverse effects. Of interest, these three pathways are regulated by a similar ubiquitination/proteasome system (38, 39, 40, 41). In our time-course study using Affymetrix technology, we observed that a large set of genes involved in the ubiquitination/degradation pathways was modulated by L. casei. In light of our results, it appears very likely that the specific modulation of the ubiquitin/proteasome system represents a main target of action of L. casei on host cells and that this effect would in turn be responsible for the alteration of several signaling pathways, such as those involving the cell cycle and hypoxia.

In agreement with our findings that L. casei affects the ubiquitin/proteasome pathway, Caco-2 cells coincubated with L. casei showed stabilization of I-κBα even after subsequent stimulation by Shigella or TNF-α. Thus, the anti-inflammatory effects of L. casei are likely mediated by its effects on the ubiquitin/proteasome system and the consequent dampening of NF-κB-driven pro-inflammatory signals.

Convergent pieces of evidence suggest that, through their ability to modulate inflammatory pathways, some commensal bacteria contribute to the homeostasis of the intestinal epithelium. Our observations, and also the results from other groups, will impact on the understanding of the mechanisms responsible for some of the beneficial effects of probiotics on IBDs. This knowledge will contribute to offer, in the near future, new therapeutic means to counteract the inflammatory disorders observed in human pathologies, such as ulcerative colitis and Crohn’s disease.

The Ab for ubiquitin detection was a gift from Dr. Claude Parsot (Pasteur Institute). We thank Drs. Armelle Phalipon, Dana Philpott, and Régis Tournebize for critical reading of this manuscript.

The authors have no financial conflict of interest.