The Heme Oxygenase-1/Carbon Monoxide Pathway Suppresses TLR4 Signaling by Regulating the Interaction of TLR4 with Caveolin-1¹

Plasma membranes contain specialized lipid microdomains occurring as planar (lipid-rafts) or flask-shaped structures (caveolae) that are enriched in glycosphingolipids and cholesterol (1). The caveolae mediate nonclathrin-dependent endocytosis, and regulate the internalization of particles such as viruses (2) and bacteria (3, 4, 5). Caveolin-1 (cav-1),³ the major structural component of caveolae, exerts pleiotropic cellular functions including the regulation of cholesterol homeostasis, vesicular transport, proliferation, and apoptosis in a diversity of cell types (1). Cav-1 null mice develop cardiac and pulmonary hypertrophy and fibrosis (6, 7). Furthermore, cardiac fibroblasts derived from cav-1 null mice display deregulated signaling pathways, with hyperactivation of ERK1/2 MAPK and NO synthase (NOS) activity (6). Such observations have indicated a critical role for caveolae and cav-1 in cellular signal transduction. Indeed, numerous transmembrane growth factor receptors (e.g., platelet-derived growth factor, epidermal growth factor, and nerve growth factor receptors) and other diverse signaling molecules (e.g., GTPases) localize to caveolae (8). Cav-1 can regulate membrane receptor signaling either by direct binding to the receptor, or downstream molecules, through interactions mediated by its scaffolding domain (8).

Recently, cav-1 has also been implicated as a modulator of innate immunity and inflammation (9, 10, 11, 12). Cav-1 null mice displayed increased susceptibility to Salmonella infection, whereas macrophages derived from these mice exhibited enhanced inflammatory responses to bacterial LPS (9). The effects of cav-1 on inflammatory processes are potentially mediated through the regulation of NF-κB and NOSII/III activities (10, 11, 12). Previously, we have demonstrated an anti-inflammatory function of cav-1 expression in vitro with respect to the inhibition of proinflammatory cytokines production during LPS-induced inflammation in macrophages (13).

Genetic studies in mice have identified TLR4 as the principle membrane receptor for LPS (14). TLRs function as primary sensors of pathogens, which activate signaling pathways leading to the expression of cytokine genes (14). Inflammatory signaling initiates when LPS binds to the acute-phase protein LPS-binding protein, which is recognized by TLR4 and by CD14, a glycosyl-phosphatidylinositol-anchored protein lacking a cytoplasmic domain. Following ligand binding, a TLR4 complex is assembled, composed of CD14, TLR4, MD-2, MyD88, and other adaptors. Complex formation is followed sequentially by the intracellular activation of signaling mediators that include IL-1R-associated kinase, Toll/IL-1R domain-containing adaptor-inducing IFN-β (TRIF), PI3K, MAPK, and NF-κB (14). We and others have recently revealed the importance of caveolae, lipid rafts, or both in TLR4 signaling. LPS stimulates the recruitment of the TLR4 and MD-2 complex to lipid rafts containing CD14 in monocytes and RAW 264.7 macrophages (15, 16), or to caveolae fractions in human aortic endothelial cells (17). Nystatin, a drug that disrupts the integrity of lipid rafts, suppressed LPS-induced secretion of TNF-α (15). A primary goal of the present study was to delineate the relationship between TLR4 and cav-1, and to explore the regulation and functional significance of this interaction.

The heme oxygenase (HO)/carbon monoxide (CO) pathway has emerged as a major inhibitor of inflammatory processes (18). HO enzymes catalyze the oxidative catabolism of heme to biliverdin-IXα, iron, and CO, and thus represent the principle endogenous source of CO in biological systems. NADPH biliverdin reductase (BVR) completes the heme degradative pathway by reducing biliverdin-IXα to bilirubin-IXα. Both the stress inducible isozyme HO-1 and exogenous CO exert anti-inflammatory activity in multiple cellular and animal models, including sepsis, experimental or transplant-associated ischemia/reperfusion injury, and oxidative lung injury (18, 19). CO inhibited LPS-induced TNF-α production and increased IL-10 production in RAW 264.7 cells and in vivo (19).

In this study, we demonstrate a unique and heretofore uncharacterized interaction of cav-1 with TLR4, with functional implications as one of the most apical regulatory events in LPS signaling with respect to the downstream production of proinflammatory cytokines. We also demonstrate a functional relationship between the HO-1/CO pathway and the attenuation of TLR4 signaling by cav-1, which provide novel mechanisms for the regulation of inflammation in macrophages.

The Abs used were as follows: anti-TLR4, anti-cav-1, anti-NADPH cytochrome p450 reductase (Santa Cruz Biotechnology); TRIF (Abcam); myeloid differentiation primary response gene 88 (MyD88; Chemicon International); HO-1 and BVR (Stressgen Biotechnologies); β-actin (Cell Signaling Technology); FLAG, IgG, and the cholera toxin B subunit (peroxidase conjugate) to detect GM1 (Sigma-Aldrich). Hemin, tin-protoporphrin-IX (Sn-PPIX; Frontier Scientific), and SB203580 (Calbiochem) were dissolved in DMSO as concentrated stock solutions. The protein kinase C (PKC) inhibitor GF109203X, brefeldin A, and all other chemicals were from Sigma-Aldrich.

Primary murine peritoneal macrophages were isolated from wild-type or genetically modified mice as previously described (13). RAW 264.7 macrophages were from American Type Culture Collection. Both cell types were maintained in DMEM containing 10% FBS and 50 μg/ml gentamicin (Life Technologies). Cav-1 expressing stable transfected RAW 264.7 cells, and corresponding vector (pcDNA3.1)-transfected control cells were generated and maintained as described (20). All cultures were maintained at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. The cells were serum starved before 100 ng/ml LPS (Escherichia coli serotype O127:B8; Frontier Scientific) stimulation. Hemin or Sn-PPIX was added to culture medium at a final concentration of 20 μM. SB203580 was added to the culture medium 1 h before other treatments at a final concentration of 10 μM.

C57BL/6, C3H/HeJ (TLR4-deficient), and cav-1-deficient (cav-1^−/−) C57BL/6 mice were purchased from The Jackson Laboratory and acclimated 1 wk before experiments. MAPK kinase-3-deficient (MKK3^−/−) mice were generated as previously described (21). All animals were housed in accordance with guidelines from the American Association for Laboratory Animal Care and Research Protocols and were approved by the Animal Care and Use Committee of University of Pittsburgh, School of Medicine.

CO exposures were performed as previously described (3). CO at a concentration of 1% corresponding to 10,000 parts per million (ppm) in compressed air was mixed with compressed air containing 5% CO₂ in a stainless steel mixing cylinder before delivery into the exposure chambers. The flow rate into the chamber used for cell culture exposures was 2 L/min. The chamber was humidified and maintained at 37°C. For animal exposures, CO was mixed with air without CO₂ supplementation. Flow into the 3.70-ft² plexiglass animal chamber was maintained at rate of 12 L/min. The final CO levels in the chambers (250 ppm) were monitored using a CO analyzer (Interscan Corporation). There were no fluctuations in the CO concentrations after the chamber had equilibrated.

Macrophages were homogenized in 2 ml MBS buffer (25 mM MES (pH 6.5), 0.15 M NaCl) with 1% Triton X-100. The homogenates were transferred into a 12-ml ultracentrifuge tube with addition of 2-ml 80% sucrose in MBS. The 4 ml of 35% sucrose and 4 ml of 5% sucrose in MBS were added, respectively, to form the discontinuous sucrose gradient. The tubes were balanced and centrifuged at 39,000 rpm for 18 h in an SW41 rotor (Beckman Instruments). The 12 subfractions were collected. For detergent-resistant fraction isolation, cells were homogenized in 0.5 ml of MBS buffer with 1% Triton X-100 and subjected to centrifugation at 12,000 rpm for 10 min. The postnuclear supernatant were collected and subjected to ultracentrifugation at 35,000 rpm for 1 h using a TOTI rotor (Beckman Instruments). The pellet was regarded as the detergent-resistant fraction.

Western immunoblot analysis was performed as previously described (13).

Cells were washed twice with cold PBS and homogenized in RIPA buffer (PBS with 1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS (pH 7.5)) with complete protease inhibitor mixture (Roche Diagnostics) for 30 min on ice. After centrifugation for 20 min at 14,000 × g at 4°C, the supernatant was collected. Protein concentrations were determined and adjusted to 1 μg/μl. For fraction samples, two volumes of MBS buffer were added. The samples were subjected to ultracentrifugation for 1 h at 29,000 rpm using a TOTI rotor (Beckman Instruments). The pellet was dissolved in 1 ml of RIPA buffer. A total of 0.5 ml was used for each immunoprecipitation. The 1 μg of Ab or IgG was added into 0.5-ml protein extracts. The mixtures were rotated at 4°C overnight. After 25 μl of protein A-agarose beads (Santa Cruz Biotechnology) were added, the mixtures were rotated for 2 h at 4°C. The beads were harvested by centrifugation and washed with RIPA buffer for five times. Loading buffer was added and boiled for 5 min. The samples were subjected to immunoblot analysis.

Cytokine (TNF-α, IL-6) levels in the medium were measured by ELISA kits (R&D Systems) according to the manufacturer’s protocol. Multiple cytokine analysis was performed using Bio-Plex Multiplex cytokine assay kit (Bio-Rad) according to the manufacturer’s protocol.

Cav-1 adenovirus was generated as described (22). For cell infection, 2 × 10⁵ cells were cultured in 6-well plates and exposed to 2 × 10⁷ PFU of each virus in 1 ml of serum-free medium for 4 h. The cells were then washed and incubated in serum-containing medium for 36 h. The cells were then subjected to other treatment as indicated.

Control siRNA (siCONTROL nontargeting siRNA) and siRNA for mouse caveolin-1 (siGENOME SMART pool siRNA) were from Dharmacon. Macrophages were transfected with siRNA as described before (13).

Cells were rinsed with PBS and fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.2). Postfixation was conducted in 1% osmium tetroxide in the same buffer. Samples were dehydrated in ethanol series and embedded in London Resin White resin. Ultrathin sections were cut and blocked with normal goat serum, which was diluted in PBS with 1% BSA. Sections were incubated with the primary Ab for cav-1 (Santa Cruz Biotechnology), or HO-1 (Stressgen Biotechnologies). After three washes, sections were incubated with secondary Ab conjugated with 5 and 10 nm diameter colloidal gold. After three washes, the sections were stained with uranyl acetate and lead citrate. The samples were visualized on a Zeiss EM 10 transmission electron microscope.

Cells were seeded on cover slips at a density of 5 × 10⁴ per well (24-well). After rinsing three times with PBS, cells were fixed in 2% paraformaldehyde. The cover slips were washed three times in 0.5% BSA and blocked for 1 h in 2% BSA. After three washes in 0.5% BSA, the slips were incubated 1 h with the Ab specific for cav-1 (Santa Cruz Biotechnology), or HO-1 (Stressgen Biotechnologies). The slips were then washed five times in 0.5% BSA and incubated for 1 h with Cy3- or Alexa 488–conjugated secondary Abs (Jackson ImmunoResearch Laboratories). The slips were washed five times with 0.5% BSA and PBS and then mounted. Cells were viewed with an Olympus Fluoview BX 61 confocal microscope, and images were collected using a DC-330S cooled CCD camera (DAGE-MTI).

pCMV-TLR4 tagged with Flag was a gift from Dr. T. Billiar (University of Pittsburgh, Pittsburgh, PA). Two primers (5′-ctt tat tca gag ccg tgc atg cat cgc tga ata tga gat tgc-3′ and 5′-gca atc tca tat tca gcg atg cat gca cgg ctc tga ata aag-3′) were designed to create mutant TLR4 with the new digestion site NsiI and SphI. PCR was used to generate mutant TLR4. The PCR products were digested with DpnI to destroy template plasmids and transformed into DH5α competent cells. Digestion and sequencing were used to confirm the mutation. All subcloning reagents were from Stratagene.

HO activity was determined as described (20, 23). Briefly, fractions were harvested and protein concentration was measured using the Coomassie (Bradford) protein assay kit (Pierce Biotechnology) with BSA as a standard. The final reaction contained 25 μM heme, 2 mM glucose 6-phosphate, 2 U of glucose 6-phosphate dehydrogenase, 1 mM β-NADPH, 1 mg/ml protein extracts, and 2 mg/ml partially purified rat liver BVR extracts. After incubation at 37°C for 60 min in the dark, two volumes of chloroform was added into the reaction for termination. The chloroform extracts were isolated after the centrifugation. OD 464 and 530 nm were measured with Beckman DU640 scanning spectrophotometer (Beckman Instruments). OD_464–530 was calculated. HO activity was expressed as picomole of bilirubin per milligram of protein per hour with the extinction coefficient of 40 mM/cm for bilirubin in chloroform (20, 23).

The differences in cytokine expression were compared using unpaired Student’s t test. The cytokine production follows a normal distribution as verified by the Anderson-Darling goodness-of-fit test. The variances are equal across groups as tested by the Bartlett test. All analyses were performed with SPSS 12.0.1 for Windows, and were considered significant at p < 0.05.

We have identified a cav-1 binding motif (24) in the mouse TLR4 amino acid sequence ((⁷³⁹)FIQSRWCIF(⁷⁴⁷)) predicted to occur in the carboxyl-terminal intracellular domain (see sequence in supplemental Fig. S1A).⁴ Therefore we hypothesized that cav-1 binds to TLR4, and consequently influences the downstream regulation of cytokines. This novel interaction was detected by coimmunoprecipitation in wild-type peritoneal macrophages (Fig. 1^,A), in macrophages transfected with TLR4 expression vector (Fig. 1^,B), and in caveolae fraction generated by sucrose density ultracentrifugation (Fig. 1^,C). We generated a mutant TLR4 containing a substitution of consensus tryptophan and phenylalanine residues for alanine (W744A/F747A) in this cav-1 binding motif (see sequence in supplemental Fig. S1A).⁴ When macrophages were transfected with mutant TLR4 (heretofore designated W744A), the interaction between TLR4 and cav-1 was abolished (Fig. 1 B). The relative expression from the TLR4 constructs is shown (see supplemental Fig. S2A).⁴

Next we examined the regulatory role of cav-1 on LPS-induced cytokine production, in stable-transfected RAW 264.7 macrophages expressing cav-1 (13, 20), using a Luminex-based multiplex cytokine assay. In addition to TNF-α and IL-6, as we previously described (13), cav-1 expression markedly suppressed LPS-inducible IL-1β, GM-CSF, and RANTES protein production and marginally suppressed IL-2, IL-4, IL-12 (p40), IL-12 (p70), and IFN-γ protein production in RAW 264.7 cells, relative to vector-transfected controls (Table I).

The effect of the TLR4/cav-1 interaction on downstream cytokines production was evaluated in macrophages isolated from C3H/HeJ mice, which express a functionally deficient TLR4 containing a missense Q712H mutation (25). Against this TLR4-deficient background, C3H/HeJ macrophages were transfected with either wild-type or mutant (W744A) TLR4, and the relative TNF-α and IL-6 production after LPS stimulation are shown in Fig. 1^,D. Although TNF-α production was elevated in C3H/HeJ macrophages transfected with wild-type TLR4 in response to LPS, these levels were further augmented in C3H/HeJ macrophages transfected with mutant (W744A) TLR4, lacking the functional cav-1 binding domain. These results indicate that transfection with W744A mutant results in increased TLR4 signaling at the same TLR4 expression level, consistent with a negative regulatory role of the cav-1 binding site in TLR4 activation. Infection with adenovirus cav-1 diminished LPS-inducible TNF-α and IL-6 production in C3H/HeJ macrophages reconstituted with wild-type TLR4, but did not affect LPS-inducible cytokines production in cells reconstituted with mutant TLR4 (Fig. 1 D). The relative expression from the TLR4 and cav-1 constructs is shown (see supplemental Fig. S2B).⁴

We have previously described that cav-1 expression suppressed the activation of NF-κB the pivotal proinflammatory pathway activated by LPS in macrophages (13). The LPS-dependent induction of NF-κB p65 phosphorylation and p65 nuclear translocation, as well as IκBα phosphorylation were also significantly suppressed in cav-1 overexpressing RAW 264.7 cells, relative to vector-transfected controls (data not shown). We also examined the effects of cav-1 on more proximal events in TLR4 signaling. Fig. 1,E shows that overexpression of cav-1 markedly attenuated the LPS-inducible protein-protein interaction between TLR4 and its binding partner, MyD88, and also delayed the association of TLR4 with TRIF (Fig. 1 E).

We sought to examine the relationship between the cav-1/TLR4 interaction and the HO-1/CO pathway, which is known to exert potent anti-inflammatory effects in macrophages with respect to LPS stimulation (19). We have recently shown that exogenous CO inhibited the trafficking of TLR4 to the lipid raft after LPS stimulation in macrophages (16). We have also reported that HO-1 localized in caveolae after LPS stimulation in endothelial cells (20) and after cadmium stimulation in mesangial cells (26), which represents a potential source of endogenous CO in caveolae. We next examined whether a similar phenomenon could be observed in macrophages.

As shown in Fig. 2 A, cav-1 was abundant in fraction nos. 4–5 obtained from sucrose density gradient ultracentrifugation of peritoneal macrophage cell lysates, which were regarded as caveolae. The expression of GM1, a marker of lipid rafts, localized primarily to fraction nos. 4–5 in unstimulated macrophages, and increased in these fractions in response to LPS stimulation. The transferrin receptor, which does not localize to lipid rafts or caveolae, localized to fraction nos. 8–12. Similar results with respect to the localization of cav-1 and GM-1 were obtained by repetition of the experiments with thioglycolate-elicited macrophages, or by repetition of the gradient in the absence of Triton X-100 (see supplemental Figs. S3 and S4).⁴

A portion of HO-1 induced by LPS localized to the caveolae fractions obtained from sucrose density gradient ultracentrifugation of peritoneal macrophage cell lysates. Cav-1 also cofractionated with HO-1 in fraction nos. 10–11 under basal or LPS-stimulated conditions. We also examined the localization of enzymes associated with HO activity in macrophages. NADPH cytochrome p450 reductase and BVR were detected in the caveolae fraction nos. 4–5 of macrophages and their relative expression in this compartment increased after LPS stimulation (Fig. 2,A). Consistent with our previous observation of TLR4 localization to lipid rafts (16), TLR4 expression was detectable in fraction nos. 4–5 isolated from RAW 264.7 cells and increased markedly upon LPS stimulation (Fig. 2 B).

To further confirm our findings, we analyzed detergent-resistant membrane fractions from peritoneal macrophages. Because caveolae are rich in glycosphingolipids, sphingomyelin, and cholesterol, they resist solubilization by mild detergent (1). Cav-1 occurred in the detergent-resistant fraction of peritoneal macrophages under basal and LPS-induced conditions (Fig. 2,C). Depletion of cholesterol in the plasma membrane by treatment with methyl-β-cyclodextrin, which disrupts caveolae structure (20), suppressed the appearance of HO-1, as well as cav-1 and GM1 in the detergent-resistant fraction (Fig. 2,C). Disruption of the Golgi complex by treatment of peritoneal macrophages with brefeldin-A inhibited the trafficking of HO-1 to caveolae (Fig. 2 D).

Compartmentalization of HO-1 and cav-1 proteins was also determined under confocal microscopy. Fig. 3,A depicts the typical punctate plasma membrane immunostaining of cav-1 in peritoneal macrophages (27). A portion of cav-1 staining occurred in the perinuclear region, which could be attributed to endoplasmic reticulum and Golgi localization. HO-1 also colocalized with cav-1 in the perinuclear region under basal conditions. After LPS stimulation, HO-1 demonstrated a similar punctate fluorescence pattern at the plasma membrane, in addition to the cytosol. Colocalization, as indicated by the yellow color in the merged images was found on either the plasma membrane or endoplasmic reticulum and Golgi (Fig. 3,A, right). The caveolae localization of the HO-1 protein was further confirmed using immunogold labeling on electron microscopy (Fig. 3,B). The HO activity in the caveolae was investigated by analyzing the caveolae fraction nos. 4–5 from sucrose gradient ultracentrifugation of cell lysates. LPS-inducible HO activity appeared in caveolae, which could be abolished by preincubation of cells with Sn-PPIX, a competitive inhibitor of HO activity (Fig. 3,C), indicating a functional compartmentalization of HO activity to the caveolae of macrophages. Under basal conditions, HO activity in caveolae was negligible, similar to that of heat-denatured protein. We also observed the caveolae localization of HO-1 in vivo. HO-1 was detected in caveolae factions isolated from lung and liver tissue derived from LPS-treated mice, but not from that of control mice (Fig. 3 D).

Because HO-1 localized to caveolae in macrophages, and is associated with anti-inflammatory effects, we examined the effect of HO-1 expression on the cav-1/TLR4 interaction. The expression of HO-1 by infection of peritoneal macrophages with adenovirus HO-1 augmented the TLR4/cav-1 interaction (Fig. 4^,A). The relative expression from the HO-1 construct is shown (see supplemental Fig. S2C).⁴ Both hemin and LPS, classical inducers of HO-1 expression, also stimulated the interaction of cav-1 and TLR4 (Fig. 4 B).

The MKK3/p38 MAPK pathway has been previously implicated in the regulation of proinflammatory signaling in macrophages (13, 19). We examined the role of this pathway in TLR4/cav-1 complex formation. SB203580, an inhibitor of p38α/β MAPK inhibited the formation of the LPS-inducible but not the hemin-inducible cav-1/TLR4 complex, suggesting a selective role for p38α/β MAPK in the LPS-dependent response (Fig. 4 B).

The subcellular trafficking of HO-1 to the cell surface is heretofore uncharacterized, therefore nothing is known about the underlying regulatory mechanisms. We examined the mechanisms by which HO-1 translocates to the cell surface upon LPS stimulation. PKC and p38 MAPK were investigated. PKC has been reported to regulate membrane trafficking (28, 29), whereas p38 MAPK is known to play a critical role in HO-1/CO-dependent anti-inflammatory functions, though nothing is known of its role in the trafficking of molecules to the lipid raft (19). Both PKC and the p38 MAPK pathway are highly activated by LPS (19, 30, 31).

We isolated peritoneal macrophages from mice genetically deficient in MKK3, the upstream kinase of p38 MAPK. In MKK3^−/− macrophages, the translocation of HO-1 to the caveolae induced by LPS was significantly diminished compared with that of the wild-type cells (Fig. 4,C). Interestingly, hemin, the substrate and inducer of HO-1, caused caveolae translocation of HO-1 independent of MKK3 status (Fig. 4,C), indicating the unique role of MKK3/p38 MAPK in mediating LPS-inducible HO-1 trafficking. Inhibition of p38 MAPK by its selective inhibitor SB203580 strikingly attenuated the appearance of HO-1 in caveolae fractions (Fig. 4,C, right). This result cannot be attributed solely to a general inhibition of total HO-1 production due to SB203580 administration because the reduction of HO-1 in caveolae was more dramatic compared with that in whole cell lysates. Conversely, inhibition of PKC by the chemical inhibitor GF109203X administered before LPS stimulation did not affect the translocation of HO-1 (Fig. 4 C, right).

CO, the metabolic byproduct of HO-1 activity, exerts similar anti-inflammatory effects as cav-1 with respect to inhibition of LPS-inducible cytokine production (e.g., TNF-α, IL-6, GM-CSF, and IL-1β) in macrophages (19, 32, 33). We have recently shown that exogenous CO inhibited the trafficking of TLR4 to the lipid raft after LPS stimulation in macrophages (16). In this experiment, we hypothesized that CO regulates inflammation initiated in caveolae microdomains. We examined the effect of CO on the association between TLR4 and cav-1 in vitro and in vivo. Exogenous CO treatment at a dose of 250 ppm markedly increased the interaction between cav-1 and TLR4 in macrophages (Fig. 5,A). Exposure of mice to CO by inhalation (250 ppm) for 24 h increased the interaction between TLR4 and cav-1 in mouse lung tissue (Fig. 5 B).

CO administration suppressed TNF-α and IL-6 production in macrophages (derived from C3H/HeJ mice) transfected with wild-type TLR4, whereas CO administration had no apparent effects on cells transfected with mutant W744A TLR4 (Fig. 5^,C). The relative expression from the TLR4 constructs is shown (see supplemental Fig. S2D).⁴ Furthermore, CO treatment (250 ppm) attenuated the LPS-inducible protein-protein interaction between TLR4 and its binding partner, MyD88 (Fig. 5,D), in macrophages infected with control siRNA. The inhibitory effect of CO on TLR4/MyD88 interaction was abolished in macrophages infected with cav-1 siRNA (Fig. 5 D). We further examined this relationship in vivo using cav-1^−/− mice. Wild-type or cav-1^−/− mice were subjected to LPS exposure following pretreatment with air or CO (250 ppm). Peritoneal macrophages were recovered by lavage and analyzed for TLR4/MyD88 interaction. CO pretreatment in vivo inhibited LPS-inducible TLR4/MyD88 complex formation in the peritoneal macrophages of wild-type mice, but did not inhibit this complex formation in cav-1^−/− mice. These results are consistent with the proposed role of cav-1 in the inhibitory effects of CO on TLR4 signaling.

We further evaluated the essential role of cav-1 in mediating the anti-inflammatory effects of CO. CO treatment (250 ppm) markedly suppressed LPS-induced TNF-α and IL-6 production in wild-type macrophages, whereas CO failed to exert an anti-inflammatory effect in cav-1^−/− macrophages (Fig. 6).

Caveolae and cav-1 exert an anti-inflammatory function by facilitating the endocytosis of bacteria (3, 4, 5). Cav-1 modulates inflammation, vascular permeability and leukocyte migration through the regulation of endothelial NOS and inhibition of its activity (10, 11, 12). NO derived from endothelial NOS is essential for maintenance of vascular permeability, inflammatory cells infiltration, and direct regulation of inflammatory proteins (34, 35). However, some controversy has arisen as to whether cav-1 exerts anti-inflammatory or proinflammatory effects with respect to endothelial NOS regulation (10, 11, 12). In vivo administration of peptides containing the scaffolding domain of cav-1 inhibited edema formation, inflammation, and vascular leakage to the same extent as glucocorticoid treatment in rats (10). In contrast, Garrean et al. (11) observed decreased polymorphonuclear leukocyte adhesion and microvascular permeability and edema formation in cav-1 null mice after LPS exposure, relative to wild-type mice. The down-regulation of cav-1 combined with up-regulation of endothelial NOS was also shown to decrease leukocyte adhesion in pial venules of ovariectomized female rats (12).

It is now widely accepted that cytokines represent key extracellular signaling proteins underlying the pathogenesis of various pulmonary diseases (36, 37, 38). For example, the proinflammatory cytokine TNF-α has been implicated in the progression of asthma, chronic bronchitis, chronic obstructive pulmonary disease, acute lung injury, and acute respiratory distress syndrome (39). IL-6, IL-1β, and TNF-α increase in the sputum of patients with chronic obstructive pulmonary disease, and contribute to small airway inflammation (37). TNF-α and IL-1β stimulate the production of matrix metalloproteinase-9, a critical factor in tissue remodeling (37). Both TNF-α and GM-CSF inhibit neutrophil apoptosis, and therefore promote neutrophil influx and accumulation. RANTES, implicated in the pathogenesis of asthma, activates eosinophils which release toxic basic proteins and lipid mediators that cause bronchial epithelial damage and airflow obstruction (40). Corticosteroids, as anti-inflammatory drugs, remain the most common therapy for pulmonary diseases such as asthma, chronic obstructive pulmonary disease, and interstitial pulmonary fibrosis (41). Therapeutic strategies targeting cytokines represent an important breakthrough in the treatment of diseases characterized by inflammation and cell death (40). For example, anti-TNF-α therapies conferred protection against ventilator-induced lung injury in rats, symptomatic corticosteroid-dependent asthma in humans, and bronchiolitis obliterans in a bone marrow transplant recipient (39). Neutralization of GM-CSF by Ab therapy ameliorates experimental chronic obstructive pulmonary disease by reducing proinflammatory cell influx in the alveoli and airways, as well as whole lung cytokine, chemokine, and protease expression (42). We demonstrate in this study that cav-1 inhibits the production of the major proinflammatory cytokines including IL-1β, IL-2, IL-4, IL-12, GM-CSF, TNF-α, and RANTES, thus identifying cav-1 as a promising novel target for therapeutic intervention.

We have previously observed that TLR4 traffics to lipid rafts upon LPS stimulation in RAW 264.7 cells (16). We show that TLR4, the key receptor for LPS, highly colocalized with cav-1 in peritoneal macrophages in caveolae (rafts) fractions, and also in high density nonraft fractions. These findings strongly suggest that regulation of TLR4 function may occur within caveolae or lipid raft microdomains. Consistently, TLR4 colocalization with cav-1 was previously shown in human aortic endothelial cells (17). We have identified a cav-1 binding motif (24) in the amino acid sequence of murine TLR4 ((⁷³⁹)FIQSRWCIF(⁷⁴⁷)). The caveolin-binding consensus motifs are as follows: ΦXΦXXXXΦ, ΦXXXXΦXXΦ, and ΦXΦXXXXΦXXΦ, where Φ indicates an aromatic residue (F, W, or Y), and X indicates any residue (1, 24). We show for the first time to our knowledge that cav-1 directly interacted with TLR4 by binding to this site and functionally suppressed TLR4 complex assembly with MyD88 or TRIF.

In addition, we demonstrate that cav-1 diminished the LPS-inducible nuclear translocation of NF-κB p65. The interaction of cav-1 with TLR4 was mediated by the cav-1 binding motif in TLR4 because the W744A mutation of this binding site abolished the interaction, as well as reversed the inhibitory effect of cav-1 on cytokines (TNF-α and IL-6) regulation. Our results are consistent with those of Medina et al. (9), who observed increased inflammatory responses to LPS in vitro, including cytokines and NO production, in macrophages derived from cav-1 null mice. In contrast to our current findings, Garrean et al. (11) reported that NF-κB DNA binding activity in response to LPS was attenuated in lung tissue from cav-1 null mice, whereas the NOS inhibitor nitro-l-arginine reversed the suppression of NF-κB activity. Inhibition of NF-κB was attributed to elevated NOS activity in cav-1 null mice (11). However, this study used LPS dosages one order of magnitude higher (1 mg/ml in cells) than those used in the current study. It is well known that low concentrations of exogenous NO up-regulate NF-κB activation, whereas high concentrations inhibit NF-κB activation (35). Our observations that cav-1 overexpression regulated NF-κB in vitro used the macrophage model, whereas Garrean et al. (11) analyzed total lung tissue from cav-1 null mice after LPS-treatment in vivo. The effects of endogenous or exogenous NO production on TLR4 trafficking and its regulation by cav-1 may be of great interest and warrant further examination.

CO, a diatomic gas with well-described toxic potential at high concentration, has been implicated as a potential endogenous signaling molecule in a variety of cellular processes including vasoregulation and neurotransmission (18). In many models, exogenously administration of CO at low concentration (250 ppm) can modulate inflammatory responses (18). Interestingly, cav-1 and CO display remarkable similarities with respect to cytokine regulation. CO inhibited LPS-inducible TNF-α and IL-6 productions and increased IL-10 production in RAW 264.7 cells and mice (19, 33). The anti-inflammatory effects of CO, as well as of cav-1 expression, depended on the MKK3/p38 MAPK pathway (13, 19). CO was also reported to regulate IL-2, IL-1β, and GM-CSF production (32, 33, 43), again similar to the effects of cav-1. Endogenous CO arises primarily from HO catalyzed heme degradation, which is thought to account for more than 85% of the CO production in the body (18). Traditionally, HO-1 has been characterized as an endoplasmic reticulum-associated protein because its activity is enriched in microsomal (104,000 g) fractions. The C-terminal hydrophobic domain of HO-1 promotes its membrane compartmentalization (18). Recently, HO-1 was also found to localize in several distinct cellular compartments, including caveolae and mitochondria of pulmonary endothelial cells (18, 20). Thus HO-1 displays a broader subcellular distribution than originally assumed, opening possibilities for organelle or compartment-specific functions.

We found that HO-1 trafficking into caveolae after LPS stimulation depended on the MKK3/p38 MAPK pathway. The caveolae trafficking of HO-1 was blocked by β-cyclodextrin and brefeldin A, indicating it is specific and requires endoplasmic reticulum and Golgi processing. Moreover, HO activity assays revealed functional HO-1 in macrophage caveolae, indicating the potential for CO generation in this compartment. We also demonstrated that exogenous administration of CO, as well as adenoviral-mediated HO-1 expression, presumably by increasing endogenous CO generation, enhanced the interaction between cav-1 and TLR4, leading to down-regulation of cytokine production. In summary, our findings propose a new model for the regulation of inflammatory signaling in the LPS model (Fig. 7). Cav-1 confers its anti-inflammatory effects through direct binding of TLR4, which prevents TLR4 association with MyD88 and TRIF, and downstream activation of the NF-κB pathway. Cav-1 was previously reported to activate p38 MAPK (13), which was required for translocation of HO-1 to caveolae. CO is generated locally in caveolae from HO activity, which increases the interaction of cav-1 with TLR4 to further attenuate the TLR4 signal.

These studies further our understanding of the mechanisms underlying the function of HO-1/CO in cellular defense mechanisms. The potential relationships between NO and cav-1 dependent regulation of TLR4, however, remain to be explored. Importantly, these results identify cav-1 and its interaction with membrane receptors (e.g., TLR4), and other caveolae-resident proteins such as HO-1, as novel regulatory targets for therapeutic intervention in inflammatory diseases.

We thank Emeka Ifedigbo for animal handling, Dr. Yingze Zhang for assistance with the Bioplex Multiplex cytokine assays, and Dr. Timothy Billiar for providing the plasmid pCMV-TLR4 tagged with Flag. We also thank Dr. Simon Watkins and the Center for Biological Imaging (Department of Cell Biology and Physiology, School of Medicine, University of Pittsburgh, PA) for valuable technical assistance.

The authors have no financial conflict of interest.