The Bacterial Fermentation Product Butyrate Influences Epithelial Signaling via Reactive Oxygen Species-Mediated Changes in Cullin-1 Neddylation¹

Virtually unique among mammalian cell types, the intestinal epithelium coexists in intimate contact with a normal flora of ∼10¹² prokaryotic organisms. The greatest numbers and diversity are in the cecum and ascending colon where bacterial density can reach 10¹¹ cells/g of contents, yielding a biomass of >1 kg. This bacterial community of ∼500–1000 species has diverse beneficial roles including vitamin synthesis, bile salt metabolism, and degradation of complex carbohydrates (1). Furthermore, through studies with germfree or gnotobiotic mice, it is also now established that the enteric flora can fundamentally affect epithelial gene transcription, ultimately affecting enterocyte proliferation and homeostasis (2, 3).

The normal flora thrives in a largely anaerobic environment, generating energy by the fermentation of luminal complex carbohydrates. The end products of fermentation are a spectrum of organic acids, including short chain fatty acids (SCFAs)³ such as butyrate, succinate, and propionate, as well as other terminal products such as lactate (4, 5). These organic compounds are an important energy source for the colonic epithelium and may influence various aspects of gut physiology. For example, butyrate and other SCFAs have well-known differentiating and growth-promoting activities in vitro and in vivo, a biological effect ascribed to histone deacetylase activity (6). Additionally, SCFAs have been noted to have immunomodulatory effects on colonic inflammation, suppressing inflammatory cytokine secretion in cultured epithelial cells, and ameliorating model colitis in mice, suggesting that these molecules contribute to the ability of the mucosa to tolerate the presence of vast quantities of living microorganisms and associated microbial-associated molecular patterns (MAMPs) (6, 7). Furthermore, luminal instillation of butyrate has been shown to be a promising experimental therapy in ulcerative colitis and related inflammatory disorders (8, 9). How these naturally occurring bacterial metabolic products influence epithelial homeostasis are unknown and are a topic of considerable research interest.

A key pathway that modulates inflammatory and/or growth survival in the intestine is the NF-κB or Rel signaling pathway. The NF-κB pathway is controlled by regulated degradation of a physically associated inhibitor molecule, IκB-α, which when phosphorylated is ubiquitinated by a specific ubiquitin ligase complex designated E3-SCF^β-TrCP. Ubiquitinated IκB-α is targeted for degradation by the proteasome (10, 11). E3-SCF^β-TrCP and other E3-SCF complexes are themselves regulated by transient covalent modifications. The ubiquitin paralog NEDD8 must be conjugated to the Cullin-1 (Cul-1) subunit of the E3-SCF complex for ubiquitin ligase activity (12, 13, 14, 15). NEDD8 modification of Cul-1 has been demonstrated to be necessary for ubiquitination of IκB-α and p100/p105, and for the subsequent activation of NF-κB in mammalian cells (16, 17, 18, 19, 20).

We have demonstrated that viable commensal bacteria can block IκB-α ubiquitination by blockade of neddylation of the Cul-1 subunit of E3-SCF^β-TrCP, accounting for the attenuation of NF-κB activation (21). We further reported that loss of Cul-1 neddylation was mediated by rapid and reversible oxidative inactivation of Ubc12, the ubiquitin-like conjugating enzyme responsible for the neddylation of Cullin subunits (22). Herein, we demonstrate that butyrate and other SCFAs supplemented to model epithelia in vitro and human tissue ex vivo also cause loss of neddylated Cul-1 resulting in consequent inhibition of the NF-κB pathway. We also show that de novo and transient reactive oxygen species (ROS) generation mediate these regulatory effects of butyrate. These results suggest that metabolic products from the normal intestinal flora (and potentially other complex microbial communities) can influence mammalian signaling pathways by modulating the activation of a rate-limiting enzymatic step.

Butyric, lactic, propionic, and succinic acid were obtained from Sigma-Aldrich. MG-262 was obtained from Affinity. TNF-α was from R&D Systems. Flagellin was purified as described previously (23). N-acetyl cysteine (NAC) and diphenyleneiodonium (DPI) were purchased from Sigma-Aldrich.

Confluent monolayers of T84 cells were grown in a polarized fashion on permeable supports as previously described (24). HeLa, Caco-2, and IEC-6 epithelial cells were grown in standard tissue culture vessels and maintained as recommended by the American Type Culture Collection. Human promyelocytic leukemia cells (HL-60) were maintained in RPMI 1640 medium supplemented with 10% (v/v) FBS and 100 U/ml penicillin and 100 U/ml streptomycin. The cultures were maintained in humidified 95% air, 5% CO₂, at 37°C by passage of 2 × 10⁵ cells/ml every other day. Differentiation was induced by treatment with PMA (20 ng/ml) for 48 h and the extent of differentiation was measured by attachment of cells to the substratum.

Healthy human colonic mucosa was obtained from the outer borders of fresh distal colonic specimens resected for neoplastic disease. Mucosa was dissected free from the underlying gut wall and 3-mm punches were incubated for 1 h in HBSS+ supplemented with butyrate at the indicated concentration. Tissue was immediately lysed in denaturing SDS-PAGE buffer.

HeLa cells were transiently transfected using Lipofectamine 2000 (Invitrogen) or FuGENE 6 (Roche) according to the manufacturer’s instructions. Briefly, plasmid DNA was mixed with serum-free DMEM and transfection reagent. After 15 min, the mixture was added in a dropwise fashion to the growth medium of ∼70–90% confluent adherent cells in 24- or 6-well plates or 100-mm dishes. The cells were allowed 16–24 h for transfection and expression of transfected genes. For luciferase reporter assays, cells were transfected with pNF-κB-Luc plasmid (Stratagene), allowed 18 h for expression, treated with TNF-α (20 ng/ml) for 6 h, and lysed according to the manufacturer’s protocol. All transfections were balanced to contain 200 ng of DNA with empty vector. Luciferase activity was determined using the Dual Luciferase Reporter Assay System (Promega).

Following experimental treatment, epithelial cells were washed in cold HBSS+ and whole cell extracts were prepared by rapid lysis in denaturing SDS-PAGE buffer. Cell lysates were electrophoresed on SDS-polyacrylamide gels and transferred to nitrocellulose using standard protocols. Immunoreactive proteins were detected with Abs to IκB-α (Santa Cruz Biotechnology), phospho-IκB-α (Cell Signaling), Cul-1 (Zymed), NEDD8 (Zymed), Ubc12 (Rockland), β-tubulin (Sigma-Aldrich), or β-actin (Sigma-Aldrich) using the enhanced chemiluminescent protocol (ECL; Amersham) and a HRP-conjugated secondary Ab. Blots were exposed to film for 1 s to 30 min.

Immunofluorescent labeling of p65 in adherent HeLa cells grown on 12-mm glass coverslips was performed as follows: cells were fixed for 20 min in 3.7% paraformaldehyde in PBS, washed in PBS, permeabilized with 0.1% Triton X-100 in PBS for 5 min, and washed again. Fixed samples were incubated in blocking solution (5% normal goat serum in PBS) overnight at 4°C. A 1-h incubation with each Ab diluted in blocking buffer followed: 1/500 rabbit anti-p65 (Rockland); 1/200 fluorescein (FITC)-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories). Cells were washed three times between each Ab. The coverslips were mounted on glass slides and stained cells were observed by laser confocal epifluorescence microscopy (Zeiss).

For monitoring ROS generation, IEC-6 cell cultures were plated onto 24-well plates and treated with butyrate for the duration indicated. Following treatment, the cells were further incubated with 5 μM each of the fluorescent ROS indicators 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate acetyl ester (DCF) (Molecular Probes) or dihydroethidium (DHE) for 15 min. After loading the ROS dyes, cultures were washed three times with warm HBSS+ and the fluorescence intensity was acquired using a confocal laser-scanning microscope (Zeiss LSM 510). For DCF imaging, images were captured with a 10× objective using an excitation wavelength of 488 nm and an emission wavelength of 513 nm. For DHE, images were captured with a 40× objective using an excitation wavelength of 490 nm and an emission wavelength of 610 nm. For determination of ROS generation in mitochondria, we used the mitochondrial superoxide dye indicator MitoSOX Red (Molecular Probes). Cells were treated with butyrate for the times indicated, after which they were loaded with 5 μM MitoSOX Red in HBSS⁺ for 10 min at 37°C. Cells were washed with warm HBSS⁺ and fluorescent images were captured with a 63× objective using an excitation wavelength of 510 nm and an emission wavelength of 580 nm.

ROS generation was measured with DCF as described by Wang et al. (25), with the following modification. Caco-2 cells grown to confluence on a 24-well plate were incubated in DMEM supplemented with 0.4% serum and containing 1 mM DCF dye. After incubation for 2 h in 5% CO₂ at 37°C, the cells were washed and treated with Lactobacillus rhamnosus GG (multiplicity of infection (MOI) = 1) or with butyrate (10 mM) in Krebs-Ringer-HEPES buffer. An increase in fluorescence units was measured by using a fluorescence microplate reader (SpectraMax M2; Molecular Devices) after various time intervals with excitation at 475 nm and emission at 525 nm.

For Thioredoxin (Trx) 1 (BD Biosciences) analysis, Caco-2 cells treated with butyrate were washed with cold 1× PBS and whole cell lysates were prepared in G-lysis buffer (6 M guanidine-HCl, 50 mM Tris (pH 8.3), 3 mM EDTA, 50 mM iodoacetamide (IAA)) as described in (26). Briefly, lysates were incubated for 30 min at 37°C, after which excess IAA was removed using a G-25 spin column (GE Healthcare). The samples were electrophoresed on a native gel, transferred to nitrocellulose membrane, probed with mouse anti-human Trx1 (BD Biosciences), and detected with an Alexa Fluor 680 anti-mouse secondary Ab. Trx2 redox states were analyzed as described in a previous report (26). Following treatment with butyrate, total cell protein was precipitated with 10% trichloroacetic acid. Samples were centrifuged, washed with acetone, and resuspended in 20 mM Tris (pH 8.0) buffer containing 15 mM 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid Molecular Probes)]. Samples were incubated for 30 min at room temperature. Oxidized and reduced forms of Trx2 were separated by nonreducing SDS-PAGE. Trx2 was detected using rabbit anti-human Trx2 Ab and Alexa Fluor 680 anti-rabbit secondary Ab. Bands corresponding to Trx1 and Trx2 were visualized using an Odyssey scanner (LI-COR) and band densometric values were applied to the Nernst equation as described elsewhere (27).

We had previously shown that colonization of epithelial cells with viable commensal bacterial strains in vitro and in vivo caused a rapid and reversible loss of epithelial Cul-1 neddylation (21). During the course of these experiments, we observed a reduction of Cul-1 neddylation if bacteria were physically separated from the epithelial cells by use of Transwell inserts mounted with a permeable membrane support (data not shown). This indicated that the bacterial effects, at least in part, were mediated by a small diffusible molecule. To identify this soluble signal, we analyzed bacterially conditioned supernatants by gas chromatography/mass spectroscopy, among other approaches. These studies revealed the presence of significant amounts of lactate in culture supernatants. Consistently, the pH of supernatants after a minimum of 30 min of coculture was 4.5–6.0, indicating the microbial production of acidic compounds. It is well known that the fermentation products of the intestinal flora include lactate, as well as SCFAs such as butyrate (C4), succinate (C4), and propionate (C3) (4, 5). Furthermore, butyrate mediates well-known effects on inflammation, including repression of NF-κB in vivo and in vitro (8, 9, 28, 29). These observations suggested the possibility that the loss of neddylated Cul-1, at least partially, resulted from the exposure of the cells to bacterial fermentation products. When fully polarized T84 epithelial cells were treated with 5 mM concentrations of butyrate, loss in Cul-1 neddylation was observed within 15 min (Fig. 1,A). Typically Cul-1 appears as a doublet of 85 and 90 kDa, with the higher molecular band representing the neddylated Cul-1. When an identical blot was probed with anti-NEDD8 Ab, the upper band reacted with the Ab, confirming the identity of this band as neddylated Cul-1 (Fig. 1,A, bottom panel). Similar kinetics of butyrate-mediated Cul-1 deneddylation was also observed in HeLa cells (Fig. 1,B, top panel) and in human promyelocytic leukemia cells (HL-60) differentiated to form macrophage-like cells (Fig. 1,B, bottom panel). Interestingly, deneddylation of Cul-1 observed upon treatment with butyrate could be reversed if the butyrate was washed off and the cells allowed to recover in HBSS+ for another 30–60 min (Fig. 1,C). The kinetics of reversibility of Cul-1 neddylation was similar as we observed with viable bacteria (21). Furthermore, loss in Cul-1 neddylation within 1 h was also observed when T84 cells were cultured in HBSS+ supplemented with at least 3 mM concentrations of other organic acids, such as lactate, propionate, and succinate (Fig. 1 D). In all of these in vitro conditions, the molarity of the organic acids was within the range reported for the physiological gut (8, 30).

In T84 cells, butyrate at a concentration of ∼3–5 mM (resulting in a pH of ∼4.5–5.0) was effective in causing loss of Cul-1 neddylation (Fig. 1,E, top panel). Interestingly, when effective concentrations of butyrate were buffered to neutral pH, the ability of this agent to affect Cul-1 neddylation was lost (Fig. 1,E, middle panel). To rule out artifacts caused by acidic pH conditions, T84 cells were similarly treated with HBSS+ supplemented with hydrochloric acid to pHs at which butyrate could induce loss of Cul-1 neddylation (Fig. 1 E, lower panel). These conditions had no effect on Cul-1 neddylation, indicating that although the inhibitory effect of butyrate was pH dependent, it was not a function of pH alone, likely because butyrate at pHs approaching its pKa (4.5) becomes protonated and is freely diffusible across cellular membranes. Thus, these data suggest that the butyrate effects observed likely do not require a receptor or active transport.

To determine whether the presence of SCFAs could influence Cul-1 neddylation in intact mucosa, we similarly treated human colonic mucosal segments ex vivo in medium supplemented with butyrate and assayed Cul-1 neddylation by immunoblot (Fig. 1 F). Reduction of Cul-1 neddylation was observed at concentrations of 1.5–3 mM butyrate. The lower concentrations required in this experiment presumably reflect the differences between intact, fully differentiated mucosa and cultured cells. Thus, by supplementing medium with SCFAs, we were able to recapitulate the effects of viable bacteria on the neddylation of Cul-1.

Butyrate is known to be a potent inflammatory modulator in vivo; it has been shown to block up-regulation of proinflammatory mediators (i.e., IL-8) and NF-κB activation in stimulated cultured intestinal epithelial cells (31). We thus undertook a mechanistic analysis of NF-κB signaling to determine whether SCFA effects on Cul-1 neddylation affected the NF-κB pathway. As shown in Fig. 2,A, in cells pretreated with 3 mM butyrate at pH 5.6 or with lactate at pH 5.0, we observed inhibition of p65 translocation, as evaluated by immunofluorescence staining of the normal TNF-α induced cytoplasmic to nuclear translocation, consistent with effects observed with intestinal epithelial cell colonization with live bacteria (32). Furthermore, as seen in Fig. 2,A, this butyrate-mediated block in p65 translocation was abrogated if butyrate was buffered to pH 7.4 or with lactate at pH 6.0, similar to the observed effects of pH on neddylation of Cul-1 in Fig. 1,D. This blockade was functionally relevant because the same or lower concentration of butyrate inhibited activation of a NF-κB- dependent reporter gene in a transient transfection assay (Fig. 2 B). Thus, butyrate pretreatment of cultured cells inhibited activation on NF-κB, consistent with previous reports of butyrate-mediated effects on inflammatory cytokine induction.

Activation of the NF-κB pathway is tightly regulated by the phosphorylation, ubiquitination, and proteolysis of its physically associated inhibitor molecule, IκB-α. Yin et al. (29) reported that prolonged exposure (24 h) of HT29 cells to butyrate suppressed the NF-κB activation via attenuation of the cellular proteasome activity. To determine whether short-term exposure (as used in our study) of intestinal epithelial cells to butyrate inhibited cellular proteasome activity, we treated intestinal epithelial Caco-2 cells with butyrate and monitored cellular proteasome activity by observing the accumulation of polyubiquitin-conjugated proteins. Pretreatment of Caco-2 cells with 2.5 mM butyrate for 1 h had no gross effects on accumulation of polyubiquitinated proteins (Fig. 2,C, compare lanes 1 and 5). We also monitored the accumulation of the ubiquitinated proteins in the presence of MG-262, an inhibitor of the 26S proteasome. In cells pretreated with butyrate and further incubated with MG-262 for several hours, we saw no gross differences in accumulation of polyubiquitinated proteins (Fig. 2 C, compare lanes 2–4 to 6–8, respectively). These data indicate that within the time frame used in this study butyrate does not suppress accumulation of cellular polyubiquitinated proteins.

Next, to determine whether butyrate-mediated inhibition of NF-κB was related to ubiquitination of IκB-α, a process controlled by Cul-1 neddylation, we undertook an analysis of IκB-α ubiquitination. Pretreatment of cells with butyrate before stimulation with proinflammatory inducers such as TNF-α and flagellin resulted in stabilization of IκB-α in contrast to near to complete degradation observed with TNF-α and flagellin, respectively (Fig. 2,D, top panel). To allow further study of phosphorylated and ubiquitinated IκB-α adducts, we pretreated our model epithelia with MG-262. Under these conditions, degradation of IκB-α in response to TNF-α and flagellin was attenuated (Fig. 2,D, bottom panel). Furthermore, very high molecular bands consistent with polyubiquitinated IκB-α was observed in cells stimulated with TNF-α and flagellin, as previously reported (33) (Fig. 2,D, bottom panel). However, these higher molecular mass species representing polyubiquitinated IκB-α were absent in cell lysates derived from butyrate-treated cells (Fig. 2,D, bottom panel). Since ubiquitination of IκB-α is dependent upon phosphorylation of IκB-α on two serine residues (Ser32 and 36), we also examined the phosphorylation of these residues with a phospho-specific Ab. Again, when we pretreated the cells with MG-262 before induction with TNF-α or flagellin, we observed an increase in high-molecular mass (HMM) ubiquitinated species that was attenuated in the presence of butyrate (Fig. 2,E). Importantly, phosphorylated IκB-α (denoted with asterisks in Fig. 2,E) was abundant in control and enhanced in butyrate-treated cell lysates (Fig. 2 E). These results are consistent with the bacterial mediated inhibition of NF-κB previously described (21, 33). Collectively, these data indicate that butyrate prevents IκB-α degradation by blocking ubiquitination of phospho-IκB-α.

Neddylation of Cul-1, required for activity of the E3-SCF^β-TrCP ubiquitin ligase, is dependent on sequential transfer of thiolester NEDD8 from the NEDD8-charging enzyme Uba3-APP/BP-1 to a catalytically active cysteine residue of the NEDD8-conjugating enzyme Ubc12. Ubc12 then catalyzes the formation of an isopeptide bond between the C-terminal glycine residue of NEDD8 and a target lysine (Lys⁷²⁰) residue present on Cul-1 (34). We have previously demonstrated that viable commensal bacteria cause transient oxidative inactivation of Ubc12 with resultant effects on neddylation of Cul-1 (22). Given that butyrate treatment resulted in potent Cul-1 deneddylation, we asked whether SCFAs were also stimulating ROS generation with resultant oxidation of Ubc12. To monitor ROS generation, IEC-6 intestinal epithelial cells were cultured with physiological concentrations of butyrate and monitored for changes in DCF fluorescence, a dye sensitive to ROS, especially H₂O₂. An increase in fluorescence of DCF was observed within 15 min of treatment with 10 mM butyrate (Fig. 3, A and D) as compared with untreated controls. An increase in ROS was also confirmed with the use of an alternative ROS sensitive dye, DHE (Fig. 3,B). DHE is freely permeable and is oxidized to fluorescent ethidium bromide by superoxides. We have previously shown that commensal bacteria elicited a rapid and transient increase in ROS within 5–30 min of contact with intestinal epithelial cells (22). In comparison to treatment with L. rhamnosus at MOI = 1, butyrate stimulated ROS with a slower kinetics that occurred over a period of 30–60 min after contact with intestinal epithelial cells (Fig. 3,D). To evaluate the sources of butyrate-mediated ROS generation, we used a mitochondria-specific ROS fluorescent dye, MitoSOX Red, that specifically detects superoxide generation in mitochondria. Using this dye, we observed mitochondrial ROS generation within 15 min of treatment with butyrate, suggesting the mitochondria as a potential source of induced ROS (Fig. 3 C).

Accumulation of intracellular levels of ROS in cells causes changes in the oxidation profiles of the antioxidants. To evaluate changes in such intracellular redox pools, we assayed the thiol/disulfide redox states of thioredoxins, a major thiol-dependent antioxidant system of cells. Steady-state E_h values for cytosolic Trx1 (Fig. 3,E) and mitochondrial Trx2 (Fig. 3 F) were calculated using the Nernst equation and show that treatment of cells with butyrate resulted in oxidation of both Trx1 and Trx2 over a 1-h time course. For Trx1, the E_h increased from –283 mV in untreated cells to −265 mV within 30 min in cells treated with 10 mM butyrate. For Trx2, the E_h increased from −359 mV to −336 mV within 15 min with 10 mM butyrate treatment. These cumulative data indicated that butyrate treatment induced ROS in cultured epithelial cells predominantly from mitochondrial sources and induced compensatory changes in the intracellular redox balance, albeit with slower kinetics and to a lesser degree as ROS induced by viable commensal bacteria.

We next sought to examine whether butyrate-mediated ROS signals were being transduced via transient oxidation of Ubc12. To evaluate the effects of butyrate on endogenous Ubc12, HeLa cells were treated with 5 and 10 mM butyrate and cell lysates were analyzed under reducing and nonreducing conditions (absence of DTT) to allow visualization of thiol modifications as previously described (22). When cells were treated with 5 and 10 mM butyrate, we observed a loss of the 30-kDa NEDD8∼Ubc12 thiolester form when cell lysates were prepared and analyzed under nonreducing conditions (Fig. 4,A, top panel, denoted with an asterisk). Significantly, this loss of NEDD8∼Ubc12 thiolester form corresponded with the appearance of a marked dose-dependent shift in the electrophoretic mobility of Ubc12 to HMM-oxidized species (Fig. 4,A, top panel, denoted with an arrow). We have previously confirmed the identity of the NEDD8∼Ubc12 thiolester form in a similar immunoblot probed with anti-NEDD8 Ab under nonreducing conditions (22). In cell lysates prepared and analyzed under standard reducing conditions (with DTT), the HMM-oxidized species were abolished (Fig. 4 A, bottom panel), indicating that these forms of the enzymes are mixed disulfides between Ubc12 and HMM proteins. Furthermore, in cell lysates analyzed under reducing conditions, we did not observe the 30-kDa NEDD8∼Ubc12 thiolester form, further confirming the identity of this form of Ubc12 as being neddylated.

Finally, to link the observed induction of intracellular ROS with deneddylation of Cul-1, HeLa cells were treated with butyrate in the presence of a ROS scavenger NAC and a flavoprotein inhibitor DPI. As shown in Fig. 4 B, pretreatment of cells with NAC or DPI attenuated butyrate-mediated deneddylation of Cul-1. Collectively, these data demonstrate that butyrate-mediated ROS generation causes attenuation of Ubc12 activity in the same manner as of live bacteria.

Herein, we have described an important eukaryotic regulatory node, namely, ubiquitin-mediated degradation by E3-SCF ubiquitin ligases, that is influenced by the product of a complex prokaryotic community. Active E3-SCF ubiquitin ligases are required for the regulation of the NF-κB, β-catenin, Snail, Twist, and Hedgehog pathways and presumably others (35). The activity of the E3-SCF ubiquitin ligases is modulated by neddylation of its Cul-1 subunit (17). Neddylation is a posttranslation modification specific to substrates of the Cullin family in which NEDD8, a small 8-kDa ubiquitin-like protein, is attached to the substrates and is emerging as a key regulatory event in cellular processes that are controlled by ubiquitin-mediated degradation of proteins. In a variety of experimental models, including yeast (36), Arabidopsis (37), Caenorhabditis elegans (38), Drosophila (39), mice (40), and human cells in vitro (41), mutations leading to loss-of-function of Cul-1 neddylation (or deneddylation) status can have profound functional consequences. However, only limited evidence is available implicating environmental signals in the physiological regulation of Cul-1 neddylation. For example, yeast Cullin was shown to be hyperneddylated in response to UV exposure (42). In Arabidopsis studies, wild-type plants showed marked Cul-1 deneddylation when reared in darkness, while mutant plants that fail to respond to light were shown to have an abnormal accumulation of neddylated Cul-1 (43). Furthermore, our previous results demonstrate that interactions of commensal bacteria with mammalian intestinal epithelia modulate Cul-1 neddylation (21) along with exogenous H₂O₂ (22), adenosine (44), and, in this report, bacterial fermentation products. We found commensal bacteria were able to mediate effects on Cul-1 neddylation via soluble bacterial fermentation products. Such SCFAs, particularly butyrate, have long been known to provide a nutritive source for the mammalian gut. Moreover, butyrate has well-known immunomodulatory and trophic effects on gut homeostasis (6). In this sense, these compounds possess the ability to act as effectors or signaling intermediates in the symbiotic cross-talk between the mammalian host and the normal flora. Unlike other intermediates in the dialog between host and microbe, such as eukaryotic pattern recognition receptors and bacterial MAMPs, butyrate-mediated signaling occurs via largely unknown processes.

Recent years have seen the establishment of the paradigm stating that inducible and “deliberate” generation of ROS have a wide variety of physiological signaling functions. Several eukaryotic signaling proteins, including growth factors, hormones, and cytokines, are capable of stimulating ROS generation in various cell types including epithelial cells (45, 46, 47, 48, 49). Physiological generation of these molecules can derive from mitochondrial sources, dedicated enzymes homologous to the phagocytic NADPH oxidase, or by 5′-lipoxygenase (50, 51). ROS are short-lived molecules and can exhibit exquisite microspatial localization within cells, allowing specific targeting of signaling effects (51). ROS signaling can be transduced by an increasingly recognized subset of enzymes that can be transiently inactivated by reversible oxidation of catalytic cysteine residues within the active sites (52, 53). Such enzymes include a variety of tyrosine phosphatases such as PTEN (54), the DUSP family of MAPK phosphatases (45), antioxidants such as thioredoxins and peroxiredoxins (55), and, as we and others have shown, members of the Ubc family of proteins (22, 56).

In the gut, sources of ROS could be exogenous, deriving from the metabolic processes of certain bacteria themselves or from infiltrating neutrophils during an acute inflammatory event. Alternatively, we have found ROS generated within epithelial cells stimulated by exogenous agents such as commensal bacteria (22), soluble fermentation products (this report), and sterile components of bacterial cell walls (N-K. Young and A. S. Neish, unpublished data). This is likely a highly conserved process, since ROS generation as a response to environmental stressors, including bacteria, is present in plants and lower metazoans (57). ROS generation by butyrate has been reported, although with far more delayed kinetics. Stimulation of ROS generation has been observed upon exposure of human tongue cancer cells to 8 and 16 mM butyrate for a period of 4–24 h (58, 59). Of note, although in our hands ROS generation elicited by butyrate was at least as strong as ROS induced by known ROS-dependant agonists such as insulin, epidermal growth factor, flagellin, or platelet-derived growth factor (A. Kumar, M.-K. Young, and A. S. Neish, unpublished results), butyrate-induced ROS was markedly less than the amounts detected after cells were stimulated with viable bacteria and occurred with noticeably slower kinetics. As suggested by our data, butyrate may induce ROS via the mitochondrial pathway rather than by receptor-mediated events.

The physiochemical environment of the intestinal lumen in immediate contact with the epithelium is largely anaerobic and is suffused with not just with a complex ecosystem of living bacteria, but also a miasma of MAMPs, small bacterially produced compounds, and a spectrum of metabolic by-products that varies markedly over the greater than 1 m length of the organ (4, 5, 8). The ascending colon contains a highly fermentative, “saccharolytic” bacterial milieu which produce high concentrations (up to 140 mM for butyrate) of organic acids including lactate, acetate, and SCFAs (succinate, propionate, butyrate), resulting in a physiological luminal pH of 5.5–6.5, as measured by a series of radiotelemetry studies in healthy human volunteers (reviewed in Ref. 30). As the carbohydrate-fermentative substrates are consumed through β oxidation and bicarbonate is secreted by the epithelia, the luminal environment shows a progressive decline in SCFA concentration (to 40 mM for butyrate) and a corresponding rise in pH, which in the descending colon is near neutral. SCFAs are weak organic acids, with pKa values generally in the range of 4.5–5.0, not coincidentally overlapping the range where loss of Cul-1 neddylation was observed (60). At these pH’s, the protonated forms of the acids are lipid soluble, membrane permeable, and exhibit maximal bioactivity (61, 62). Thus, different regions of the colon are exposed to significantly different concentrations of bioavailable SCFAs and no doubt other chemical parameters as well. Possibly, regions of the colon exposed to maximal bioavailable SCFA concentrations may have attenuated inflammatory potential. Intriguingly, this anatomical distribution of the gradient of SCFA concentrations is the precise inverse to the anatomical distribution of the inflammatory lesions of acute ulcerative colitis, which is invariably seen in the distal colon and progressively attenuates proximally (63).

Excessively high concentrations of SCFAs may have deleterious clinical consequences. Necrotizing enterocolitis is a spontaneous complication of prematurity that has been suggested to be mediated by excessive levels of SCFAs (64, 65). Empirically, the pHs of luminal contents in patients with this disorder have been measured at 3.8–4.6. Other workers have induced intestinal injury in rodents with very high doses of butyrate (300 mM/pH 4.0 with normal being 70–140 mM/pH 5.5 in the proximal colon) and other SCFAs (66). Interestingly, it was also shown that butyrate anion at neutral or alkaline pH had no effect nor did low pH alone, consistent with our observations in vitro (67). The transient nature of Cul-1 deneddylation indicates that enterocytes can compensate and adapt to dynamic changes; however, extreme conditions may limit the ability of enterocytes to adapt. Given the wide variety of cellular processes other than NF-κB and the Wnt signaling pathway activation that are regulated by SCF-mediated protein-degradative processes, it is not surprising that excessive exposure to very high levels of compounds that cause loss of Cul-1 neddylation could inhibit many enzymatic processes and be rapidly toxic to the cell.

Luminal physiochemical properties are dynamic and may be manipulated in a therapeutic fashion by high-starch dietary supplements (prebiotics) or oral supplementation with live bacteria (probiotics) (68, 69). Interestingly, one group of microbes with therapeutic value are the lactic acid bacteria, well-known Gram-positive enteric commensals that include the genera Lactobacillus, Enterococcus, and Bifidobacteria, among others, and are defined by a common ability to produce lactic acid under microaerophilic to anaerobic conditions (70). As part of the human enteric flora, they contribute to the physiological acidic pH of the intraluminal colonic environment (68) and lactate produced by this subset of organisms may also serve as substrate for butyrate production by others, a process termed metabolic cross-feeding (71). More recently, these organisms have been recognized as possessing marked probiotic activity both clinically (reviewed in Ref. 69) and in animal models (72). Conditioned medium from L. rhamnosis GG has been shown to repress proinflammatory gene expression in cultured epithelial cells (73). It has also been reported that conditioned medium from two of these bacteria, Bifidobacterium breve and Streptococcus thermophilus, exhibited inhibitory effects on NF-κB activation in vitro (74). Furthermore, this activity was attributed to a small (<3 kDa), soluble metabolite. These workers quantified lactate in the conditioned medium at 4.5–9 mM, but were unable to show NF-κB inhibition at these concentrations. Finally, the bacterial metabolites butyrate, propionate, and acetate have been shown to have inhibitory activity on proinflammatory signaling pathways. These compounds have been used therapeutically (by intraluminal instillation) to dampen intestinal inflammation in inflammatory bowl diseases (8, 9, 75) and have been shown to block NF-κB activation in vitro (28, 29).

Our observations are likely limited to effects of large bacterial communities with substantial metabolic outputs necessary to ferment sufficient quantities of organic acids and perhaps other products. As such, our observations probably do not reflect events in most acute enteric infections. Invasive enteric pathogens such as Salmonella, Shigella, etc., do not reach biomass densities of 10¹¹ organisms/ml before executing their effects on host cells, although pathogens that exist in microbe-dense biofilms could plausibly generate local microenvironments of higher metabolite concentrations.

The relationship of the normal flora to the intestinal epithelia is a prime example of a complex microbial community associated with a higher eukaryote (76). It may be an oversimplification to view this system as a binary host-pathogen interaction, with exchanges of individual and discrete biochemical signals, such as secreted effector proteins, toxins, or even small molecules with specificity toward eukaryotic cellular processes. The intestinal luminal flora produces a massive and diverse metabolic output that varies according to anatomical location, nutrient composition, and host health. It may be more illuminating to view the normal flora as a holistic community rather than as a collection of semi-independent organisms mediating “beneficial” or “detrimental” effects. Our results suggest a novel mechanism by which this complex community can collectively influence key epithelial signaling pathways. Metabolic products/small molecules produced at the eukaryotic/prokaryotic interface may account for some of the widely known effects of the bacterial flora on normal intestinal function (3) and may influence a range of eukaryotic regulatory processes.

Human Biopsy Samples were provided by Dr. Shanti Srinivasan. We thank Kirsten Gerner-Smidt for help with the confocal microscope.

The authors have no financial conflict of interest.